diff options
| author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-05-04 12:47:55 +0000 |
|---|---|---|
| committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-05-04 12:47:55 +0000 |
| commit | 2aadc03ef15cb5ca5cc2af8a7c08e070742f0ac4 (patch) | |
| tree | 033cc839730fda84ff08db877037977be94e5e3a /crates | |
| parent | Initial commit. (diff) | |
| download | cargo-2aadc03ef15cb5ca5cc2af8a7c08e070742f0ac4.tar.xz cargo-2aadc03ef15cb5ca5cc2af8a7c08e070742f0ac4.zip | |
Adding upstream version 0.70.1+ds1.upstream/0.70.1+ds1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'crates')
110 files changed, 16882 insertions, 0 deletions
diff --git a/crates/cargo-platform/Cargo.toml b/crates/cargo-platform/Cargo.toml new file mode 100644 index 0000000..9a31170 --- /dev/null +++ b/crates/cargo-platform/Cargo.toml @@ -0,0 +1,12 @@ +[package] +name = "cargo-platform" +version = "0.1.2" +edition = "2021" +license = "MIT OR Apache-2.0" +homepage = "https://github.com/rust-lang/cargo" +repository = "https://github.com/rust-lang/cargo" +documentation = "https://docs.rs/cargo-platform" +description = "Cargo's representation of a target platform." + +[dependencies] +serde = { version = "1.0.82", features = ['derive'] } diff --git a/crates/cargo-platform/LICENSE-APACHE b/crates/cargo-platform/LICENSE-APACHE new file mode 120000 index 0000000..1cd601d --- /dev/null +++ b/crates/cargo-platform/LICENSE-APACHE @@ -0,0 +1 @@ +../../LICENSE-APACHE
\ No newline at end of file diff --git a/crates/cargo-platform/LICENSE-MIT b/crates/cargo-platform/LICENSE-MIT new file mode 120000 index 0000000..b2cfbdc --- /dev/null +++ b/crates/cargo-platform/LICENSE-MIT @@ -0,0 +1 @@ +../../LICENSE-MIT
\ No newline at end of file diff --git a/crates/cargo-platform/examples/matches.rs b/crates/cargo-platform/examples/matches.rs new file mode 100644 index 0000000..9ad5d10 --- /dev/null +++ b/crates/cargo-platform/examples/matches.rs @@ -0,0 +1,55 @@ +//! This example demonstrates how to filter a Platform based on the current +//! host target. + +use cargo_platform::{Cfg, Platform}; +use std::process::Command; +use std::str::FromStr; + +static EXAMPLES: &[&str] = &[ + "cfg(windows)", + "cfg(unix)", + "cfg(target_os=\"macos\")", + "cfg(target_os=\"linux\")", + "cfg(any(target_arch=\"x86\", target_arch=\"x86_64\"))", +]; + +fn main() { + let target = get_target(); + let cfgs = get_cfgs(); + println!("host target={} cfgs:", target); + for cfg in &cfgs { + println!(" {}", cfg); + } + let mut examples: Vec<&str> = EXAMPLES.iter().copied().collect(); + examples.push(target.as_str()); + for example in examples { + let p = Platform::from_str(example).unwrap(); + println!("{:?} matches: {:?}", example, p.matches(&target, &cfgs)); + } +} + +fn get_target() -> String { + let output = Command::new("rustc") + .arg("-Vv") + .output() + .expect("rustc failed to run"); + let stdout = String::from_utf8(output.stdout).unwrap(); + for line in stdout.lines() { + if line.starts_with("host: ") { + return String::from(&line[6..]); + } + } + panic!("Failed to find host: {}", stdout); +} + +fn get_cfgs() -> Vec<Cfg> { + let output = Command::new("rustc") + .arg("--print=cfg") + .output() + .expect("rustc failed to run"); + let stdout = String::from_utf8(output.stdout).unwrap(); + stdout + .lines() + .map(|line| Cfg::from_str(line).unwrap()) + .collect() +} diff --git a/crates/cargo-platform/src/cfg.rs b/crates/cargo-platform/src/cfg.rs new file mode 100644 index 0000000..c3ddb69 --- /dev/null +++ b/crates/cargo-platform/src/cfg.rs @@ -0,0 +1,319 @@ +use crate::error::{ParseError, ParseErrorKind::*}; +use std::fmt; +use std::iter; +use std::str::{self, FromStr}; + +/// A cfg expression. +#[derive(Eq, PartialEq, Hash, Ord, PartialOrd, Clone, Debug)] +pub enum CfgExpr { + Not(Box<CfgExpr>), + All(Vec<CfgExpr>), + Any(Vec<CfgExpr>), + Value(Cfg), +} + +/// A cfg value. +#[derive(Eq, PartialEq, Hash, Ord, PartialOrd, Clone, Debug)] +pub enum Cfg { + /// A named cfg value, like `unix`. + Name(String), + /// A key/value cfg pair, like `target_os = "linux"`. + KeyPair(String, String), +} + +#[derive(PartialEq)] +enum Token<'a> { + LeftParen, + RightParen, + Ident(&'a str), + Comma, + Equals, + String(&'a str), +} + +#[derive(Clone)] +struct Tokenizer<'a> { + s: iter::Peekable<str::CharIndices<'a>>, + orig: &'a str, +} + +struct Parser<'a> { + t: Tokenizer<'a>, +} + +impl FromStr for Cfg { + type Err = ParseError; + + fn from_str(s: &str) -> Result<Cfg, Self::Err> { + let mut p = Parser::new(s); + let e = p.cfg()?; + if let Some(rest) = p.rest() { + return Err(ParseError::new( + p.t.orig, + UnterminatedExpression(rest.to_string()), + )); + } + Ok(e) + } +} + +impl fmt::Display for Cfg { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + Cfg::Name(ref s) => s.fmt(f), + Cfg::KeyPair(ref k, ref v) => write!(f, "{} = \"{}\"", k, v), + } + } +} + +impl CfgExpr { + /// Utility function to check if the key, "cfg(..)" matches the `target_cfg` + pub fn matches_key(key: &str, target_cfg: &[Cfg]) -> bool { + if key.starts_with("cfg(") && key.ends_with(')') { + let cfg = &key[4..key.len() - 1]; + + CfgExpr::from_str(cfg) + .ok() + .map(|ce| ce.matches(target_cfg)) + .unwrap_or(false) + } else { + false + } + } + + pub fn matches(&self, cfg: &[Cfg]) -> bool { + match *self { + CfgExpr::Not(ref e) => !e.matches(cfg), + CfgExpr::All(ref e) => e.iter().all(|e| e.matches(cfg)), + CfgExpr::Any(ref e) => e.iter().any(|e| e.matches(cfg)), + CfgExpr::Value(ref e) => cfg.contains(e), + } + } +} + +impl FromStr for CfgExpr { + type Err = ParseError; + + fn from_str(s: &str) -> Result<CfgExpr, Self::Err> { + let mut p = Parser::new(s); + let e = p.expr()?; + if let Some(rest) = p.rest() { + return Err(ParseError::new( + p.t.orig, + UnterminatedExpression(rest.to_string()), + )); + } + Ok(e) + } +} + +impl fmt::Display for CfgExpr { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + CfgExpr::Not(ref e) => write!(f, "not({})", e), + CfgExpr::All(ref e) => write!(f, "all({})", CommaSep(e)), + CfgExpr::Any(ref e) => write!(f, "any({})", CommaSep(e)), + CfgExpr::Value(ref e) => write!(f, "{}", e), + } + } +} + +struct CommaSep<'a, T>(&'a [T]); + +impl<'a, T: fmt::Display> fmt::Display for CommaSep<'a, T> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + for (i, v) in self.0.iter().enumerate() { + if i > 0 { + write!(f, ", ")?; + } + write!(f, "{}", v)?; + } + Ok(()) + } +} + +impl<'a> Parser<'a> { + fn new(s: &'a str) -> Parser<'a> { + Parser { + t: Tokenizer { + s: s.char_indices().peekable(), + orig: s, + }, + } + } + + fn expr(&mut self) -> Result<CfgExpr, ParseError> { + match self.peek() { + Some(Ok(Token::Ident(op @ "all"))) | Some(Ok(Token::Ident(op @ "any"))) => { + self.t.next(); + let mut e = Vec::new(); + self.eat(&Token::LeftParen)?; + while !self.r#try(&Token::RightParen) { + e.push(self.expr()?); + if !self.r#try(&Token::Comma) { + self.eat(&Token::RightParen)?; + break; + } + } + if op == "all" { + Ok(CfgExpr::All(e)) + } else { + Ok(CfgExpr::Any(e)) + } + } + Some(Ok(Token::Ident("not"))) => { + self.t.next(); + self.eat(&Token::LeftParen)?; + let e = self.expr()?; + self.eat(&Token::RightParen)?; + Ok(CfgExpr::Not(Box::new(e))) + } + Some(Ok(..)) => self.cfg().map(CfgExpr::Value), + Some(Err(..)) => Err(self.t.next().unwrap().err().unwrap()), + None => Err(ParseError::new( + self.t.orig, + IncompleteExpr("start of a cfg expression"), + )), + } + } + + fn cfg(&mut self) -> Result<Cfg, ParseError> { + match self.t.next() { + Some(Ok(Token::Ident(name))) => { + let e = if self.r#try(&Token::Equals) { + let val = match self.t.next() { + Some(Ok(Token::String(s))) => s, + Some(Ok(t)) => { + return Err(ParseError::new( + self.t.orig, + UnexpectedToken { + expected: "a string", + found: t.classify(), + }, + )) + } + Some(Err(e)) => return Err(e), + None => { + return Err(ParseError::new(self.t.orig, IncompleteExpr("a string"))) + } + }; + Cfg::KeyPair(name.to_string(), val.to_string()) + } else { + Cfg::Name(name.to_string()) + }; + Ok(e) + } + Some(Ok(t)) => Err(ParseError::new( + self.t.orig, + UnexpectedToken { + expected: "identifier", + found: t.classify(), + }, + )), + Some(Err(e)) => Err(e), + None => Err(ParseError::new(self.t.orig, IncompleteExpr("identifier"))), + } + } + + fn peek(&mut self) -> Option<Result<Token<'a>, ParseError>> { + self.t.clone().next() + } + + fn r#try(&mut self, token: &Token<'a>) -> bool { + match self.peek() { + Some(Ok(ref t)) if token == t => {} + _ => return false, + } + self.t.next(); + true + } + + fn eat(&mut self, token: &Token<'a>) -> Result<(), ParseError> { + match self.t.next() { + Some(Ok(ref t)) if token == t => Ok(()), + Some(Ok(t)) => Err(ParseError::new( + self.t.orig, + UnexpectedToken { + expected: token.classify(), + found: t.classify(), + }, + )), + Some(Err(e)) => Err(e), + None => Err(ParseError::new( + self.t.orig, + IncompleteExpr(token.classify()), + )), + } + } + + /// Returns the rest of the input from the current location. + fn rest(&self) -> Option<&str> { + let mut s = self.t.s.clone(); + loop { + match s.next() { + Some((_, ' ')) => {} + Some((start, _ch)) => return Some(&self.t.orig[start..]), + None => return None, + } + } + } +} + +impl<'a> Iterator for Tokenizer<'a> { + type Item = Result<Token<'a>, ParseError>; + + fn next(&mut self) -> Option<Result<Token<'a>, ParseError>> { + loop { + match self.s.next() { + Some((_, ' ')) => {} + Some((_, '(')) => return Some(Ok(Token::LeftParen)), + Some((_, ')')) => return Some(Ok(Token::RightParen)), + Some((_, ',')) => return Some(Ok(Token::Comma)), + Some((_, '=')) => return Some(Ok(Token::Equals)), + Some((start, '"')) => { + while let Some((end, ch)) = self.s.next() { + if ch == '"' { + return Some(Ok(Token::String(&self.orig[start + 1..end]))); + } + } + return Some(Err(ParseError::new(self.orig, UnterminatedString))); + } + Some((start, ch)) if is_ident_start(ch) => { + while let Some(&(end, ch)) = self.s.peek() { + if !is_ident_rest(ch) { + return Some(Ok(Token::Ident(&self.orig[start..end]))); + } else { + self.s.next(); + } + } + return Some(Ok(Token::Ident(&self.orig[start..]))); + } + Some((_, ch)) => { + return Some(Err(ParseError::new(self.orig, UnexpectedChar(ch)))); + } + None => return None, + } + } + } +} + +fn is_ident_start(ch: char) -> bool { + ch == '_' || ch.is_ascii_alphabetic() +} + +fn is_ident_rest(ch: char) -> bool { + is_ident_start(ch) || ch.is_ascii_digit() +} + +impl<'a> Token<'a> { + fn classify(&self) -> &'static str { + match *self { + Token::LeftParen => "`(`", + Token::RightParen => "`)`", + Token::Ident(..) => "an identifier", + Token::Comma => "`,`", + Token::Equals => "`=`", + Token::String(..) => "a string", + } + } +} diff --git a/crates/cargo-platform/src/error.rs b/crates/cargo-platform/src/error.rs new file mode 100644 index 0000000..bf4b35f --- /dev/null +++ b/crates/cargo-platform/src/error.rs @@ -0,0 +1,67 @@ +use std::fmt; + +#[derive(Debug)] +pub struct ParseError { + kind: ParseErrorKind, + orig: String, +} + +#[non_exhaustive] +#[derive(Debug)] +pub enum ParseErrorKind { + UnterminatedString, + UnexpectedChar(char), + UnexpectedToken { + expected: &'static str, + found: &'static str, + }, + IncompleteExpr(&'static str), + UnterminatedExpression(String), + InvalidTarget(String), +} + +impl fmt::Display for ParseError { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + write!( + f, + "failed to parse `{}` as a cfg expression: {}", + self.orig, self.kind + ) + } +} + +impl fmt::Display for ParseErrorKind { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + use ParseErrorKind::*; + match self { + UnterminatedString => write!(f, "unterminated string in cfg"), + UnexpectedChar(ch) => write!( + f, + "unexpected character `{}` in cfg, expected parens, a comma, \ + an identifier, or a string", + ch + ), + UnexpectedToken { expected, found } => { + write!(f, "expected {}, found {}", expected, found) + } + IncompleteExpr(expected) => { + write!(f, "expected {}, but cfg expression ended", expected) + } + UnterminatedExpression(s) => { + write!(f, "unexpected content `{}` found after cfg expression", s) + } + InvalidTarget(s) => write!(f, "invalid target specifier: {}", s), + } + } +} + +impl std::error::Error for ParseError {} + +impl ParseError { + pub fn new(orig: &str, kind: ParseErrorKind) -> ParseError { + ParseError { + kind, + orig: orig.to_string(), + } + } +} diff --git a/crates/cargo-platform/src/lib.rs b/crates/cargo-platform/src/lib.rs new file mode 100644 index 0000000..0a3dcf1 --- /dev/null +++ b/crates/cargo-platform/src/lib.rs @@ -0,0 +1,146 @@ +//! Platform definition used by Cargo. +//! +//! This defines a [`Platform`] type which is used in Cargo to specify a target platform. +//! There are two kinds, a named target like `x86_64-apple-darwin`, and a "cfg expression" +//! like `cfg(any(target_os = "macos", target_os = "ios"))`. +//! +//! See `examples/matches.rs` for an example of how to match against a `Platform`. +//! +//! [`Platform`]: enum.Platform.html + +use std::fmt; +use std::str::FromStr; + +mod cfg; +mod error; + +pub use cfg::{Cfg, CfgExpr}; +pub use error::{ParseError, ParseErrorKind}; + +/// Platform definition. +#[derive(Eq, PartialEq, Hash, Ord, PartialOrd, Clone, Debug)] +pub enum Platform { + /// A named platform, like `x86_64-apple-darwin`. + Name(String), + /// A cfg expression, like `cfg(windows)`. + Cfg(CfgExpr), +} + +impl Platform { + /// Returns whether the Platform matches the given target and cfg. + /// + /// The named target and cfg values should be obtained from `rustc`. + pub fn matches(&self, name: &str, cfg: &[Cfg]) -> bool { + match *self { + Platform::Name(ref p) => p == name, + Platform::Cfg(ref p) => p.matches(cfg), + } + } + + fn validate_named_platform(name: &str) -> Result<(), ParseError> { + if let Some(ch) = name + .chars() + .find(|&c| !(c.is_alphanumeric() || c == '_' || c == '-' || c == '.')) + { + if name.chars().any(|c| c == '(') { + return Err(ParseError::new( + name, + ParseErrorKind::InvalidTarget( + "unexpected `(` character, cfg expressions must start with `cfg(`" + .to_string(), + ), + )); + } + return Err(ParseError::new( + name, + ParseErrorKind::InvalidTarget(format!( + "unexpected character {} in target name", + ch + )), + )); + } + Ok(()) + } + + pub fn check_cfg_attributes(&self, warnings: &mut Vec<String>) { + fn check_cfg_expr(expr: &CfgExpr, warnings: &mut Vec<String>) { + match *expr { + CfgExpr::Not(ref e) => check_cfg_expr(e, warnings), + CfgExpr::All(ref e) | CfgExpr::Any(ref e) => { + for e in e { + check_cfg_expr(e, warnings); + } + } + CfgExpr::Value(ref e) => match e { + Cfg::Name(name) => match name.as_str() { + "test" | "debug_assertions" | "proc_macro" => + warnings.push(format!( + "Found `{}` in `target.'cfg(...)'.dependencies`. \ + This value is not supported for selecting dependencies \ + and will not work as expected. \ + To learn more visit \ + https://doc.rust-lang.org/cargo/reference/specifying-dependencies.html#platform-specific-dependencies", + name + )), + _ => (), + }, + Cfg::KeyPair(name, _) => if name.as_str() == "feature" { + warnings.push(String::from( + "Found `feature = ...` in `target.'cfg(...)'.dependencies`. \ + This key is not supported for selecting dependencies \ + and will not work as expected. \ + Use the [features] section instead: \ + https://doc.rust-lang.org/cargo/reference/features.html" + )) + }, + } + } + } + + if let Platform::Cfg(cfg) = self { + check_cfg_expr(cfg, warnings); + } + } +} + +impl serde::Serialize for Platform { + fn serialize<S>(&self, s: S) -> Result<S::Ok, S::Error> + where + S: serde::Serializer, + { + self.to_string().serialize(s) + } +} + +impl<'de> serde::Deserialize<'de> for Platform { + fn deserialize<D>(deserializer: D) -> Result<Self, D::Error> + where + D: serde::Deserializer<'de>, + { + let s = String::deserialize(deserializer)?; + FromStr::from_str(&s).map_err(serde::de::Error::custom) + } +} + +impl FromStr for Platform { + type Err = ParseError; + + fn from_str(s: &str) -> Result<Platform, ParseError> { + if s.starts_with("cfg(") && s.ends_with(')') { + let s = &s[4..s.len() - 1]; + s.parse().map(Platform::Cfg) + } else { + Platform::validate_named_platform(s)?; + Ok(Platform::Name(s.to_string())) + } + } +} + +impl fmt::Display for Platform { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + Platform::Name(ref n) => n.fmt(f), + Platform::Cfg(ref e) => write!(f, "cfg({})", e), + } + } +} diff --git a/crates/cargo-platform/tests/test_cfg.rs b/crates/cargo-platform/tests/test_cfg.rs new file mode 100644 index 0000000..dd99d9a --- /dev/null +++ b/crates/cargo-platform/tests/test_cfg.rs @@ -0,0 +1,251 @@ +use cargo_platform::{Cfg, CfgExpr, Platform}; +use std::fmt; +use std::str::FromStr; + +macro_rules! c { + ($a:ident) => { + Cfg::Name(stringify!($a).to_string()) + }; + ($a:ident = $e:expr) => { + Cfg::KeyPair(stringify!($a).to_string(), $e.to_string()) + }; +} + +macro_rules! e { + (any($($t:tt),*)) => (CfgExpr::Any(vec![$(e!($t)),*])); + (all($($t:tt),*)) => (CfgExpr::All(vec![$(e!($t)),*])); + (not($($t:tt)*)) => (CfgExpr::Not(Box::new(e!($($t)*)))); + (($($t:tt)*)) => (e!($($t)*)); + ($($t:tt)*) => (CfgExpr::Value(c!($($t)*))); +} + +fn good<T>(s: &str, expected: T) +where + T: FromStr + PartialEq + fmt::Debug, + T::Err: fmt::Display, +{ + let c = match T::from_str(s) { + Ok(c) => c, + Err(e) => panic!("failed to parse `{}`: {}", s, e), + }; + assert_eq!(c, expected); +} + +fn bad<T>(s: &str, err: &str) +where + T: FromStr + fmt::Display, + T::Err: fmt::Display, +{ + let e = match T::from_str(s) { + Ok(cfg) => panic!("expected `{}` to not parse but got {}", s, cfg), + Err(e) => e.to_string(), + }; + assert!( + e.contains(err), + "when parsing `{}`,\n\"{}\" not contained \ + inside: {}", + s, + err, + e + ); +} + +#[test] +fn cfg_syntax() { + good("foo", c!(foo)); + good("_bar", c!(_bar)); + good(" foo", c!(foo)); + good(" foo ", c!(foo)); + good(" foo = \"bar\"", c!(foo = "bar")); + good("foo=\"\"", c!(foo = "")); + good(" foo=\"3\" ", c!(foo = "3")); + good("foo = \"3 e\"", c!(foo = "3 e")); +} + +#[test] +fn cfg_syntax_bad() { + bad::<Cfg>("", "but cfg expression ended"); + bad::<Cfg>(" ", "but cfg expression ended"); + bad::<Cfg>("\t", "unexpected character"); + bad::<Cfg>("7", "unexpected character"); + bad::<Cfg>("=", "expected identifier"); + bad::<Cfg>(",", "expected identifier"); + bad::<Cfg>("(", "expected identifier"); + bad::<Cfg>("foo (", "unexpected content `(` found after cfg expression"); + bad::<Cfg>("bar =", "expected a string"); + bad::<Cfg>("bar = \"", "unterminated string"); + bad::<Cfg>( + "foo, bar", + "unexpected content `, bar` found after cfg expression", + ); +} + +#[test] +fn cfg_expr() { + good("foo", e!(foo)); + good("_bar", e!(_bar)); + good(" foo", e!(foo)); + good(" foo ", e!(foo)); + good(" foo = \"bar\"", e!(foo = "bar")); + good("foo=\"\"", e!(foo = "")); + good(" foo=\"3\" ", e!(foo = "3")); + good("foo = \"3 e\"", e!(foo = "3 e")); + + good("all()", e!(all())); + good("all(a)", e!(all(a))); + good("all(a, b)", e!(all(a, b))); + good("all(a, )", e!(all(a))); + good("not(a = \"b\")", e!(not(a = "b"))); + good("not(all(a))", e!(not(all(a)))); +} + +#[test] +fn cfg_expr_bad() { + bad::<CfgExpr>(" ", "but cfg expression ended"); + bad::<CfgExpr>(" all", "expected `(`"); + bad::<CfgExpr>("all(a", "expected `)`"); + bad::<CfgExpr>("not", "expected `(`"); + bad::<CfgExpr>("not(a", "expected `)`"); + bad::<CfgExpr>("a = ", "expected a string"); + bad::<CfgExpr>("all(not())", "expected identifier"); + bad::<CfgExpr>( + "foo(a)", + "unexpected content `(a)` found after cfg expression", + ); +} + +#[test] +fn cfg_matches() { + assert!(e!(foo).matches(&[c!(bar), c!(foo), c!(baz)])); + assert!(e!(any(foo)).matches(&[c!(bar), c!(foo), c!(baz)])); + assert!(e!(any(foo, bar)).matches(&[c!(bar)])); + assert!(e!(any(foo, bar)).matches(&[c!(foo)])); + assert!(e!(all(foo, bar)).matches(&[c!(foo), c!(bar)])); + assert!(e!(all(foo, bar)).matches(&[c!(foo), c!(bar)])); + assert!(e!(not(foo)).matches(&[c!(bar)])); + assert!(e!(not(foo)).matches(&[])); + assert!(e!(any((not(foo)), (all(foo, bar)))).matches(&[c!(bar)])); + assert!(e!(any((not(foo)), (all(foo, bar)))).matches(&[c!(foo), c!(bar)])); + + assert!(!e!(foo).matches(&[])); + assert!(!e!(foo).matches(&[c!(bar)])); + assert!(!e!(foo).matches(&[c!(fo)])); + assert!(!e!(any(foo)).matches(&[])); + assert!(!e!(any(foo)).matches(&[c!(bar)])); + assert!(!e!(any(foo)).matches(&[c!(bar), c!(baz)])); + assert!(!e!(all(foo)).matches(&[c!(bar), c!(baz)])); + assert!(!e!(all(foo, bar)).matches(&[c!(bar)])); + assert!(!e!(all(foo, bar)).matches(&[c!(foo)])); + assert!(!e!(all(foo, bar)).matches(&[])); + assert!(!e!(not(bar)).matches(&[c!(bar)])); + assert!(!e!(not(bar)).matches(&[c!(baz), c!(bar)])); + assert!(!e!(any((not(foo)), (all(foo, bar)))).matches(&[c!(foo)])); +} + +#[test] +fn bad_target_name() { + bad::<Platform>( + "any(cfg(unix), cfg(windows))", + "failed to parse `any(cfg(unix), cfg(windows))` as a cfg expression: \ + invalid target specifier: unexpected `(` character, \ + cfg expressions must start with `cfg(`", + ); + bad::<Platform>( + "!foo", + "failed to parse `!foo` as a cfg expression: \ + invalid target specifier: unexpected character ! in target name", + ); +} + +#[test] +fn round_trip_platform() { + fn rt(s: &str) { + let p = Platform::from_str(s).unwrap(); + let s2 = p.to_string(); + let p2 = Platform::from_str(&s2).unwrap(); + assert_eq!(p, p2); + } + rt("x86_64-apple-darwin"); + rt("foo"); + rt("cfg(windows)"); + rt("cfg(target_os = \"windows\")"); + rt( + "cfg(any(all(any(target_os = \"android\", target_os = \"linux\"), \ + any(target_arch = \"aarch64\", target_arch = \"arm\", target_arch = \"powerpc64\", \ + target_arch = \"x86\", target_arch = \"x86_64\")), \ + all(target_os = \"freebsd\", target_arch = \"x86_64\")))", + ); +} + +#[test] +fn check_cfg_attributes() { + fn ok(s: &str) { + let p = Platform::Cfg(s.parse().unwrap()); + let mut warnings = Vec::new(); + p.check_cfg_attributes(&mut warnings); + assert!( + warnings.is_empty(), + "Expected no warnings but got: {:?}", + warnings, + ); + } + + fn warn(s: &str, names: &[&str]) { + let p = Platform::Cfg(s.parse().unwrap()); + let mut warnings = Vec::new(); + p.check_cfg_attributes(&mut warnings); + assert_eq!( + warnings.len(), + names.len(), + "Expecter warnings about {:?} but got {:?}", + names, + warnings, + ); + for (name, warning) in names.iter().zip(warnings.iter()) { + assert!( + warning.contains(name), + "Expected warning about '{}' but got: {}", + name, + warning, + ); + } + } + + ok("unix"); + ok("windows"); + ok("any(not(unix), windows)"); + ok("foo"); + + ok("target_arch = \"abc\""); + ok("target_feature = \"abc\""); + ok("target_os = \"abc\""); + ok("target_family = \"abc\""); + ok("target_env = \"abc\""); + ok("target_endian = \"abc\""); + ok("target_pointer_width = \"abc\""); + ok("target_vendor = \"abc\""); + ok("bar = \"def\""); + + warn("test", &["test"]); + warn("debug_assertions", &["debug_assertions"]); + warn("proc_macro", &["proc_macro"]); + warn("feature = \"abc\"", &["feature"]); + + warn("any(not(debug_assertions), windows)", &["debug_assertions"]); + warn( + "any(not(feature = \"def\"), target_arch = \"abc\")", + &["feature"], + ); + warn( + "any(not(target_os = \"windows\"), proc_macro)", + &["proc_macro"], + ); + warn( + "any(not(feature = \"windows\"), proc_macro)", + &["feature", "proc_macro"], + ); + warn( + "all(not(debug_assertions), any(windows, proc_macro))", + &["debug_assertions", "proc_macro"], + ); +} diff --git a/crates/cargo-test-macro/Cargo.toml b/crates/cargo-test-macro/Cargo.toml new file mode 100644 index 0000000..04dafc0 --- /dev/null +++ b/crates/cargo-test-macro/Cargo.toml @@ -0,0 +1,12 @@ +[package] +name = "cargo-test-macro" +version = "0.1.0" +edition = "2021" +license = "MIT OR Apache-2.0" +homepage = "https://github.com/rust-lang/cargo" +repository = "https://github.com/rust-lang/cargo" +documentation = "https://github.com/rust-lang/cargo" +description = "Helper proc-macro for Cargo's testsuite." + +[lib] +proc-macro = true diff --git a/crates/cargo-test-macro/src/lib.rs b/crates/cargo-test-macro/src/lib.rs new file mode 100644 index 0000000..aa06f47 --- /dev/null +++ b/crates/cargo-test-macro/src/lib.rs @@ -0,0 +1,245 @@ +extern crate proc_macro; + +use proc_macro::*; +use std::process::Command; +use std::sync::Once; + +#[proc_macro_attribute] +pub fn cargo_test(attr: TokenStream, item: TokenStream) -> TokenStream { + // Ideally these options would be embedded in the test itself. However, I + // find it very helpful to have the test clearly state whether or not it + // is ignored. It would be nice to have some kind of runtime ignore + // support (such as + // https://internals.rust-lang.org/t/pre-rfc-skippable-tests/14611). + // + // Unfortunately a big drawback here is that if the environment changes + // (such as the existence of the `git` CLI), this will not trigger a + // rebuild and the test will still be ignored. In theory, something like + // `tracked_env` or `tracked_path` + // (https://github.com/rust-lang/rust/issues/99515) could help with this, + // but they don't really handle the absence of files well. + let mut ignore = false; + let mut requires_reason = false; + let mut explicit_reason = None; + let mut implicit_reasons = Vec::new(); + macro_rules! set_ignore { + ($predicate:expr, $($arg:tt)*) => { + let p = $predicate; + ignore |= p; + if p { + implicit_reasons.push(std::fmt::format(format_args!($($arg)*))); + } + }; + } + let is_not_nightly = !version().1; + for rule in split_rules(attr) { + match rule.as_str() { + "build_std_real" => { + // Only run the "real" build-std tests on nightly and with an + // explicit opt-in (these generally only work on linux, and + // have some extra requirements, and are slow, and can pollute + // the environment since it downloads dependencies). + set_ignore!(is_not_nightly, "requires nightly"); + set_ignore!( + option_env!("CARGO_RUN_BUILD_STD_TESTS").is_none(), + "CARGO_RUN_BUILD_STD_TESTS must be set" + ); + } + "build_std_mock" => { + // Only run the "mock" build-std tests on nightly and disable + // for windows-gnu which is missing object files (see + // https://github.com/rust-lang/wg-cargo-std-aware/issues/46). + set_ignore!(is_not_nightly, "requires nightly"); + set_ignore!( + cfg!(all(target_os = "windows", target_env = "gnu")), + "does not work on windows-gnu" + ); + } + "container_test" => { + // These tests must be opt-in because they require docker. + set_ignore!( + option_env!("CARGO_CONTAINER_TESTS").is_none(), + "CARGO_CONTAINER_TESTS must be set" + ); + } + "public_network_test" => { + // These tests must be opt-in because they touch the public + // network. The use of these should be **EXTREMELY RARE**, and + // should only touch things which would nearly certainly work + // in CI (like github.com). + set_ignore!( + option_env!("CARGO_PUBLIC_NETWORK_TESTS").is_none(), + "CARGO_PUBLIC_NETWORK_TESTS must be set" + ); + } + "nightly" => { + requires_reason = true; + set_ignore!(is_not_nightly, "requires nightly"); + } + s if s.starts_with("requires_") => { + let command = &s[9..]; + set_ignore!(!has_command(command), "{command} not installed"); + } + s if s.starts_with(">=1.") => { + requires_reason = true; + let min_minor = s[4..].parse().unwrap(); + let minor = version().0; + set_ignore!(minor < min_minor, "requires rustc 1.{minor} or newer"); + } + s if s.starts_with("reason=") => { + explicit_reason = Some(s[7..].parse().unwrap()); + } + s if s.starts_with("ignore_windows=") => { + set_ignore!(cfg!(windows), "{}", &s[16..s.len() - 1]); + } + _ => panic!("unknown rule {:?}", rule), + } + } + if requires_reason && explicit_reason.is_none() { + panic!( + "#[cargo_test] with a rule also requires a reason, \ + such as #[cargo_test(nightly, reason = \"needs -Z unstable-thing\")]" + ); + } + + // Construct the appropriate attributes. + let span = Span::call_site(); + let mut ret = TokenStream::new(); + let add_attr = |ret: &mut TokenStream, attr_name, attr_input| { + ret.extend(Some(TokenTree::from(Punct::new('#', Spacing::Alone)))); + let attr = TokenTree::from(Ident::new(attr_name, span)); + let mut attr_stream: TokenStream = attr.into(); + if let Some(input) = attr_input { + attr_stream.extend(input); + } + ret.extend(Some(TokenTree::from(Group::new( + Delimiter::Bracket, + attr_stream, + )))); + }; + add_attr(&mut ret, "test", None); + if ignore { + let reason = explicit_reason + .or_else(|| { + (!implicit_reasons.is_empty()) + .then(|| TokenTree::from(Literal::string(&implicit_reasons.join(", "))).into()) + }) + .map(|reason: TokenStream| { + let mut stream = TokenStream::new(); + stream.extend(Some(TokenTree::from(Punct::new('=', Spacing::Alone)))); + stream.extend(Some(reason)); + stream + }); + add_attr(&mut ret, "ignore", reason); + } + + // Find where the function body starts, and add the boilerplate at the start. + for token in item { + let group = match token { + TokenTree::Group(g) => { + if g.delimiter() == Delimiter::Brace { + g + } else { + ret.extend(Some(TokenTree::Group(g))); + continue; + } + } + other => { + ret.extend(Some(other)); + continue; + } + }; + + let mut new_body = to_token_stream( + r#"let _test_guard = { + let tmp_dir = option_env!("CARGO_TARGET_TMPDIR"); + cargo_test_support::paths::init_root(tmp_dir) + };"#, + ); + + new_body.extend(group.stream()); + ret.extend(Some(TokenTree::from(Group::new( + group.delimiter(), + new_body, + )))); + } + + ret +} + +fn split_rules(t: TokenStream) -> Vec<String> { + let tts: Vec<_> = t.into_iter().collect(); + tts.split(|tt| match tt { + TokenTree::Punct(p) => p.as_char() == ',', + _ => false, + }) + .filter(|parts| !parts.is_empty()) + .map(|parts| { + parts + .into_iter() + .map(|part| part.to_string()) + .collect::<String>() + }) + .collect() +} + +fn to_token_stream(code: &str) -> TokenStream { + code.parse().unwrap() +} + +static mut VERSION: (u32, bool) = (0, false); + +fn version() -> &'static (u32, bool) { + static INIT: Once = Once::new(); + INIT.call_once(|| { + let output = Command::new("rustc") + .arg("-V") + .output() + .expect("rustc should run"); + let stdout = std::str::from_utf8(&output.stdout).expect("utf8"); + let vers = stdout.split_whitespace().skip(1).next().unwrap(); + let is_nightly = option_env!("CARGO_TEST_DISABLE_NIGHTLY").is_none() + && (vers.contains("-nightly") || vers.contains("-dev")); + let minor = vers.split('.').skip(1).next().unwrap().parse().unwrap(); + unsafe { VERSION = (minor, is_nightly) } + }); + unsafe { &VERSION } +} + +fn has_command(command: &str) -> bool { + let output = match Command::new(command).arg("--version").output() { + Ok(output) => output, + Err(e) => { + // hg is not installed on GitHub macOS or certain constrained + // environments like Docker. Consider installing it if Cargo gains + // more hg support, but otherwise it isn't critical. + if is_ci() && command != "hg" { + panic!( + "expected command `{}` to be somewhere in PATH: {}", + command, e + ); + } + return false; + } + }; + if !output.status.success() { + panic!( + "expected command `{}` to be runnable, got error {}:\n\ + stderr:{}\n\ + stdout:{}\n", + command, + output.status, + String::from_utf8_lossy(&output.stderr), + String::from_utf8_lossy(&output.stdout) + ); + } + true +} + +/// Whether or not this running in a Continuous Integration environment. +fn is_ci() -> bool { + // Consider using `tracked_env` instead of option_env! when it is stabilized. + // `tracked_env` will handle changes, but not require rebuilding the macro + // itself like option_env does. + option_env!("CI").is_some() || option_env!("TF_BUILD").is_some() +} diff --git a/crates/cargo-test-support/Cargo.toml b/crates/cargo-test-support/Cargo.toml new file mode 100644 index 0000000..6c4b251 --- /dev/null +++ b/crates/cargo-test-support/Cargo.toml @@ -0,0 +1,32 @@ +[package] +name = "cargo-test-support" +version = "0.1.0" +license = "MIT OR Apache-2.0" +edition = "2021" + +[lib] +doctest = false + +[dependencies] +anyhow = "1.0.34" +cargo-test-macro = { path = "../cargo-test-macro" } +cargo-util = { path = "../cargo-util" } +crates-io = { path = "../crates-io" } +filetime = "0.2" +flate2 = { version = "1.0", default-features = false, features = ["zlib"] } +git2 = "0.16.0" +glob = "0.3" +itertools = "0.10.0" +lazy_static = "1.0" +pasetors = { version = "0.6.4", features = ["v3", "paserk", "std", "serde"] } +serde = { version = "1.0.123", features = ["derive"] } +serde_json = "1.0" +snapbox = { version = "0.4.0", features = ["diff", "path"] } +tar = { version = "0.4.38", default-features = false } +termcolor = "1.1.2" +time = { version = "0.3", features = ["parsing", "formatting"]} +toml = "0.7.0" +url = "2.2.2" + +[target.'cfg(windows)'.dependencies] +windows-sys = { version = "0.45.0", features = ["Win32_Storage_FileSystem"] } diff --git a/crates/cargo-test-support/build.rs b/crates/cargo-test-support/build.rs new file mode 100644 index 0000000..478da7d --- /dev/null +++ b/crates/cargo-test-support/build.rs @@ -0,0 +1,7 @@ +fn main() { + println!( + "cargo:rustc-env=NATIVE_ARCH={}", + std::env::var("TARGET").unwrap() + ); + println!("cargo:rerun-if-changed=build.rs"); +} diff --git a/crates/cargo-test-support/containers/apache/Dockerfile b/crates/cargo-test-support/containers/apache/Dockerfile new file mode 100644 index 0000000..8726024 --- /dev/null +++ b/crates/cargo-test-support/containers/apache/Dockerfile @@ -0,0 +1,26 @@ +FROM httpd:2.4-alpine + +RUN apk add --no-cache git git-daemon openssl + +COPY bar /repos/bar +WORKDIR /repos/bar +RUN git config --global user.email "testuser@example.com" &&\ + git config --global user.name "Test User" &&\ + git init -b master . &&\ + git add Cargo.toml src &&\ + git commit -m "Initial commit" &&\ + mv .git ../bar.git &&\ + cd ../bar.git &&\ + git config --bool core.bare true &&\ + rm -rf ../bar +WORKDIR / + +EXPOSE 443 + +WORKDIR /usr/local/apache2/conf +COPY httpd-cargo.conf . +RUN cat httpd-cargo.conf >> httpd.conf +RUN openssl req -x509 -nodes -days 3650 -newkey rsa:2048 \ + -keyout server.key -out server.crt \ + -subj "/emailAddress=webmaster@example.com/C=US/ST=California/L=San Francisco/O=Rust/OU=Cargo/CN=127.0.0.1" +WORKDIR / diff --git a/crates/cargo-test-support/containers/apache/bar/Cargo.toml b/crates/cargo-test-support/containers/apache/bar/Cargo.toml new file mode 100644 index 0000000..84fd5d8 --- /dev/null +++ b/crates/cargo-test-support/containers/apache/bar/Cargo.toml @@ -0,0 +1,4 @@ +[package] +name = "bar" +version = "1.0.0" +edition = "2021" diff --git a/crates/cargo-test-support/containers/apache/bar/src/lib.rs b/crates/cargo-test-support/containers/apache/bar/src/lib.rs new file mode 100644 index 0000000..ca74e3a --- /dev/null +++ b/crates/cargo-test-support/containers/apache/bar/src/lib.rs @@ -0,0 +1 @@ +// Intentionally blank. diff --git a/crates/cargo-test-support/containers/apache/httpd-cargo.conf b/crates/cargo-test-support/containers/apache/httpd-cargo.conf new file mode 100644 index 0000000..a4ba7d5 --- /dev/null +++ b/crates/cargo-test-support/containers/apache/httpd-cargo.conf @@ -0,0 +1,12 @@ +SetEnv GIT_PROJECT_ROOT /repos +SetEnv GIT_HTTP_EXPORT_ALL +ScriptAlias /repos /usr/libexec/git-core/git-http-backend/ +LoadModule cgid_module modules/mod_cgid.so + +<Files "git-http-backend"> + Require all granted +</Files> + +Include conf/extra/httpd-ssl.conf +LoadModule ssl_module modules/mod_ssl.so +LoadModule socache_shmcb_module modules/mod_socache_shmcb.so diff --git a/crates/cargo-test-support/containers/sshd/Dockerfile b/crates/cargo-test-support/containers/sshd/Dockerfile new file mode 100644 index 0000000..b52eefb --- /dev/null +++ b/crates/cargo-test-support/containers/sshd/Dockerfile @@ -0,0 +1,29 @@ +FROM alpine:3.17 + +RUN apk add --no-cache openssh git +RUN ssh-keygen -A + +RUN addgroup -S testuser && adduser -S testuser -G testuser -s /bin/ash +# NOTE: Ideally the password should be set to *, but I am uncertain how to do +# that in alpine. It shouldn't matter since PermitEmptyPasswords is "no". +RUN passwd -u testuser + +RUN mkdir /repos && chown testuser /repos +COPY --chown=testuser:testuser bar /repos/bar +USER testuser +WORKDIR /repos/bar +RUN git config --global user.email "testuser@example.com" &&\ + git config --global user.name "Test User" &&\ + git init -b master . &&\ + git add Cargo.toml src &&\ + git commit -m "Initial commit" &&\ + mv .git ../bar.git &&\ + cd ../bar.git &&\ + git config --bool core.bare true &&\ + rm -rf ../bar +WORKDIR / +USER root + +EXPOSE 22 + +ENTRYPOINT ["/usr/sbin/sshd", "-D", "-E", "/var/log/auth.log"] diff --git a/crates/cargo-test-support/containers/sshd/bar/Cargo.toml b/crates/cargo-test-support/containers/sshd/bar/Cargo.toml new file mode 100644 index 0000000..84fd5d8 --- /dev/null +++ b/crates/cargo-test-support/containers/sshd/bar/Cargo.toml @@ -0,0 +1,4 @@ +[package] +name = "bar" +version = "1.0.0" +edition = "2021" diff --git a/crates/cargo-test-support/containers/sshd/bar/src/lib.rs b/crates/cargo-test-support/containers/sshd/bar/src/lib.rs new file mode 100644 index 0000000..ca74e3a --- /dev/null +++ b/crates/cargo-test-support/containers/sshd/bar/src/lib.rs @@ -0,0 +1 @@ +// Intentionally blank. diff --git a/crates/cargo-test-support/src/compare.rs b/crates/cargo-test-support/src/compare.rs new file mode 100644 index 0000000..da1d099 --- /dev/null +++ b/crates/cargo-test-support/src/compare.rs @@ -0,0 +1,781 @@ +//! Routines for comparing and diffing output. +//! +//! # Patterns +//! +//! Many of these functions support special markup to assist with comparing +//! text that may vary or is otherwise uninteresting for the test at hand. The +//! supported patterns are: +//! +//! - `[..]` is a wildcard that matches 0 or more characters on the same line +//! (similar to `.*` in a regex). It is non-greedy. +//! - `[EXE]` optionally adds `.exe` on Windows (empty string on other +//! platforms). +//! - `[ROOT]` is the path to the test directory's root. +//! - `[CWD]` is the working directory of the process that was run. +//! - There is a wide range of substitutions (such as `[COMPILING]` or +//! `[WARNING]`) to match cargo's "status" output and allows you to ignore +//! the alignment. See the source of `substitute_macros` for a complete list +//! of substitutions. +//! - `[DIRTY-MSVC]` (only when the line starts with it) would be replaced by +//! `[DIRTY]` when `cfg(target_env = "msvc")` or the line will be ignored otherwise. +//! Tests that work around [issue 7358](https://github.com/rust-lang/cargo/issues/7358) +//! can use this to avoid duplicating the `with_stderr` call like: +//! `if cfg!(target_env = "msvc") {e.with_stderr("...[DIRTY]...");} else {e.with_stderr("...");}`. +//! +//! # Normalization +//! +//! In addition to the patterns described above, the strings are normalized +//! in such a way to avoid unwanted differences. The normalizations are: +//! +//! - Raw tab characters are converted to the string `<tab>`. This is helpful +//! so that raw tabs do not need to be written in the expected string, and +//! to avoid confusion of tabs vs spaces. +//! - Backslashes are converted to forward slashes to deal with Windows paths. +//! This helps so that all tests can be written assuming forward slashes. +//! Other heuristics are applied to try to ensure Windows-style paths aren't +//! a problem. +//! - Carriage returns are removed, which can help when running on Windows. + +use crate::diff; +use crate::paths; +use anyhow::{bail, Context, Result}; +use serde_json::Value; +use std::env; +use std::fmt; +use std::path::Path; +use std::str; +use url::Url; + +/// Default `snapbox` Assertions +/// +/// # Snapshots +/// +/// Updating of snapshots is controlled with the `SNAPSHOTS` environment variable: +/// +/// - `skip`: do not run the tests +/// - `ignore`: run the tests but ignore their failure +/// - `verify`: run the tests +/// - `overwrite`: update the snapshots based on the output of the tests +/// +/// # Patterns +/// +/// - `[..]` is a character wildcard, stopping at line breaks +/// - `\n...\n` is a multi-line wildcard +/// - `[EXE]` matches the exe suffix for the current platform +/// - `[ROOT]` matches [`paths::root()`][crate::paths::root] +/// - `[ROOTURL]` matches [`paths::root()`][crate::paths::root] as a URL +/// +/// # Normalization +/// +/// In addition to the patterns described above, text is normalized +/// in such a way to avoid unwanted differences. The normalizations are: +/// +/// - Backslashes are converted to forward slashes to deal with Windows paths. +/// This helps so that all tests can be written assuming forward slashes. +/// Other heuristics are applied to try to ensure Windows-style paths aren't +/// a problem. +/// - Carriage returns are removed, which can help when running on Windows. +pub fn assert_ui() -> snapbox::Assert { + let root = paths::root(); + // Use `from_file_path` instead of `from_dir_path` so the trailing slash is + // put in the users output, rather than hidden in the variable + let root_url = url::Url::from_file_path(&root).unwrap().to_string(); + let root = root.display().to_string(); + + let mut subs = snapbox::Substitutions::new(); + subs.extend([ + ( + "[EXE]", + std::borrow::Cow::Borrowed(std::env::consts::EXE_SUFFIX), + ), + ("[ROOT]", std::borrow::Cow::Owned(root)), + ("[ROOTURL]", std::borrow::Cow::Owned(root_url)), + ]) + .unwrap(); + snapbox::Assert::new() + .action_env(snapbox::DEFAULT_ACTION_ENV) + .substitutions(subs) +} + +/// Normalizes the output so that it can be compared against the expected value. +fn normalize_actual(actual: &str, cwd: Option<&Path>) -> String { + // It's easier to read tabs in outputs if they don't show up as literal + // hidden characters + let actual = actual.replace('\t', "<tab>"); + if cfg!(windows) { + // Let's not deal with \r\n vs \n on windows... + let actual = actual.replace('\r', ""); + normalize_windows(&actual, cwd) + } else { + actual + } +} + +/// Normalizes the expected string so that it can be compared against the actual output. +fn normalize_expected(expected: &str, cwd: Option<&Path>) -> String { + let expected = replace_dirty_msvc(expected); + let expected = substitute_macros(&expected); + + if cfg!(windows) { + normalize_windows(&expected, cwd) + } else { + let expected = match cwd { + None => expected, + Some(cwd) => expected.replace("[CWD]", &cwd.display().to_string()), + }; + let expected = expected.replace("[ROOT]", &paths::root().display().to_string()); + expected + } +} + +fn replace_dirty_msvc_impl(s: &str, is_msvc: bool) -> String { + if is_msvc { + s.replace("[DIRTY-MSVC]", "[DIRTY]") + } else { + use itertools::Itertools; + + let mut new = s + .lines() + .filter(|it| !it.starts_with("[DIRTY-MSVC]")) + .join("\n"); + + if s.ends_with("\n") { + new.push_str("\n"); + } + + new + } +} + +fn replace_dirty_msvc(s: &str) -> String { + replace_dirty_msvc_impl(s, cfg!(target_env = "msvc")) +} + +/// Normalizes text for both actual and expected strings on Windows. +fn normalize_windows(text: &str, cwd: Option<&Path>) -> String { + // Let's not deal with / vs \ (windows...) + let text = text.replace('\\', "/"); + + // Weirdness for paths on Windows extends beyond `/` vs `\` apparently. + // Namely paths like `c:\` and `C:\` are equivalent and that can cause + // issues. The return value of `env::current_dir()` may return a + // lowercase drive name, but we round-trip a lot of values through `Url` + // which will auto-uppercase the drive name. To just ignore this + // distinction we try to canonicalize as much as possible, taking all + // forms of a path and canonicalizing them to one. + let replace_path = |s: &str, path: &Path, with: &str| { + let path_through_url = Url::from_file_path(path).unwrap().to_file_path().unwrap(); + let path1 = path.display().to_string().replace('\\', "/"); + let path2 = path_through_url.display().to_string().replace('\\', "/"); + s.replace(&path1, with) + .replace(&path2, with) + .replace(with, &path1) + }; + + let text = match cwd { + None => text, + Some(p) => replace_path(&text, p, "[CWD]"), + }; + + // Similar to cwd above, perform similar treatment to the root path + // which in theory all of our paths should otherwise get rooted at. + let root = paths::root(); + let text = replace_path(&text, &root, "[ROOT]"); + + text +} + +fn substitute_macros(input: &str) -> String { + let macros = [ + ("[RUNNING]", " Running"), + ("[COMPILING]", " Compiling"), + ("[CHECKING]", " Checking"), + ("[COMPLETED]", " Completed"), + ("[CREATED]", " Created"), + ("[FINISHED]", " Finished"), + ("[ERROR]", "error:"), + ("[WARNING]", "warning:"), + ("[NOTE]", "note:"), + ("[HELP]", "help:"), + ("[DOCUMENTING]", " Documenting"), + ("[SCRAPING]", " Scraping"), + ("[FRESH]", " Fresh"), + ("[DIRTY]", " Dirty"), + ("[UPDATING]", " Updating"), + ("[ADDING]", " Adding"), + ("[REMOVING]", " Removing"), + ("[DOCTEST]", " Doc-tests"), + ("[PACKAGING]", " Packaging"), + ("[PACKAGED]", " Packaged"), + ("[DOWNLOADING]", " Downloading"), + ("[DOWNLOADED]", " Downloaded"), + ("[UPLOADING]", " Uploading"), + ("[VERIFYING]", " Verifying"), + ("[ARCHIVING]", " Archiving"), + ("[INSTALLING]", " Installing"), + ("[REPLACING]", " Replacing"), + ("[UNPACKING]", " Unpacking"), + ("[SUMMARY]", " Summary"), + ("[FIXED]", " Fixed"), + ("[FIXING]", " Fixing"), + ("[EXE]", env::consts::EXE_SUFFIX), + ("[IGNORED]", " Ignored"), + ("[INSTALLED]", " Installed"), + ("[REPLACED]", " Replaced"), + ("[BUILDING]", " Building"), + ("[LOGIN]", " Login"), + ("[LOGOUT]", " Logout"), + ("[YANK]", " Yank"), + ("[OWNER]", " Owner"), + ("[MIGRATING]", " Migrating"), + ("[EXECUTABLE]", " Executable"), + ("[SKIPPING]", " Skipping"), + ("[WAITING]", " Waiting"), + ]; + let mut result = input.to_owned(); + for &(pat, subst) in ¯os { + result = result.replace(pat, subst) + } + result +} + +/// Compares one string against another, checking that they both match. +/// +/// See [Patterns](index.html#patterns) for more information on pattern matching. +/// +/// - `description` explains where the output is from (usually "stdout" or "stderr"). +/// - `other_output` is other output to display in the error (usually stdout or stderr). +pub fn match_exact( + expected: &str, + actual: &str, + description: &str, + other_output: &str, + cwd: Option<&Path>, +) -> Result<()> { + let expected = normalize_expected(expected, cwd); + let actual = normalize_actual(actual, cwd); + let e: Vec<_> = expected.lines().map(WildStr::new).collect(); + let a: Vec<_> = actual.lines().map(WildStr::new).collect(); + if e == a { + return Ok(()); + } + let diff = diff::colored_diff(&e, &a); + bail!( + "{} did not match:\n\ + {}\n\n\ + other output:\n\ + {}\n", + description, + diff, + other_output, + ); +} + +/// Convenience wrapper around [`match_exact`] which will panic on error. +#[track_caller] +pub fn assert_match_exact(expected: &str, actual: &str) { + if let Err(e) = match_exact(expected, actual, "", "", None) { + crate::panic_error("", e); + } +} + +/// Checks that the given string contains the given lines, ignoring the order +/// of the lines. +/// +/// See [Patterns](index.html#patterns) for more information on pattern matching. +pub fn match_unordered(expected: &str, actual: &str, cwd: Option<&Path>) -> Result<()> { + let expected = normalize_expected(expected, cwd); + let actual = normalize_actual(actual, cwd); + let e: Vec<_> = expected.lines().map(|line| WildStr::new(line)).collect(); + let mut a: Vec<_> = actual.lines().map(|line| WildStr::new(line)).collect(); + // match more-constrained lines first, although in theory we'll + // need some sort of recursive match here. This handles the case + // that you expect "a\n[..]b" and two lines are printed out, + // "ab\n"a", where technically we do match unordered but a naive + // search fails to find this. This simple sort at least gets the + // test suite to pass for now, but we may need to get more fancy + // if tests start failing again. + a.sort_by_key(|s| s.line.len()); + let mut changes = Vec::new(); + let mut a_index = 0; + let mut failure = false; + + use crate::diff::Change; + for (e_i, e_line) in e.into_iter().enumerate() { + match a.iter().position(|a_line| e_line == *a_line) { + Some(index) => { + let a_line = a.remove(index); + changes.push(Change::Keep(e_i, index, a_line)); + a_index += 1; + } + None => { + failure = true; + changes.push(Change::Remove(e_i, e_line)); + } + } + } + for unmatched in a { + failure = true; + changes.push(Change::Add(a_index, unmatched)); + a_index += 1; + } + if failure { + bail!( + "Expected lines did not match (ignoring order):\n{}\n", + diff::render_colored_changes(&changes) + ); + } else { + Ok(()) + } +} + +/// Checks that the given string contains the given contiguous lines +/// somewhere. +/// +/// See [Patterns](index.html#patterns) for more information on pattern matching. +pub fn match_contains(expected: &str, actual: &str, cwd: Option<&Path>) -> Result<()> { + let expected = normalize_expected(expected, cwd); + let actual = normalize_actual(actual, cwd); + let e: Vec<_> = expected.lines().map(|line| WildStr::new(line)).collect(); + let a: Vec<_> = actual.lines().map(|line| WildStr::new(line)).collect(); + if e.len() == 0 { + bail!("expected length must not be zero"); + } + for window in a.windows(e.len()) { + if window == e { + return Ok(()); + } + } + bail!( + "expected to find:\n\ + {}\n\n\ + did not find in output:\n\ + {}", + expected, + actual + ); +} + +/// Checks that the given string does not contain the given contiguous lines +/// anywhere. +/// +/// See [Patterns](index.html#patterns) for more information on pattern matching. +pub fn match_does_not_contain(expected: &str, actual: &str, cwd: Option<&Path>) -> Result<()> { + if match_contains(expected, actual, cwd).is_ok() { + bail!( + "expected not to find:\n\ + {}\n\n\ + but found in output:\n\ + {}", + expected, + actual + ); + } else { + Ok(()) + } +} + +/// Checks that the given string contains the given contiguous lines +/// somewhere, and should be repeated `number` times. +/// +/// See [Patterns](index.html#patterns) for more information on pattern matching. +pub fn match_contains_n( + expected: &str, + number: usize, + actual: &str, + cwd: Option<&Path>, +) -> Result<()> { + let expected = normalize_expected(expected, cwd); + let actual = normalize_actual(actual, cwd); + let e: Vec<_> = expected.lines().map(|line| WildStr::new(line)).collect(); + let a: Vec<_> = actual.lines().map(|line| WildStr::new(line)).collect(); + if e.len() == 0 { + bail!("expected length must not be zero"); + } + let matches = a.windows(e.len()).filter(|window| *window == e).count(); + if matches == number { + Ok(()) + } else { + bail!( + "expected to find {} occurrences of:\n\ + {}\n\n\ + but found {} matches in the output:\n\ + {}", + number, + expected, + matches, + actual + ) + } +} + +/// Checks that the given string has a line that contains the given patterns, +/// and that line also does not contain the `without` patterns. +/// +/// See [Patterns](index.html#patterns) for more information on pattern matching. +/// +/// See [`crate::Execs::with_stderr_line_without`] for an example and cautions +/// against using. +pub fn match_with_without( + actual: &str, + with: &[String], + without: &[String], + cwd: Option<&Path>, +) -> Result<()> { + let actual = normalize_actual(actual, cwd); + let norm = |s: &String| format!("[..]{}[..]", normalize_expected(s, cwd)); + let with: Vec<_> = with.iter().map(norm).collect(); + let without: Vec<_> = without.iter().map(norm).collect(); + let with_wild: Vec<_> = with.iter().map(|w| WildStr::new(w)).collect(); + let without_wild: Vec<_> = without.iter().map(|w| WildStr::new(w)).collect(); + + let matches: Vec<_> = actual + .lines() + .map(WildStr::new) + .filter(|line| with_wild.iter().all(|with| with == line)) + .filter(|line| !without_wild.iter().any(|without| without == line)) + .collect(); + match matches.len() { + 0 => bail!( + "Could not find expected line in output.\n\ + With contents: {:?}\n\ + Without contents: {:?}\n\ + Actual stderr:\n\ + {}\n", + with, + without, + actual + ), + 1 => Ok(()), + _ => bail!( + "Found multiple matching lines, but only expected one.\n\ + With contents: {:?}\n\ + Without contents: {:?}\n\ + Matching lines:\n\ + {}\n", + with, + without, + itertools::join(matches, "\n") + ), + } +} + +/// Checks that the given string of JSON objects match the given set of +/// expected JSON objects. +/// +/// See [`crate::Execs::with_json`] for more details. +pub fn match_json(expected: &str, actual: &str, cwd: Option<&Path>) -> Result<()> { + let (exp_objs, act_objs) = collect_json_objects(expected, actual)?; + if exp_objs.len() != act_objs.len() { + bail!( + "expected {} json lines, got {}, stdout:\n{}", + exp_objs.len(), + act_objs.len(), + actual + ); + } + for (exp_obj, act_obj) in exp_objs.iter().zip(act_objs) { + find_json_mismatch(exp_obj, &act_obj, cwd)?; + } + Ok(()) +} + +/// Checks that the given string of JSON objects match the given set of +/// expected JSON objects, ignoring their order. +/// +/// See [`crate::Execs::with_json_contains_unordered`] for more details and +/// cautions when using. +pub fn match_json_contains_unordered( + expected: &str, + actual: &str, + cwd: Option<&Path>, +) -> Result<()> { + let (exp_objs, mut act_objs) = collect_json_objects(expected, actual)?; + for exp_obj in exp_objs { + match act_objs + .iter() + .position(|act_obj| find_json_mismatch(&exp_obj, act_obj, cwd).is_ok()) + { + Some(index) => act_objs.remove(index), + None => { + bail!( + "Did not find expected JSON:\n\ + {}\n\ + Remaining available output:\n\ + {}\n", + serde_json::to_string_pretty(&exp_obj).unwrap(), + itertools::join( + act_objs.iter().map(|o| serde_json::to_string(o).unwrap()), + "\n" + ) + ); + } + }; + } + Ok(()) +} + +fn collect_json_objects( + expected: &str, + actual: &str, +) -> Result<(Vec<serde_json::Value>, Vec<serde_json::Value>)> { + let expected_objs: Vec<_> = expected + .split("\n\n") + .map(|expect| { + expect + .parse() + .with_context(|| format!("failed to parse expected JSON object:\n{}", expect)) + }) + .collect::<Result<_>>()?; + let actual_objs: Vec<_> = actual + .lines() + .filter(|line| line.starts_with('{')) + .map(|line| { + line.parse() + .with_context(|| format!("failed to parse JSON object:\n{}", line)) + }) + .collect::<Result<_>>()?; + Ok((expected_objs, actual_objs)) +} + +/// Compares JSON object for approximate equality. +/// You can use `[..]` wildcard in strings (useful for OS-dependent things such +/// as paths). You can use a `"{...}"` string literal as a wildcard for +/// arbitrary nested JSON (useful for parts of object emitted by other programs +/// (e.g., rustc) rather than Cargo itself). +pub fn find_json_mismatch(expected: &Value, actual: &Value, cwd: Option<&Path>) -> Result<()> { + match find_json_mismatch_r(expected, actual, cwd) { + Some((expected_part, actual_part)) => bail!( + "JSON mismatch\nExpected:\n{}\nWas:\n{}\nExpected part:\n{}\nActual part:\n{}\n", + serde_json::to_string_pretty(expected).unwrap(), + serde_json::to_string_pretty(&actual).unwrap(), + serde_json::to_string_pretty(expected_part).unwrap(), + serde_json::to_string_pretty(actual_part).unwrap(), + ), + None => Ok(()), + } +} + +fn find_json_mismatch_r<'a>( + expected: &'a Value, + actual: &'a Value, + cwd: Option<&Path>, +) -> Option<(&'a Value, &'a Value)> { + use serde_json::Value::*; + match (expected, actual) { + (&Number(ref l), &Number(ref r)) if l == r => None, + (&Bool(l), &Bool(r)) if l == r => None, + (&String(ref l), _) if l == "{...}" => None, + (&String(ref l), &String(ref r)) => { + if match_exact(l, r, "", "", cwd).is_err() { + Some((expected, actual)) + } else { + None + } + } + (&Array(ref l), &Array(ref r)) => { + if l.len() != r.len() { + return Some((expected, actual)); + } + + l.iter() + .zip(r.iter()) + .filter_map(|(l, r)| find_json_mismatch_r(l, r, cwd)) + .next() + } + (&Object(ref l), &Object(ref r)) => { + let same_keys = l.len() == r.len() && l.keys().all(|k| r.contains_key(k)); + if !same_keys { + return Some((expected, actual)); + } + + l.values() + .zip(r.values()) + .filter_map(|(l, r)| find_json_mismatch_r(l, r, cwd)) + .next() + } + (&Null, &Null) => None, + // Magic string literal `"{...}"` acts as wildcard for any sub-JSON. + _ => Some((expected, actual)), + } +} + +/// A single line string that supports `[..]` wildcard matching. +pub struct WildStr<'a> { + has_meta: bool, + line: &'a str, +} + +impl<'a> WildStr<'a> { + pub fn new(line: &'a str) -> WildStr<'a> { + WildStr { + has_meta: line.contains("[..]"), + line, + } + } +} + +impl<'a> PartialEq for WildStr<'a> { + fn eq(&self, other: &Self) -> bool { + match (self.has_meta, other.has_meta) { + (false, false) => self.line == other.line, + (true, false) => meta_cmp(self.line, other.line), + (false, true) => meta_cmp(other.line, self.line), + (true, true) => panic!("both lines cannot have [..]"), + } + } +} + +fn meta_cmp(a: &str, mut b: &str) -> bool { + for (i, part) in a.split("[..]").enumerate() { + match b.find(part) { + Some(j) => { + if i == 0 && j != 0 { + return false; + } + b = &b[j + part.len()..]; + } + None => return false, + } + } + b.is_empty() || a.ends_with("[..]") +} + +impl fmt::Display for WildStr<'_> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.write_str(&self.line) + } +} + +impl fmt::Debug for WildStr<'_> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "{:?}", self.line) + } +} + +#[test] +fn wild_str_cmp() { + for (a, b) in &[ + ("a b", "a b"), + ("a[..]b", "a b"), + ("a[..]", "a b"), + ("[..]", "a b"), + ("[..]b", "a b"), + ] { + assert_eq!(WildStr::new(a), WildStr::new(b)); + } + for (a, b) in &[("[..]b", "c"), ("b", "c"), ("b", "cb")] { + assert_ne!(WildStr::new(a), WildStr::new(b)); + } +} + +#[test] +fn dirty_msvc() { + let case = |expected: &str, wild: &str, msvc: bool| { + assert_eq!(expected, &replace_dirty_msvc_impl(wild, msvc)); + }; + + // no replacements + case("aa", "aa", false); + case("aa", "aa", true); + + // with replacements + case( + "\ +[DIRTY] a", + "\ +[DIRTY-MSVC] a", + true, + ); + case( + "", + "\ +[DIRTY-MSVC] a", + false, + ); + case( + "\ +[DIRTY] a +[COMPILING] a", + "\ +[DIRTY-MSVC] a +[COMPILING] a", + true, + ); + case( + "\ +[COMPILING] a", + "\ +[DIRTY-MSVC] a +[COMPILING] a", + false, + ); + + // test trailing newline behavior + case( + "\ +A +B +", "\ +A +B +", true, + ); + + case( + "\ +A +B +", "\ +A +B +", false, + ); + + case( + "\ +A +B", "\ +A +B", true, + ); + + case( + "\ +A +B", "\ +A +B", false, + ); + + case( + "\ +[DIRTY] a +", + "\ +[DIRTY-MSVC] a +", + true, + ); + case( + "\n", + "\ +[DIRTY-MSVC] a +", + false, + ); + + case( + "\ +[DIRTY] a", + "\ +[DIRTY-MSVC] a", + true, + ); + case( + "", + "\ +[DIRTY-MSVC] a", + false, + ); +} diff --git a/crates/cargo-test-support/src/containers.rs b/crates/cargo-test-support/src/containers.rs new file mode 100644 index 0000000..17040d8 --- /dev/null +++ b/crates/cargo-test-support/src/containers.rs @@ -0,0 +1,285 @@ +//! Support for testing using Docker containers. +//! +//! The [`Container`] type is a builder for configuring a container to run. +//! After you call `launch`, you can use the [`ContainerHandle`] to interact +//! with the running container. +//! +//! Tests using containers must use `#[cargo_test(container_test)]` to disable +//! them unless the CARGO_CONTAINER_TESTS environment variable is set. + +use cargo_util::ProcessBuilder; +use std::collections::HashMap; +use std::io::Read; +use std::path::PathBuf; +use std::process::Command; +use std::sync::atomic::{AtomicUsize, Ordering}; +use std::sync::Mutex; +use tar::Header; + +/// A builder for configuring a container to run. +pub struct Container { + /// The host directory that forms the basis of the Docker image. + build_context: PathBuf, + /// Files to copy over to the image. + files: Vec<MkFile>, +} + +/// A handle to a running container. +/// +/// You can use this to interact with the container. +pub struct ContainerHandle { + /// The name of the container. + name: String, + /// The IP address of the container. + /// + /// NOTE: This is currently unused, but may be useful so I left it in. + /// This can only be used on Linux. macOS and Windows docker doesn't allow + /// direct connection to the container. + pub ip_address: String, + /// Port mappings of container_port to host_port for ports exposed via EXPOSE. + pub port_mappings: HashMap<u16, u16>, +} + +impl Container { + pub fn new(context_dir: &str) -> Container { + assert!(std::env::var_os("CARGO_CONTAINER_TESTS").is_some()); + let mut build_context = PathBuf::from(env!("CARGO_MANIFEST_DIR")); + build_context.push("containers"); + build_context.push(context_dir); + Container { + build_context, + files: Vec::new(), + } + } + + /// Adds a file to be copied into the container. + pub fn file(mut self, file: MkFile) -> Self { + self.files.push(file); + self + } + + /// Starts the container. + pub fn launch(mut self) -> ContainerHandle { + static NEXT_ID: AtomicUsize = AtomicUsize::new(0); + + let id = NEXT_ID.fetch_add(1, Ordering::SeqCst); + let name = format!("cargo_test_{id}"); + remove_if_exists(&name); + self.create_container(&name); + self.copy_files(&name); + self.start_container(&name); + let info = self.container_inspect(&name); + let ip_address = if cfg!(target_os = "linux") { + info[0]["NetworkSettings"]["IPAddress"] + .as_str() + .unwrap() + .to_string() + } else { + // macOS and Windows can't make direct connections to the + // container. It only works through exposed ports or mapped ports. + "127.0.0.1".to_string() + }; + let port_mappings = self.port_mappings(&info); + self.wait_till_ready(&port_mappings); + + ContainerHandle { + name, + ip_address, + port_mappings, + } + } + + fn create_container(&self, name: &str) { + static BUILD_LOCK: Mutex<()> = Mutex::new(()); + + let image_base = self.build_context.file_name().unwrap(); + let image_name = format!("cargo-test-{}", image_base.to_str().unwrap()); + let _lock = BUILD_LOCK.lock().unwrap(); + ProcessBuilder::new("docker") + .args(&["build", "--tag", image_name.as_str()]) + .arg(&self.build_context) + .exec_with_output() + .unwrap(); + + ProcessBuilder::new("docker") + .args(&[ + "container", + "create", + "--publish-all", + "--rm", + "--name", + name, + ]) + .arg(image_name) + .exec_with_output() + .unwrap(); + } + + fn copy_files(&mut self, name: &str) { + if self.files.is_empty() { + return; + } + let mut ar = tar::Builder::new(Vec::new()); + let files = std::mem::replace(&mut self.files, Vec::new()); + for mut file in files { + ar.append_data(&mut file.header, &file.path, file.contents.as_slice()) + .unwrap(); + } + let ar = ar.into_inner().unwrap(); + ProcessBuilder::new("docker") + .args(&["cp", "-"]) + .arg(format!("{name}:/")) + .stdin(ar) + .exec_with_output() + .unwrap(); + } + + fn start_container(&self, name: &str) { + ProcessBuilder::new("docker") + .args(&["container", "start"]) + .arg(name) + .exec_with_output() + .unwrap(); + } + + fn container_inspect(&self, name: &str) -> serde_json::Value { + let output = ProcessBuilder::new("docker") + .args(&["inspect", name]) + .exec_with_output() + .unwrap(); + serde_json::from_slice(&output.stdout).unwrap() + } + + /// Returns the mapping of container_port->host_port for ports that were + /// exposed with EXPOSE. + fn port_mappings(&self, info: &serde_json::Value) -> HashMap<u16, u16> { + info[0]["NetworkSettings"]["Ports"] + .as_object() + .unwrap() + .iter() + .map(|(key, value)| { + let key = key + .strip_suffix("/tcp") + .expect("expected TCP only ports") + .parse() + .unwrap(); + let values = value.as_array().unwrap(); + let value = values + .iter() + .find(|value| value["HostIp"].as_str().unwrap() == "0.0.0.0") + .expect("expected localhost IP"); + let host_port = value["HostPort"].as_str().unwrap().parse().unwrap(); + (key, host_port) + }) + .collect() + } + + fn wait_till_ready(&self, port_mappings: &HashMap<u16, u16>) { + for port in port_mappings.values() { + let mut ok = false; + for _ in 0..30 { + match std::net::TcpStream::connect(format!("127.0.0.1:{port}")) { + Ok(_) => { + ok = true; + break; + } + Err(e) => { + if e.kind() != std::io::ErrorKind::ConnectionRefused { + panic!("unexpected localhost connection error: {e:?}"); + } + std::thread::sleep(std::time::Duration::new(1, 0)); + } + } + } + if !ok { + panic!("no listener on localhost port {port}"); + } + } + } +} + +impl ContainerHandle { + /// Executes a program inside a running container. + pub fn exec(&self, args: &[&str]) -> std::process::Output { + ProcessBuilder::new("docker") + .args(&["container", "exec", &self.name]) + .args(args) + .exec_with_output() + .unwrap() + } + + /// Returns the contents of a file inside the container. + pub fn read_file(&self, path: &str) -> String { + let output = ProcessBuilder::new("docker") + .args(&["cp", &format!("{}:{}", self.name, path), "-"]) + .exec_with_output() + .unwrap(); + let mut ar = tar::Archive::new(output.stdout.as_slice()); + let mut entry = ar.entries().unwrap().next().unwrap().unwrap(); + let mut contents = String::new(); + entry.read_to_string(&mut contents).unwrap(); + contents + } +} + +impl Drop for ContainerHandle { + fn drop(&mut self) { + // To help with debugging, this will keep the container alive. + if std::env::var_os("CARGO_CONTAINER_TEST_KEEP").is_some() { + return; + } + remove_if_exists(&self.name); + } +} + +fn remove_if_exists(name: &str) { + if let Err(e) = Command::new("docker") + .args(&["container", "rm", "--force", name]) + .output() + { + panic!("failed to run docker: {e}"); + } +} + +/// Builder for configuring a file to copy into a container. +pub struct MkFile { + path: String, + contents: Vec<u8>, + header: Header, +} + +impl MkFile { + /// Defines a file to add to the container. + /// + /// This should be passed to `Container::file`. + /// + /// The path is the path inside the container to create the file. + pub fn path(path: &str) -> MkFile { + MkFile { + path: path.to_string(), + contents: Vec::new(), + header: Header::new_gnu(), + } + } + + pub fn contents(mut self, contents: impl Into<Vec<u8>>) -> Self { + self.contents = contents.into(); + self.header.set_size(self.contents.len() as u64); + self + } + + pub fn mode(mut self, mode: u32) -> Self { + self.header.set_mode(mode); + self + } + + pub fn uid(mut self, uid: u64) -> Self { + self.header.set_uid(uid); + self + } + + pub fn gid(mut self, gid: u64) -> Self { + self.header.set_gid(gid); + self + } +} diff --git a/crates/cargo-test-support/src/cross_compile.rs b/crates/cargo-test-support/src/cross_compile.rs new file mode 100644 index 0000000..a2daf88 --- /dev/null +++ b/crates/cargo-test-support/src/cross_compile.rs @@ -0,0 +1,264 @@ +//! Support for cross-compile tests with the `--target` flag. +//! +//! Note that cross-testing is very limited. You need to install the +//! "alternate" target to the host (32-bit for 64-bit hosts or vice-versa). +//! +//! Set CFG_DISABLE_CROSS_TESTS=1 environment variable to disable these tests +//! if you are unable to use the alternate target. Unfortunately 32-bit +//! support on macOS is going away, so macOS users are out of luck. +//! +//! These tests are all disabled on rust-lang/rust's CI, but run in Cargo's CI. + +use crate::{basic_manifest, main_file, project}; +use cargo_util::ProcessError; +use std::env; +use std::fmt::Write; +use std::process::{Command, Output}; +use std::sync::atomic::{AtomicBool, Ordering}; +use std::sync::Once; + +/// Whether or not the resulting cross binaries can run on the host. +static CAN_RUN_ON_HOST: AtomicBool = AtomicBool::new(false); + +pub fn disabled() -> bool { + // First, disable if requested. + match env::var("CFG_DISABLE_CROSS_TESTS") { + Ok(ref s) if *s == "1" => return true, + _ => {} + } + + // Cross tests are only tested to work on macos, linux, and MSVC windows. + if !(cfg!(target_os = "macos") || cfg!(target_os = "linux") || cfg!(target_env = "msvc")) { + return true; + } + + // It's not particularly common to have a cross-compilation setup, so + // try to detect that before we fail a bunch of tests through no fault + // of the user. + static CAN_BUILD_CROSS_TESTS: AtomicBool = AtomicBool::new(false); + static CHECK: Once = Once::new(); + + let cross_target = alternate(); + + let run_cross_test = || -> anyhow::Result<Output> { + let p = project() + .at("cross_test") + .file("Cargo.toml", &basic_manifest("cross_test", "1.0.0")) + .file("src/main.rs", &main_file(r#""testing!""#, &[])) + .build(); + + let build_result = p + .cargo("build --target") + .arg(&cross_target) + .exec_with_output(); + + if build_result.is_ok() { + CAN_BUILD_CROSS_TESTS.store(true, Ordering::SeqCst); + } + + let result = p + .cargo("run --target") + .arg(&cross_target) + .exec_with_output(); + + if result.is_ok() { + CAN_RUN_ON_HOST.store(true, Ordering::SeqCst); + } + build_result + }; + + CHECK.call_once(|| { + drop(run_cross_test()); + }); + + if CAN_BUILD_CROSS_TESTS.load(Ordering::SeqCst) { + // We were able to compile a simple project, so the user has the + // necessary `std::` bits installed. Therefore, tests should not + // be disabled. + return false; + } + + // We can't compile a simple cross project. We want to warn the user + // by failing a single test and having the remainder of the cross tests + // pass. We don't use `std::sync::Once` here because panicking inside its + // `call_once` method would poison the `Once` instance, which is not what + // we want. + static HAVE_WARNED: AtomicBool = AtomicBool::new(false); + + if HAVE_WARNED.swap(true, Ordering::SeqCst) { + // We are some other test and somebody else is handling the warning. + // Just disable the current test. + return true; + } + + // We are responsible for warning the user, which we do by panicking. + let mut message = format!( + " +Cannot cross compile to {}. + +This failure can be safely ignored. If you would prefer to not see this +failure, you can set the environment variable CFG_DISABLE_CROSS_TESTS to \"1\". + +Alternatively, you can install the necessary libraries to enable cross +compilation tests. Cross compilation tests depend on your host platform. +", + cross_target + ); + + if cfg!(target_os = "linux") { + message.push_str( + " +Linux cross tests target i686-unknown-linux-gnu, which requires the ability to +build and run 32-bit targets. This requires the 32-bit libraries to be +installed. For example, on Ubuntu, run `sudo apt install gcc-multilib` to +install the necessary libraries. +", + ); + } else if cfg!(all(target_os = "macos", target_arch = "aarch64")) { + message.push_str( + " +macOS on aarch64 cross tests to target x86_64-apple-darwin. +This should be natively supported via Xcode, nothing additional besides the +rustup target should be needed. +", + ); + } else if cfg!(target_os = "macos") { + message.push_str( + " +macOS on x86_64 cross tests to target x86_64-apple-ios, which requires the iOS +SDK to be installed. This should be included with Xcode automatically. If you +are using the Xcode command line tools, you'll need to install the full Xcode +app (from the Apple App Store), and switch to it with this command: + + sudo xcode-select --switch /Applications/Xcode.app/Contents/Developer + +Some cross-tests want to *run* the executables on the host. These tests will +be ignored if this is not possible. On macOS, this means you need an iOS +simulator installed to run these tests. To install a simulator, open Xcode, go +to preferences > Components, and download the latest iOS simulator. +", + ); + } else if cfg!(target_os = "windows") { + message.push_str( + " +Windows cross tests target i686-pc-windows-msvc, which requires the ability +to build and run 32-bit targets. This should work automatically if you have +properly installed Visual Studio build tools. +", + ); + } else { + // The check at the top should prevent this. + panic!("platform should have been skipped"); + } + + let rustup_available = Command::new("rustup").output().is_ok(); + if rustup_available { + write!( + message, + " +Make sure that the appropriate `rustc` target is installed with rustup: + + rustup target add {} +", + cross_target + ) + .unwrap(); + } else { + write!( + message, + " +rustup does not appear to be installed. Make sure that the appropriate +`rustc` target is installed for the target `{}`. +", + cross_target + ) + .unwrap(); + } + + // Show the actual error message. + match run_cross_test() { + Ok(_) => message.push_str("\nUh oh, second run succeeded?\n"), + Err(err) => match err.downcast_ref::<ProcessError>() { + Some(proc_err) => write!(message, "\nTest error: {}\n", proc_err).unwrap(), + None => write!(message, "\nUnexpected non-process error: {}\n", err).unwrap(), + }, + } + + panic!("{}", message); +} + +/// The arch triple of the test-running host. +pub fn native() -> &'static str { + env!("NATIVE_ARCH") +} + +pub fn native_arch() -> &'static str { + match native() + .split("-") + .next() + .expect("Target triple has unexpected format") + { + "x86_64" => "x86_64", + "aarch64" => "aarch64", + "i686" => "x86", + _ => panic!("This test should be gated on cross_compile::disabled."), + } +} + +/// The alternate target-triple to build with. +/// +/// Only use this function on tests that check `cross_compile::disabled`. +pub fn alternate() -> &'static str { + if cfg!(all(target_os = "macos", target_arch = "aarch64")) { + "x86_64-apple-darwin" + } else if cfg!(target_os = "macos") { + "x86_64-apple-ios" + } else if cfg!(target_os = "linux") { + "i686-unknown-linux-gnu" + } else if cfg!(all(target_os = "windows", target_env = "msvc")) { + "i686-pc-windows-msvc" + } else if cfg!(all(target_os = "windows", target_env = "gnu")) { + "i686-pc-windows-gnu" + } else { + panic!("This test should be gated on cross_compile::disabled."); + } +} + +pub fn alternate_arch() -> &'static str { + if cfg!(target_os = "macos") { + "x86_64" + } else { + "x86" + } +} + +/// A target-triple that is neither the host nor the target. +/// +/// Rustc may not work with it and it's alright, apart from being a +/// valid target triple it is supposed to be used only as a +/// placeholder for targets that should not be considered. +pub fn unused() -> &'static str { + "wasm32-unknown-unknown" +} + +/// Whether or not the host can run cross-compiled executables. +pub fn can_run_on_host() -> bool { + if disabled() { + return false; + } + // macos is currently configured to cross compile to x86_64-apple-ios + // which requires a simulator to run. Azure's CI image appears to have the + // SDK installed, but are not configured to launch iOS images with a + // simulator. + if cfg!(target_os = "macos") { + if CAN_RUN_ON_HOST.load(Ordering::SeqCst) { + return true; + } else { + println!("Note: Cannot run on host, skipping."); + return false; + } + } else { + assert!(CAN_RUN_ON_HOST.load(Ordering::SeqCst)); + return true; + } +} diff --git a/crates/cargo-test-support/src/diff.rs b/crates/cargo-test-support/src/diff.rs new file mode 100644 index 0000000..f3b283b --- /dev/null +++ b/crates/cargo-test-support/src/diff.rs @@ -0,0 +1,174 @@ +//! A simple Myers diff implementation. +//! +//! This focuses on being short and simple, and the expense of being +//! inefficient. A key characteristic here is that this supports cargotest's +//! `[..]` wildcard matching. That means things like hashing can't be used. +//! Since Cargo's output tends to be small, this should be sufficient. + +use std::fmt; +use std::io::Write; +use termcolor::{Ansi, Color, ColorSpec, NoColor, WriteColor}; + +/// A single line change to be applied to the original. +#[derive(Debug, Eq, PartialEq)] +pub enum Change<T> { + Add(usize, T), + Remove(usize, T), + Keep(usize, usize, T), +} + +pub fn diff<'a, T>(a: &'a [T], b: &'a [T]) -> Vec<Change<&'a T>> +where + T: PartialEq, +{ + if a.is_empty() && b.is_empty() { + return vec![]; + } + let mut diff = vec![]; + for (prev_x, prev_y, x, y) in backtrack(&a, &b) { + if x == prev_x { + diff.push(Change::Add(prev_y + 1, &b[prev_y])); + } else if y == prev_y { + diff.push(Change::Remove(prev_x + 1, &a[prev_x])); + } else { + diff.push(Change::Keep(prev_x + 1, prev_y + 1, &a[prev_x])); + } + } + diff.reverse(); + diff +} + +fn shortest_edit<T>(a: &[T], b: &[T]) -> Vec<Vec<usize>> +where + T: PartialEq, +{ + let max = a.len() + b.len(); + let mut v = vec![0; 2 * max + 1]; + let mut trace = vec![]; + for d in 0..=max { + trace.push(v.clone()); + for k in (0..=(2 * d)).step_by(2) { + let mut x = if k == 0 || (k != 2 * d && v[max - d + k - 1] < v[max - d + k + 1]) { + // Move down + v[max - d + k + 1] + } else { + // Move right + v[max - d + k - 1] + 1 + }; + let mut y = x + d - k; + // Step diagonally as far as possible. + while x < a.len() && y < b.len() && a[x] == b[y] { + x += 1; + y += 1; + } + v[max - d + k] = x; + // Return if reached the bottom-right position. + if x >= a.len() && y >= b.len() { + return trace; + } + } + } + panic!("finished without hitting end?"); +} + +fn backtrack<T>(a: &[T], b: &[T]) -> Vec<(usize, usize, usize, usize)> +where + T: PartialEq, +{ + let mut result = vec![]; + let mut x = a.len(); + let mut y = b.len(); + let max = x + y; + for (d, v) in shortest_edit(a, b).iter().enumerate().rev() { + let k = x + d - y; + let prev_k = if k == 0 || (k != 2 * d && v[max - d + k - 1] < v[max - d + k + 1]) { + k + 1 + } else { + k - 1 + }; + let prev_x = v[max - d + prev_k]; + let prev_y = (prev_x + d).saturating_sub(prev_k); + while x > prev_x && y > prev_y { + result.push((x - 1, y - 1, x, y)); + x -= 1; + y -= 1; + } + if d > 0 { + result.push((prev_x, prev_y, x, y)); + } + x = prev_x; + y = prev_y; + } + return result; +} + +pub fn colored_diff<'a, T>(a: &'a [T], b: &'a [T]) -> String +where + T: PartialEq + fmt::Display, +{ + let changes = diff(a, b); + render_colored_changes(&changes) +} + +pub fn render_colored_changes<T: fmt::Display>(changes: &[Change<T>]) -> String { + // termcolor is not very ergonomic, but I don't want to bring in another dependency. + let mut red = ColorSpec::new(); + red.set_fg(Some(Color::Red)); + let mut green = ColorSpec::new(); + green.set_fg(Some(Color::Green)); + let mut dim = ColorSpec::new(); + dim.set_dimmed(true); + let mut v = Vec::new(); + let mut result: Box<dyn WriteColor> = if crate::is_ci() { + // Don't use color on CI. Even though GitHub can display colors, it + // makes reading the raw logs more difficult. + Box::new(NoColor::new(&mut v)) + } else { + Box::new(Ansi::new(&mut v)) + }; + + for change in changes { + let (nums, sign, color, text) = match change { + Change::Add(i, s) => (format!(" {:<4} ", i), '+', &green, s), + Change::Remove(i, s) => (format!("{:<4} ", i), '-', &red, s), + Change::Keep(x, y, s) => (format!("{:<4}{:<4} ", x, y), ' ', &dim, s), + }; + result.set_color(&dim).unwrap(); + write!(result, "{}", nums).unwrap(); + let mut bold = color.clone(); + bold.set_bold(true); + result.set_color(&bold).unwrap(); + write!(result, "{}", sign).unwrap(); + result.reset().unwrap(); + result.set_color(&color).unwrap(); + write!(result, "{}", text).unwrap(); + result.reset().unwrap(); + writeln!(result).unwrap(); + } + drop(result); + String::from_utf8(v).unwrap() +} + +#[cfg(test)] +pub fn compare(a: &str, b: &str) { + let a: Vec<_> = a.chars().collect(); + let b: Vec<_> = b.chars().collect(); + let changes = diff(&a, &b); + let mut result = vec![]; + for change in changes { + match change { + Change::Add(_, s) => result.push(*s), + Change::Remove(_, _s) => {} + Change::Keep(_, _, s) => result.push(*s), + } + } + assert_eq!(b, result); +} + +#[test] +fn basic_tests() { + compare("", ""); + compare("A", ""); + compare("", "B"); + compare("ABCABBA", "CBABAC"); +} diff --git a/crates/cargo-test-support/src/git.rs b/crates/cargo-test-support/src/git.rs new file mode 100644 index 0000000..18c4646 --- /dev/null +++ b/crates/cargo-test-support/src/git.rs @@ -0,0 +1,249 @@ +/* +# Git Testing Support + +## Creating a git dependency +`git::new()` is an easy way to create a new git repository containing a +project that you can then use as a dependency. It will automatically add all +the files you specify in the project and commit them to the repository. +Example: + +``` +let git_project = git::new("dep1", |project| { + project + .file("Cargo.toml", &basic_manifest("dep1")) + .file("src/lib.rs", r#"pub fn f() { println!("hi!"); } "#) +}); + +// Use the `url()` method to get the file url to the new repository. +let p = project() + .file("Cargo.toml", &format!(r#" + [package] + name = "a" + version = "1.0.0" + + [dependencies] + dep1 = {{ git = '{}' }} + "#, git_project.url())) + .file("src/lib.rs", "extern crate dep1;") + .build(); +``` + +## Manually creating repositories +`git::repo()` can be used to create a `RepoBuilder` which provides a way of +adding files to a blank repository and committing them. + +If you want to then manipulate the repository (such as adding new files or +tags), you can use `git2::Repository::open()` to open the repository and then +use some of the helper functions in this file to interact with the repository. + +*/ + +use crate::{path2url, project, Project, ProjectBuilder}; +use std::fs; +use std::path::{Path, PathBuf}; +use std::sync::Once; +use url::Url; + +#[must_use] +pub struct RepoBuilder { + repo: git2::Repository, + files: Vec<PathBuf>, +} + +pub struct Repository(git2::Repository); + +/// Create a `RepoBuilder` to build a new git repository. +/// +/// Call `build()` to finalize and create the repository. +pub fn repo(p: &Path) -> RepoBuilder { + RepoBuilder::init(p) +} + +impl RepoBuilder { + pub fn init(p: &Path) -> RepoBuilder { + t!(fs::create_dir_all(p.parent().unwrap())); + let repo = init(p); + RepoBuilder { + repo, + files: Vec::new(), + } + } + + /// Add a file to the repository. + pub fn file(self, path: &str, contents: &str) -> RepoBuilder { + let mut me = self.nocommit_file(path, contents); + me.files.push(PathBuf::from(path)); + me + } + + /// Add a file that will be left in the working directory, but not added + /// to the repository. + pub fn nocommit_file(self, path: &str, contents: &str) -> RepoBuilder { + let dst = self.repo.workdir().unwrap().join(path); + t!(fs::create_dir_all(dst.parent().unwrap())); + t!(fs::write(&dst, contents)); + self + } + + /// Create the repository and commit the new files. + pub fn build(self) -> Repository { + { + let mut index = t!(self.repo.index()); + for file in self.files.iter() { + t!(index.add_path(file)); + } + t!(index.write()); + let id = t!(index.write_tree()); + let tree = t!(self.repo.find_tree(id)); + let sig = t!(self.repo.signature()); + t!(self + .repo + .commit(Some("HEAD"), &sig, &sig, "Initial commit", &tree, &[])); + } + let RepoBuilder { repo, .. } = self; + Repository(repo) + } +} + +impl Repository { + pub fn root(&self) -> &Path { + self.0.workdir().unwrap() + } + + pub fn url(&self) -> Url { + path2url(self.0.workdir().unwrap().to_path_buf()) + } + + pub fn revparse_head(&self) -> String { + self.0 + .revparse_single("HEAD") + .expect("revparse HEAD") + .id() + .to_string() + } +} + +/// Initialize a new repository at the given path. +pub fn init(path: &Path) -> git2::Repository { + default_search_path(); + let repo = t!(git2::Repository::init(path)); + default_repo_cfg(&repo); + repo +} + +fn default_search_path() { + use crate::paths::global_root; + use git2::{opts::set_search_path, ConfigLevel}; + + static INIT: Once = Once::new(); + INIT.call_once(|| unsafe { + let path = global_root().join("blank_git_search_path"); + t!(set_search_path(ConfigLevel::System, &path)); + t!(set_search_path(ConfigLevel::Global, &path)); + t!(set_search_path(ConfigLevel::XDG, &path)); + t!(set_search_path(ConfigLevel::ProgramData, &path)); + }) +} + +fn default_repo_cfg(repo: &git2::Repository) { + let mut cfg = t!(repo.config()); + t!(cfg.set_str("user.email", "foo@bar.com")); + t!(cfg.set_str("user.name", "Foo Bar")); +} + +/// Create a new git repository with a project. +pub fn new<F>(name: &str, callback: F) -> Project +where + F: FnOnce(ProjectBuilder) -> ProjectBuilder, +{ + new_repo(name, callback).0 +} + +/// Create a new git repository with a project. +/// Returns both the Project and the git Repository. +pub fn new_repo<F>(name: &str, callback: F) -> (Project, git2::Repository) +where + F: FnOnce(ProjectBuilder) -> ProjectBuilder, +{ + let mut git_project = project().at(name); + git_project = callback(git_project); + let git_project = git_project.build(); + + let repo = init(&git_project.root()); + add(&repo); + commit(&repo); + (git_project, repo) +} + +/// Add all files in the working directory to the git index. +pub fn add(repo: &git2::Repository) { + // FIXME(libgit2/libgit2#2514): apparently, `add_all` will add all submodules + // as well, and then fail because they're directories. As a stop-gap, we just + // ignore all submodules. + let mut s = t!(repo.submodules()); + for submodule in s.iter_mut() { + t!(submodule.add_to_index(false)); + } + let mut index = t!(repo.index()); + t!(index.add_all( + ["*"].iter(), + git2::IndexAddOption::DEFAULT, + Some( + &mut (|a, _b| if s.iter().any(|s| a.starts_with(s.path())) { + 1 + } else { + 0 + }) + ) + )); + t!(index.write()); +} + +/// Add a git submodule to the repository. +pub fn add_submodule<'a>( + repo: &'a git2::Repository, + url: &str, + path: &Path, +) -> git2::Submodule<'a> { + let path = path.to_str().unwrap().replace(r"\", "/"); + let mut s = t!(repo.submodule(url, Path::new(&path), false)); + let subrepo = t!(s.open()); + default_repo_cfg(&subrepo); + t!(subrepo.remote_add_fetch("origin", "refs/heads/*:refs/heads/*")); + let mut origin = t!(subrepo.find_remote("origin")); + t!(origin.fetch(&Vec::<String>::new(), None, None)); + t!(subrepo.checkout_head(None)); + t!(s.add_finalize()); + s +} + +/// Commit changes to the git repository. +pub fn commit(repo: &git2::Repository) -> git2::Oid { + let tree_id = t!(t!(repo.index()).write_tree()); + let sig = t!(repo.signature()); + let mut parents = Vec::new(); + if let Some(parent) = repo.head().ok().map(|h| h.target().unwrap()) { + parents.push(t!(repo.find_commit(parent))) + } + let parents = parents.iter().collect::<Vec<_>>(); + t!(repo.commit( + Some("HEAD"), + &sig, + &sig, + "test", + &t!(repo.find_tree(tree_id)), + &parents + )) +} + +/// Create a new tag in the git repository. +pub fn tag(repo: &git2::Repository, name: &str) { + let head = repo.head().unwrap().target().unwrap(); + t!(repo.tag( + name, + &t!(repo.find_object(head, None)), + &t!(repo.signature()), + "make a new tag", + false + )); +} diff --git a/crates/cargo-test-support/src/install.rs b/crates/cargo-test-support/src/install.rs new file mode 100644 index 0000000..478b482 --- /dev/null +++ b/crates/cargo-test-support/src/install.rs @@ -0,0 +1,29 @@ +use crate::paths; +use std::env::consts::EXE_SUFFIX; +use std::path::{Path, PathBuf}; + +/// Used by `cargo install` tests to assert an executable binary +/// has been installed. Example usage: +/// +/// assert_has_installed_exe(cargo_home(), "foo"); +#[track_caller] +pub fn assert_has_installed_exe<P: AsRef<Path>>(path: P, name: &'static str) { + assert!(check_has_installed_exe(path, name)); +} + +#[track_caller] +pub fn assert_has_not_installed_exe<P: AsRef<Path>>(path: P, name: &'static str) { + assert!(!check_has_installed_exe(path, name)); +} + +fn check_has_installed_exe<P: AsRef<Path>>(path: P, name: &'static str) -> bool { + path.as_ref().join("bin").join(exe(name)).is_file() +} + +pub fn cargo_home() -> PathBuf { + paths::home().join(".cargo") +} + +pub fn exe(name: &str) -> String { + format!("{}{}", name, EXE_SUFFIX) +} diff --git a/crates/cargo-test-support/src/lib.rs b/crates/cargo-test-support/src/lib.rs new file mode 100644 index 0000000..04d6ce9 --- /dev/null +++ b/crates/cargo-test-support/src/lib.rs @@ -0,0 +1,1424 @@ +//! # Cargo test support. +//! +//! See <https://rust-lang.github.io/cargo/contrib/> for a guide on writing tests. + +#![allow(clippy::all)] + +use std::env; +use std::ffi::OsStr; +use std::fmt::Write; +use std::fs; +use std::os; +use std::path::{Path, PathBuf}; +use std::process::{Command, Output}; +use std::str; +use std::time::{self, Duration}; + +use anyhow::{bail, Result}; +use cargo_util::{is_ci, ProcessBuilder, ProcessError}; +use serde_json; +use url::Url; + +use self::paths::CargoPathExt; + +#[macro_export] +macro_rules! t { + ($e:expr) => { + match $e { + Ok(e) => e, + Err(e) => $crate::panic_error(&format!("failed running {}", stringify!($e)), e), + } + }; +} + +#[macro_export] +macro_rules! curr_dir { + () => { + $crate::_curr_dir(std::path::Path::new(file!())); + }; +} + +#[doc(hidden)] +pub fn _curr_dir(mut file_path: &'static Path) -> &'static Path { + if !file_path.exists() { + // HACK: Must be running in the rust-lang/rust workspace, adjust the paths accordingly. + let prefix = PathBuf::from("src").join("tools").join("cargo"); + if let Ok(crate_relative) = file_path.strip_prefix(prefix) { + file_path = crate_relative + } + } + assert!(file_path.exists(), "{} does not exist", file_path.display()); + file_path.parent().unwrap() +} + +#[track_caller] +pub fn panic_error(what: &str, err: impl Into<anyhow::Error>) -> ! { + let err = err.into(); + pe(what, err); + #[track_caller] + fn pe(what: &str, err: anyhow::Error) -> ! { + let mut result = format!("{}\nerror: {}", what, err); + for cause in err.chain().skip(1) { + drop(writeln!(result, "\nCaused by:")); + drop(write!(result, "{}", cause)); + } + panic!("\n{}", result); + } +} + +pub use cargo_test_macro::cargo_test; + +pub mod compare; +pub mod containers; +pub mod cross_compile; +mod diff; +pub mod git; +pub mod install; +pub mod paths; +pub mod publish; +pub mod registry; +pub mod tools; + +pub mod prelude { + pub use crate::ArgLine; + pub use crate::CargoCommand; + pub use crate::ChannelChanger; + pub use crate::TestEnv; +} + +/* + * + * ===== Builders ===== + * + */ + +#[derive(PartialEq, Clone)] +struct FileBuilder { + path: PathBuf, + body: String, + executable: bool, +} + +impl FileBuilder { + pub fn new(path: PathBuf, body: &str, executable: bool) -> FileBuilder { + FileBuilder { + path, + body: body.to_string(), + executable: executable, + } + } + + fn mk(&mut self) { + if self.executable { + self.path.set_extension(env::consts::EXE_EXTENSION); + } + + self.dirname().mkdir_p(); + fs::write(&self.path, &self.body) + .unwrap_or_else(|e| panic!("could not create file {}: {}", self.path.display(), e)); + + #[cfg(unix)] + if self.executable { + use std::os::unix::fs::PermissionsExt; + + let mut perms = fs::metadata(&self.path).unwrap().permissions(); + let mode = perms.mode(); + perms.set_mode(mode | 0o111); + fs::set_permissions(&self.path, perms).unwrap(); + } + } + + fn dirname(&self) -> &Path { + self.path.parent().unwrap() + } +} + +#[derive(PartialEq, Clone)] +struct SymlinkBuilder { + dst: PathBuf, + src: PathBuf, + src_is_dir: bool, +} + +impl SymlinkBuilder { + pub fn new(dst: PathBuf, src: PathBuf) -> SymlinkBuilder { + SymlinkBuilder { + dst, + src, + src_is_dir: false, + } + } + + pub fn new_dir(dst: PathBuf, src: PathBuf) -> SymlinkBuilder { + SymlinkBuilder { + dst, + src, + src_is_dir: true, + } + } + + #[cfg(unix)] + fn mk(&self) { + self.dirname().mkdir_p(); + t!(os::unix::fs::symlink(&self.dst, &self.src)); + } + + #[cfg(windows)] + fn mk(&mut self) { + self.dirname().mkdir_p(); + if self.src_is_dir { + t!(os::windows::fs::symlink_dir(&self.dst, &self.src)); + } else { + if let Some(ext) = self.dst.extension() { + if ext == env::consts::EXE_EXTENSION { + self.src.set_extension(ext); + } + } + t!(os::windows::fs::symlink_file(&self.dst, &self.src)); + } + } + + fn dirname(&self) -> &Path { + self.src.parent().unwrap() + } +} + +/// A cargo project to run tests against. +/// +/// See [`ProjectBuilder`] or [`Project::from_template`] to get started. +pub struct Project { + root: PathBuf, +} + +/// Create a project to run tests against +/// +/// The project can be constructed programmatically or from the filesystem with [`Project::from_template`] +#[must_use] +pub struct ProjectBuilder { + root: Project, + files: Vec<FileBuilder>, + symlinks: Vec<SymlinkBuilder>, + no_manifest: bool, +} + +impl ProjectBuilder { + /// Root of the project, ex: `/path/to/cargo/target/cit/t0/foo` + pub fn root(&self) -> PathBuf { + self.root.root() + } + + /// Project's debug dir, ex: `/path/to/cargo/target/cit/t0/foo/target/debug` + pub fn target_debug_dir(&self) -> PathBuf { + self.root.target_debug_dir() + } + + pub fn new(root: PathBuf) -> ProjectBuilder { + ProjectBuilder { + root: Project { root }, + files: vec![], + symlinks: vec![], + no_manifest: false, + } + } + + pub fn at<P: AsRef<Path>>(mut self, path: P) -> Self { + self.root = Project { + root: paths::root().join(path), + }; + self + } + + /// Adds a file to the project. + pub fn file<B: AsRef<Path>>(mut self, path: B, body: &str) -> Self { + self._file(path.as_ref(), body, false); + self + } + + /// Adds an executable file to the project. + pub fn executable<B: AsRef<Path>>(mut self, path: B, body: &str) -> Self { + self._file(path.as_ref(), body, true); + self + } + + fn _file(&mut self, path: &Path, body: &str, executable: bool) { + self.files.push(FileBuilder::new( + self.root.root().join(path), + body, + executable, + )); + } + + /// Adds a symlink to a file to the project. + pub fn symlink<T: AsRef<Path>>(mut self, dst: T, src: T) -> Self { + self.symlinks.push(SymlinkBuilder::new( + self.root.root().join(dst), + self.root.root().join(src), + )); + self + } + + /// Create a symlink to a directory + pub fn symlink_dir<T: AsRef<Path>>(mut self, dst: T, src: T) -> Self { + self.symlinks.push(SymlinkBuilder::new_dir( + self.root.root().join(dst), + self.root.root().join(src), + )); + self + } + + pub fn no_manifest(mut self) -> Self { + self.no_manifest = true; + self + } + + /// Creates the project. + pub fn build(mut self) -> Project { + // First, clean the directory if it already exists + self.rm_root(); + + // Create the empty directory + self.root.root().mkdir_p(); + + let manifest_path = self.root.root().join("Cargo.toml"); + if !self.no_manifest && self.files.iter().all(|fb| fb.path != manifest_path) { + self._file( + Path::new("Cargo.toml"), + &basic_manifest("foo", "0.0.1"), + false, + ) + } + + let past = time::SystemTime::now() - Duration::new(1, 0); + let ftime = filetime::FileTime::from_system_time(past); + + for file in self.files.iter_mut() { + file.mk(); + if is_coarse_mtime() { + // Place the entire project 1 second in the past to ensure + // that if cargo is called multiple times, the 2nd call will + // see targets as "fresh". Without this, if cargo finishes in + // under 1 second, the second call will see the mtime of + // source == mtime of output and consider it dirty. + filetime::set_file_times(&file.path, ftime, ftime).unwrap(); + } + } + + for symlink in self.symlinks.iter_mut() { + symlink.mk(); + } + + let ProjectBuilder { root, .. } = self; + root + } + + fn rm_root(&self) { + self.root.root().rm_rf() + } +} + +impl Project { + /// Copy the test project from a fixed state + pub fn from_template(template_path: impl AsRef<std::path::Path>) -> Self { + let root = paths::root(); + let project_root = root.join("case"); + snapbox::path::copy_template(template_path.as_ref(), &project_root).unwrap(); + Self { root: project_root } + } + + /// Root of the project, ex: `/path/to/cargo/target/cit/t0/foo` + pub fn root(&self) -> PathBuf { + self.root.clone() + } + + /// Project's target dir, ex: `/path/to/cargo/target/cit/t0/foo/target` + pub fn build_dir(&self) -> PathBuf { + self.root().join("target") + } + + /// Project's debug dir, ex: `/path/to/cargo/target/cit/t0/foo/target/debug` + pub fn target_debug_dir(&self) -> PathBuf { + self.build_dir().join("debug") + } + + /// File url for root, ex: `file:///path/to/cargo/target/cit/t0/foo` + pub fn url(&self) -> Url { + path2url(self.root()) + } + + /// Path to an example built as a library. + /// `kind` should be one of: "lib", "rlib", "staticlib", "dylib", "proc-macro" + /// ex: `/path/to/cargo/target/cit/t0/foo/target/debug/examples/libex.rlib` + pub fn example_lib(&self, name: &str, kind: &str) -> PathBuf { + self.target_debug_dir() + .join("examples") + .join(paths::get_lib_filename(name, kind)) + } + + /// Path to a debug binary. + /// ex: `/path/to/cargo/target/cit/t0/foo/target/debug/foo` + pub fn bin(&self, b: &str) -> PathBuf { + self.build_dir() + .join("debug") + .join(&format!("{}{}", b, env::consts::EXE_SUFFIX)) + } + + /// Path to a release binary. + /// ex: `/path/to/cargo/target/cit/t0/foo/target/release/foo` + pub fn release_bin(&self, b: &str) -> PathBuf { + self.build_dir() + .join("release") + .join(&format!("{}{}", b, env::consts::EXE_SUFFIX)) + } + + /// Path to a debug binary for a specific target triple. + /// ex: `/path/to/cargo/target/cit/t0/foo/target/i686-apple-darwin/debug/foo` + pub fn target_bin(&self, target: &str, b: &str) -> PathBuf { + self.build_dir().join(target).join("debug").join(&format!( + "{}{}", + b, + env::consts::EXE_SUFFIX + )) + } + + /// Returns an iterator of paths matching the glob pattern, which is + /// relative to the project root. + pub fn glob<P: AsRef<Path>>(&self, pattern: P) -> glob::Paths { + let pattern = self.root().join(pattern); + glob::glob(pattern.to_str().expect("failed to convert pattern to str")) + .expect("failed to glob") + } + + /// Changes the contents of an existing file. + pub fn change_file(&self, path: &str, body: &str) { + FileBuilder::new(self.root().join(path), body, false).mk() + } + + /// Creates a `ProcessBuilder` to run a program in the project + /// and wrap it in an Execs to assert on the execution. + /// Example: + /// p.process(&p.bin("foo")) + /// .with_stdout("bar\n") + /// .run(); + pub fn process<T: AsRef<OsStr>>(&self, program: T) -> Execs { + let mut p = process(program); + p.cwd(self.root()); + execs().with_process_builder(p) + } + + /// Creates a `ProcessBuilder` to run cargo. + /// Arguments can be separated by spaces. + /// Example: + /// p.cargo("build --bin foo").run(); + pub fn cargo(&self, cmd: &str) -> Execs { + let cargo = cargo_exe(); + let mut execs = self.process(&cargo); + if let Some(ref mut p) = execs.process_builder { + p.env("CARGO", cargo); + p.arg_line(cmd); + } + execs + } + + /// Safely run a process after `cargo build`. + /// + /// Windows has a problem where a process cannot be reliably + /// be replaced, removed, or renamed immediately after executing it. + /// The action may fail (with errors like Access is denied), or + /// it may succeed, but future attempts to use the same filename + /// will fail with "Already Exists". + /// + /// If you have a test that needs to do `cargo run` multiple + /// times, you should instead use `cargo build` and use this + /// method to run the executable. Each time you call this, + /// use a new name for `dst`. + /// See rust-lang/cargo#5481. + pub fn rename_run(&self, src: &str, dst: &str) -> Execs { + let src = self.bin(src); + let dst = self.bin(dst); + fs::rename(&src, &dst) + .unwrap_or_else(|e| panic!("Failed to rename `{:?}` to `{:?}`: {}", src, dst, e)); + self.process(dst) + } + + /// Returns the contents of `Cargo.lock`. + pub fn read_lockfile(&self) -> String { + self.read_file("Cargo.lock") + } + + /// Returns the contents of a path in the project root + pub fn read_file(&self, path: &str) -> String { + let full = self.root().join(path); + fs::read_to_string(&full) + .unwrap_or_else(|e| panic!("could not read file {}: {}", full.display(), e)) + } + + /// Modifies `Cargo.toml` to remove all commented lines. + pub fn uncomment_root_manifest(&self) { + let contents = self.read_file("Cargo.toml").replace("#", ""); + fs::write(self.root().join("Cargo.toml"), contents).unwrap(); + } + + pub fn symlink(&self, src: impl AsRef<Path>, dst: impl AsRef<Path>) { + let src = self.root().join(src.as_ref()); + let dst = self.root().join(dst.as_ref()); + #[cfg(unix)] + { + if let Err(e) = os::unix::fs::symlink(&src, &dst) { + panic!("failed to symlink {:?} to {:?}: {:?}", src, dst, e); + } + } + #[cfg(windows)] + { + if src.is_dir() { + if let Err(e) = os::windows::fs::symlink_dir(&src, &dst) { + panic!("failed to symlink {:?} to {:?}: {:?}", src, dst, e); + } + } else { + if let Err(e) = os::windows::fs::symlink_file(&src, &dst) { + panic!("failed to symlink {:?} to {:?}: {:?}", src, dst, e); + } + } + } + } +} + +// Generates a project layout +pub fn project() -> ProjectBuilder { + ProjectBuilder::new(paths::root().join("foo")) +} + +// Generates a project layout in given directory +pub fn project_in(dir: &str) -> ProjectBuilder { + ProjectBuilder::new(paths::root().join(dir).join("foo")) +} + +// Generates a project layout inside our fake home dir +pub fn project_in_home(name: &str) -> ProjectBuilder { + ProjectBuilder::new(paths::home().join(name)) +} + +// === Helpers === + +pub fn main_file(println: &str, deps: &[&str]) -> String { + let mut buf = String::new(); + + for dep in deps.iter() { + buf.push_str(&format!("extern crate {};\n", dep)); + } + + buf.push_str("fn main() { println!("); + buf.push_str(println); + buf.push_str("); }\n"); + + buf +} + +pub fn cargo_exe() -> PathBuf { + snapbox::cmd::cargo_bin("cargo") +} + +/// This is the raw output from the process. +/// +/// This is similar to `std::process::Output`, however the `status` is +/// translated to the raw `code`. This is necessary because `ProcessError` +/// does not have access to the raw `ExitStatus` because `ProcessError` needs +/// to be serializable (for the Rustc cache), and `ExitStatus` does not +/// provide a constructor. +pub struct RawOutput { + pub code: Option<i32>, + pub stdout: Vec<u8>, + pub stderr: Vec<u8>, +} + +#[must_use] +#[derive(Clone)] +pub struct Execs { + ran: bool, + process_builder: Option<ProcessBuilder>, + expect_stdout: Option<String>, + expect_stdin: Option<String>, + expect_stderr: Option<String>, + expect_exit_code: Option<i32>, + expect_stdout_contains: Vec<String>, + expect_stderr_contains: Vec<String>, + expect_stdout_contains_n: Vec<(String, usize)>, + expect_stdout_not_contains: Vec<String>, + expect_stderr_not_contains: Vec<String>, + expect_stderr_unordered: Vec<String>, + expect_stderr_with_without: Vec<(Vec<String>, Vec<String>)>, + expect_json: Option<String>, + expect_json_contains_unordered: Option<String>, + stream_output: bool, +} + +impl Execs { + pub fn with_process_builder(mut self, p: ProcessBuilder) -> Execs { + self.process_builder = Some(p); + self + } + + /// Verifies that stdout is equal to the given lines. + /// See [`compare`] for supported patterns. + pub fn with_stdout<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stdout = Some(expected.to_string()); + self + } + + /// Verifies that stderr is equal to the given lines. + /// See [`compare`] for supported patterns. + pub fn with_stderr<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stderr = Some(expected.to_string()); + self + } + + /// Writes the given lines to stdin. + pub fn with_stdin<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stdin = Some(expected.to_string()); + self + } + + /// Verifies the exit code from the process. + /// + /// This is not necessary if the expected exit code is `0`. + pub fn with_status(&mut self, expected: i32) -> &mut Self { + self.expect_exit_code = Some(expected); + self + } + + /// Removes exit code check for the process. + /// + /// By default, the expected exit code is `0`. + pub fn without_status(&mut self) -> &mut Self { + self.expect_exit_code = None; + self + } + + /// Verifies that stdout contains the given contiguous lines somewhere in + /// its output. + /// + /// See [`compare`] for supported patterns. + pub fn with_stdout_contains<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stdout_contains.push(expected.to_string()); + self + } + + /// Verifies that stderr contains the given contiguous lines somewhere in + /// its output. + /// + /// See [`compare`] for supported patterns. + pub fn with_stderr_contains<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stderr_contains.push(expected.to_string()); + self + } + + /// Verifies that stdout contains the given contiguous lines somewhere in + /// its output, and should be repeated `number` times. + /// + /// See [`compare`] for supported patterns. + pub fn with_stdout_contains_n<S: ToString>(&mut self, expected: S, number: usize) -> &mut Self { + self.expect_stdout_contains_n + .push((expected.to_string(), number)); + self + } + + /// Verifies that stdout does not contain the given contiguous lines. + /// + /// See [`compare`] for supported patterns. + /// + /// See note on [`Self::with_stderr_does_not_contain`]. + pub fn with_stdout_does_not_contain<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stdout_not_contains.push(expected.to_string()); + self + } + + /// Verifies that stderr does not contain the given contiguous lines. + /// + /// See [`compare`] for supported patterns. + /// + /// Care should be taken when using this method because there is a + /// limitless number of possible things that *won't* appear. A typo means + /// your test will pass without verifying the correct behavior. If + /// possible, write the test first so that it fails, and then implement + /// your fix/feature to make it pass. + pub fn with_stderr_does_not_contain<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stderr_not_contains.push(expected.to_string()); + self + } + + /// Verifies that all of the stderr output is equal to the given lines, + /// ignoring the order of the lines. + /// + /// See [`compare`] for supported patterns. + /// + /// This is useful when checking the output of `cargo build -v` since + /// the order of the output is not always deterministic. + /// Recommend use `with_stderr_contains` instead unless you really want to + /// check *every* line of output. + /// + /// Be careful when using patterns such as `[..]`, because you may end up + /// with multiple lines that might match, and this is not smart enough to + /// do anything like longest-match. For example, avoid something like: + /// + /// ```text + /// [RUNNING] `rustc [..] + /// [RUNNING] `rustc --crate-name foo [..] + /// ``` + /// + /// This will randomly fail if the other crate name is `bar`, and the + /// order changes. + pub fn with_stderr_unordered<S: ToString>(&mut self, expected: S) -> &mut Self { + self.expect_stderr_unordered.push(expected.to_string()); + self + } + + /// Verify that a particular line appears in stderr with and without the + /// given substrings. Exactly one line must match. + /// + /// The substrings are matched as `contains`. Example: + /// + /// ```no_run + /// execs.with_stderr_line_without( + /// &[ + /// "[RUNNING] `rustc --crate-name build_script_build", + /// "-C opt-level=3", + /// ], + /// &["-C debuginfo", "-C incremental"], + /// ) + /// ``` + /// + /// This will check that a build line includes `-C opt-level=3` but does + /// not contain `-C debuginfo` or `-C incremental`. + /// + /// Be careful writing the `without` fragments, see note in + /// `with_stderr_does_not_contain`. + pub fn with_stderr_line_without<S: ToString>( + &mut self, + with: &[S], + without: &[S], + ) -> &mut Self { + let with = with.iter().map(|s| s.to_string()).collect(); + let without = without.iter().map(|s| s.to_string()).collect(); + self.expect_stderr_with_without.push((with, without)); + self + } + + /// Verifies the JSON output matches the given JSON. + /// + /// This is typically used when testing cargo commands that emit JSON. + /// Each separate JSON object should be separated by a blank line. + /// Example: + /// + /// ```rust,ignore + /// assert_that( + /// p.cargo("metadata"), + /// execs().with_json(r#" + /// {"example": "abc"} + /// + /// {"example": "def"} + /// "#) + /// ); + /// ``` + /// + /// - Objects should match in the order given. + /// - The order of arrays is ignored. + /// - Strings support patterns described in [`compare`]. + /// - Use `"{...}"` to match any object. + pub fn with_json(&mut self, expected: &str) -> &mut Self { + self.expect_json = Some(expected.to_string()); + self + } + + /// Verifies JSON output contains the given objects (in any order) somewhere + /// in its output. + /// + /// CAUTION: Be very careful when using this. Make sure every object is + /// unique (not a subset of one another). Also avoid using objects that + /// could possibly match multiple output lines unless you're very sure of + /// what you are doing. + /// + /// See `with_json` for more detail. + pub fn with_json_contains_unordered(&mut self, expected: &str) -> &mut Self { + match &mut self.expect_json_contains_unordered { + None => self.expect_json_contains_unordered = Some(expected.to_string()), + Some(e) => { + e.push_str("\n\n"); + e.push_str(expected); + } + } + self + } + + /// Forward subordinate process stdout/stderr to the terminal. + /// Useful for printf debugging of the tests. + /// CAUTION: CI will fail if you leave this in your test! + #[allow(unused)] + pub fn stream(&mut self) -> &mut Self { + self.stream_output = true; + self + } + + pub fn arg<T: AsRef<OsStr>>(&mut self, arg: T) -> &mut Self { + if let Some(ref mut p) = self.process_builder { + p.arg(arg); + } + self + } + + pub fn cwd<T: AsRef<OsStr>>(&mut self, path: T) -> &mut Self { + if let Some(ref mut p) = self.process_builder { + if let Some(cwd) = p.get_cwd() { + let new_path = cwd.join(path.as_ref()); + p.cwd(new_path); + } else { + p.cwd(path); + } + } + self + } + + fn get_cwd(&self) -> Option<&Path> { + self.process_builder.as_ref().and_then(|p| p.get_cwd()) + } + + pub fn env<T: AsRef<OsStr>>(&mut self, key: &str, val: T) -> &mut Self { + if let Some(ref mut p) = self.process_builder { + p.env(key, val); + } + self + } + + pub fn env_remove(&mut self, key: &str) -> &mut Self { + if let Some(ref mut p) = self.process_builder { + p.env_remove(key); + } + self + } + + pub fn exec_with_output(&mut self) -> Result<Output> { + self.ran = true; + // TODO avoid unwrap + let p = (&self.process_builder).clone().unwrap(); + p.exec_with_output() + } + + pub fn build_command(&mut self) -> Command { + self.ran = true; + // TODO avoid unwrap + let p = (&self.process_builder).clone().unwrap(); + p.build_command() + } + + /// Enables nightly features for testing + /// + /// The list of reasons should be why nightly cargo is needed. If it is + /// becuase of an unstable feature put the name of the feature as the reason, + /// e.g. `&["print-im-a-teapot"]` + pub fn masquerade_as_nightly_cargo(&mut self, reasons: &[&str]) -> &mut Self { + if let Some(ref mut p) = self.process_builder { + p.masquerade_as_nightly_cargo(reasons); + } + self + } + + /// Overrides the crates.io URL for testing. + /// + /// Can be used for testing crates-io functionality where alt registries + /// cannot be used. + pub fn replace_crates_io(&mut self, url: &Url) -> &mut Self { + if let Some(ref mut p) = self.process_builder { + p.env("__CARGO_TEST_CRATES_IO_URL_DO_NOT_USE_THIS", url.as_str()); + } + self + } + + pub fn enable_split_debuginfo_packed(&mut self) -> &mut Self { + self.env("CARGO_PROFILE_DEV_SPLIT_DEBUGINFO", "packed") + .env("CARGO_PROFILE_TEST_SPLIT_DEBUGINFO", "packed") + .env("CARGO_PROFILE_RELEASE_SPLIT_DEBUGINFO", "packed") + .env("CARGO_PROFILE_BENCH_SPLIT_DEBUGINFO", "packed"); + self + } + + pub fn enable_mac_dsym(&mut self) -> &mut Self { + if cfg!(target_os = "macos") { + return self.enable_split_debuginfo_packed(); + } + self + } + + #[track_caller] + pub fn run(&mut self) { + self.ran = true; + let mut p = (&self.process_builder).clone().unwrap(); + if let Some(stdin) = self.expect_stdin.take() { + p.stdin(stdin); + } + if let Err(e) = self.match_process(&p) { + panic_error(&format!("test failed running {}", p), e); + } + } + + #[track_caller] + pub fn run_expect_error(&mut self) { + self.ran = true; + let p = (&self.process_builder).clone().unwrap(); + if self.match_process(&p).is_ok() { + panic!("test was expected to fail, but succeeded running {}", p); + } + } + + /// Runs the process, checks the expected output, and returns the first + /// JSON object on stdout. + #[track_caller] + pub fn run_json(&mut self) -> serde_json::Value { + self.ran = true; + let p = (&self.process_builder).clone().unwrap(); + match self.match_process(&p) { + Err(e) => panic_error(&format!("test failed running {}", p), e), + Ok(output) => serde_json::from_slice(&output.stdout).unwrap_or_else(|e| { + panic!( + "\nfailed to parse JSON: {}\n\ + output was:\n{}\n", + e, + String::from_utf8_lossy(&output.stdout) + ); + }), + } + } + + #[track_caller] + pub fn run_output(&mut self, output: &Output) { + self.ran = true; + if let Err(e) = self.match_output(output.status.code(), &output.stdout, &output.stderr) { + panic_error("process did not return the expected result", e) + } + } + + fn verify_checks_output(&self, stdout: &[u8], stderr: &[u8]) { + if self.expect_exit_code.unwrap_or(0) != 0 + && self.expect_stdout.is_none() + && self.expect_stdin.is_none() + && self.expect_stderr.is_none() + && self.expect_stdout_contains.is_empty() + && self.expect_stderr_contains.is_empty() + && self.expect_stdout_contains_n.is_empty() + && self.expect_stdout_not_contains.is_empty() + && self.expect_stderr_not_contains.is_empty() + && self.expect_stderr_unordered.is_empty() + && self.expect_stderr_with_without.is_empty() + && self.expect_json.is_none() + && self.expect_json_contains_unordered.is_none() + { + panic!( + "`with_status()` is used, but no output is checked.\n\ + The test must check the output to ensure the correct error is triggered.\n\ + --- stdout\n{}\n--- stderr\n{}", + String::from_utf8_lossy(stdout), + String::from_utf8_lossy(stderr), + ); + } + } + + fn match_process(&self, process: &ProcessBuilder) -> Result<RawOutput> { + println!("running {}", process); + let res = if self.stream_output { + if is_ci() { + panic!("`.stream()` is for local debugging") + } + process.exec_with_streaming( + &mut |out| { + println!("{}", out); + Ok(()) + }, + &mut |err| { + eprintln!("{}", err); + Ok(()) + }, + true, + ) + } else { + process.exec_with_output() + }; + + match res { + Ok(out) => { + self.match_output(out.status.code(), &out.stdout, &out.stderr)?; + return Ok(RawOutput { + stdout: out.stdout, + stderr: out.stderr, + code: out.status.code(), + }); + } + Err(e) => { + if let Some(ProcessError { + stdout: Some(stdout), + stderr: Some(stderr), + code, + .. + }) = e.downcast_ref::<ProcessError>() + { + self.match_output(*code, stdout, stderr)?; + return Ok(RawOutput { + stdout: stdout.to_vec(), + stderr: stderr.to_vec(), + code: *code, + }); + } + bail!("could not exec process {}: {:?}", process, e) + } + } + } + + fn match_output(&self, code: Option<i32>, stdout: &[u8], stderr: &[u8]) -> Result<()> { + self.verify_checks_output(stdout, stderr); + let stdout = str::from_utf8(stdout).expect("stdout is not utf8"); + let stderr = str::from_utf8(stderr).expect("stderr is not utf8"); + let cwd = self.get_cwd(); + + match self.expect_exit_code { + None => {} + Some(expected) if code == Some(expected) => {} + Some(expected) => bail!( + "process exited with code {} (expected {})\n--- stdout\n{}\n--- stderr\n{}", + code.unwrap_or(-1), + expected, + stdout, + stderr + ), + } + + if let Some(expect_stdout) = &self.expect_stdout { + compare::match_exact(expect_stdout, stdout, "stdout", stderr, cwd)?; + } + if let Some(expect_stderr) = &self.expect_stderr { + compare::match_exact(expect_stderr, stderr, "stderr", stdout, cwd)?; + } + for expect in self.expect_stdout_contains.iter() { + compare::match_contains(expect, stdout, cwd)?; + } + for expect in self.expect_stderr_contains.iter() { + compare::match_contains(expect, stderr, cwd)?; + } + for &(ref expect, number) in self.expect_stdout_contains_n.iter() { + compare::match_contains_n(expect, number, stdout, cwd)?; + } + for expect in self.expect_stdout_not_contains.iter() { + compare::match_does_not_contain(expect, stdout, cwd)?; + } + for expect in self.expect_stderr_not_contains.iter() { + compare::match_does_not_contain(expect, stderr, cwd)?; + } + for expect in self.expect_stderr_unordered.iter() { + compare::match_unordered(expect, stderr, cwd)?; + } + for (with, without) in self.expect_stderr_with_without.iter() { + compare::match_with_without(stderr, with, without, cwd)?; + } + + if let Some(ref expect_json) = self.expect_json { + compare::match_json(expect_json, stdout, cwd)?; + } + + if let Some(ref expected) = self.expect_json_contains_unordered { + compare::match_json_contains_unordered(expected, stdout, cwd)?; + } + Ok(()) + } +} + +impl Drop for Execs { + fn drop(&mut self) { + if !self.ran && !std::thread::panicking() { + panic!("forgot to run this command"); + } + } +} + +pub fn execs() -> Execs { + Execs { + ran: false, + process_builder: None, + expect_stdout: None, + expect_stderr: None, + expect_stdin: None, + expect_exit_code: Some(0), + expect_stdout_contains: Vec::new(), + expect_stderr_contains: Vec::new(), + expect_stdout_contains_n: Vec::new(), + expect_stdout_not_contains: Vec::new(), + expect_stderr_not_contains: Vec::new(), + expect_stderr_unordered: Vec::new(), + expect_stderr_with_without: Vec::new(), + expect_json: None, + expect_json_contains_unordered: None, + stream_output: false, + } +} + +pub fn basic_manifest(name: &str, version: &str) -> String { + format!( + r#" + [package] + name = "{}" + version = "{}" + authors = [] + "#, + name, version + ) +} + +pub fn basic_bin_manifest(name: &str) -> String { + format!( + r#" + [package] + + name = "{}" + version = "0.5.0" + authors = ["wycats@example.com"] + + [[bin]] + + name = "{}" + "#, + name, name + ) +} + +pub fn basic_lib_manifest(name: &str) -> String { + format!( + r#" + [package] + + name = "{}" + version = "0.5.0" + authors = ["wycats@example.com"] + + [lib] + + name = "{}" + "#, + name, name + ) +} + +pub fn path2url<P: AsRef<Path>>(p: P) -> Url { + Url::from_file_path(p).ok().unwrap() +} + +struct RustcInfo { + verbose_version: String, + host: String, +} + +impl RustcInfo { + fn new() -> RustcInfo { + let output = ProcessBuilder::new("rustc") + .arg("-vV") + .exec_with_output() + .expect("rustc should exec"); + let verbose_version = String::from_utf8(output.stdout).expect("utf8 output"); + let host = verbose_version + .lines() + .filter_map(|line| line.strip_prefix("host: ")) + .next() + .expect("verbose version has host: field") + .to_string(); + RustcInfo { + verbose_version, + host, + } + } +} + +lazy_static::lazy_static! { + static ref RUSTC_INFO: RustcInfo = RustcInfo::new(); +} + +/// The rustc host such as `x86_64-unknown-linux-gnu`. +pub fn rustc_host() -> &'static str { + &RUSTC_INFO.host +} + +/// The host triple suitable for use in a cargo environment variable (uppercased). +pub fn rustc_host_env() -> String { + rustc_host().to_uppercase().replace('-', "_") +} + +pub fn is_nightly() -> bool { + let vv = &RUSTC_INFO.verbose_version; + // CARGO_TEST_DISABLE_NIGHTLY is set in rust-lang/rust's CI so that all + // nightly-only tests are disabled there. Otherwise, it could make it + // difficult to land changes which would need to be made simultaneously in + // rust-lang/cargo and rust-lan/rust, which isn't possible. + env::var("CARGO_TEST_DISABLE_NIGHTLY").is_err() + && (vv.contains("-nightly") || vv.contains("-dev")) +} + +pub fn process<T: AsRef<OsStr>>(t: T) -> ProcessBuilder { + _process(t.as_ref()) +} + +fn _process(t: &OsStr) -> ProcessBuilder { + let mut p = ProcessBuilder::new(t); + p.cwd(&paths::root()).test_env(); + p +} + +/// Enable nightly features for testing +pub trait ChannelChanger { + /// The list of reasons should be why nightly cargo is needed. If it is + /// becuase of an unstable feature put the name of the feature as the reason, + /// e.g. `&["print-im-a-teapot"]`. + fn masquerade_as_nightly_cargo(self, _reasons: &[&str]) -> Self; +} + +impl ChannelChanger for &mut ProcessBuilder { + fn masquerade_as_nightly_cargo(self, _reasons: &[&str]) -> Self { + self.env("__CARGO_TEST_CHANNEL_OVERRIDE_DO_NOT_USE_THIS", "nightly") + } +} + +impl ChannelChanger for snapbox::cmd::Command { + fn masquerade_as_nightly_cargo(self, _reasons: &[&str]) -> Self { + self.env("__CARGO_TEST_CHANNEL_OVERRIDE_DO_NOT_USE_THIS", "nightly") + } +} + +/// Establish a process's test environment +pub trait TestEnv: Sized { + fn test_env(mut self) -> Self { + // In general just clear out all cargo-specific configuration already in the + // environment. Our tests all assume a "default configuration" unless + // specified otherwise. + for (k, _v) in env::vars() { + if k.starts_with("CARGO_") { + self = self.env_remove(&k); + } + } + if env::var_os("RUSTUP_TOOLCHAIN").is_some() { + // Override the PATH to avoid executing the rustup wrapper thousands + // of times. This makes the testsuite run substantially faster. + lazy_static::lazy_static! { + static ref RUSTC_DIR: PathBuf = { + match ProcessBuilder::new("rustup") + .args(&["which", "rustc"]) + .exec_with_output() + { + Ok(output) => { + let s = str::from_utf8(&output.stdout).expect("utf8").trim(); + let mut p = PathBuf::from(s); + p.pop(); + p + } + Err(e) => { + panic!("RUSTUP_TOOLCHAIN was set, but could not run rustup: {}", e); + } + } + }; + } + let path = env::var_os("PATH").unwrap_or_default(); + let paths = env::split_paths(&path); + let new_path = + env::join_paths(std::iter::once(RUSTC_DIR.clone()).chain(paths)).unwrap(); + self = self.env("PATH", new_path); + } + + self = self + .current_dir(&paths::root()) + .env("HOME", paths::home()) + .env("CARGO_HOME", paths::home().join(".cargo")) + .env("__CARGO_TEST_ROOT", paths::global_root()) + // Force Cargo to think it's on the stable channel for all tests, this + // should hopefully not surprise us as we add cargo features over time and + // cargo rides the trains. + .env("__CARGO_TEST_CHANNEL_OVERRIDE_DO_NOT_USE_THIS", "stable") + // Keeps cargo within its sandbox. + .env("__CARGO_TEST_DISABLE_GLOBAL_KNOWN_HOST", "1") + // Incremental generates a huge amount of data per test, which we + // don't particularly need. Tests that specifically need to check + // the incremental behavior should turn this back on. + .env("CARGO_INCREMENTAL", "0") + // Don't read the system git config which is out of our control. + .env("GIT_CONFIG_NOSYSTEM", "1") + .env_remove("__CARGO_DEFAULT_LIB_METADATA") + .env_remove("ALL_PROXY") + .env_remove("EMAIL") + .env_remove("GIT_AUTHOR_EMAIL") + .env_remove("GIT_AUTHOR_NAME") + .env_remove("GIT_COMMITTER_EMAIL") + .env_remove("GIT_COMMITTER_NAME") + .env_remove("http_proxy") + .env_remove("HTTPS_PROXY") + .env_remove("https_proxy") + .env_remove("MAKEFLAGS") + .env_remove("MFLAGS") + .env_remove("MSYSTEM") // assume cmd.exe everywhere on windows + .env_remove("RUSTC") + .env_remove("RUSTC_WORKSPACE_WRAPPER") + .env_remove("RUSTC_WRAPPER") + .env_remove("RUSTDOC") + .env_remove("RUSTDOCFLAGS") + .env_remove("RUSTFLAGS") + .env_remove("SSH_AUTH_SOCK") // ensure an outer agent is never contacted + .env_remove("USER") // not set on some rust-lang docker images + .env_remove("XDG_CONFIG_HOME"); // see #2345 + if cfg!(target_os = "macos") { + // Work-around a bug in macOS 10.15, see `link_or_copy` for details. + self = self.env("__CARGO_COPY_DONT_LINK_DO_NOT_USE_THIS", "1"); + } + if cfg!(windows) { + self = self.env("USERPROFILE", paths::home()); + } + self + } + + fn current_dir<S: AsRef<std::path::Path>>(self, path: S) -> Self; + fn env<S: AsRef<std::ffi::OsStr>>(self, key: &str, value: S) -> Self; + fn env_remove(self, key: &str) -> Self; +} + +impl TestEnv for &mut ProcessBuilder { + fn current_dir<S: AsRef<std::path::Path>>(self, path: S) -> Self { + let path = path.as_ref(); + self.cwd(path) + } + fn env<S: AsRef<std::ffi::OsStr>>(self, key: &str, value: S) -> Self { + self.env(key, value) + } + fn env_remove(self, key: &str) -> Self { + self.env_remove(key) + } +} + +impl TestEnv for snapbox::cmd::Command { + fn current_dir<S: AsRef<std::path::Path>>(self, path: S) -> Self { + self.current_dir(path) + } + fn env<S: AsRef<std::ffi::OsStr>>(self, key: &str, value: S) -> Self { + self.env(key, value) + } + fn env_remove(self, key: &str) -> Self { + self.env_remove(key) + } +} + +/// Test the cargo command +pub trait CargoCommand { + fn cargo_ui() -> Self; +} + +impl CargoCommand for snapbox::cmd::Command { + fn cargo_ui() -> Self { + Self::new(cargo_exe()) + .with_assert(compare::assert_ui()) + .test_env() + } +} + +/// Add a list of arguments as a line +pub trait ArgLine: Sized { + fn arg_line(mut self, s: &str) -> Self { + for mut arg in s.split_whitespace() { + if (arg.starts_with('"') && arg.ends_with('"')) + || (arg.starts_with('\'') && arg.ends_with('\'')) + { + arg = &arg[1..(arg.len() - 1).max(1)]; + } else if arg.contains(&['"', '\''][..]) { + panic!("shell-style argument parsing is not supported") + } + self = self.arg(arg); + } + self + } + + fn arg<S: AsRef<std::ffi::OsStr>>(self, s: S) -> Self; +} + +impl ArgLine for &mut ProcessBuilder { + fn arg<S: AsRef<std::ffi::OsStr>>(self, s: S) -> Self { + self.arg(s) + } +} + +impl ArgLine for snapbox::cmd::Command { + fn arg<S: AsRef<std::ffi::OsStr>>(self, s: S) -> Self { + self.arg(s) + } +} + +pub fn cargo_process(s: &str) -> Execs { + let cargo = cargo_exe(); + let mut p = process(&cargo); + p.env("CARGO", cargo); + p.arg_line(s); + execs().with_process_builder(p) +} + +pub fn git_process(s: &str) -> ProcessBuilder { + let mut p = process("git"); + p.arg_line(s); + p +} + +pub fn sleep_ms(ms: u64) { + ::std::thread::sleep(Duration::from_millis(ms)); +} + +/// Returns `true` if the local filesystem has low-resolution mtimes. +pub fn is_coarse_mtime() -> bool { + // If the filetime crate is being used to emulate HFS then + // return `true`, without looking at the actual hardware. + cfg!(emulate_second_only_system) || + // This should actually be a test that `$CARGO_TARGET_DIR` is on an HFS + // filesystem, (or any filesystem with low-resolution mtimes). However, + // that's tricky to detect, so for now just deal with CI. + cfg!(target_os = "macos") && is_ci() +} + +/// Some CI setups are much slower then the equipment used by Cargo itself. +/// Architectures that do not have a modern processor, hardware emulation, etc. +/// This provides a way for those setups to increase the cut off for all the time based test. +pub fn slow_cpu_multiplier(main: u64) -> Duration { + lazy_static::lazy_static! { + static ref SLOW_CPU_MULTIPLIER: u64 = + env::var("CARGO_TEST_SLOW_CPU_MULTIPLIER").ok().and_then(|m| m.parse().ok()).unwrap_or(1); + } + Duration::from_secs(*SLOW_CPU_MULTIPLIER * main) +} + +#[cfg(windows)] +pub fn symlink_supported() -> bool { + if is_ci() { + // We want to be absolutely sure this runs on CI. + return true; + } + let src = paths::root().join("symlink_src"); + fs::write(&src, "").unwrap(); + let dst = paths::root().join("symlink_dst"); + let result = match os::windows::fs::symlink_file(&src, &dst) { + Ok(_) => { + fs::remove_file(&dst).unwrap(); + true + } + Err(e) => { + eprintln!( + "symlinks not supported: {:?}\n\ + Windows 10 users should enable developer mode.", + e + ); + false + } + }; + fs::remove_file(&src).unwrap(); + return result; +} + +#[cfg(not(windows))] +pub fn symlink_supported() -> bool { + true +} + +/// The error message for ENOENT. +pub fn no_such_file_err_msg() -> String { + std::io::Error::from_raw_os_error(2).to_string() +} diff --git a/crates/cargo-test-support/src/paths.rs b/crates/cargo-test-support/src/paths.rs new file mode 100644 index 0000000..ef1fddb --- /dev/null +++ b/crates/cargo-test-support/src/paths.rs @@ -0,0 +1,347 @@ +use filetime::{self, FileTime}; +use lazy_static::lazy_static; +use std::cell::RefCell; +use std::collections::HashMap; +use std::env; +use std::fs; +use std::io::{self, ErrorKind}; +use std::path::{Path, PathBuf}; +use std::process::Command; +use std::sync::atomic::{AtomicUsize, Ordering}; +use std::sync::Mutex; + +static CARGO_INTEGRATION_TEST_DIR: &str = "cit"; + +lazy_static! { + // TODO: Use `SyncOnceCell` when stable + static ref GLOBAL_ROOT: Mutex<Option<PathBuf>> = Mutex::new(None); + + static ref TEST_ROOTS: Mutex<HashMap<String, PathBuf>> = Default::default(); +} + +/// This is used when running cargo is pre-CARGO_TARGET_TMPDIR +/// TODO: Remove when CARGO_TARGET_TMPDIR grows old enough. +fn global_root_legacy() -> PathBuf { + let mut path = t!(env::current_exe()); + path.pop(); // chop off exe name + path.pop(); // chop off "deps" + path.push("tmp"); + path.mkdir_p(); + path +} + +fn set_global_root(tmp_dir: Option<&'static str>) { + let mut lock = GLOBAL_ROOT.lock().unwrap(); + if lock.is_none() { + let mut root = match tmp_dir { + Some(tmp_dir) => PathBuf::from(tmp_dir), + None => global_root_legacy(), + }; + + root.push(CARGO_INTEGRATION_TEST_DIR); + *lock = Some(root); + } +} + +pub fn global_root() -> PathBuf { + let lock = GLOBAL_ROOT.lock().unwrap(); + match lock.as_ref() { + Some(p) => p.clone(), + None => unreachable!("GLOBAL_ROOT not set yet"), + } +} + +// We need to give each test a unique id. The test name could serve this +// purpose, but the `test` crate doesn't have a way to obtain the current test +// name.[*] Instead, we used the `cargo-test-macro` crate to automatically +// insert an init function for each test that sets the test name in a thread +// local variable. +// +// [*] It does set the thread name, but only when running concurrently. If not +// running concurrently, all tests are run on the main thread. +thread_local! { + static TEST_ID: RefCell<Option<usize>> = RefCell::new(None); +} + +pub struct TestIdGuard { + _private: (), +} + +pub fn init_root(tmp_dir: Option<&'static str>) -> TestIdGuard { + static NEXT_ID: AtomicUsize = AtomicUsize::new(0); + + let id = NEXT_ID.fetch_add(1, Ordering::SeqCst); + TEST_ID.with(|n| *n.borrow_mut() = Some(id)); + + let guard = TestIdGuard { _private: () }; + + set_global_root(tmp_dir); + let r = root(); + r.rm_rf(); + r.mkdir_p(); + + guard +} + +impl Drop for TestIdGuard { + fn drop(&mut self) { + TEST_ID.with(|n| *n.borrow_mut() = None); + } +} + +pub fn root() -> PathBuf { + let id = TEST_ID.with(|n| { + n.borrow().expect( + "Tests must use the `#[cargo_test]` attribute in \ + order to be able to use the crate root.", + ) + }); + + let mut root = global_root(); + root.push(&format!("t{}", id)); + root +} + +pub fn home() -> PathBuf { + let mut path = root(); + path.push("home"); + path.mkdir_p(); + path +} + +pub trait CargoPathExt { + fn rm_rf(&self); + fn mkdir_p(&self); + + fn move_into_the_past(&self) { + self.move_in_time(|sec, nsec| (sec - 3600, nsec)) + } + + fn move_into_the_future(&self) { + self.move_in_time(|sec, nsec| (sec + 3600, nsec)) + } + + fn move_in_time<F>(&self, travel_amount: F) + where + F: Fn(i64, u32) -> (i64, u32); +} + +impl CargoPathExt for Path { + fn rm_rf(&self) { + let meta = match self.symlink_metadata() { + Ok(meta) => meta, + Err(e) => { + if e.kind() == ErrorKind::NotFound { + return; + } + panic!("failed to remove {:?}, could not read: {:?}", self, e); + } + }; + // There is a race condition between fetching the metadata and + // actually performing the removal, but we don't care all that much + // for our tests. + if meta.is_dir() { + if let Err(e) = fs::remove_dir_all(self) { + panic!("failed to remove {:?}: {:?}", self, e) + } + } else if let Err(e) = fs::remove_file(self) { + panic!("failed to remove {:?}: {:?}", self, e) + } + } + + fn mkdir_p(&self) { + fs::create_dir_all(self) + .unwrap_or_else(|e| panic!("failed to mkdir_p {}: {}", self.display(), e)) + } + + fn move_in_time<F>(&self, travel_amount: F) + where + F: Fn(i64, u32) -> (i64, u32), + { + if self.is_file() { + time_travel(self, &travel_amount); + } else { + recurse(self, &self.join("target"), &travel_amount); + } + + fn recurse<F>(p: &Path, bad: &Path, travel_amount: &F) + where + F: Fn(i64, u32) -> (i64, u32), + { + if p.is_file() { + time_travel(p, travel_amount) + } else if !p.starts_with(bad) { + for f in t!(fs::read_dir(p)) { + let f = t!(f).path(); + recurse(&f, bad, travel_amount); + } + } + } + + fn time_travel<F>(path: &Path, travel_amount: &F) + where + F: Fn(i64, u32) -> (i64, u32), + { + let stat = t!(path.symlink_metadata()); + + let mtime = FileTime::from_last_modification_time(&stat); + + let (sec, nsec) = travel_amount(mtime.unix_seconds(), mtime.nanoseconds()); + let newtime = FileTime::from_unix_time(sec, nsec); + + // Sadly change_file_times has a failure mode where a readonly file + // cannot have its times changed on windows. + do_op(path, "set file times", |path| { + filetime::set_file_times(path, newtime, newtime) + }); + } + } +} + +fn do_op<F>(path: &Path, desc: &str, mut f: F) +where + F: FnMut(&Path) -> io::Result<()>, +{ + match f(path) { + Ok(()) => {} + Err(ref e) if e.kind() == ErrorKind::PermissionDenied => { + let mut p = t!(path.metadata()).permissions(); + p.set_readonly(false); + t!(fs::set_permissions(path, p)); + + // Unix also requires the parent to not be readonly for example when + // removing files + let parent = path.parent().unwrap(); + let mut p = t!(parent.metadata()).permissions(); + p.set_readonly(false); + t!(fs::set_permissions(parent, p)); + + f(path).unwrap_or_else(|e| { + panic!("failed to {} {}: {}", desc, path.display(), e); + }) + } + Err(e) => { + panic!("failed to {} {}: {}", desc, path.display(), e); + } + } +} + +/// Get the filename for a library. +/// +/// `kind` should be one of: "lib", "rlib", "staticlib", "dylib", "proc-macro" +/// +/// For example, dynamic library named "foo" would return: +/// - macOS: "libfoo.dylib" +/// - Windows: "foo.dll" +/// - Unix: "libfoo.so" +pub fn get_lib_filename(name: &str, kind: &str) -> String { + let prefix = get_lib_prefix(kind); + let extension = get_lib_extension(kind); + format!("{}{}.{}", prefix, name, extension) +} + +pub fn get_lib_prefix(kind: &str) -> &str { + match kind { + "lib" | "rlib" => "lib", + "staticlib" | "dylib" | "proc-macro" => { + if cfg!(windows) { + "" + } else { + "lib" + } + } + _ => unreachable!(), + } +} + +pub fn get_lib_extension(kind: &str) -> &str { + match kind { + "lib" | "rlib" => "rlib", + "staticlib" => { + if cfg!(windows) { + "lib" + } else { + "a" + } + } + "dylib" | "proc-macro" => { + if cfg!(windows) { + "dll" + } else if cfg!(target_os = "macos") { + "dylib" + } else { + "so" + } + } + _ => unreachable!(), + } +} + +/// Returns the sysroot as queried from rustc. +pub fn sysroot() -> String { + let output = Command::new("rustc") + .arg("--print=sysroot") + .output() + .expect("rustc to run"); + assert!(output.status.success()); + let sysroot = String::from_utf8(output.stdout).unwrap(); + sysroot.trim().to_string() +} + +/// Returns true if names such as aux.* are allowed. +/// +/// Traditionally, Windows did not allow a set of file names (see `is_windows_reserved` +/// for a list). More recent versions of Windows have relaxed this restriction. This test +/// determines whether we are running in a mode that allows Windows reserved names. +#[cfg(windows)] +pub fn windows_reserved_names_are_allowed() -> bool { + use cargo_util::is_ci; + + // Ensure tests still run in CI until we need to migrate. + if is_ci() { + return false; + } + + use std::ffi::OsStr; + use std::os::windows::ffi::OsStrExt; + use std::ptr; + use windows_sys::Win32::Storage::FileSystem::GetFullPathNameW; + + let test_file_name: Vec<_> = OsStr::new("aux.rs").encode_wide().collect(); + + let buffer_length = + unsafe { GetFullPathNameW(test_file_name.as_ptr(), 0, ptr::null_mut(), ptr::null_mut()) }; + + if buffer_length == 0 { + // This means the call failed, so we'll conservatively assume reserved names are not allowed. + return false; + } + + let mut buffer = vec![0u16; buffer_length as usize]; + + let result = unsafe { + GetFullPathNameW( + test_file_name.as_ptr(), + buffer_length, + buffer.as_mut_ptr(), + ptr::null_mut(), + ) + }; + + if result == 0 { + // Once again, conservatively assume reserved names are not allowed if the + // GetFullPathNameW call failed. + return false; + } + + // Under the old rules, a file name like aux.rs would get converted into \\.\aux, so + // we detect this case by checking if the string starts with \\.\ + // + // Otherwise, the filename will be something like C:\Users\Foo\Documents\aux.rs + let prefix: Vec<_> = OsStr::new("\\\\.\\").encode_wide().collect(); + if buffer.starts_with(&prefix) { + false + } else { + true + } +} diff --git a/crates/cargo-test-support/src/publish.rs b/crates/cargo-test-support/src/publish.rs new file mode 100644 index 0000000..85bc93c --- /dev/null +++ b/crates/cargo-test-support/src/publish.rs @@ -0,0 +1,245 @@ +use crate::compare::{assert_match_exact, find_json_mismatch}; +use crate::registry::{self, alt_api_path, FeatureMap}; +use flate2::read::GzDecoder; +use std::collections::{HashMap, HashSet}; +use std::fs; +use std::fs::File; +use std::io::{self, prelude::*, SeekFrom}; +use std::path::{Path, PathBuf}; +use tar::Archive; + +fn read_le_u32<R>(mut reader: R) -> io::Result<u32> +where + R: Read, +{ + let mut buf = [0; 4]; + reader.read_exact(&mut buf)?; + Ok(u32::from_le_bytes(buf)) +} + +/// Checks the result of a crate publish. +pub fn validate_upload(expected_json: &str, expected_crate_name: &str, expected_files: &[&str]) { + let new_path = registry::api_path().join("api/v1/crates/new"); + _validate_upload( + &new_path, + expected_json, + expected_crate_name, + expected_files, + &[], + ); +} + +/// Checks the result of a crate publish, along with the contents of the files. +pub fn validate_upload_with_contents( + expected_json: &str, + expected_crate_name: &str, + expected_files: &[&str], + expected_contents: &[(&str, &str)], +) { + let new_path = registry::api_path().join("api/v1/crates/new"); + _validate_upload( + &new_path, + expected_json, + expected_crate_name, + expected_files, + expected_contents, + ); +} + +/// Checks the result of a crate publish to an alternative registry. +pub fn validate_alt_upload( + expected_json: &str, + expected_crate_name: &str, + expected_files: &[&str], +) { + let new_path = alt_api_path().join("api/v1/crates/new"); + _validate_upload( + &new_path, + expected_json, + expected_crate_name, + expected_files, + &[], + ); +} + +fn _validate_upload( + new_path: &Path, + expected_json: &str, + expected_crate_name: &str, + expected_files: &[&str], + expected_contents: &[(&str, &str)], +) { + let mut f = File::open(new_path).unwrap(); + // 32-bit little-endian integer of length of JSON data. + let json_sz = read_le_u32(&mut f).expect("read json length"); + let mut json_bytes = vec![0; json_sz as usize]; + f.read_exact(&mut json_bytes).expect("read JSON data"); + let actual_json = serde_json::from_slice(&json_bytes).expect("uploaded JSON should be valid"); + let expected_json = serde_json::from_str(expected_json).expect("expected JSON does not parse"); + + if let Err(e) = find_json_mismatch(&expected_json, &actual_json, None) { + panic!("{}", e); + } + + // 32-bit little-endian integer of length of crate file. + let crate_sz = read_le_u32(&mut f).expect("read crate length"); + let mut krate_bytes = vec![0; crate_sz as usize]; + f.read_exact(&mut krate_bytes).expect("read crate data"); + // Check at end. + let current = f.seek(SeekFrom::Current(0)).unwrap(); + assert_eq!(f.seek(SeekFrom::End(0)).unwrap(), current); + + // Verify the tarball. + validate_crate_contents( + &krate_bytes[..], + expected_crate_name, + expected_files, + expected_contents, + ); +} + +/// Checks the contents of a `.crate` file. +/// +/// - `expected_crate_name` should be something like `foo-0.0.1.crate`. +/// - `expected_files` should be a complete list of files in the crate +/// (relative to expected_crate_name). +/// - `expected_contents` should be a list of `(file_name, contents)` tuples +/// to validate the contents of the given file. Only the listed files will +/// be checked (others will be ignored). +pub fn validate_crate_contents( + reader: impl Read, + expected_crate_name: &str, + expected_files: &[&str], + expected_contents: &[(&str, &str)], +) { + let mut rdr = GzDecoder::new(reader); + assert_eq!( + rdr.header().unwrap().filename().unwrap(), + expected_crate_name.as_bytes() + ); + let mut contents = Vec::new(); + rdr.read_to_end(&mut contents).unwrap(); + let mut ar = Archive::new(&contents[..]); + let files: HashMap<PathBuf, String> = ar + .entries() + .unwrap() + .map(|entry| { + let mut entry = entry.unwrap(); + let name = entry.path().unwrap().into_owned(); + let mut contents = String::new(); + entry.read_to_string(&mut contents).unwrap(); + (name, contents) + }) + .collect(); + assert!(expected_crate_name.ends_with(".crate")); + let base_crate_name = Path::new(&expected_crate_name[..expected_crate_name.len() - 6]); + let actual_files: HashSet<PathBuf> = files.keys().cloned().collect(); + let expected_files: HashSet<PathBuf> = expected_files + .iter() + .map(|name| base_crate_name.join(name)) + .collect(); + let missing: Vec<&PathBuf> = expected_files.difference(&actual_files).collect(); + let extra: Vec<&PathBuf> = actual_files.difference(&expected_files).collect(); + if !missing.is_empty() || !extra.is_empty() { + panic!( + "uploaded archive does not match.\nMissing: {:?}\nExtra: {:?}\n", + missing, extra + ); + } + if !expected_contents.is_empty() { + for (e_file_name, e_file_contents) in expected_contents { + let full_e_name = base_crate_name.join(e_file_name); + let actual_contents = files + .get(&full_e_name) + .unwrap_or_else(|| panic!("file `{}` missing in archive", e_file_name)); + assert_match_exact(e_file_contents, actual_contents); + } + } +} + +pub(crate) fn create_index_line( + name: serde_json::Value, + vers: &str, + deps: Vec<serde_json::Value>, + cksum: &str, + features: crate::registry::FeatureMap, + yanked: bool, + links: Option<String>, + v: Option<u32>, +) -> String { + // This emulates what crates.io does to retain backwards compatibility. + let (features, features2) = split_index_features(features.clone()); + let mut json = serde_json::json!({ + "name": name, + "vers": vers, + "deps": deps, + "cksum": cksum, + "features": features, + "yanked": yanked, + "links": links, + }); + if let Some(f2) = &features2 { + json["features2"] = serde_json::json!(f2); + json["v"] = serde_json::json!(2); + } + if let Some(v) = v { + json["v"] = serde_json::json!(v); + } + + json.to_string() +} + +pub(crate) fn write_to_index(registry_path: &PathBuf, name: &str, line: String, local: bool) { + let file = cargo_util::registry::make_dep_path(name, false); + + // Write file/line in the index. + let dst = if local { + registry_path.join("index").join(&file) + } else { + registry_path.join(&file) + }; + let prev = fs::read_to_string(&dst).unwrap_or_default(); + t!(fs::create_dir_all(dst.parent().unwrap())); + t!(fs::write(&dst, prev + &line[..] + "\n")); + + // Add the new file to the index. + if !local { + let repo = t!(git2::Repository::open(®istry_path)); + let mut index = t!(repo.index()); + t!(index.add_path(Path::new(&file))); + t!(index.write()); + let id = t!(index.write_tree()); + + // Commit this change. + let tree = t!(repo.find_tree(id)); + let sig = t!(repo.signature()); + let parent = t!(repo.refname_to_id("refs/heads/master")); + let parent = t!(repo.find_commit(parent)); + t!(repo.commit( + Some("HEAD"), + &sig, + &sig, + "Another commit", + &tree, + &[&parent] + )); + } +} + +fn split_index_features(mut features: FeatureMap) -> (FeatureMap, Option<FeatureMap>) { + let mut features2 = FeatureMap::new(); + for (feat, values) in features.iter_mut() { + if values + .iter() + .any(|value| value.starts_with("dep:") || value.contains("?/")) + { + let new_values = values.drain(..).collect(); + features2.insert(feat.clone(), new_values); + } + } + if features2.is_empty() { + (features, None) + } else { + (features, Some(features2)) + } +} diff --git a/crates/cargo-test-support/src/registry.rs b/crates/cargo-test-support/src/registry.rs new file mode 100644 index 0000000..7b1dc54 --- /dev/null +++ b/crates/cargo-test-support/src/registry.rs @@ -0,0 +1,1581 @@ +use crate::git::repo; +use crate::paths; +use crate::publish::{create_index_line, write_to_index}; +use cargo_util::paths::append; +use cargo_util::Sha256; +use flate2::write::GzEncoder; +use flate2::Compression; +use pasetors::keys::{AsymmetricPublicKey, AsymmetricSecretKey}; +use pasetors::paserk::FormatAsPaserk; +use pasetors::token::UntrustedToken; +use std::collections::{BTreeMap, HashMap}; +use std::fmt; +use std::fs::{self, File}; +use std::io::{BufRead, BufReader, Read, Write}; +use std::net::{SocketAddr, TcpListener, TcpStream}; +use std::path::PathBuf; +use std::thread::{self, JoinHandle}; +use tar::{Builder, Header}; +use time::format_description::well_known::Rfc3339; +use time::{Duration, OffsetDateTime}; +use url::Url; + +/// Gets the path to the local index pretending to be crates.io. This is a Git repo +/// initialized with a `config.json` file pointing to `dl_path` for downloads +/// and `api_path` for uploads. +pub fn registry_path() -> PathBuf { + generate_path("registry") +} +/// Gets the path for local web API uploads. Cargo will place the contents of a web API +/// request here. For example, `api/v1/crates/new` is the result of publishing a crate. +pub fn api_path() -> PathBuf { + generate_path("api") +} +/// Gets the path where crates can be downloaded using the web API endpoint. Crates +/// should be organized as `{name}/{version}/download` to match the web API +/// endpoint. This is rarely used and must be manually set up. +fn dl_path() -> PathBuf { + generate_path("dl") +} +/// Gets the alternative-registry version of `registry_path`. +fn alt_registry_path() -> PathBuf { + generate_path("alternative-registry") +} +/// Gets the alternative-registry version of `registry_url`. +fn alt_registry_url() -> Url { + generate_url("alternative-registry") +} +/// Gets the alternative-registry version of `dl_path`. +pub fn alt_dl_path() -> PathBuf { + generate_path("alternative-dl") +} +/// Gets the alternative-registry version of `api_path`. +pub fn alt_api_path() -> PathBuf { + generate_path("alternative-api") +} +fn generate_path(name: &str) -> PathBuf { + paths::root().join(name) +} +fn generate_url(name: &str) -> Url { + Url::from_file_path(generate_path(name)).ok().unwrap() +} + +#[derive(Clone)] +pub enum Token { + Plaintext(String), + Keys(String, Option<String>), +} + +impl Token { + /// This is a valid PASETO secret key. + /// This one is already publicly available as part of the text of the RFC so is safe to use for tests. + pub fn rfc_key() -> Token { + Token::Keys( + "k3.secret.fNYVuMvBgOlljt9TDohnaYLblghqaHoQquVZwgR6X12cBFHZLFsaU3q7X3k1Zn36" + .to_string(), + Some("sub".to_string()), + ) + } +} + +/// A builder for initializing registries. +pub struct RegistryBuilder { + /// If set, configures an alternate registry with the given name. + alternative: Option<String>, + /// The authorization token for the registry. + token: Option<Token>, + /// If set, the registry requires authorization for all operations. + auth_required: bool, + /// If set, serves the index over http. + http_index: bool, + /// If set, serves the API over http. + http_api: bool, + /// If set, config.json includes 'api' + api: bool, + /// Write the token in the configuration. + configure_token: bool, + /// Write the registry in configuration. + configure_registry: bool, + /// API responders. + custom_responders: HashMap<&'static str, Box<dyn Send + Fn(&Request, &HttpServer) -> Response>>, +} + +pub struct TestRegistry { + server: Option<HttpServerHandle>, + index_url: Url, + path: PathBuf, + api_url: Url, + dl_url: Url, + token: Token, +} + +impl TestRegistry { + pub fn index_url(&self) -> &Url { + &self.index_url + } + + pub fn api_url(&self) -> &Url { + &self.api_url + } + + pub fn token(&self) -> &str { + match &self.token { + Token::Plaintext(s) => s, + Token::Keys(_, _) => panic!("registry was not configured with a plaintext token"), + } + } + + pub fn key(&self) -> &str { + match &self.token { + Token::Plaintext(_) => panic!("registry was not configured with a secret key"), + Token::Keys(s, _) => s, + } + } + + /// Shutdown the server thread and wait for it to stop. + /// `Drop` automatically stops the server, but this additionally + /// waits for the thread to stop. + pub fn join(self) { + if let Some(mut server) = self.server { + server.stop(); + let handle = server.handle.take().unwrap(); + handle.join().unwrap(); + } + } +} + +impl RegistryBuilder { + #[must_use] + pub fn new() -> RegistryBuilder { + RegistryBuilder { + alternative: None, + token: None, + auth_required: false, + http_api: false, + http_index: false, + api: true, + configure_registry: true, + configure_token: true, + custom_responders: HashMap::new(), + } + } + + /// Adds a custom HTTP response for a specific url + #[must_use] + pub fn add_responder<R: 'static + Send + Fn(&Request, &HttpServer) -> Response>( + mut self, + url: &'static str, + responder: R, + ) -> Self { + self.custom_responders.insert(url, Box::new(responder)); + self + } + + /// Sets whether or not to initialize as an alternative registry. + #[must_use] + pub fn alternative_named(mut self, alt: &str) -> Self { + self.alternative = Some(alt.to_string()); + self + } + + /// Sets whether or not to initialize as an alternative registry. + #[must_use] + pub fn alternative(self) -> Self { + self.alternative_named("alternative") + } + + /// Prevents placing a token in the configuration + #[must_use] + pub fn no_configure_token(mut self) -> Self { + self.configure_token = false; + self + } + + /// Prevents adding the registry to the configuration. + #[must_use] + pub fn no_configure_registry(mut self) -> Self { + self.configure_registry = false; + self + } + + /// Sets the token value + #[must_use] + pub fn token(mut self, token: Token) -> Self { + self.token = Some(token); + self + } + + /// Sets this registry to require the authentication token for + /// all operations. + #[must_use] + pub fn auth_required(mut self) -> Self { + self.auth_required = true; + self + } + + /// Operate the index over http + #[must_use] + pub fn http_index(mut self) -> Self { + self.http_index = true; + self + } + + /// Operate the api over http + #[must_use] + pub fn http_api(mut self) -> Self { + self.http_api = true; + self + } + + /// The registry has no api. + #[must_use] + pub fn no_api(mut self) -> Self { + self.api = false; + self + } + + /// Initializes the registry. + #[must_use] + pub fn build(self) -> TestRegistry { + let config_path = paths::home().join(".cargo/config"); + t!(fs::create_dir_all(config_path.parent().unwrap())); + let prefix = if let Some(alternative) = &self.alternative { + format!("{alternative}-") + } else { + String::new() + }; + let registry_path = generate_path(&format!("{prefix}registry")); + let index_url = generate_url(&format!("{prefix}registry")); + let api_url = generate_url(&format!("{prefix}api")); + let dl_url = generate_url(&format!("{prefix}dl")); + let dl_path = generate_path(&format!("{prefix}dl")); + let api_path = generate_path(&format!("{prefix}api")); + let token = self + .token + .unwrap_or_else(|| Token::Plaintext(format!("{prefix}sekrit"))); + + let (server, index_url, api_url, dl_url) = if !self.http_index && !self.http_api { + // No need to start the HTTP server. + (None, index_url, api_url, dl_url) + } else { + let server = HttpServer::new( + registry_path.clone(), + dl_path, + api_path.clone(), + token.clone(), + self.auth_required, + self.custom_responders, + ); + let index_url = if self.http_index { + server.index_url() + } else { + index_url + }; + let api_url = if self.http_api { + server.api_url() + } else { + api_url + }; + let dl_url = server.dl_url(); + (Some(server), index_url, api_url, dl_url) + }; + + let registry = TestRegistry { + api_url, + index_url, + server, + dl_url, + path: registry_path, + token, + }; + + if self.configure_registry { + if let Some(alternative) = &self.alternative { + append( + &config_path, + format!( + " + [registries.{alternative}] + index = '{}'", + registry.index_url + ) + .as_bytes(), + ) + .unwrap(); + } else { + append( + &config_path, + format!( + " + [source.crates-io] + replace-with = 'dummy-registry' + + [registries.dummy-registry] + index = '{}'", + registry.index_url + ) + .as_bytes(), + ) + .unwrap(); + } + } + + if self.configure_token { + let credentials = paths::home().join(".cargo/credentials.toml"); + match ®istry.token { + Token::Plaintext(token) => { + if let Some(alternative) = &self.alternative { + append( + &credentials, + format!( + r#" + [registries.{alternative}] + token = "{token}" + "# + ) + .as_bytes(), + ) + .unwrap(); + } else { + append( + &credentials, + format!( + r#" + [registry] + token = "{token}" + "# + ) + .as_bytes(), + ) + .unwrap(); + } + } + Token::Keys(key, subject) => { + let mut out = if let Some(alternative) = &self.alternative { + format!("\n[registries.{alternative}]\n") + } else { + format!("\n[registry]\n") + }; + out += &format!("secret-key = \"{key}\"\n"); + if let Some(subject) = subject { + out += &format!("secret-key-subject = \"{subject}\"\n"); + } + + append(&credentials, out.as_bytes()).unwrap(); + } + } + } + + let auth = if self.auth_required { + r#","auth-required":true"# + } else { + "" + }; + let api = if self.api { + format!(r#","api":"{}""#, registry.api_url) + } else { + String::new() + }; + // Initialize a new registry. + repo(®istry.path) + .file( + "config.json", + &format!(r#"{{"dl":"{}"{api}{auth}}}"#, registry.dl_url), + ) + .build(); + fs::create_dir_all(api_path.join("api/v1/crates")).unwrap(); + + registry + } +} + +/// A builder for creating a new package in a registry. +/// +/// This uses "source replacement" using an automatically generated +/// `.cargo/config` file to ensure that dependencies will use these packages +/// instead of contacting crates.io. See `source-replacement.md` for more +/// details on how source replacement works. +/// +/// Call `publish` to finalize and create the package. +/// +/// If no files are specified, an empty `lib.rs` file is automatically created. +/// +/// The `Cargo.toml` file is automatically generated based on the methods +/// called on `Package` (for example, calling `dep()` will add to the +/// `[dependencies]` automatically). You may also specify a `Cargo.toml` file +/// to override the generated one. +/// +/// This supports different registry types: +/// - Regular source replacement that replaces `crates.io` (the default). +/// - A "local registry" which is a subset for vendoring (see +/// `Package::local`). +/// - An "alternative registry" which requires specifying the registry name +/// (see `Package::alternative`). +/// +/// This does not support "directory sources". See `directory.rs` for +/// `VendorPackage` which implements directory sources. +/// +/// # Example +/// ``` +/// // Publish package "a" depending on "b". +/// Package::new("a", "1.0.0") +/// .dep("b", "1.0.0") +/// .file("src/lib.rs", r#" +/// extern crate b; +/// pub fn f() -> i32 { b::f() * 2 } +/// "#) +/// .publish(); +/// +/// // Publish package "b". +/// Package::new("b", "1.0.0") +/// .file("src/lib.rs", r#" +/// pub fn f() -> i32 { 12 } +/// "#) +/// .publish(); +/// +/// // Create a project that uses package "a". +/// let p = project() +/// .file("Cargo.toml", r#" +/// [package] +/// name = "foo" +/// version = "0.0.1" +/// +/// [dependencies] +/// a = "1.0" +/// "#) +/// .file("src/main.rs", r#" +/// extern crate a; +/// fn main() { println!("{}", a::f()); } +/// "#) +/// .build(); +/// +/// p.cargo("run").with_stdout("24").run(); +/// ``` +#[must_use] +pub struct Package { + name: String, + vers: String, + deps: Vec<Dependency>, + files: Vec<PackageFile>, + yanked: bool, + features: FeatureMap, + local: bool, + alternative: bool, + invalid_json: bool, + proc_macro: bool, + links: Option<String>, + rust_version: Option<String>, + cargo_features: Vec<String>, + v: Option<u32>, +} + +pub(crate) type FeatureMap = BTreeMap<String, Vec<String>>; + +#[derive(Clone)] +pub struct Dependency { + name: String, + vers: String, + kind: String, + artifact: Option<(String, Option<String>)>, + target: Option<String>, + features: Vec<String>, + registry: Option<String>, + package: Option<String>, + optional: bool, +} + +/// Entry with data that corresponds to [`tar::EntryType`]. +#[non_exhaustive] +enum EntryData { + Regular(String), + Symlink(PathBuf), +} + +/// A file to be created in a package. +struct PackageFile { + path: String, + contents: EntryData, + /// The Unix mode for the file. Note that when extracted on Windows, this + /// is mostly ignored since it doesn't have the same style of permissions. + mode: u32, + /// If `true`, the file is created in the root of the tarfile, used for + /// testing invalid packages. + extra: bool, +} + +const DEFAULT_MODE: u32 = 0o644; + +/// Initializes the on-disk registry and sets up the config so that crates.io +/// is replaced with the one on disk. +pub fn init() -> TestRegistry { + RegistryBuilder::new().build() +} + +/// Variant of `init` that initializes the "alternative" registry and crates.io +/// replacement. +pub fn alt_init() -> TestRegistry { + init(); + RegistryBuilder::new().alternative().build() +} + +pub struct HttpServerHandle { + addr: SocketAddr, + handle: Option<JoinHandle<()>>, +} + +impl HttpServerHandle { + pub fn index_url(&self) -> Url { + Url::parse(&format!("sparse+http://{}/index/", self.addr.to_string())).unwrap() + } + + pub fn api_url(&self) -> Url { + Url::parse(&format!("http://{}/", self.addr.to_string())).unwrap() + } + + pub fn dl_url(&self) -> Url { + Url::parse(&format!("http://{}/dl", self.addr.to_string())).unwrap() + } + + fn stop(&self) { + if let Ok(mut stream) = TcpStream::connect(self.addr) { + // shutdown the server + let _ = stream.write_all(b"stop"); + let _ = stream.flush(); + } + } +} + +impl Drop for HttpServerHandle { + fn drop(&mut self) { + self.stop(); + } +} + +/// Request to the test http server +#[derive(Clone)] +pub struct Request { + pub url: Url, + pub method: String, + pub body: Option<Vec<u8>>, + pub authorization: Option<String>, + pub if_modified_since: Option<String>, + pub if_none_match: Option<String>, +} + +impl fmt::Debug for Request { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + // body is not included as it can produce long debug outputs + f.debug_struct("Request") + .field("url", &self.url) + .field("method", &self.method) + .field("authorization", &self.authorization) + .field("if_modified_since", &self.if_modified_since) + .field("if_none_match", &self.if_none_match) + .finish() + } +} + +/// Response from the test http server +pub struct Response { + pub code: u32, + pub headers: Vec<String>, + pub body: Vec<u8>, +} + +pub struct HttpServer { + listener: TcpListener, + registry_path: PathBuf, + dl_path: PathBuf, + api_path: PathBuf, + addr: SocketAddr, + token: Token, + auth_required: bool, + custom_responders: HashMap<&'static str, Box<dyn Send + Fn(&Request, &HttpServer) -> Response>>, +} + +/// A helper struct that collects the arguments for [HttpServer::check_authorized]. +/// Based on looking at the request, these are the fields that the authentication header should attest to. +pub struct Mutation<'a> { + pub mutation: &'a str, + pub name: Option<&'a str>, + pub vers: Option<&'a str>, + pub cksum: Option<&'a str>, +} + +impl HttpServer { + pub fn new( + registry_path: PathBuf, + dl_path: PathBuf, + api_path: PathBuf, + token: Token, + auth_required: bool, + api_responders: HashMap< + &'static str, + Box<dyn Send + Fn(&Request, &HttpServer) -> Response>, + >, + ) -> HttpServerHandle { + let listener = TcpListener::bind("127.0.0.1:0").unwrap(); + let addr = listener.local_addr().unwrap(); + let server = HttpServer { + listener, + registry_path, + dl_path, + api_path, + addr, + token, + auth_required, + custom_responders: api_responders, + }; + let handle = Some(thread::spawn(move || server.start())); + HttpServerHandle { addr, handle } + } + + fn start(&self) { + let mut line = String::new(); + 'server: loop { + let (socket, _) = self.listener.accept().unwrap(); + let mut buf = BufReader::new(socket); + line.clear(); + if buf.read_line(&mut line).unwrap() == 0 { + // Connection terminated. + continue; + } + // Read the "GET path HTTP/1.1" line. + let mut parts = line.split_ascii_whitespace(); + let method = parts.next().unwrap().to_ascii_lowercase(); + if method == "stop" { + // Shutdown the server. + return; + } + let addr = self.listener.local_addr().unwrap(); + let url = format!( + "http://{}/{}", + addr, + parts.next().unwrap().trim_start_matches('/') + ); + let url = Url::parse(&url).unwrap(); + + // Grab headers we care about. + let mut if_modified_since = None; + let mut if_none_match = None; + let mut authorization = None; + let mut content_len = None; + loop { + line.clear(); + if buf.read_line(&mut line).unwrap() == 0 { + continue 'server; + } + if line == "\r\n" { + // End of headers. + line.clear(); + break; + } + let (name, value) = line.split_once(':').unwrap(); + let name = name.trim().to_ascii_lowercase(); + let value = value.trim().to_string(); + match name.as_str() { + "if-modified-since" => if_modified_since = Some(value), + "if-none-match" => if_none_match = Some(value), + "authorization" => authorization = Some(value), + "content-length" => content_len = Some(value), + _ => {} + } + } + + let mut body = None; + if let Some(con_len) = content_len { + let len = con_len.parse::<u64>().unwrap(); + let mut content = vec![0u8; len as usize]; + buf.read_exact(&mut content).unwrap(); + body = Some(content) + } + + let req = Request { + authorization, + if_modified_since, + if_none_match, + method, + url, + body, + }; + println!("req: {:#?}", req); + let response = self.route(&req); + let buf = buf.get_mut(); + write!(buf, "HTTP/1.1 {}\r\n", response.code).unwrap(); + write!(buf, "Content-Length: {}\r\n", response.body.len()).unwrap(); + for header in response.headers { + write!(buf, "{}\r\n", header).unwrap(); + } + write!(buf, "\r\n").unwrap(); + buf.write_all(&response.body).unwrap(); + buf.flush().unwrap(); + } + } + + fn check_authorized(&self, req: &Request, mutation: Option<Mutation>) -> bool { + let (private_key, private_key_subject) = if mutation.is_some() || self.auth_required { + match &self.token { + Token::Plaintext(token) => return Some(token) == req.authorization.as_ref(), + Token::Keys(private_key, private_key_subject) => { + (private_key.as_str(), private_key_subject) + } + } + } else { + assert!(req.authorization.is_none(), "unexpected token"); + return true; + }; + + macro_rules! t { + ($e:expr) => { + match $e { + Some(e) => e, + None => return false, + } + }; + } + + let secret: AsymmetricSecretKey<pasetors::version3::V3> = private_key.try_into().unwrap(); + let public: AsymmetricPublicKey<pasetors::version3::V3> = (&secret).try_into().unwrap(); + let pub_key_id: pasetors::paserk::Id = (&public).into(); + let mut paserk_pub_key_id = String::new(); + FormatAsPaserk::fmt(&pub_key_id, &mut paserk_pub_key_id).unwrap(); + // https://github.com/rust-lang/rfcs/blob/master/text/3231-cargo-asymmetric-tokens.md#how-the-registry-server-will-validate-an-asymmetric-token + + // - The PASETO is in v3.public format. + let authorization = t!(&req.authorization); + let untrusted_token = t!( + UntrustedToken::<pasetors::Public, pasetors::version3::V3>::try_from(authorization) + .ok() + ); + + // - The PASETO validates using the public key it looked up based on the key ID. + #[derive(serde::Deserialize, Debug)] + struct Footer<'a> { + url: &'a str, + kip: &'a str, + } + let footer: Footer = t!(serde_json::from_slice(untrusted_token.untrusted_footer()).ok()); + if footer.kip != paserk_pub_key_id { + return false; + } + let trusted_token = + t!( + pasetors::version3::PublicToken::verify(&public, &untrusted_token, None, None,) + .ok() + ); + + // - The URL matches the registry base URL + if footer.url != "https://github.com/rust-lang/crates.io-index" + && footer.url != &format!("sparse+http://{}/index/", self.addr.to_string()) + { + dbg!(footer.url); + return false; + } + + // - The PASETO is still within its valid time period. + #[derive(serde::Deserialize)] + struct Message<'a> { + iat: &'a str, + sub: Option<&'a str>, + mutation: Option<&'a str>, + name: Option<&'a str>, + vers: Option<&'a str>, + cksum: Option<&'a str>, + _challenge: Option<&'a str>, // todo: PASETO with challenges + v: Option<u8>, + } + let message: Message = t!(serde_json::from_str(trusted_token.payload()).ok()); + let token_time = t!(OffsetDateTime::parse(message.iat, &Rfc3339).ok()); + let now = OffsetDateTime::now_utc(); + if (now - token_time) > Duration::MINUTE { + return false; + } + if private_key_subject.as_deref() != message.sub { + dbg!(message.sub); + return false; + } + // - If the claim v is set, that it has the value of 1. + if let Some(v) = message.v { + if v != 1 { + dbg!(message.v); + return false; + } + } + // - If the server issues challenges, that the challenge has not yet been answered. + // todo: PASETO with challenges + // - If the operation is a mutation: + if let Some(mutation) = mutation { + // - That the operation matches the mutation field and is one of publish, yank, or unyank. + if message.mutation != Some(mutation.mutation) { + dbg!(message.mutation); + return false; + } + // - That the package, and version match the request. + if message.name != mutation.name { + dbg!(message.name); + return false; + } + if message.vers != mutation.vers { + dbg!(message.vers); + return false; + } + // - If the mutation is publish, that the version has not already been published, and that the hash matches the request. + if mutation.mutation == "publish" { + if message.cksum != mutation.cksum { + dbg!(message.cksum); + return false; + } + } + } else { + // - If the operation is a read, that the mutation field is not set. + if message.mutation.is_some() + || message.name.is_some() + || message.vers.is_some() + || message.cksum.is_some() + { + return false; + } + } + true + } + + /// Route the request + fn route(&self, req: &Request) -> Response { + // Check for custom responder + if let Some(responder) = self.custom_responders.get(req.url.path()) { + return responder(&req, self); + } + let path: Vec<_> = req.url.path()[1..].split('/').collect(); + match (req.method.as_str(), path.as_slice()) { + ("get", ["index", ..]) => { + if !self.check_authorized(req, None) { + self.unauthorized(req) + } else { + self.index(&req) + } + } + ("get", ["dl", ..]) => { + if !self.check_authorized(req, None) { + self.unauthorized(req) + } else { + self.dl(&req) + } + } + // publish + ("put", ["api", "v1", "crates", "new"]) => self.check_authorized_publish(req), + // The remainder of the operators in the test framework do nothing other than responding 'ok'. + // + // Note: We don't need to support anything real here because there are no tests that + // currently require anything other than publishing via the http api. + + // yank / unyank + ("delete" | "put", ["api", "v1", "crates", crate_name, version, mutation]) => { + if !self.check_authorized( + req, + Some(Mutation { + mutation, + name: Some(crate_name), + vers: Some(version), + cksum: None, + }), + ) { + self.unauthorized(req) + } else { + self.ok(&req) + } + } + // owners + ("get" | "put" | "delete", ["api", "v1", "crates", crate_name, "owners"]) => { + if !self.check_authorized( + req, + Some(Mutation { + mutation: "owners", + name: Some(crate_name), + vers: None, + cksum: None, + }), + ) { + self.unauthorized(req) + } else { + self.ok(&req) + } + } + _ => self.not_found(&req), + } + } + + /// Unauthorized response + pub fn unauthorized(&self, _req: &Request) -> Response { + Response { + code: 401, + headers: vec![ + r#"WWW-Authenticate: Cargo login_url="https://test-registry-login/me""#.to_string(), + ], + body: b"Unauthorized message from server.".to_vec(), + } + } + + /// Not found response + pub fn not_found(&self, _req: &Request) -> Response { + Response { + code: 404, + headers: vec![], + body: b"not found".to_vec(), + } + } + + /// Respond OK without doing anything + pub fn ok(&self, _req: &Request) -> Response { + Response { + code: 200, + headers: vec![], + body: br#"{"ok": true, "msg": "completed!"}"#.to_vec(), + } + } + + /// Return an internal server error (HTTP 500) + pub fn internal_server_error(&self, _req: &Request) -> Response { + Response { + code: 500, + headers: vec![], + body: br#"internal server error"#.to_vec(), + } + } + + /// Serve the download endpoint + pub fn dl(&self, req: &Request) -> Response { + let file = self + .dl_path + .join(req.url.path().strip_prefix("/dl/").unwrap()); + println!("{}", file.display()); + if !file.exists() { + return self.not_found(req); + } + return Response { + body: fs::read(&file).unwrap(), + code: 200, + headers: vec![], + }; + } + + /// Serve the registry index + pub fn index(&self, req: &Request) -> Response { + let file = self + .registry_path + .join(req.url.path().strip_prefix("/index/").unwrap()); + if !file.exists() { + return self.not_found(req); + } else { + // Now grab info about the file. + let data = fs::read(&file).unwrap(); + let etag = Sha256::new().update(&data).finish_hex(); + let last_modified = format!("{:?}", file.metadata().unwrap().modified().unwrap()); + + // Start to construct our response: + let mut any_match = false; + let mut all_match = true; + if let Some(expected) = &req.if_none_match { + if &etag != expected { + all_match = false; + } else { + any_match = true; + } + } + if let Some(expected) = &req.if_modified_since { + // NOTE: Equality comparison is good enough for tests. + if &last_modified != expected { + all_match = false; + } else { + any_match = true; + } + } + + if any_match && all_match { + return Response { + body: Vec::new(), + code: 304, + headers: vec![], + }; + } else { + return Response { + body: data, + code: 200, + headers: vec![ + format!("ETag: \"{}\"", etag), + format!("Last-Modified: {}", last_modified), + ], + }; + } + } + } + + pub fn check_authorized_publish(&self, req: &Request) -> Response { + if let Some(body) = &req.body { + // Mimic the publish behavior for local registries by writing out the request + // so tests can verify publishes made to either registry type. + let path = self.api_path.join("api/v1/crates/new"); + t!(fs::create_dir_all(path.parent().unwrap())); + t!(fs::write(&path, body)); + + // Get the metadata of the package + let (len, remaining) = body.split_at(4); + let json_len = u32::from_le_bytes(len.try_into().unwrap()); + let (json, remaining) = remaining.split_at(json_len as usize); + let new_crate = serde_json::from_slice::<crates_io::NewCrate>(json).unwrap(); + // Get the `.crate` file + let (len, remaining) = remaining.split_at(4); + let file_len = u32::from_le_bytes(len.try_into().unwrap()); + let (file, _remaining) = remaining.split_at(file_len as usize); + let file_cksum = cksum(&file); + + if !self.check_authorized( + req, + Some(Mutation { + mutation: "publish", + name: Some(&new_crate.name), + vers: Some(&new_crate.vers), + cksum: Some(&file_cksum), + }), + ) { + return self.unauthorized(req); + } + + // Write the `.crate` + let dst = self + .dl_path + .join(&new_crate.name) + .join(&new_crate.vers) + .join("download"); + t!(fs::create_dir_all(dst.parent().unwrap())); + t!(fs::write(&dst, file)); + + let deps = new_crate + .deps + .iter() + .map(|dep| { + let (name, package) = match &dep.explicit_name_in_toml { + Some(explicit) => (explicit.to_string(), Some(dep.name.to_string())), + None => (dep.name.to_string(), None), + }; + serde_json::json!({ + "name": name, + "req": dep.version_req, + "features": dep.features, + "default_features": true, + "target": dep.target, + "optional": dep.optional, + "kind": dep.kind, + "registry": dep.registry, + "package": package, + }) + }) + .collect::<Vec<_>>(); + + let line = create_index_line( + serde_json::json!(new_crate.name), + &new_crate.vers, + deps, + &file_cksum, + new_crate.features, + false, + new_crate.links, + None, + ); + + write_to_index(&self.registry_path, &new_crate.name, line, false); + + self.ok(&req) + } else { + Response { + code: 400, + headers: vec![], + body: b"The request was missing a body".to_vec(), + } + } + } +} + +impl Package { + /// Creates a new package builder. + /// Call `publish()` to finalize and build the package. + pub fn new(name: &str, vers: &str) -> Package { + let config = paths::home().join(".cargo/config"); + if !config.exists() { + init(); + } + Package { + name: name.to_string(), + vers: vers.to_string(), + deps: Vec::new(), + files: Vec::new(), + yanked: false, + features: BTreeMap::new(), + local: false, + alternative: false, + invalid_json: false, + proc_macro: false, + links: None, + rust_version: None, + cargo_features: Vec::new(), + v: None, + } + } + + /// Call with `true` to publish in a "local registry". + /// + /// See `source-replacement.html#local-registry-sources` for more details + /// on local registries. See `local_registry.rs` for the tests that use + /// this. + pub fn local(&mut self, local: bool) -> &mut Package { + self.local = local; + self + } + + /// Call with `true` to publish in an "alternative registry". + /// + /// The name of the alternative registry is called "alternative". + /// + /// See `src/doc/src/reference/registries.md` for more details on + /// alternative registries. See `alt_registry.rs` for the tests that use + /// this. + pub fn alternative(&mut self, alternative: bool) -> &mut Package { + self.alternative = alternative; + self + } + + /// Adds a file to the package. + pub fn file(&mut self, name: &str, contents: &str) -> &mut Package { + self.file_with_mode(name, DEFAULT_MODE, contents) + } + + /// Adds a file with a specific Unix mode. + pub fn file_with_mode(&mut self, path: &str, mode: u32, contents: &str) -> &mut Package { + self.files.push(PackageFile { + path: path.to_string(), + contents: EntryData::Regular(contents.into()), + mode, + extra: false, + }); + self + } + + /// Adds a symlink to a path to the package. + pub fn symlink(&mut self, dst: &str, src: &str) -> &mut Package { + self.files.push(PackageFile { + path: dst.to_string(), + contents: EntryData::Symlink(src.into()), + mode: DEFAULT_MODE, + extra: false, + }); + self + } + + /// Adds an "extra" file that is not rooted within the package. + /// + /// Normal files are automatically placed within a directory named + /// `$PACKAGE-$VERSION`. This allows you to override that behavior, + /// typically for testing invalid behavior. + pub fn extra_file(&mut self, path: &str, contents: &str) -> &mut Package { + self.files.push(PackageFile { + path: path.to_string(), + contents: EntryData::Regular(contents.to_string()), + mode: DEFAULT_MODE, + extra: true, + }); + self + } + + /// Adds a normal dependency. Example: + /// ``` + /// [dependencies] + /// foo = {version = "1.0"} + /// ``` + pub fn dep(&mut self, name: &str, vers: &str) -> &mut Package { + self.add_dep(&Dependency::new(name, vers)) + } + + /// Adds a dependency with the given feature. Example: + /// ``` + /// [dependencies] + /// foo = {version = "1.0", "features": ["feat1", "feat2"]} + /// ``` + pub fn feature_dep(&mut self, name: &str, vers: &str, features: &[&str]) -> &mut Package { + self.add_dep(Dependency::new(name, vers).enable_features(features)) + } + + /// Adds a platform-specific dependency. Example: + /// ``` + /// [target.'cfg(windows)'.dependencies] + /// foo = {version = "1.0"} + /// ``` + pub fn target_dep(&mut self, name: &str, vers: &str, target: &str) -> &mut Package { + self.add_dep(Dependency::new(name, vers).target(target)) + } + + /// Adds a dependency to the alternative registry. + pub fn registry_dep(&mut self, name: &str, vers: &str) -> &mut Package { + self.add_dep(Dependency::new(name, vers).registry("alternative")) + } + + /// Adds a dev-dependency. Example: + /// ``` + /// [dev-dependencies] + /// foo = {version = "1.0"} + /// ``` + pub fn dev_dep(&mut self, name: &str, vers: &str) -> &mut Package { + self.add_dep(Dependency::new(name, vers).dev()) + } + + /// Adds a build-dependency. Example: + /// ``` + /// [build-dependencies] + /// foo = {version = "1.0"} + /// ``` + pub fn build_dep(&mut self, name: &str, vers: &str) -> &mut Package { + self.add_dep(Dependency::new(name, vers).build()) + } + + pub fn add_dep(&mut self, dep: &Dependency) -> &mut Package { + self.deps.push(dep.clone()); + self + } + + /// Specifies whether or not the package is "yanked". + pub fn yanked(&mut self, yanked: bool) -> &mut Package { + self.yanked = yanked; + self + } + + /// Specifies whether or not this is a proc macro. + pub fn proc_macro(&mut self, proc_macro: bool) -> &mut Package { + self.proc_macro = proc_macro; + self + } + + /// Adds an entry in the `[features]` section. + pub fn feature(&mut self, name: &str, deps: &[&str]) -> &mut Package { + let deps = deps.iter().map(|s| s.to_string()).collect(); + self.features.insert(name.to_string(), deps); + self + } + + /// Specify a minimal Rust version. + pub fn rust_version(&mut self, rust_version: &str) -> &mut Package { + self.rust_version = Some(rust_version.into()); + self + } + + /// Causes the JSON line emitted in the index to be invalid, presumably + /// causing Cargo to skip over this version. + pub fn invalid_json(&mut self, invalid: bool) -> &mut Package { + self.invalid_json = invalid; + self + } + + pub fn links(&mut self, links: &str) -> &mut Package { + self.links = Some(links.to_string()); + self + } + + pub fn cargo_feature(&mut self, feature: &str) -> &mut Package { + self.cargo_features.push(feature.to_owned()); + self + } + + /// Sets the index schema version for this package. + /// + /// See `cargo::sources::registry::RegistryPackage` for more information. + pub fn schema_version(&mut self, version: u32) -> &mut Package { + self.v = Some(version); + self + } + + /// Creates the package and place it in the registry. + /// + /// This does not actually use Cargo's publishing system, but instead + /// manually creates the entry in the registry on the filesystem. + /// + /// Returns the checksum for the package. + pub fn publish(&self) -> String { + self.make_archive(); + + // Figure out what we're going to write into the index. + let deps = self + .deps + .iter() + .map(|dep| { + // In the index, the `registry` is null if it is from the same registry. + // In Cargo.toml, it is None if it is from crates.io. + let registry_url = match (self.alternative, dep.registry.as_deref()) { + (false, None) => None, + (false, Some("alternative")) => Some(alt_registry_url().to_string()), + (true, None) => { + Some("https://github.com/rust-lang/crates.io-index".to_string()) + } + (true, Some("alternative")) => None, + _ => panic!("registry_dep currently only supports `alternative`"), + }; + serde_json::json!({ + "name": dep.name, + "req": dep.vers, + "features": dep.features, + "default_features": true, + "target": dep.target, + "artifact": dep.artifact, + "optional": dep.optional, + "kind": dep.kind, + "registry": registry_url, + "package": dep.package, + }) + }) + .collect::<Vec<_>>(); + let cksum = { + let c = t!(fs::read(&self.archive_dst())); + cksum(&c) + }; + let name = if self.invalid_json { + serde_json::json!(1) + } else { + serde_json::json!(self.name) + }; + let line = create_index_line( + name, + &self.vers, + deps, + &cksum, + self.features.clone(), + self.yanked, + self.links.clone(), + self.v, + ); + + let registry_path = if self.alternative { + alt_registry_path() + } else { + registry_path() + }; + + write_to_index(®istry_path, &self.name, line, self.local); + + cksum + } + + fn make_archive(&self) { + let dst = self.archive_dst(); + t!(fs::create_dir_all(dst.parent().unwrap())); + let f = t!(File::create(&dst)); + let mut a = Builder::new(GzEncoder::new(f, Compression::default())); + + if !self + .files + .iter() + .any(|PackageFile { path, .. }| path == "Cargo.toml") + { + self.append_manifest(&mut a); + } + if self.files.is_empty() { + self.append( + &mut a, + "src/lib.rs", + DEFAULT_MODE, + &EntryData::Regular("".into()), + ); + } else { + for PackageFile { + path, + contents, + mode, + extra, + } in &self.files + { + if *extra { + self.append_raw(&mut a, path, *mode, contents); + } else { + self.append(&mut a, path, *mode, contents); + } + } + } + } + + fn append_manifest<W: Write>(&self, ar: &mut Builder<W>) { + let mut manifest = String::new(); + + if !self.cargo_features.is_empty() { + let mut features = String::new(); + serde::Serialize::serialize( + &self.cargo_features, + toml::ser::ValueSerializer::new(&mut features), + ) + .unwrap(); + manifest.push_str(&format!("cargo-features = {}\n\n", features)); + } + + manifest.push_str(&format!( + r#" + [package] + name = "{}" + version = "{}" + authors = [] + "#, + self.name, self.vers + )); + + if let Some(version) = &self.rust_version { + manifest.push_str(&format!("rust-version = \"{}\"", version)); + } + + for dep in self.deps.iter() { + let target = match dep.target { + None => String::new(), + Some(ref s) => format!("target.'{}'.", s), + }; + let kind = match &dep.kind[..] { + "build" => "build-", + "dev" => "dev-", + _ => "", + }; + manifest.push_str(&format!( + r#" + [{}{}dependencies.{}] + version = "{}" + "#, + target, kind, dep.name, dep.vers + )); + if let Some((artifact, target)) = &dep.artifact { + manifest.push_str(&format!("artifact = \"{}\"\n", artifact)); + if let Some(target) = &target { + manifest.push_str(&format!("target = \"{}\"\n", target)) + } + } + if let Some(registry) = &dep.registry { + assert_eq!(registry, "alternative"); + manifest.push_str(&format!("registry-index = \"{}\"", alt_registry_url())); + } + } + if self.proc_macro { + manifest.push_str("[lib]\nproc-macro = true\n"); + } + + self.append( + ar, + "Cargo.toml", + DEFAULT_MODE, + &EntryData::Regular(manifest.into()), + ); + } + + fn append<W: Write>(&self, ar: &mut Builder<W>, file: &str, mode: u32, contents: &EntryData) { + self.append_raw( + ar, + &format!("{}-{}/{}", self.name, self.vers, file), + mode, + contents, + ); + } + + fn append_raw<W: Write>( + &self, + ar: &mut Builder<W>, + path: &str, + mode: u32, + contents: &EntryData, + ) { + let mut header = Header::new_ustar(); + let contents = match contents { + EntryData::Regular(contents) => contents.as_str(), + EntryData::Symlink(src) => { + header.set_entry_type(tar::EntryType::Symlink); + t!(header.set_link_name(src)); + "" // Symlink has no contents. + } + }; + header.set_size(contents.len() as u64); + t!(header.set_path(path)); + header.set_mode(mode); + header.set_cksum(); + t!(ar.append(&header, contents.as_bytes())); + } + + /// Returns the path to the compressed package file. + pub fn archive_dst(&self) -> PathBuf { + if self.local { + registry_path().join(format!("{}-{}.crate", self.name, self.vers)) + } else if self.alternative { + alt_dl_path() + .join(&self.name) + .join(&self.vers) + .join("download") + } else { + dl_path().join(&self.name).join(&self.vers).join("download") + } + } +} + +pub fn cksum(s: &[u8]) -> String { + Sha256::new().update(s).finish_hex() +} + +impl Dependency { + pub fn new(name: &str, vers: &str) -> Dependency { + Dependency { + name: name.to_string(), + vers: vers.to_string(), + kind: "normal".to_string(), + artifact: None, + target: None, + features: Vec::new(), + package: None, + optional: false, + registry: None, + } + } + + /// Changes this to `[build-dependencies]`. + pub fn build(&mut self) -> &mut Self { + self.kind = "build".to_string(); + self + } + + /// Changes this to `[dev-dependencies]`. + pub fn dev(&mut self) -> &mut Self { + self.kind = "dev".to_string(); + self + } + + /// Changes this to `[target.$target.dependencies]`. + pub fn target(&mut self, target: &str) -> &mut Self { + self.target = Some(target.to_string()); + self + } + + /// Change the artifact to be of the given kind, like "bin", or "staticlib", + /// along with a specific target triple if provided. + pub fn artifact(&mut self, kind: &str, target: Option<String>) -> &mut Self { + self.artifact = Some((kind.to_string(), target)); + self + } + + /// Adds `registry = $registry` to this dependency. + pub fn registry(&mut self, registry: &str) -> &mut Self { + self.registry = Some(registry.to_string()); + self + } + + /// Adds `features = [ ... ]` to this dependency. + pub fn enable_features(&mut self, features: &[&str]) -> &mut Self { + self.features.extend(features.iter().map(|s| s.to_string())); + self + } + + /// Adds `package = ...` to this dependency. + pub fn package(&mut self, pkg: &str) -> &mut Self { + self.package = Some(pkg.to_string()); + self + } + + /// Changes this to an optional dependency. + pub fn optional(&mut self, optional: bool) -> &mut Self { + self.optional = optional; + self + } +} diff --git a/crates/cargo-test-support/src/tools.rs b/crates/cargo-test-support/src/tools.rs new file mode 100644 index 0000000..7c056b6 --- /dev/null +++ b/crates/cargo-test-support/src/tools.rs @@ -0,0 +1,108 @@ +//! Common executables that can be reused by various tests. + +use crate::{basic_manifest, paths, project, Project}; +use lazy_static::lazy_static; +use std::path::{Path, PathBuf}; +use std::sync::Mutex; + +lazy_static! { + static ref ECHO_WRAPPER: Mutex<Option<PathBuf>> = Mutex::new(None); + static ref ECHO: Mutex<Option<PathBuf>> = Mutex::new(None); +} + +/// Returns the path to an executable that works as a wrapper around rustc. +/// +/// The wrapper will echo the command line it was called with to stderr. +pub fn echo_wrapper() -> PathBuf { + let mut lock = ECHO_WRAPPER.lock().unwrap(); + if let Some(path) = &*lock { + return path.clone(); + } + let p = project() + .at(paths::global_root().join("rustc-echo-wrapper")) + .file("Cargo.toml", &basic_manifest("rustc-echo-wrapper", "1.0.0")) + .file( + "src/main.rs", + r#" + use std::fs::read_to_string; + use std::path::PathBuf; + fn main() { + // Handle args from `@path` argfile for rustc + let args = std::env::args() + .flat_map(|p| if let Some(p) = p.strip_prefix("@") { + read_to_string(p).unwrap().lines().map(String::from).collect() + } else { + vec![p] + }) + .collect::<Vec<_>>(); + eprintln!("WRAPPER CALLED: {}", args[1..].join(" ")); + let status = std::process::Command::new(&args[1]) + .args(&args[2..]).status().unwrap(); + std::process::exit(status.code().unwrap_or(1)); + } + "#, + ) + .build(); + p.cargo("build").run(); + let path = p.bin("rustc-echo-wrapper"); + *lock = Some(path.clone()); + path +} + +/// Returns the path to an executable that prints its arguments. +/// +/// Do not expect this to be anything fancy. +pub fn echo() -> PathBuf { + let mut lock = ECHO.lock().unwrap(); + if let Some(path) = &*lock { + return path.clone(); + } + if let Ok(path) = cargo_util::paths::resolve_executable(Path::new("echo")) { + *lock = Some(path.clone()); + return path; + } + // Often on Windows, `echo` is not available. + let p = project() + .at(paths::global_root().join("basic-echo")) + .file("Cargo.toml", &basic_manifest("basic-echo", "1.0.0")) + .file( + "src/main.rs", + r#" + fn main() { + let mut s = String::new(); + let mut it = std::env::args().skip(1).peekable(); + while let Some(n) = it.next() { + s.push_str(&n); + if it.peek().is_some() { + s.push(' '); + } + } + println!("{}", s); + } + "#, + ) + .build(); + p.cargo("build").run(); + let path = p.bin("basic-echo"); + *lock = Some(path.clone()); + path +} + +/// Returns a project which builds a cargo-echo simple subcommand +pub fn echo_subcommand() -> Project { + let p = project() + .at("cargo-echo") + .file("Cargo.toml", &basic_manifest("cargo-echo", "0.0.1")) + .file( + "src/main.rs", + r#" + fn main() { + let args: Vec<_> = ::std::env::args().skip(1).collect(); + println!("{}", args.join(" ")); + } + "#, + ) + .build(); + p.cargo("build").run(); + p +} diff --git a/crates/cargo-util/Cargo.toml b/crates/cargo-util/Cargo.toml new file mode 100644 index 0000000..aa25c13 --- /dev/null +++ b/crates/cargo-util/Cargo.toml @@ -0,0 +1,28 @@ +[package] +name = "cargo-util" +version = "0.2.3" +edition = "2021" +license = "MIT OR Apache-2.0" +homepage = "https://github.com/rust-lang/cargo" +repository = "https://github.com/rust-lang/cargo" +description = "Miscellaneous support code used by Cargo." + +[dependencies] +anyhow = "1.0.34" +crypto-hash = "0.3.1" +filetime = "0.2.9" +hex = "0.4.2" +jobserver = "0.1.26" +libc = "0.2.88" +log = "0.4.6" +same-file = "1.0.6" +shell-escape = "0.1.4" +tempfile = "3.1.0" +walkdir = "2.3.1" + +[target.'cfg(target_os = "macos")'.dependencies] +core-foundation = { version = "0.9.0", features = ["mac_os_10_7_support"] } + +[target.'cfg(windows)'.dependencies] +miow = "0.5.0" +windows-sys = { version = "0.45.0", features = ["Win32_Storage_FileSystem", "Win32_Foundation", "Win32_System_Console"] } diff --git a/crates/cargo-util/LICENSE-APACHE b/crates/cargo-util/LICENSE-APACHE new file mode 120000 index 0000000..1cd601d --- /dev/null +++ b/crates/cargo-util/LICENSE-APACHE @@ -0,0 +1 @@ +../../LICENSE-APACHE
\ No newline at end of file diff --git a/crates/cargo-util/LICENSE-MIT b/crates/cargo-util/LICENSE-MIT new file mode 120000 index 0000000..b2cfbdc --- /dev/null +++ b/crates/cargo-util/LICENSE-MIT @@ -0,0 +1 @@ +../../LICENSE-MIT
\ No newline at end of file diff --git a/crates/cargo-util/src/lib.rs b/crates/cargo-util/src/lib.rs new file mode 100644 index 0000000..0cbc920 --- /dev/null +++ b/crates/cargo-util/src/lib.rs @@ -0,0 +1,18 @@ +//! Miscellaneous support code used by Cargo. + +pub use self::read2::read2; +pub use process_builder::ProcessBuilder; +pub use process_error::{exit_status_to_string, is_simple_exit_code, ProcessError}; +pub use sha256::Sha256; + +pub mod paths; +mod process_builder; +mod process_error; +mod read2; +pub mod registry; +mod sha256; + +/// Whether or not this running in a Continuous Integration environment. +pub fn is_ci() -> bool { + std::env::var("CI").is_ok() || std::env::var("TF_BUILD").is_ok() +} diff --git a/crates/cargo-util/src/paths.rs b/crates/cargo-util/src/paths.rs new file mode 100644 index 0000000..69df7a2 --- /dev/null +++ b/crates/cargo-util/src/paths.rs @@ -0,0 +1,788 @@ +//! Various utilities for working with files and paths. + +use anyhow::{Context, Result}; +use filetime::FileTime; +use std::env; +use std::ffi::{OsStr, OsString}; +use std::fs::{self, File, OpenOptions}; +use std::io; +use std::io::prelude::*; +use std::iter; +use std::path::{Component, Path, PathBuf}; +use tempfile::Builder as TempFileBuilder; + +/// Joins paths into a string suitable for the `PATH` environment variable. +/// +/// This is equivalent to [`std::env::join_paths`], but includes a more +/// detailed error message. The given `env` argument is the name of the +/// environment variable this is will be used for, which is included in the +/// error message. +pub fn join_paths<T: AsRef<OsStr>>(paths: &[T], env: &str) -> Result<OsString> { + env::join_paths(paths.iter()).with_context(|| { + let mut message = format!( + "failed to join paths from `${env}` together\n\n\ + Check if any of path segments listed below contain an \ + unterminated quote character or path separator:" + ); + for path in paths { + use std::fmt::Write; + write!(&mut message, "\n {:?}", Path::new(path)).unwrap(); + } + + message + }) +} + +/// Returns the name of the environment variable used for searching for +/// dynamic libraries. +pub fn dylib_path_envvar() -> &'static str { + if cfg!(windows) { + "PATH" + } else if cfg!(target_os = "macos") { + // When loading and linking a dynamic library or bundle, dlopen + // searches in LD_LIBRARY_PATH, DYLD_LIBRARY_PATH, PWD, and + // DYLD_FALLBACK_LIBRARY_PATH. + // In the Mach-O format, a dynamic library has an "install path." + // Clients linking against the library record this path, and the + // dynamic linker, dyld, uses it to locate the library. + // dyld searches DYLD_LIBRARY_PATH *before* the install path. + // dyld searches DYLD_FALLBACK_LIBRARY_PATH only if it cannot + // find the library in the install path. + // Setting DYLD_LIBRARY_PATH can easily have unintended + // consequences. + // + // Also, DYLD_LIBRARY_PATH appears to have significant performance + // penalty starting in 10.13. Cargo's testsuite ran more than twice as + // slow with it on CI. + "DYLD_FALLBACK_LIBRARY_PATH" + } else { + "LD_LIBRARY_PATH" + } +} + +/// Returns a list of directories that are searched for dynamic libraries. +/// +/// Note that some operating systems will have defaults if this is empty that +/// will need to be dealt with. +pub fn dylib_path() -> Vec<PathBuf> { + match env::var_os(dylib_path_envvar()) { + Some(var) => env::split_paths(&var).collect(), + None => Vec::new(), + } +} + +/// Normalize a path, removing things like `.` and `..`. +/// +/// CAUTION: This does not resolve symlinks (unlike +/// [`std::fs::canonicalize`]). This may cause incorrect or surprising +/// behavior at times. This should be used carefully. Unfortunately, +/// [`std::fs::canonicalize`] can be hard to use correctly, since it can often +/// fail, or on Windows returns annoying device paths. This is a problem Cargo +/// needs to improve on. +pub fn normalize_path(path: &Path) -> PathBuf { + let mut components = path.components().peekable(); + let mut ret = if let Some(c @ Component::Prefix(..)) = components.peek().cloned() { + components.next(); + PathBuf::from(c.as_os_str()) + } else { + PathBuf::new() + }; + + for component in components { + match component { + Component::Prefix(..) => unreachable!(), + Component::RootDir => { + ret.push(component.as_os_str()); + } + Component::CurDir => {} + Component::ParentDir => { + ret.pop(); + } + Component::Normal(c) => { + ret.push(c); + } + } + } + ret +} + +/// Returns the absolute path of where the given executable is located based +/// on searching the `PATH` environment variable. +/// +/// Returns an error if it cannot be found. +pub fn resolve_executable(exec: &Path) -> Result<PathBuf> { + if exec.components().count() == 1 { + let paths = env::var_os("PATH").ok_or_else(|| anyhow::format_err!("no PATH"))?; + let candidates = env::split_paths(&paths).flat_map(|path| { + let candidate = path.join(&exec); + let with_exe = if env::consts::EXE_EXTENSION.is_empty() { + None + } else { + Some(candidate.with_extension(env::consts::EXE_EXTENSION)) + }; + iter::once(candidate).chain(with_exe) + }); + for candidate in candidates { + if candidate.is_file() { + return Ok(candidate); + } + } + + anyhow::bail!("no executable for `{}` found in PATH", exec.display()) + } else { + Ok(exec.into()) + } +} + +/// Reads a file to a string. +/// +/// Equivalent to [`std::fs::read_to_string`] with better error messages. +pub fn read(path: &Path) -> Result<String> { + match String::from_utf8(read_bytes(path)?) { + Ok(s) => Ok(s), + Err(_) => anyhow::bail!("path at `{}` was not valid utf-8", path.display()), + } +} + +/// Reads a file into a bytes vector. +/// +/// Equivalent to [`std::fs::read`] with better error messages. +pub fn read_bytes(path: &Path) -> Result<Vec<u8>> { + fs::read(path).with_context(|| format!("failed to read `{}`", path.display())) +} + +/// Writes a file to disk. +/// +/// Equivalent to [`std::fs::write`] with better error messages. +pub fn write<P: AsRef<Path>, C: AsRef<[u8]>>(path: P, contents: C) -> Result<()> { + let path = path.as_ref(); + fs::write(path, contents.as_ref()) + .with_context(|| format!("failed to write `{}`", path.display())) +} + +/// Equivalent to [`write()`], but does not write anything if the file contents +/// are identical to the given contents. +pub fn write_if_changed<P: AsRef<Path>, C: AsRef<[u8]>>(path: P, contents: C) -> Result<()> { + (|| -> Result<()> { + let contents = contents.as_ref(); + let mut f = OpenOptions::new() + .read(true) + .write(true) + .create(true) + .open(&path)?; + let mut orig = Vec::new(); + f.read_to_end(&mut orig)?; + if orig != contents { + f.set_len(0)?; + f.seek(io::SeekFrom::Start(0))?; + f.write_all(contents)?; + } + Ok(()) + })() + .with_context(|| format!("failed to write `{}`", path.as_ref().display()))?; + Ok(()) +} + +/// Equivalent to [`write()`], but appends to the end instead of replacing the +/// contents. +pub fn append(path: &Path, contents: &[u8]) -> Result<()> { + (|| -> Result<()> { + let mut f = OpenOptions::new() + .write(true) + .append(true) + .create(true) + .open(path)?; + + f.write_all(contents)?; + Ok(()) + })() + .with_context(|| format!("failed to write `{}`", path.display()))?; + Ok(()) +} + +/// Creates a new file. +pub fn create<P: AsRef<Path>>(path: P) -> Result<File> { + let path = path.as_ref(); + File::create(path).with_context(|| format!("failed to create file `{}`", path.display())) +} + +/// Opens an existing file. +pub fn open<P: AsRef<Path>>(path: P) -> Result<File> { + let path = path.as_ref(); + File::open(path).with_context(|| format!("failed to open file `{}`", path.display())) +} + +/// Returns the last modification time of a file. +pub fn mtime(path: &Path) -> Result<FileTime> { + let meta = + fs::metadata(path).with_context(|| format!("failed to stat `{}`", path.display()))?; + Ok(FileTime::from_last_modification_time(&meta)) +} + +/// Returns the maximum mtime of the given path, recursing into +/// subdirectories, and following symlinks. +pub fn mtime_recursive(path: &Path) -> Result<FileTime> { + let meta = + fs::metadata(path).with_context(|| format!("failed to stat `{}`", path.display()))?; + if !meta.is_dir() { + return Ok(FileTime::from_last_modification_time(&meta)); + } + let max_meta = walkdir::WalkDir::new(path) + .follow_links(true) + .into_iter() + .filter_map(|e| match e { + Ok(e) => Some(e), + Err(e) => { + // Ignore errors while walking. If Cargo can't access it, the + // build script probably can't access it, either. + log::debug!("failed to determine mtime while walking directory: {}", e); + None + } + }) + .filter_map(|e| { + if e.path_is_symlink() { + // Use the mtime of both the symlink and its target, to + // handle the case where the symlink is modified to a + // different target. + let sym_meta = match std::fs::symlink_metadata(e.path()) { + Ok(m) => m, + Err(err) => { + // I'm not sure when this is really possible (maybe a + // race with unlinking?). Regardless, if Cargo can't + // read it, the build script probably can't either. + log::debug!( + "failed to determine mtime while fetching symlink metadata of {}: {}", + e.path().display(), + err + ); + return None; + } + }; + let sym_mtime = FileTime::from_last_modification_time(&sym_meta); + // Walkdir follows symlinks. + match e.metadata() { + Ok(target_meta) => { + let target_mtime = FileTime::from_last_modification_time(&target_meta); + Some(sym_mtime.max(target_mtime)) + } + Err(err) => { + // Can't access the symlink target. If Cargo can't + // access it, the build script probably can't access + // it either. + log::debug!( + "failed to determine mtime of symlink target for {}: {}", + e.path().display(), + err + ); + Some(sym_mtime) + } + } + } else { + let meta = match e.metadata() { + Ok(m) => m, + Err(err) => { + // I'm not sure when this is really possible (maybe a + // race with unlinking?). Regardless, if Cargo can't + // read it, the build script probably can't either. + log::debug!( + "failed to determine mtime while fetching metadata of {}: {}", + e.path().display(), + err + ); + return None; + } + }; + Some(FileTime::from_last_modification_time(&meta)) + } + }) + .max() + // or_else handles the case where there are no files in the directory. + .unwrap_or_else(|| FileTime::from_last_modification_time(&meta)); + Ok(max_meta) +} + +/// Record the current time on the filesystem (using the filesystem's clock) +/// using a file at the given directory. Returns the current time. +pub fn set_invocation_time(path: &Path) -> Result<FileTime> { + // note that if `FileTime::from_system_time(SystemTime::now());` is determined to be sufficient, + // then this can be removed. + let timestamp = path.join("invoked.timestamp"); + write( + ×tamp, + "This file has an mtime of when this was started.", + )?; + let ft = mtime(×tamp)?; + log::debug!("invocation time for {:?} is {}", path, ft); + Ok(ft) +} + +/// Converts a path to UTF-8 bytes. +pub fn path2bytes(path: &Path) -> Result<&[u8]> { + #[cfg(unix)] + { + use std::os::unix::prelude::*; + Ok(path.as_os_str().as_bytes()) + } + #[cfg(windows)] + { + match path.as_os_str().to_str() { + Some(s) => Ok(s.as_bytes()), + None => Err(anyhow::format_err!( + "invalid non-unicode path: {}", + path.display() + )), + } + } +} + +/// Converts UTF-8 bytes to a path. +pub fn bytes2path(bytes: &[u8]) -> Result<PathBuf> { + #[cfg(unix)] + { + use std::os::unix::prelude::*; + Ok(PathBuf::from(OsStr::from_bytes(bytes))) + } + #[cfg(windows)] + { + use std::str; + match str::from_utf8(bytes) { + Ok(s) => Ok(PathBuf::from(s)), + Err(..) => Err(anyhow::format_err!("invalid non-unicode path")), + } + } +} + +/// Returns an iterator that walks up the directory hierarchy towards the root. +/// +/// Each item is a [`Path`]. It will start with the given path, finishing at +/// the root. If the `stop_root_at` parameter is given, it will stop at the +/// given path (which will be the last item). +pub fn ancestors<'a>(path: &'a Path, stop_root_at: Option<&Path>) -> PathAncestors<'a> { + PathAncestors::new(path, stop_root_at) +} + +pub struct PathAncestors<'a> { + current: Option<&'a Path>, + stop_at: Option<PathBuf>, +} + +impl<'a> PathAncestors<'a> { + fn new(path: &'a Path, stop_root_at: Option<&Path>) -> PathAncestors<'a> { + let stop_at = env::var("__CARGO_TEST_ROOT") + .ok() + .map(PathBuf::from) + .or_else(|| stop_root_at.map(|p| p.to_path_buf())); + PathAncestors { + current: Some(path), + //HACK: avoid reading `~/.cargo/config` when testing Cargo itself. + stop_at, + } + } +} + +impl<'a> Iterator for PathAncestors<'a> { + type Item = &'a Path; + + fn next(&mut self) -> Option<&'a Path> { + if let Some(path) = self.current { + self.current = path.parent(); + + if let Some(ref stop_at) = self.stop_at { + if path == stop_at { + self.current = None; + } + } + + Some(path) + } else { + None + } + } +} + +/// Equivalent to [`std::fs::create_dir_all`] with better error messages. +pub fn create_dir_all(p: impl AsRef<Path>) -> Result<()> { + _create_dir_all(p.as_ref()) +} + +fn _create_dir_all(p: &Path) -> Result<()> { + fs::create_dir_all(p) + .with_context(|| format!("failed to create directory `{}`", p.display()))?; + Ok(()) +} + +/// Recursively remove all files and directories at the given directory. +/// +/// This does *not* follow symlinks. +pub fn remove_dir_all<P: AsRef<Path>>(p: P) -> Result<()> { + _remove_dir_all(p.as_ref()) +} + +fn _remove_dir_all(p: &Path) -> Result<()> { + if p.symlink_metadata() + .with_context(|| format!("could not get metadata for `{}` to remove", p.display()))? + .is_symlink() + { + return remove_file(p); + } + let entries = p + .read_dir() + .with_context(|| format!("failed to read directory `{}`", p.display()))?; + for entry in entries { + let entry = entry?; + let path = entry.path(); + if entry.file_type()?.is_dir() { + remove_dir_all(&path)?; + } else { + remove_file(&path)?; + } + } + remove_dir(&p) +} + +/// Equivalent to [`std::fs::remove_dir`] with better error messages. +pub fn remove_dir<P: AsRef<Path>>(p: P) -> Result<()> { + _remove_dir(p.as_ref()) +} + +fn _remove_dir(p: &Path) -> Result<()> { + fs::remove_dir(p).with_context(|| format!("failed to remove directory `{}`", p.display()))?; + Ok(()) +} + +/// Equivalent to [`std::fs::remove_file`] with better error messages. +/// +/// If the file is readonly, this will attempt to change the permissions to +/// force the file to be deleted. +pub fn remove_file<P: AsRef<Path>>(p: P) -> Result<()> { + _remove_file(p.as_ref()) +} + +fn _remove_file(p: &Path) -> Result<()> { + let mut err = match fs::remove_file(p) { + Ok(()) => return Ok(()), + Err(e) => e, + }; + + if err.kind() == io::ErrorKind::PermissionDenied && set_not_readonly(p).unwrap_or(false) { + match fs::remove_file(p) { + Ok(()) => return Ok(()), + Err(e) => err = e, + } + } + + Err(err).with_context(|| format!("failed to remove file `{}`", p.display()))?; + Ok(()) +} + +fn set_not_readonly(p: &Path) -> io::Result<bool> { + let mut perms = p.metadata()?.permissions(); + if !perms.readonly() { + return Ok(false); + } + perms.set_readonly(false); + fs::set_permissions(p, perms)?; + Ok(true) +} + +/// Hardlink (file) or symlink (dir) src to dst if possible, otherwise copy it. +/// +/// If the destination already exists, it is removed before linking. +pub fn link_or_copy(src: impl AsRef<Path>, dst: impl AsRef<Path>) -> Result<()> { + let src = src.as_ref(); + let dst = dst.as_ref(); + _link_or_copy(src, dst) +} + +fn _link_or_copy(src: &Path, dst: &Path) -> Result<()> { + log::debug!("linking {} to {}", src.display(), dst.display()); + if same_file::is_same_file(src, dst).unwrap_or(false) { + return Ok(()); + } + + // NB: we can't use dst.exists(), as if dst is a broken symlink, + // dst.exists() will return false. This is problematic, as we still need to + // unlink dst in this case. symlink_metadata(dst).is_ok() will tell us + // whether dst exists *without* following symlinks, which is what we want. + if fs::symlink_metadata(dst).is_ok() { + remove_file(&dst)?; + } + + let link_result = if src.is_dir() { + #[cfg(target_os = "redox")] + use std::os::redox::fs::symlink; + #[cfg(unix)] + use std::os::unix::fs::symlink; + #[cfg(windows)] + // FIXME: This should probably panic or have a copy fallback. Symlinks + // are not supported in all windows environments. Currently symlinking + // is only used for .dSYM directories on macos, but this shouldn't be + // accidentally relied upon. + use std::os::windows::fs::symlink_dir as symlink; + + let dst_dir = dst.parent().unwrap(); + let src = if src.starts_with(dst_dir) { + src.strip_prefix(dst_dir).unwrap() + } else { + src + }; + symlink(src, dst) + } else if env::var_os("__CARGO_COPY_DONT_LINK_DO_NOT_USE_THIS").is_some() { + // This is a work-around for a bug in macOS 10.15. When running on + // APFS, there seems to be a strange race condition with + // Gatekeeper where it will forcefully kill a process launched via + // `cargo run` with SIGKILL. Copying seems to avoid the problem. + // This shouldn't affect anyone except Cargo's test suite because + // it is very rare, and only seems to happen under heavy load and + // rapidly creating lots of executables and running them. + // See https://github.com/rust-lang/cargo/issues/7821 for the + // gory details. + fs::copy(src, dst).map(|_| ()) + } else { + if cfg!(target_os = "macos") { + // This is a work-around for a bug on macos. There seems to be a race condition + // with APFS when hard-linking binaries. Gatekeeper does not have signing or + // hash information stored in kernel when running the process. Therefore killing it. + // This problem does not appear when copying files as kernel has time to process it. + // Note that: fs::copy on macos is using CopyOnWrite (syscall fclonefileat) which should be + // as fast as hardlinking. + // See https://github.com/rust-lang/cargo/issues/10060 for the details + fs::copy(src, dst).map(|_| ()) + } else { + fs::hard_link(src, dst) + } + }; + link_result + .or_else(|err| { + log::debug!("link failed {}. falling back to fs::copy", err); + fs::copy(src, dst).map(|_| ()) + }) + .with_context(|| { + format!( + "failed to link or copy `{}` to `{}`", + src.display(), + dst.display() + ) + })?; + Ok(()) +} + +/// Copies a file from one location to another. +/// +/// Equivalent to [`std::fs::copy`] with better error messages. +pub fn copy<P: AsRef<Path>, Q: AsRef<Path>>(from: P, to: Q) -> Result<u64> { + let from = from.as_ref(); + let to = to.as_ref(); + fs::copy(from, to) + .with_context(|| format!("failed to copy `{}` to `{}`", from.display(), to.display())) +} + +/// Changes the filesystem mtime (and atime if possible) for the given file. +/// +/// This intentionally does not return an error, as this is sometimes not +/// supported on network filesystems. For the current uses in Cargo, this is a +/// "best effort" approach, and errors shouldn't be propagated. +pub fn set_file_time_no_err<P: AsRef<Path>>(path: P, time: FileTime) { + let path = path.as_ref(); + match filetime::set_file_times(path, time, time) { + Ok(()) => log::debug!("set file mtime {} to {}", path.display(), time), + Err(e) => log::warn!( + "could not set mtime of {} to {}: {:?}", + path.display(), + time, + e + ), + } +} + +/// Strips `base` from `path`. +/// +/// This canonicalizes both paths before stripping. This is useful if the +/// paths are obtained in different ways, and one or the other may or may not +/// have been normalized in some way. +pub fn strip_prefix_canonical<P: AsRef<Path>>( + path: P, + base: P, +) -> Result<PathBuf, std::path::StripPrefixError> { + // Not all filesystems support canonicalize. Just ignore if it doesn't work. + let safe_canonicalize = |path: &Path| match path.canonicalize() { + Ok(p) => p, + Err(e) => { + log::warn!("cannot canonicalize {:?}: {:?}", path, e); + path.to_path_buf() + } + }; + let canon_path = safe_canonicalize(path.as_ref()); + let canon_base = safe_canonicalize(base.as_ref()); + canon_path.strip_prefix(canon_base).map(|p| p.to_path_buf()) +} + +/// Creates an excluded from cache directory atomically with its parents as needed. +/// +/// The atomicity only covers creating the leaf directory and exclusion from cache. Any missing +/// parent directories will not be created in an atomic manner. +/// +/// This function is idempotent and in addition to that it won't exclude ``p`` from cache if it +/// already exists. +pub fn create_dir_all_excluded_from_backups_atomic(p: impl AsRef<Path>) -> Result<()> { + let path = p.as_ref(); + if path.is_dir() { + return Ok(()); + } + + let parent = path.parent().unwrap(); + let base = path.file_name().unwrap(); + create_dir_all(parent)?; + // We do this in two steps (first create a temporary directory and exclude + // it from backups, then rename it to the desired name. If we created the + // directory directly where it should be and then excluded it from backups + // we would risk a situation where cargo is interrupted right after the directory + // creation but before the exclusion the directory would remain non-excluded from + // backups because we only perform exclusion right after we created the directory + // ourselves. + // + // We need the tempdir created in parent instead of $TMP, because only then we can be + // easily sure that rename() will succeed (the new name needs to be on the same mount + // point as the old one). + let tempdir = TempFileBuilder::new().prefix(base).tempdir_in(parent)?; + exclude_from_backups(tempdir.path()); + exclude_from_content_indexing(tempdir.path()); + // Previously std::fs::create_dir_all() (through paths::create_dir_all()) was used + // here to create the directory directly and fs::create_dir_all() explicitly treats + // the directory being created concurrently by another thread or process as success, + // hence the check below to follow the existing behavior. If we get an error at + // rename() and suddenly the directory (which didn't exist a moment earlier) exists + // we can infer from it's another cargo process doing work. + if let Err(e) = fs::rename(tempdir.path(), path) { + if !path.exists() { + return Err(anyhow::Error::from(e)); + } + } + Ok(()) +} + +/// Mark an existing directory as excluded from backups and indexing. +/// +/// Errors in marking it are ignored. +pub fn exclude_from_backups_and_indexing(p: impl AsRef<Path>) { + let path = p.as_ref(); + exclude_from_backups(path); + exclude_from_content_indexing(path); +} + +/// Marks the directory as excluded from archives/backups. +/// +/// This is recommended to prevent derived/temporary files from bloating backups. There are two +/// mechanisms used to achieve this right now: +/// +/// * A dedicated resource property excluding from Time Machine backups on macOS +/// * CACHEDIR.TAG files supported by various tools in a platform-independent way +fn exclude_from_backups(path: &Path) { + exclude_from_time_machine(path); + let _ = std::fs::write( + path.join("CACHEDIR.TAG"), + "Signature: 8a477f597d28d172789f06886806bc55 +# This file is a cache directory tag created by cargo. +# For information about cache directory tags see https://bford.info/cachedir/ +", + ); + // Similarly to exclude_from_time_machine() we ignore errors here as it's an optional feature. +} + +/// Marks the directory as excluded from content indexing. +/// +/// This is recommended to prevent the content of derived/temporary files from being indexed. +/// This is very important for Windows users, as the live content indexing significantly slows +/// cargo's I/O operations. +/// +/// This is currently a no-op on non-Windows platforms. +fn exclude_from_content_indexing(path: &Path) { + #[cfg(windows)] + { + use std::iter::once; + use std::os::windows::prelude::OsStrExt; + use windows_sys::Win32::Storage::FileSystem::{ + GetFileAttributesW, SetFileAttributesW, FILE_ATTRIBUTE_NOT_CONTENT_INDEXED, + }; + + let path: Vec<u16> = path.as_os_str().encode_wide().chain(once(0)).collect(); + unsafe { + SetFileAttributesW( + path.as_ptr(), + GetFileAttributesW(path.as_ptr()) | FILE_ATTRIBUTE_NOT_CONTENT_INDEXED, + ); + } + } + #[cfg(not(windows))] + { + let _ = path; + } +} + +#[cfg(not(target_os = "macos"))] +fn exclude_from_time_machine(_: &Path) {} + +#[cfg(target_os = "macos")] +/// Marks files or directories as excluded from Time Machine on macOS +fn exclude_from_time_machine(path: &Path) { + use core_foundation::base::TCFType; + use core_foundation::{number, string, url}; + use std::ptr; + + // For compatibility with 10.7 a string is used instead of global kCFURLIsExcludedFromBackupKey + let is_excluded_key: Result<string::CFString, _> = "NSURLIsExcludedFromBackupKey".parse(); + let path = url::CFURL::from_path(path, false); + if let (Some(path), Ok(is_excluded_key)) = (path, is_excluded_key) { + unsafe { + url::CFURLSetResourcePropertyForKey( + path.as_concrete_TypeRef(), + is_excluded_key.as_concrete_TypeRef(), + number::kCFBooleanTrue as *const _, + ptr::null_mut(), + ); + } + } + // Errors are ignored, since it's an optional feature and failure + // doesn't prevent Cargo from working +} + +#[cfg(test)] +mod tests { + use super::join_paths; + + #[test] + fn join_paths_lists_paths_on_error() { + let valid_paths = vec!["/testing/one", "/testing/two"]; + // does not fail on valid input + let _joined = join_paths(&valid_paths, "TESTING1").unwrap(); + + #[cfg(unix)] + { + let invalid_paths = vec!["/testing/one", "/testing/t:wo/three"]; + let err = join_paths(&invalid_paths, "TESTING2").unwrap_err(); + assert_eq!( + err.to_string(), + "failed to join paths from `$TESTING2` together\n\n\ + Check if any of path segments listed below contain an \ + unterminated quote character or path separator:\ + \n \"/testing/one\"\ + \n \"/testing/t:wo/three\"\ + " + ); + } + #[cfg(windows)] + { + let invalid_paths = vec!["/testing/one", "/testing/t\"wo/three"]; + let err = join_paths(&invalid_paths, "TESTING2").unwrap_err(); + assert_eq!( + err.to_string(), + "failed to join paths from `$TESTING2` together\n\n\ + Check if any of path segments listed below contain an \ + unterminated quote character or path separator:\ + \n \"/testing/one\"\ + \n \"/testing/t\\\"wo/three\"\ + " + ); + } + } +} diff --git a/crates/cargo-util/src/process_builder.rs b/crates/cargo-util/src/process_builder.rs new file mode 100644 index 0000000..76392f2 --- /dev/null +++ b/crates/cargo-util/src/process_builder.rs @@ -0,0 +1,689 @@ +use crate::process_error::ProcessError; +use crate::read2; + +use anyhow::{bail, Context, Result}; +use jobserver::Client; +use shell_escape::escape; +use tempfile::NamedTempFile; + +use std::collections::BTreeMap; +use std::env; +use std::ffi::{OsStr, OsString}; +use std::fmt; +use std::io::{self, Write}; +use std::iter::once; +use std::path::Path; +use std::process::{Command, ExitStatus, Output, Stdio}; + +/// A builder object for an external process, similar to [`std::process::Command`]. +#[derive(Clone, Debug)] +pub struct ProcessBuilder { + /// The program to execute. + program: OsString, + /// A list of arguments to pass to the program. + args: Vec<OsString>, + /// Any environment variables that should be set for the program. + env: BTreeMap<String, Option<OsString>>, + /// The directory to run the program from. + cwd: Option<OsString>, + /// A list of wrappers that wrap the original program when calling + /// [`ProcessBuilder::wrapped`]. The last one is the outermost one. + wrappers: Vec<OsString>, + /// The `make` jobserver. See the [jobserver crate] for + /// more information. + /// + /// [jobserver crate]: https://docs.rs/jobserver/ + jobserver: Option<Client>, + /// `true` to include environment variable in display. + display_env_vars: bool, + /// `true` to retry with an argfile if hitting "command line too big" error. + /// See [`ProcessBuilder::retry_with_argfile`] for more information. + retry_with_argfile: bool, + /// Data to write to stdin. + stdin: Option<Vec<u8>>, +} + +impl fmt::Display for ProcessBuilder { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "`")?; + + if self.display_env_vars { + for (key, val) in self.env.iter() { + if let Some(val) = val { + let val = escape(val.to_string_lossy()); + if cfg!(windows) { + write!(f, "set {}={}&& ", key, val)?; + } else { + write!(f, "{}={} ", key, val)?; + } + } + } + } + + write!(f, "{}", self.get_program().to_string_lossy())?; + + for arg in self.get_args() { + write!(f, " {}", escape(arg.to_string_lossy()))?; + } + + write!(f, "`") + } +} + +impl ProcessBuilder { + /// Creates a new [`ProcessBuilder`] with the given executable path. + pub fn new<T: AsRef<OsStr>>(cmd: T) -> ProcessBuilder { + ProcessBuilder { + program: cmd.as_ref().to_os_string(), + args: Vec::new(), + cwd: None, + env: BTreeMap::new(), + wrappers: Vec::new(), + jobserver: None, + display_env_vars: false, + retry_with_argfile: false, + stdin: None, + } + } + + /// (chainable) Sets the executable for the process. + pub fn program<T: AsRef<OsStr>>(&mut self, program: T) -> &mut ProcessBuilder { + self.program = program.as_ref().to_os_string(); + self + } + + /// (chainable) Adds `arg` to the args list. + pub fn arg<T: AsRef<OsStr>>(&mut self, arg: T) -> &mut ProcessBuilder { + self.args.push(arg.as_ref().to_os_string()); + self + } + + /// (chainable) Adds multiple `args` to the args list. + pub fn args<T: AsRef<OsStr>>(&mut self, args: &[T]) -> &mut ProcessBuilder { + self.args + .extend(args.iter().map(|t| t.as_ref().to_os_string())); + self + } + + /// (chainable) Replaces the args list with the given `args`. + pub fn args_replace<T: AsRef<OsStr>>(&mut self, args: &[T]) -> &mut ProcessBuilder { + if let Some(program) = self.wrappers.pop() { + // User intend to replace all args, so we + // - use the outermost wrapper as the main program, and + // - cleanup other inner wrappers. + self.program = program; + self.wrappers = Vec::new(); + } + self.args = args.iter().map(|t| t.as_ref().to_os_string()).collect(); + self + } + + /// (chainable) Sets the current working directory of the process. + pub fn cwd<T: AsRef<OsStr>>(&mut self, path: T) -> &mut ProcessBuilder { + self.cwd = Some(path.as_ref().to_os_string()); + self + } + + /// (chainable) Sets an environment variable for the process. + pub fn env<T: AsRef<OsStr>>(&mut self, key: &str, val: T) -> &mut ProcessBuilder { + self.env + .insert(key.to_string(), Some(val.as_ref().to_os_string())); + self + } + + /// (chainable) Unsets an environment variable for the process. + pub fn env_remove(&mut self, key: &str) -> &mut ProcessBuilder { + self.env.insert(key.to_string(), None); + self + } + + /// Gets the executable name. + pub fn get_program(&self) -> &OsString { + self.wrappers.last().unwrap_or(&self.program) + } + + /// Gets the program arguments. + pub fn get_args(&self) -> impl Iterator<Item = &OsString> { + self.wrappers + .iter() + .rev() + .chain(once(&self.program)) + .chain(self.args.iter()) + .skip(1) // Skip the main `program + } + + /// Gets the current working directory for the process. + pub fn get_cwd(&self) -> Option<&Path> { + self.cwd.as_ref().map(Path::new) + } + + /// Gets an environment variable as the process will see it (will inherit from environment + /// unless explicitally unset). + pub fn get_env(&self, var: &str) -> Option<OsString> { + self.env + .get(var) + .cloned() + .or_else(|| Some(env::var_os(var))) + .and_then(|s| s) + } + + /// Gets all environment variables explicitly set or unset for the process (not inherited + /// vars). + pub fn get_envs(&self) -> &BTreeMap<String, Option<OsString>> { + &self.env + } + + /// Sets the `make` jobserver. See the [jobserver crate][jobserver_docs] for + /// more information. + /// + /// [jobserver_docs]: https://docs.rs/jobserver/0.1.6/jobserver/ + pub fn inherit_jobserver(&mut self, jobserver: &Client) -> &mut Self { + self.jobserver = Some(jobserver.clone()); + self + } + + /// Enables environment variable display. + pub fn display_env_vars(&mut self) -> &mut Self { + self.display_env_vars = true; + self + } + + /// Enables retrying with an argfile if hitting "command line too big" error + /// + /// This is primarily for the `@path` arg of rustc and rustdoc, which treat + /// each line as an command-line argument, so `LF` and `CRLF` bytes are not + /// valid as an argument for argfile at this moment. + /// For example, `RUSTDOCFLAGS="--crate-version foo\nbar" cargo doc` is + /// valid when invoking from command-line but not from argfile. + /// + /// To sum up, the limitations of the argfile are: + /// + /// - Must be valid UTF-8 encoded. + /// - Must not contain any newlines in each argument. + /// + /// Ref: + /// + /// - <https://doc.rust-lang.org/rustdoc/command-line-arguments.html#path-load-command-line-flags-from-a-path> + /// - <https://doc.rust-lang.org/rustc/command-line-arguments.html#path-load-command-line-flags-from-a-path> + pub fn retry_with_argfile(&mut self, enabled: bool) -> &mut Self { + self.retry_with_argfile = enabled; + self + } + + /// Sets a value that will be written to stdin of the process on launch. + pub fn stdin<T: Into<Vec<u8>>>(&mut self, stdin: T) -> &mut Self { + self.stdin = Some(stdin.into()); + self + } + + fn should_retry_with_argfile(&self, err: &io::Error) -> bool { + self.retry_with_argfile && imp::command_line_too_big(err) + } + + /// Like [`Command::status`] but with a better error message. + pub fn status(&self) -> Result<ExitStatus> { + self._status() + .with_context(|| ProcessError::could_not_execute(self)) + } + + fn _status(&self) -> io::Result<ExitStatus> { + if !debug_force_argfile(self.retry_with_argfile) { + let mut cmd = self.build_command(); + match cmd.spawn() { + Err(ref e) if self.should_retry_with_argfile(e) => {} + Err(e) => return Err(e), + Ok(mut child) => return child.wait(), + } + } + let (mut cmd, argfile) = self.build_command_with_argfile()?; + let status = cmd.spawn()?.wait(); + close_tempfile_and_log_error(argfile); + status + } + + /// Runs the process, waiting for completion, and mapping non-success exit codes to an error. + pub fn exec(&self) -> Result<()> { + let exit = self.status()?; + if exit.success() { + Ok(()) + } else { + Err(ProcessError::new( + &format!("process didn't exit successfully: {}", self), + Some(exit), + None, + ) + .into()) + } + } + + /// Replaces the current process with the target process. + /// + /// On Unix, this executes the process using the Unix syscall `execvp`, which will block + /// this process, and will only return if there is an error. + /// + /// On Windows this isn't technically possible. Instead we emulate it to the best of our + /// ability. One aspect we fix here is that we specify a handler for the Ctrl-C handler. + /// In doing so (and by effectively ignoring it) we should emulate proxying Ctrl-C + /// handling to the application at hand, which will either terminate or handle it itself. + /// According to Microsoft's documentation at + /// <https://docs.microsoft.com/en-us/windows/console/ctrl-c-and-ctrl-break-signals>. + /// the Ctrl-C signal is sent to all processes attached to a terminal, which should + /// include our child process. If the child terminates then we'll reap them in Cargo + /// pretty quickly, and if the child handles the signal then we won't terminate + /// (and we shouldn't!) until the process itself later exits. + pub fn exec_replace(&self) -> Result<()> { + imp::exec_replace(self) + } + + /// Like [`Command::output`] but with a better error message. + pub fn output(&self) -> Result<Output> { + self._output() + .with_context(|| ProcessError::could_not_execute(self)) + } + + fn _output(&self) -> io::Result<Output> { + if !debug_force_argfile(self.retry_with_argfile) { + let mut cmd = self.build_command(); + match piped(&mut cmd, self.stdin.is_some()).spawn() { + Err(ref e) if self.should_retry_with_argfile(e) => {} + Err(e) => return Err(e), + Ok(mut child) => { + if let Some(stdin) = &self.stdin { + child.stdin.take().unwrap().write_all(stdin)?; + } + return child.wait_with_output(); + } + } + } + let (mut cmd, argfile) = self.build_command_with_argfile()?; + let mut child = piped(&mut cmd, self.stdin.is_some()).spawn()?; + if let Some(stdin) = &self.stdin { + child.stdin.take().unwrap().write_all(stdin)?; + } + let output = child.wait_with_output(); + close_tempfile_and_log_error(argfile); + output + } + + /// Executes the process, returning the stdio output, or an error if non-zero exit status. + pub fn exec_with_output(&self) -> Result<Output> { + let output = self.output()?; + if output.status.success() { + Ok(output) + } else { + Err(ProcessError::new( + &format!("process didn't exit successfully: {}", self), + Some(output.status), + Some(&output), + ) + .into()) + } + } + + /// Executes a command, passing each line of stdout and stderr to the supplied callbacks, which + /// can mutate the string data. + /// + /// If any invocations of these function return an error, it will be propagated. + /// + /// If `capture_output` is true, then all the output will also be buffered + /// and stored in the returned `Output` object. If it is false, no caching + /// is done, and the callbacks are solely responsible for handling the + /// output. + pub fn exec_with_streaming( + &self, + on_stdout_line: &mut dyn FnMut(&str) -> Result<()>, + on_stderr_line: &mut dyn FnMut(&str) -> Result<()>, + capture_output: bool, + ) -> Result<Output> { + let mut stdout = Vec::new(); + let mut stderr = Vec::new(); + + let mut callback_error = None; + let mut stdout_pos = 0; + let mut stderr_pos = 0; + + let spawn = |mut cmd| { + if !debug_force_argfile(self.retry_with_argfile) { + match piped(&mut cmd, false).spawn() { + Err(ref e) if self.should_retry_with_argfile(e) => {} + Err(e) => return Err(e), + Ok(child) => return Ok((child, None)), + } + } + let (mut cmd, argfile) = self.build_command_with_argfile()?; + Ok((piped(&mut cmd, false).spawn()?, Some(argfile))) + }; + + let status = (|| { + let cmd = self.build_command(); + let (mut child, argfile) = spawn(cmd)?; + let out = child.stdout.take().unwrap(); + let err = child.stderr.take().unwrap(); + read2(out, err, &mut |is_out, data, eof| { + let pos = if is_out { + &mut stdout_pos + } else { + &mut stderr_pos + }; + let idx = if eof { + data.len() + } else { + match data[*pos..].iter().rposition(|b| *b == b'\n') { + Some(i) => *pos + i + 1, + None => { + *pos = data.len(); + return; + } + } + }; + + let new_lines = &data[..idx]; + + for line in String::from_utf8_lossy(new_lines).lines() { + if callback_error.is_some() { + break; + } + let callback_result = if is_out { + on_stdout_line(line) + } else { + on_stderr_line(line) + }; + if let Err(e) = callback_result { + callback_error = Some(e); + break; + } + } + + if capture_output { + let dst = if is_out { &mut stdout } else { &mut stderr }; + dst.extend(new_lines); + } + + data.drain(..idx); + *pos = 0; + })?; + let status = child.wait(); + if let Some(argfile) = argfile { + close_tempfile_and_log_error(argfile); + } + status + })() + .with_context(|| ProcessError::could_not_execute(self))?; + let output = Output { + status, + stdout, + stderr, + }; + + { + let to_print = if capture_output { Some(&output) } else { None }; + if let Some(e) = callback_error { + let cx = ProcessError::new( + &format!("failed to parse process output: {}", self), + Some(output.status), + to_print, + ); + bail!(anyhow::Error::new(cx).context(e)); + } else if !output.status.success() { + bail!(ProcessError::new( + &format!("process didn't exit successfully: {}", self), + Some(output.status), + to_print, + )); + } + } + + Ok(output) + } + + /// Builds the command with an `@<path>` argfile that contains all the + /// arguments. This is primarily served for rustc/rustdoc command family. + fn build_command_with_argfile(&self) -> io::Result<(Command, NamedTempFile)> { + use std::io::Write as _; + + let mut tmp = tempfile::Builder::new() + .prefix("cargo-argfile.") + .tempfile()?; + + let mut arg = OsString::from("@"); + arg.push(tmp.path()); + let mut cmd = self.build_command_without_args(); + cmd.arg(arg); + log::debug!("created argfile at {} for {self}", tmp.path().display()); + + let cap = self.get_args().map(|arg| arg.len() + 1).sum::<usize>(); + let mut buf = Vec::with_capacity(cap); + for arg in &self.args { + let arg = arg.to_str().ok_or_else(|| { + io::Error::new( + io::ErrorKind::Other, + format!( + "argument for argfile contains invalid UTF-8 characters: `{}`", + arg.to_string_lossy() + ), + ) + })?; + if arg.contains('\n') { + return Err(io::Error::new( + io::ErrorKind::Other, + format!("argument for argfile contains newlines: `{arg}`"), + )); + } + writeln!(buf, "{arg}")?; + } + tmp.write_all(&mut buf)?; + Ok((cmd, tmp)) + } + + /// Builds a command from `ProcessBuilder` for everything but not `args`. + fn build_command_without_args(&self) -> Command { + let mut command = { + let mut iter = self.wrappers.iter().rev().chain(once(&self.program)); + let mut cmd = Command::new(iter.next().expect("at least one `program` exists")); + cmd.args(iter); + cmd + }; + if let Some(cwd) = self.get_cwd() { + command.current_dir(cwd); + } + for (k, v) in &self.env { + match *v { + Some(ref v) => { + command.env(k, v); + } + None => { + command.env_remove(k); + } + } + } + if let Some(ref c) = self.jobserver { + c.configure(&mut command); + } + command + } + + /// Converts `ProcessBuilder` into a `std::process::Command`, and handles + /// the jobserver, if present. + /// + /// Note that this method doesn't take argfile fallback into account. The + /// caller should handle it by themselves. + pub fn build_command(&self) -> Command { + let mut command = self.build_command_without_args(); + for arg in &self.args { + command.arg(arg); + } + command + } + + /// Wraps an existing command with the provided wrapper, if it is present and valid. + /// + /// # Examples + /// + /// ```rust + /// use cargo_util::ProcessBuilder; + /// // Running this would execute `rustc` + /// let cmd = ProcessBuilder::new("rustc"); + /// + /// // Running this will execute `sccache rustc` + /// let cmd = cmd.wrapped(Some("sccache")); + /// ``` + pub fn wrapped(mut self, wrapper: Option<impl AsRef<OsStr>>) -> Self { + if let Some(wrapper) = wrapper.as_ref() { + let wrapper = wrapper.as_ref(); + if !wrapper.is_empty() { + self.wrappers.push(wrapper.to_os_string()); + } + } + self + } +} + +/// Forces the command to use `@path` argfile. +/// +/// You should set `__CARGO_TEST_FORCE_ARGFILE` to enable this. +fn debug_force_argfile(retry_enabled: bool) -> bool { + cfg!(debug_assertions) && env::var("__CARGO_TEST_FORCE_ARGFILE").is_ok() && retry_enabled +} + +/// Creates new pipes for stderr, stdout, and optionally stdin. +fn piped(cmd: &mut Command, pipe_stdin: bool) -> &mut Command { + cmd.stdout(Stdio::piped()) + .stderr(Stdio::piped()) + .stdin(if pipe_stdin { + Stdio::piped() + } else { + Stdio::null() + }) +} + +fn close_tempfile_and_log_error(file: NamedTempFile) { + file.close().unwrap_or_else(|e| { + log::warn!("failed to close temporary file: {e}"); + }); +} + +#[cfg(unix)] +mod imp { + use super::{close_tempfile_and_log_error, debug_force_argfile, ProcessBuilder, ProcessError}; + use anyhow::Result; + use std::io; + use std::os::unix::process::CommandExt; + + pub fn exec_replace(process_builder: &ProcessBuilder) -> Result<()> { + let mut error; + let mut file = None; + if debug_force_argfile(process_builder.retry_with_argfile) { + let (mut command, argfile) = process_builder.build_command_with_argfile()?; + file = Some(argfile); + error = command.exec() + } else { + let mut command = process_builder.build_command(); + error = command.exec(); + if process_builder.should_retry_with_argfile(&error) { + let (mut command, argfile) = process_builder.build_command_with_argfile()?; + file = Some(argfile); + error = command.exec() + } + } + if let Some(file) = file { + close_tempfile_and_log_error(file); + } + + Err(anyhow::Error::from(error).context(ProcessError::new( + &format!("could not execute process {}", process_builder), + None, + None, + ))) + } + + pub fn command_line_too_big(err: &io::Error) -> bool { + err.raw_os_error() == Some(libc::E2BIG) + } +} + +#[cfg(windows)] +mod imp { + use super::{ProcessBuilder, ProcessError}; + use anyhow::Result; + use std::io; + use windows_sys::Win32::Foundation::{BOOL, FALSE, TRUE}; + use windows_sys::Win32::System::Console::SetConsoleCtrlHandler; + + unsafe extern "system" fn ctrlc_handler(_: u32) -> BOOL { + // Do nothing; let the child process handle it. + TRUE + } + + pub fn exec_replace(process_builder: &ProcessBuilder) -> Result<()> { + unsafe { + if SetConsoleCtrlHandler(Some(ctrlc_handler), TRUE) == FALSE { + return Err(ProcessError::new("Could not set Ctrl-C handler.", None, None).into()); + } + } + + // Just execute the process as normal. + process_builder.exec() + } + + pub fn command_line_too_big(err: &io::Error) -> bool { + use windows_sys::Win32::Foundation::ERROR_FILENAME_EXCED_RANGE; + err.raw_os_error() == Some(ERROR_FILENAME_EXCED_RANGE as i32) + } +} + +#[cfg(test)] +mod tests { + use super::ProcessBuilder; + use std::fs; + + #[test] + fn argfile_build_succeeds() { + let mut cmd = ProcessBuilder::new("echo"); + cmd.args(["foo", "bar"].as_slice()); + let (cmd, argfile) = cmd.build_command_with_argfile().unwrap(); + + assert_eq!(cmd.get_program(), "echo"); + let cmd_args: Vec<_> = cmd.get_args().map(|s| s.to_str().unwrap()).collect(); + assert_eq!(cmd_args.len(), 1); + assert!(cmd_args[0].starts_with("@")); + assert!(cmd_args[0].contains("cargo-argfile.")); + + let buf = fs::read_to_string(argfile.path()).unwrap(); + assert_eq!(buf, "foo\nbar\n"); + } + + #[test] + fn argfile_build_fails_if_arg_contains_newline() { + let mut cmd = ProcessBuilder::new("echo"); + cmd.arg("foo\n"); + let err = cmd.build_command_with_argfile().unwrap_err(); + assert_eq!( + err.to_string(), + "argument for argfile contains newlines: `foo\n`" + ); + } + + #[test] + fn argfile_build_fails_if_arg_contains_invalid_utf8() { + let mut cmd = ProcessBuilder::new("echo"); + + #[cfg(windows)] + let invalid_arg = { + use std::os::windows::prelude::*; + std::ffi::OsString::from_wide(&[0x0066, 0x006f, 0xD800, 0x006f]) + }; + + #[cfg(unix)] + let invalid_arg = { + use std::os::unix::ffi::OsStrExt; + std::ffi::OsStr::from_bytes(&[0x66, 0x6f, 0x80, 0x6f]).to_os_string() + }; + + cmd.arg(invalid_arg); + let err = cmd.build_command_with_argfile().unwrap_err(); + assert_eq!( + err.to_string(), + "argument for argfile contains invalid UTF-8 characters: `fo�o`" + ); + } +} diff --git a/crates/cargo-util/src/process_error.rs b/crates/cargo-util/src/process_error.rs new file mode 100644 index 0000000..9b4a38c --- /dev/null +++ b/crates/cargo-util/src/process_error.rs @@ -0,0 +1,200 @@ +//! Error value for [`crate::ProcessBuilder`] when a process fails. + +use std::fmt; +use std::process::{ExitStatus, Output}; +use std::str; + +#[derive(Debug)] +pub struct ProcessError { + /// A detailed description to show to the user why the process failed. + pub desc: String, + + /// The exit status of the process. + /// + /// This can be `None` if the process failed to launch (like process not + /// found) or if the exit status wasn't a code but was instead something + /// like termination via a signal. + pub code: Option<i32>, + + /// The stdout from the process. + /// + /// This can be `None` if the process failed to launch, or the output was + /// not captured. + pub stdout: Option<Vec<u8>>, + + /// The stderr from the process. + /// + /// This can be `None` if the process failed to launch, or the output was + /// not captured. + pub stderr: Option<Vec<u8>>, +} + +impl fmt::Display for ProcessError { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.desc.fmt(f) + } +} + +impl std::error::Error for ProcessError {} + +impl ProcessError { + /// Creates a new [`ProcessError`]. + /// + /// * `status` can be `None` if the process did not launch. + /// * `output` can be `None` if the process did not launch, or output was not captured. + pub fn new(msg: &str, status: Option<ExitStatus>, output: Option<&Output>) -> ProcessError { + let exit = match status { + Some(s) => exit_status_to_string(s), + None => "never executed".to_string(), + }; + + Self::new_raw( + msg, + status.and_then(|s| s.code()), + &exit, + output.map(|s| s.stdout.as_slice()), + output.map(|s| s.stderr.as_slice()), + ) + } + + /// Creates a new [`ProcessError`] with the raw output data. + /// + /// * `code` can be `None` for situations like being killed by a signal on unix. + pub fn new_raw( + msg: &str, + code: Option<i32>, + status: &str, + stdout: Option<&[u8]>, + stderr: Option<&[u8]>, + ) -> ProcessError { + let mut desc = format!("{} ({})", msg, status); + + if let Some(out) = stdout { + match str::from_utf8(out) { + Ok(s) if !s.trim().is_empty() => { + desc.push_str("\n--- stdout\n"); + desc.push_str(s); + } + Ok(..) | Err(..) => {} + } + } + if let Some(out) = stderr { + match str::from_utf8(out) { + Ok(s) if !s.trim().is_empty() => { + desc.push_str("\n--- stderr\n"); + desc.push_str(s); + } + Ok(..) | Err(..) => {} + } + } + + ProcessError { + desc, + code, + stdout: stdout.map(|s| s.to_vec()), + stderr: stderr.map(|s| s.to_vec()), + } + } + + /// Creates a [`ProcessError`] with "could not execute process {cmd}". + /// + /// * `cmd` is usually but not limited to [`std::process::Command`]. + pub fn could_not_execute(cmd: impl fmt::Display) -> ProcessError { + ProcessError::new(&format!("could not execute process {cmd}"), None, None) + } +} + +/// Converts an [`ExitStatus`] to a human-readable string suitable for +/// displaying to a user. +pub fn exit_status_to_string(status: ExitStatus) -> String { + return status_to_string(status); + + #[cfg(unix)] + fn status_to_string(status: ExitStatus) -> String { + use std::os::unix::process::*; + + if let Some(signal) = status.signal() { + let name = match signal as libc::c_int { + libc::SIGABRT => ", SIGABRT: process abort signal", + libc::SIGALRM => ", SIGALRM: alarm clock", + libc::SIGFPE => ", SIGFPE: erroneous arithmetic operation", + libc::SIGHUP => ", SIGHUP: hangup", + libc::SIGILL => ", SIGILL: illegal instruction", + libc::SIGINT => ", SIGINT: terminal interrupt signal", + libc::SIGKILL => ", SIGKILL: kill", + libc::SIGPIPE => ", SIGPIPE: write on a pipe with no one to read", + libc::SIGQUIT => ", SIGQUIT: terminal quit signal", + libc::SIGSEGV => ", SIGSEGV: invalid memory reference", + libc::SIGTERM => ", SIGTERM: termination signal", + libc::SIGBUS => ", SIGBUS: access to undefined memory", + #[cfg(not(target_os = "haiku"))] + libc::SIGSYS => ", SIGSYS: bad system call", + libc::SIGTRAP => ", SIGTRAP: trace/breakpoint trap", + _ => "", + }; + format!("signal: {}{}", signal, name) + } else { + status.to_string() + } + } + + #[cfg(windows)] + fn status_to_string(status: ExitStatus) -> String { + use windows_sys::Win32::Foundation::*; + + let mut base = status.to_string(); + let extra = match status.code().unwrap() as i32 { + STATUS_ACCESS_VIOLATION => "STATUS_ACCESS_VIOLATION", + STATUS_IN_PAGE_ERROR => "STATUS_IN_PAGE_ERROR", + STATUS_INVALID_HANDLE => "STATUS_INVALID_HANDLE", + STATUS_INVALID_PARAMETER => "STATUS_INVALID_PARAMETER", + STATUS_NO_MEMORY => "STATUS_NO_MEMORY", + STATUS_ILLEGAL_INSTRUCTION => "STATUS_ILLEGAL_INSTRUCTION", + STATUS_NONCONTINUABLE_EXCEPTION => "STATUS_NONCONTINUABLE_EXCEPTION", + STATUS_INVALID_DISPOSITION => "STATUS_INVALID_DISPOSITION", + STATUS_ARRAY_BOUNDS_EXCEEDED => "STATUS_ARRAY_BOUNDS_EXCEEDED", + STATUS_FLOAT_DENORMAL_OPERAND => "STATUS_FLOAT_DENORMAL_OPERAND", + STATUS_FLOAT_DIVIDE_BY_ZERO => "STATUS_FLOAT_DIVIDE_BY_ZERO", + STATUS_FLOAT_INEXACT_RESULT => "STATUS_FLOAT_INEXACT_RESULT", + STATUS_FLOAT_INVALID_OPERATION => "STATUS_FLOAT_INVALID_OPERATION", + STATUS_FLOAT_OVERFLOW => "STATUS_FLOAT_OVERFLOW", + STATUS_FLOAT_STACK_CHECK => "STATUS_FLOAT_STACK_CHECK", + STATUS_FLOAT_UNDERFLOW => "STATUS_FLOAT_UNDERFLOW", + STATUS_INTEGER_DIVIDE_BY_ZERO => "STATUS_INTEGER_DIVIDE_BY_ZERO", + STATUS_INTEGER_OVERFLOW => "STATUS_INTEGER_OVERFLOW", + STATUS_PRIVILEGED_INSTRUCTION => "STATUS_PRIVILEGED_INSTRUCTION", + STATUS_STACK_OVERFLOW => "STATUS_STACK_OVERFLOW", + STATUS_DLL_NOT_FOUND => "STATUS_DLL_NOT_FOUND", + STATUS_ORDINAL_NOT_FOUND => "STATUS_ORDINAL_NOT_FOUND", + STATUS_ENTRYPOINT_NOT_FOUND => "STATUS_ENTRYPOINT_NOT_FOUND", + STATUS_CONTROL_C_EXIT => "STATUS_CONTROL_C_EXIT", + STATUS_DLL_INIT_FAILED => "STATUS_DLL_INIT_FAILED", + STATUS_FLOAT_MULTIPLE_FAULTS => "STATUS_FLOAT_MULTIPLE_FAULTS", + STATUS_FLOAT_MULTIPLE_TRAPS => "STATUS_FLOAT_MULTIPLE_TRAPS", + STATUS_REG_NAT_CONSUMPTION => "STATUS_REG_NAT_CONSUMPTION", + STATUS_HEAP_CORRUPTION => "STATUS_HEAP_CORRUPTION", + STATUS_STACK_BUFFER_OVERRUN => "STATUS_STACK_BUFFER_OVERRUN", + STATUS_ASSERTION_FAILURE => "STATUS_ASSERTION_FAILURE", + _ => return base, + }; + base.push_str(", "); + base.push_str(extra); + base + } +} + +/// Returns `true` if the given process exit code is something a normal +/// process would exit with. +/// +/// This helps differentiate from abnormal termination codes, such as +/// segmentation faults or signals. +pub fn is_simple_exit_code(code: i32) -> bool { + // Typical unix exit codes are 0 to 127. + // Windows doesn't have anything "typical", and is a + // 32-bit number (which appears signed here, but is really + // unsigned). However, most of the interesting NTSTATUS + // codes are very large. This is just a rough + // approximation of which codes are "normal" and which + // ones are abnormal termination. + code >= 0 && code <= 127 +} diff --git a/crates/cargo-util/src/read2.rs b/crates/cargo-util/src/read2.rs new file mode 100644 index 0000000..742dc1d --- /dev/null +++ b/crates/cargo-util/src/read2.rs @@ -0,0 +1,178 @@ +pub use self::imp::read2; + +#[cfg(unix)] +mod imp { + use std::io; + use std::io::prelude::*; + use std::mem; + use std::os::unix::prelude::*; + use std::process::{ChildStderr, ChildStdout}; + + pub fn read2( + mut out_pipe: ChildStdout, + mut err_pipe: ChildStderr, + data: &mut dyn FnMut(bool, &mut Vec<u8>, bool), + ) -> io::Result<()> { + unsafe { + libc::fcntl(out_pipe.as_raw_fd(), libc::F_SETFL, libc::O_NONBLOCK); + libc::fcntl(err_pipe.as_raw_fd(), libc::F_SETFL, libc::O_NONBLOCK); + } + + let mut out_done = false; + let mut err_done = false; + let mut out = Vec::new(); + let mut err = Vec::new(); + + let mut fds: [libc::pollfd; 2] = unsafe { mem::zeroed() }; + fds[0].fd = out_pipe.as_raw_fd(); + fds[0].events = libc::POLLIN; + fds[1].fd = err_pipe.as_raw_fd(); + fds[1].events = libc::POLLIN; + let mut nfds = 2; + let mut errfd = 1; + + while nfds > 0 { + // wait for either pipe to become readable using `select` + let r = unsafe { libc::poll(fds.as_mut_ptr(), nfds, -1) }; + if r == -1 { + let err = io::Error::last_os_error(); + if err.kind() == io::ErrorKind::Interrupted { + continue; + } + return Err(err); + } + + // Read as much as we can from each pipe, ignoring EWOULDBLOCK or + // EAGAIN. If we hit EOF, then this will happen because the underlying + // reader will return Ok(0), in which case we'll see `Ok` ourselves. In + // this case we flip the other fd back into blocking mode and read + // whatever's leftover on that file descriptor. + let handle = |res: io::Result<_>| match res { + Ok(_) => Ok(true), + Err(e) => { + if e.kind() == io::ErrorKind::WouldBlock { + Ok(false) + } else { + Err(e) + } + } + }; + if !err_done && fds[errfd].revents != 0 && handle(err_pipe.read_to_end(&mut err))? { + err_done = true; + nfds -= 1; + } + data(false, &mut err, err_done); + if !out_done && fds[0].revents != 0 && handle(out_pipe.read_to_end(&mut out))? { + out_done = true; + fds[0].fd = err_pipe.as_raw_fd(); + errfd = 0; + nfds -= 1; + } + data(true, &mut out, out_done); + } + Ok(()) + } +} + +#[cfg(windows)] +mod imp { + use std::io; + use std::os::windows::prelude::*; + use std::process::{ChildStderr, ChildStdout}; + use std::slice; + + use miow::iocp::{CompletionPort, CompletionStatus}; + use miow::pipe::NamedPipe; + use miow::Overlapped; + use windows_sys::Win32::Foundation::ERROR_BROKEN_PIPE; + + struct Pipe<'a> { + dst: &'a mut Vec<u8>, + overlapped: Overlapped, + pipe: NamedPipe, + done: bool, + } + + pub fn read2( + out_pipe: ChildStdout, + err_pipe: ChildStderr, + data: &mut dyn FnMut(bool, &mut Vec<u8>, bool), + ) -> io::Result<()> { + let mut out = Vec::new(); + let mut err = Vec::new(); + + let port = CompletionPort::new(1)?; + port.add_handle(0, &out_pipe)?; + port.add_handle(1, &err_pipe)?; + + unsafe { + let mut out_pipe = Pipe::new(out_pipe, &mut out); + let mut err_pipe = Pipe::new(err_pipe, &mut err); + + out_pipe.read()?; + err_pipe.read()?; + + let mut status = [CompletionStatus::zero(), CompletionStatus::zero()]; + + while !out_pipe.done || !err_pipe.done { + for status in port.get_many(&mut status, None)? { + if status.token() == 0 { + out_pipe.complete(status); + data(true, out_pipe.dst, out_pipe.done); + out_pipe.read()?; + } else { + err_pipe.complete(status); + data(false, err_pipe.dst, err_pipe.done); + err_pipe.read()?; + } + } + } + + Ok(()) + } + } + + impl<'a> Pipe<'a> { + unsafe fn new<P: IntoRawHandle>(p: P, dst: &'a mut Vec<u8>) -> Pipe<'a> { + Pipe { + dst, + pipe: NamedPipe::from_raw_handle(p.into_raw_handle()), + overlapped: Overlapped::zero(), + done: false, + } + } + + unsafe fn read(&mut self) -> io::Result<()> { + let dst = slice_to_end(self.dst); + match self.pipe.read_overlapped(dst, self.overlapped.raw()) { + Ok(_) => Ok(()), + Err(e) => { + if e.raw_os_error() == Some(ERROR_BROKEN_PIPE as i32) { + self.done = true; + Ok(()) + } else { + Err(e) + } + } + } + } + + unsafe fn complete(&mut self, status: &CompletionStatus) { + let prev = self.dst.len(); + self.dst.set_len(prev + status.bytes_transferred() as usize); + if status.bytes_transferred() == 0 { + self.done = true; + } + } + } + + unsafe fn slice_to_end(v: &mut Vec<u8>) -> &mut [u8] { + if v.capacity() == 0 { + v.reserve(16); + } + if v.capacity() == v.len() { + v.reserve(1); + } + slice::from_raw_parts_mut(v.as_mut_ptr().add(v.len()), v.capacity() - v.len()) + } +} diff --git a/crates/cargo-util/src/registry.rs b/crates/cargo-util/src/registry.rs new file mode 100644 index 0000000..6b1ccd2 --- /dev/null +++ b/crates/cargo-util/src/registry.rs @@ -0,0 +1,45 @@ +/// Make a path to a dependency, which aligns to +/// +/// - [index from of Cargo's index on filesystem][1], and +/// - [index from Crates.io][2]. +/// +/// [1]: https://docs.rs/cargo/latest/cargo/sources/registry/index.html#the-format-of-the-index +/// [2]: https://github.com/rust-lang/crates.io-index +pub fn make_dep_path(dep_name: &str, prefix_only: bool) -> String { + let (slash, name) = if prefix_only { + ("", "") + } else { + ("/", dep_name) + }; + match dep_name.len() { + 1 => format!("1{}{}", slash, name), + 2 => format!("2{}{}", slash, name), + 3 => format!("3/{}{}{}", &dep_name[..1], slash, name), + _ => format!("{}/{}{}{}", &dep_name[0..2], &dep_name[2..4], slash, name), + } +} + +#[cfg(test)] +mod tests { + use super::make_dep_path; + + #[test] + fn prefix_only() { + assert_eq!(make_dep_path("a", true), "1"); + assert_eq!(make_dep_path("ab", true), "2"); + assert_eq!(make_dep_path("abc", true), "3/a"); + assert_eq!(make_dep_path("Abc", true), "3/A"); + assert_eq!(make_dep_path("AbCd", true), "Ab/Cd"); + assert_eq!(make_dep_path("aBcDe", true), "aB/cD"); + } + + #[test] + fn full() { + assert_eq!(make_dep_path("a", false), "1/a"); + assert_eq!(make_dep_path("ab", false), "2/ab"); + assert_eq!(make_dep_path("abc", false), "3/a/abc"); + assert_eq!(make_dep_path("Abc", false), "3/A/Abc"); + assert_eq!(make_dep_path("AbCd", false), "Ab/Cd/AbCd"); + assert_eq!(make_dep_path("aBcDe", false), "aB/cD/aBcDe"); + } +} diff --git a/crates/cargo-util/src/sha256.rs b/crates/cargo-util/src/sha256.rs new file mode 100644 index 0000000..58821f4 --- /dev/null +++ b/crates/cargo-util/src/sha256.rs @@ -0,0 +1,56 @@ +use super::paths; +use anyhow::{Context, Result}; +use crypto_hash::{Algorithm, Hasher}; +use std::fs::File; +use std::io::{self, Read, Write}; +use std::path::Path; + +pub struct Sha256(Hasher); + +impl Sha256 { + pub fn new() -> Sha256 { + let hasher = Hasher::new(Algorithm::SHA256); + Sha256(hasher) + } + + pub fn update(&mut self, bytes: &[u8]) -> &mut Sha256 { + let _ = self.0.write_all(bytes); + self + } + + pub fn update_file(&mut self, mut file: &File) -> io::Result<&mut Sha256> { + let mut buf = [0; 64 * 1024]; + loop { + let n = file.read(&mut buf)?; + if n == 0 { + break Ok(self); + } + self.update(&buf[..n]); + } + } + + pub fn update_path<P: AsRef<Path>>(&mut self, path: P) -> Result<&mut Sha256> { + let path = path.as_ref(); + let file = paths::open(path)?; + self.update_file(&file) + .with_context(|| format!("failed to read `{}`", path.display()))?; + Ok(self) + } + + pub fn finish(&mut self) -> [u8; 32] { + let mut ret = [0u8; 32]; + let data = self.0.finish(); + ret.copy_from_slice(&data[..]); + ret + } + + pub fn finish_hex(&mut self) -> String { + hex::encode(self.finish()) + } +} + +impl Default for Sha256 { + fn default() -> Self { + Self::new() + } +} diff --git a/crates/crates-io/Cargo.toml b/crates/crates-io/Cargo.toml new file mode 100644 index 0000000..004e2da --- /dev/null +++ b/crates/crates-io/Cargo.toml @@ -0,0 +1,21 @@ +[package] +name = "crates-io" +version = "0.36.0" +edition = "2021" +license = "MIT OR Apache-2.0" +repository = "https://github.com/rust-lang/cargo" +description = """ +Helpers for interacting with crates.io +""" + +[lib] +name = "crates_io" +path = "lib.rs" + +[dependencies] +anyhow = "1.0.34" +curl = "0.4" +percent-encoding = "2.0" +serde = { version = "1.0", features = ['derive'] } +serde_json = "1.0" +url = "2.0" diff --git a/crates/crates-io/LICENSE-APACHE b/crates/crates-io/LICENSE-APACHE new file mode 120000 index 0000000..1cd601d --- /dev/null +++ b/crates/crates-io/LICENSE-APACHE @@ -0,0 +1 @@ +../../LICENSE-APACHE
\ No newline at end of file diff --git a/crates/crates-io/LICENSE-MIT b/crates/crates-io/LICENSE-MIT new file mode 120000 index 0000000..b2cfbdc --- /dev/null +++ b/crates/crates-io/LICENSE-MIT @@ -0,0 +1 @@ +../../LICENSE-MIT
\ No newline at end of file diff --git a/crates/crates-io/lib.rs b/crates/crates-io/lib.rs new file mode 100644 index 0000000..ad3ea76 --- /dev/null +++ b/crates/crates-io/lib.rs @@ -0,0 +1,537 @@ +#![allow(clippy::all)] + +use std::collections::BTreeMap; +use std::fmt; +use std::fs::File; +use std::io::prelude::*; +use std::io::{Cursor, SeekFrom}; +use std::time::Instant; + +use anyhow::{bail, format_err, Context, Result}; +use curl::easy::{Easy, List}; +use percent_encoding::{percent_encode, NON_ALPHANUMERIC}; +use serde::{Deserialize, Serialize}; +use url::Url; + +pub struct Registry { + /// The base URL for issuing API requests. + host: String, + /// Optional authorization token. + /// If None, commands requiring authorization will fail. + token: Option<String>, + /// Curl handle for issuing requests. + handle: Easy, + /// Whether to include the authorization token with all requests. + auth_required: bool, +} + +#[derive(PartialEq, Clone, Copy)] +pub enum Auth { + Authorized, + Unauthorized, +} + +#[derive(Deserialize)] +pub struct Crate { + pub name: String, + pub description: Option<String>, + pub max_version: String, +} + +#[derive(Serialize, Deserialize)] +pub struct NewCrate { + pub name: String, + pub vers: String, + pub deps: Vec<NewCrateDependency>, + pub features: BTreeMap<String, Vec<String>>, + pub authors: Vec<String>, + pub description: Option<String>, + pub documentation: Option<String>, + pub homepage: Option<String>, + pub readme: Option<String>, + pub readme_file: Option<String>, + pub keywords: Vec<String>, + pub categories: Vec<String>, + pub license: Option<String>, + pub license_file: Option<String>, + pub repository: Option<String>, + pub badges: BTreeMap<String, BTreeMap<String, String>>, + pub links: Option<String>, +} + +#[derive(Serialize, Deserialize)] +pub struct NewCrateDependency { + pub optional: bool, + pub default_features: bool, + pub name: String, + pub features: Vec<String>, + pub version_req: String, + pub target: Option<String>, + pub kind: String, + #[serde(skip_serializing_if = "Option::is_none")] + pub registry: Option<String>, + #[serde(skip_serializing_if = "Option::is_none")] + pub explicit_name_in_toml: Option<String>, +} + +#[derive(Deserialize)] +pub struct User { + pub id: u32, + pub login: String, + pub avatar: Option<String>, + pub email: Option<String>, + pub name: Option<String>, +} + +pub struct Warnings { + pub invalid_categories: Vec<String>, + pub invalid_badges: Vec<String>, + pub other: Vec<String>, +} + +#[derive(Deserialize)] +struct R { + ok: bool, +} +#[derive(Deserialize)] +struct OwnerResponse { + ok: bool, + msg: String, +} +#[derive(Deserialize)] +struct ApiErrorList { + errors: Vec<ApiError>, +} +#[derive(Deserialize)] +struct ApiError { + detail: String, +} +#[derive(Serialize)] +struct OwnersReq<'a> { + users: &'a [&'a str], +} +#[derive(Deserialize)] +struct Users { + users: Vec<User>, +} +#[derive(Deserialize)] +struct TotalCrates { + total: u32, +} +#[derive(Deserialize)] +struct Crates { + crates: Vec<Crate>, + meta: TotalCrates, +} + +#[derive(Debug)] +pub enum ResponseError { + Curl(curl::Error), + Api { + code: u32, + errors: Vec<String>, + }, + Code { + code: u32, + headers: Vec<String>, + body: String, + }, + Other(anyhow::Error), +} + +impl std::error::Error for ResponseError { + fn source(&self) -> Option<&(dyn std::error::Error + 'static)> { + match self { + ResponseError::Curl(..) => None, + ResponseError::Api { .. } => None, + ResponseError::Code { .. } => None, + ResponseError::Other(e) => Some(e.as_ref()), + } + } +} + +impl fmt::Display for ResponseError { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + match self { + ResponseError::Curl(e) => write!(f, "{}", e), + ResponseError::Api { code, errors } => { + f.write_str("the remote server responded with an error")?; + if *code != 200 { + write!(f, " (status {} {})", code, reason(*code))?; + }; + write!(f, ": {}", errors.join(", ")) + } + ResponseError::Code { + code, + headers, + body, + } => write!( + f, + "failed to get a 200 OK response, got {}\n\ + headers:\n\ + \t{}\n\ + body:\n\ + {}", + code, + headers.join("\n\t"), + body + ), + ResponseError::Other(..) => write!(f, "invalid response from server"), + } + } +} + +impl From<curl::Error> for ResponseError { + fn from(error: curl::Error) -> Self { + ResponseError::Curl(error) + } +} + +impl Registry { + /// Creates a new `Registry`. + /// + /// ## Example + /// + /// ```rust + /// use curl::easy::Easy; + /// use crates_io::Registry; + /// + /// let mut handle = Easy::new(); + /// // If connecting to crates.io, a user-agent is required. + /// handle.useragent("my_crawler (example.com/info)"); + /// let mut reg = Registry::new_handle(String::from("https://crates.io"), None, handle); + /// ``` + pub fn new_handle( + host: String, + token: Option<String>, + handle: Easy, + auth_required: bool, + ) -> Registry { + Registry { + host, + token, + handle, + auth_required, + } + } + + pub fn set_token(&mut self, token: Option<String>) { + self.token = token; + } + + pub fn host(&self) -> &str { + &self.host + } + + pub fn host_is_crates_io(&self) -> bool { + is_url_crates_io(&self.host) + } + + pub fn add_owners(&mut self, krate: &str, owners: &[&str]) -> Result<String> { + let body = serde_json::to_string(&OwnersReq { users: owners })?; + let body = self.put(&format!("/crates/{}/owners", krate), body.as_bytes())?; + assert!(serde_json::from_str::<OwnerResponse>(&body)?.ok); + Ok(serde_json::from_str::<OwnerResponse>(&body)?.msg) + } + + pub fn remove_owners(&mut self, krate: &str, owners: &[&str]) -> Result<()> { + let body = serde_json::to_string(&OwnersReq { users: owners })?; + let body = self.delete(&format!("/crates/{}/owners", krate), Some(body.as_bytes()))?; + assert!(serde_json::from_str::<OwnerResponse>(&body)?.ok); + Ok(()) + } + + pub fn list_owners(&mut self, krate: &str) -> Result<Vec<User>> { + let body = self.get(&format!("/crates/{}/owners", krate))?; + Ok(serde_json::from_str::<Users>(&body)?.users) + } + + pub fn publish(&mut self, krate: &NewCrate, mut tarball: &File) -> Result<Warnings> { + let json = serde_json::to_string(krate)?; + // Prepare the body. The format of the upload request is: + // + // <le u32 of json> + // <json request> (metadata for the package) + // <le u32 of tarball> + // <source tarball> + + // NOTE: This can be replaced with `stream_len` if it is ever stabilized. + // + // This checks the length using seeking instead of metadata, because + // on some filesystems, getting the metadata will fail because + // the file was renamed in ops::package. + let tarball_len = tarball + .seek(SeekFrom::End(0)) + .with_context(|| "failed to seek tarball")?; + tarball + .seek(SeekFrom::Start(0)) + .with_context(|| "failed to seek tarball")?; + let header = { + let mut w = Vec::new(); + w.extend(&(json.len() as u32).to_le_bytes()); + w.extend(json.as_bytes().iter().cloned()); + w.extend(&(tarball_len as u32).to_le_bytes()); + w + }; + let size = tarball_len as usize + header.len(); + let mut body = Cursor::new(header).chain(tarball); + + let url = format!("{}/api/v1/crates/new", self.host); + + let token = match self.token.as_ref() { + Some(s) => s, + None => bail!("no upload token found, please run `cargo login`"), + }; + self.handle.put(true)?; + self.handle.url(&url)?; + self.handle.in_filesize(size as u64)?; + let mut headers = List::new(); + headers.append("Accept: application/json")?; + headers.append(&format!("Authorization: {}", token))?; + self.handle.http_headers(headers)?; + + let started = Instant::now(); + let body = self + .handle(&mut |buf| body.read(buf).unwrap_or(0)) + .map_err(|e| match e { + ResponseError::Code { code, .. } + if code == 503 + && started.elapsed().as_secs() >= 29 + && self.host_is_crates_io() => + { + format_err!( + "Request timed out after 30 seconds. If you're trying to \ + upload a crate it may be too large. If the crate is under \ + 10MB in size, you can email help@crates.io for assistance.\n\ + Total size was {}.", + tarball_len + ) + } + _ => e.into(), + })?; + + let response = if body.is_empty() { + "{}".parse()? + } else { + body.parse::<serde_json::Value>()? + }; + + let invalid_categories: Vec<String> = response + .get("warnings") + .and_then(|j| j.get("invalid_categories")) + .and_then(|j| j.as_array()) + .map(|x| x.iter().flat_map(|j| j.as_str()).map(Into::into).collect()) + .unwrap_or_else(Vec::new); + + let invalid_badges: Vec<String> = response + .get("warnings") + .and_then(|j| j.get("invalid_badges")) + .and_then(|j| j.as_array()) + .map(|x| x.iter().flat_map(|j| j.as_str()).map(Into::into).collect()) + .unwrap_or_else(Vec::new); + + let other: Vec<String> = response + .get("warnings") + .and_then(|j| j.get("other")) + .and_then(|j| j.as_array()) + .map(|x| x.iter().flat_map(|j| j.as_str()).map(Into::into).collect()) + .unwrap_or_else(Vec::new); + + Ok(Warnings { + invalid_categories, + invalid_badges, + other, + }) + } + + pub fn search(&mut self, query: &str, limit: u32) -> Result<(Vec<Crate>, u32)> { + let formatted_query = percent_encode(query.as_bytes(), NON_ALPHANUMERIC); + let body = self.req( + &format!("/crates?q={}&per_page={}", formatted_query, limit), + None, + Auth::Unauthorized, + )?; + + let crates = serde_json::from_str::<Crates>(&body)?; + Ok((crates.crates, crates.meta.total)) + } + + pub fn yank(&mut self, krate: &str, version: &str) -> Result<()> { + let body = self.delete(&format!("/crates/{}/{}/yank", krate, version), None)?; + assert!(serde_json::from_str::<R>(&body)?.ok); + Ok(()) + } + + pub fn unyank(&mut self, krate: &str, version: &str) -> Result<()> { + let body = self.put(&format!("/crates/{}/{}/unyank", krate, version), &[])?; + assert!(serde_json::from_str::<R>(&body)?.ok); + Ok(()) + } + + fn put(&mut self, path: &str, b: &[u8]) -> Result<String> { + self.handle.put(true)?; + self.req(path, Some(b), Auth::Authorized) + } + + fn get(&mut self, path: &str) -> Result<String> { + self.handle.get(true)?; + self.req(path, None, Auth::Authorized) + } + + fn delete(&mut self, path: &str, b: Option<&[u8]>) -> Result<String> { + self.handle.custom_request("DELETE")?; + self.req(path, b, Auth::Authorized) + } + + fn req(&mut self, path: &str, body: Option<&[u8]>, authorized: Auth) -> Result<String> { + self.handle.url(&format!("{}/api/v1{}", self.host, path))?; + let mut headers = List::new(); + headers.append("Accept: application/json")?; + headers.append("Content-Type: application/json")?; + + if self.auth_required || authorized == Auth::Authorized { + let token = match self.token.as_ref() { + Some(s) => s, + None => bail!("no upload token found, please run `cargo login`"), + }; + check_token(token)?; + headers.append(&format!("Authorization: {}", token))?; + } + self.handle.http_headers(headers)?; + match body { + Some(mut body) => { + self.handle.upload(true)?; + self.handle.in_filesize(body.len() as u64)?; + self.handle(&mut |buf| body.read(buf).unwrap_or(0)) + .map_err(|e| e.into()) + } + None => self.handle(&mut |_| 0).map_err(|e| e.into()), + } + } + + fn handle( + &mut self, + read: &mut dyn FnMut(&mut [u8]) -> usize, + ) -> std::result::Result<String, ResponseError> { + let mut headers = Vec::new(); + let mut body = Vec::new(); + { + let mut handle = self.handle.transfer(); + handle.read_function(|buf| Ok(read(buf)))?; + handle.write_function(|data| { + body.extend_from_slice(data); + Ok(data.len()) + })?; + handle.header_function(|data| { + // Headers contain trailing \r\n, trim them to make it easier + // to work with. + let s = String::from_utf8_lossy(data).trim().to_string(); + headers.push(s); + true + })?; + handle.perform()?; + } + + let body = match String::from_utf8(body) { + Ok(body) => body, + Err(..) => { + return Err(ResponseError::Other(format_err!( + "response body was not valid utf-8" + ))) + } + }; + let errors = serde_json::from_str::<ApiErrorList>(&body) + .ok() + .map(|s| s.errors.into_iter().map(|s| s.detail).collect::<Vec<_>>()); + + match (self.handle.response_code()?, errors) { + (0, None) | (200, None) => Ok(body), + (code, Some(errors)) => Err(ResponseError::Api { code, errors }), + (code, None) => Err(ResponseError::Code { + code, + headers, + body, + }), + } + } +} + +fn reason(code: u32) -> &'static str { + // Taken from https://developer.mozilla.org/en-US/docs/Web/HTTP/Status + match code { + 100 => "Continue", + 101 => "Switching Protocol", + 103 => "Early Hints", + 200 => "OK", + 201 => "Created", + 202 => "Accepted", + 203 => "Non-Authoritative Information", + 204 => "No Content", + 205 => "Reset Content", + 206 => "Partial Content", + 300 => "Multiple Choice", + 301 => "Moved Permanently", + 302 => "Found", + 303 => "See Other", + 304 => "Not Modified", + 307 => "Temporary Redirect", + 308 => "Permanent Redirect", + 400 => "Bad Request", + 401 => "Unauthorized", + 402 => "Payment Required", + 403 => "Forbidden", + 404 => "Not Found", + 405 => "Method Not Allowed", + 406 => "Not Acceptable", + 407 => "Proxy Authentication Required", + 408 => "Request Timeout", + 409 => "Conflict", + 410 => "Gone", + 411 => "Length Required", + 412 => "Precondition Failed", + 413 => "Payload Too Large", + 414 => "URI Too Long", + 415 => "Unsupported Media Type", + 416 => "Request Range Not Satisfiable", + 417 => "Expectation Failed", + 429 => "Too Many Requests", + 431 => "Request Header Fields Too Large", + 500 => "Internal Server Error", + 501 => "Not Implemented", + 502 => "Bad Gateway", + 503 => "Service Unavailable", + 504 => "Gateway Timeout", + _ => "<unknown>", + } +} + +/// Returns `true` if the host of the given URL is "crates.io". +pub fn is_url_crates_io(url: &str) -> bool { + Url::parse(url) + .map(|u| u.host_str() == Some("crates.io")) + .unwrap_or(false) +} + +/// Checks if a token is valid or malformed. +/// +/// This check is necessary to prevent sending tokens which create an invalid HTTP request. +/// It would be easier to check just for alphanumeric tokens, but we can't be sure that all +/// registries only create tokens in that format so that is as less restricted as possible. +pub fn check_token(token: &str) -> Result<()> { + if token.is_empty() { + bail!("please provide a non-empty token"); + } + if token.bytes().all(|b| { + b >= 32 // undefined in ISO-8859-1, in ASCII/ UTF-8 not-printable character + && b < 128 // utf-8: the first bit signals a multi-byte character + && b != 127 // 127 is a control character in ascii and not in ISO 8859-1 + || b == b't' // tab is also allowed (even when < 32) + }) { + Ok(()) + } else { + Err(anyhow::anyhow!( + "token contains invalid characters.\nOnly printable ISO-8859-1 characters \ + are allowed as it is sent in a HTTPS header." + )) + } +} diff --git a/crates/credential/README.md b/crates/credential/README.md new file mode 100644 index 0000000..168cc71 --- /dev/null +++ b/crates/credential/README.md @@ -0,0 +1,8 @@ +# Cargo Credential Packages + +This directory contains Cargo packages for handling storage of tokens in a +secure manner. + +`cargo-credential` is a generic library to assist writing a credential +process. The other directories contain implementations that integrate with +specific credential systems. diff --git a/crates/credential/cargo-credential-1password/Cargo.toml b/crates/credential/cargo-credential-1password/Cargo.toml new file mode 100644 index 0000000..093fde8 --- /dev/null +++ b/crates/credential/cargo-credential-1password/Cargo.toml @@ -0,0 +1,12 @@ +[package] +name = "cargo-credential-1password" +version = "0.2.0" +edition = "2021" +license = "MIT OR Apache-2.0" +repository = "https://github.com/rust-lang/cargo" +description = "A Cargo credential process that stores tokens in a 1password vault." + +[dependencies] +cargo-credential = { version = "0.2.0", path = "../cargo-credential" } +serde = { version = "1.0.117", features = ["derive"] } +serde_json = "1.0.59" diff --git a/crates/credential/cargo-credential-1password/src/main.rs b/crates/credential/cargo-credential-1password/src/main.rs new file mode 100644 index 0000000..4f512b7 --- /dev/null +++ b/crates/credential/cargo-credential-1password/src/main.rs @@ -0,0 +1,314 @@ +//! Cargo registry 1password credential process. + +use cargo_credential::{Credential, Error}; +use serde::Deserialize; +use std::io::Read; +use std::process::{Command, Stdio}; + +const CARGO_TAG: &str = "cargo-registry"; + +/// Implementation of 1password keychain access for Cargo registries. +struct OnePasswordKeychain { + account: Option<String>, + vault: Option<String>, +} + +/// 1password Login item type, used for the JSON output of `op item get`. +#[derive(Deserialize)] +struct Login { + fields: Vec<Field>, +} + +#[derive(Deserialize)] +struct Field { + id: String, + value: Option<String>, +} + +/// 1password item from `op items list`. +#[derive(Deserialize)] +struct ListItem { + id: String, + urls: Vec<Url>, +} + +#[derive(Deserialize)] +struct Url { + href: String, +} + +impl OnePasswordKeychain { + fn new() -> Result<OnePasswordKeychain, Error> { + let mut args = std::env::args().skip(1); + let mut action = false; + let mut account = None; + let mut vault = None; + while let Some(arg) = args.next() { + match arg.as_str() { + "--account" => { + account = Some(args.next().ok_or("--account needs an arg")?); + } + "--vault" => { + vault = Some(args.next().ok_or("--vault needs an arg")?); + } + s if s.starts_with('-') => { + return Err(format!("unknown option {}", s).into()); + } + _ => { + if action { + return Err("too many arguments".into()); + } else { + action = true; + } + } + } + } + Ok(OnePasswordKeychain { account, vault }) + } + + fn signin(&self) -> Result<Option<String>, Error> { + // If there are any session env vars, we'll assume that this is the + // correct account, and that the user knows what they are doing. + if std::env::vars().any(|(name, _)| name.starts_with("OP_SESSION_")) { + return Ok(None); + } + let mut cmd = Command::new("op"); + cmd.args(&["signin", "--raw"]); + cmd.stdout(Stdio::piped()); + self.with_tty(&mut cmd)?; + let mut child = cmd + .spawn() + .map_err(|e| format!("failed to spawn `op`: {}", e))?; + let mut buffer = String::new(); + child + .stdout + .as_mut() + .unwrap() + .read_to_string(&mut buffer) + .map_err(|e| format!("failed to get session from `op`: {}", e))?; + if let Some(end) = buffer.find('\n') { + buffer.truncate(end); + } + let status = child + .wait() + .map_err(|e| format!("failed to wait for `op`: {}", e))?; + if !status.success() { + return Err(format!("failed to run `op signin`: {}", status).into()); + } + if buffer.is_empty() { + // When using CLI integration, `op signin` returns no output, + // so there is no need to set the session. + return Ok(None); + } + Ok(Some(buffer)) + } + + fn make_cmd(&self, session: &Option<String>, args: &[&str]) -> Command { + let mut cmd = Command::new("op"); + cmd.args(args); + if let Some(account) = &self.account { + cmd.arg("--account"); + cmd.arg(account); + } + if let Some(vault) = &self.vault { + cmd.arg("--vault"); + cmd.arg(vault); + } + if let Some(session) = session { + cmd.arg("--session"); + cmd.arg(session); + } + cmd + } + + fn with_tty(&self, cmd: &mut Command) -> Result<(), Error> { + #[cfg(unix)] + const IN_DEVICE: &str = "/dev/tty"; + #[cfg(windows)] + const IN_DEVICE: &str = "CONIN$"; + let stdin = std::fs::OpenOptions::new() + .read(true) + .write(true) + .open(IN_DEVICE)?; + cmd.stdin(stdin); + Ok(()) + } + + fn run_cmd(&self, mut cmd: Command) -> Result<String, Error> { + cmd.stdout(Stdio::piped()); + let mut child = cmd + .spawn() + .map_err(|e| format!("failed to spawn `op`: {}", e))?; + let mut buffer = String::new(); + child + .stdout + .as_mut() + .unwrap() + .read_to_string(&mut buffer) + .map_err(|e| format!("failed to read `op` output: {}", e))?; + let status = child + .wait() + .map_err(|e| format!("failed to wait for `op`: {}", e))?; + if !status.success() { + return Err(format!("`op` command exit error: {}", status).into()); + } + Ok(buffer) + } + + fn search(&self, session: &Option<String>, index_url: &str) -> Result<Option<String>, Error> { + let cmd = self.make_cmd( + session, + &[ + "items", + "list", + "--categories", + "Login", + "--tags", + CARGO_TAG, + "--format", + "json", + ], + ); + let buffer = self.run_cmd(cmd)?; + let items: Vec<ListItem> = serde_json::from_str(&buffer) + .map_err(|e| format!("failed to deserialize JSON from 1password list: {}", e))?; + let mut matches = items + .into_iter() + .filter(|item| item.urls.iter().any(|url| url.href == index_url)); + match matches.next() { + Some(login) => { + // Should this maybe just sort on `updatedAt` and return the newest one? + if matches.next().is_some() { + return Err(format!( + "too many 1password logins match registry `{}`, \ + consider deleting the excess entries", + index_url + ) + .into()); + } + Ok(Some(login.id)) + } + None => Ok(None), + } + } + + fn modify( + &self, + session: &Option<String>, + id: &str, + token: &str, + _name: Option<&str>, + ) -> Result<(), Error> { + let cmd = self.make_cmd( + session, + &["item", "edit", id, &format!("password={}", token)], + ); + self.run_cmd(cmd)?; + Ok(()) + } + + fn create( + &self, + session: &Option<String>, + index_url: &str, + token: &str, + name: Option<&str>, + ) -> Result<(), Error> { + let title = match name { + Some(name) => format!("Cargo registry token for {}", name), + None => "Cargo registry token".to_string(), + }; + let mut cmd = self.make_cmd( + session, + &[ + "item", + "create", + "--category", + "Login", + &format!("password={}", token), + &format!("url={}", index_url), + "--title", + &title, + "--tags", + CARGO_TAG, + ], + ); + // For unknown reasons, `op item create` seems to not be happy if + // stdin is not a tty. Otherwise it returns with a 0 exit code without + // doing anything. + self.with_tty(&mut cmd)?; + self.run_cmd(cmd)?; + Ok(()) + } + + fn get_token(&self, session: &Option<String>, id: &str) -> Result<String, Error> { + let cmd = self.make_cmd(session, &["item", "get", "--format=json", id]); + let buffer = self.run_cmd(cmd)?; + let item: Login = serde_json::from_str(&buffer) + .map_err(|e| format!("failed to deserialize JSON from 1password get: {}", e))?; + let password = item.fields.into_iter().find(|item| item.id == "password"); + match password { + Some(password) => password + .value + .ok_or_else(|| format!("missing password value for entry").into()), + None => Err("could not find password field".into()), + } + } + + fn delete(&self, session: &Option<String>, id: &str) -> Result<(), Error> { + let cmd = self.make_cmd(session, &["item", "delete", id]); + self.run_cmd(cmd)?; + Ok(()) + } +} + +impl Credential for OnePasswordKeychain { + fn name(&self) -> &'static str { + env!("CARGO_PKG_NAME") + } + + fn get(&self, index_url: &str) -> Result<String, Error> { + let session = self.signin()?; + if let Some(id) = self.search(&session, index_url)? { + self.get_token(&session, &id) + } else { + return Err(format!( + "no 1password entry found for registry `{}`, try `cargo login` to add a token", + index_url + ) + .into()); + } + } + + fn store(&self, index_url: &str, token: &str, name: Option<&str>) -> Result<(), Error> { + let session = self.signin()?; + // Check if an item already exists. + if let Some(id) = self.search(&session, index_url)? { + self.modify(&session, &id, token, name) + } else { + self.create(&session, index_url, token, name) + } + } + + fn erase(&self, index_url: &str) -> Result<(), Error> { + let session = self.signin()?; + // Check if an item already exists. + if let Some(id) = self.search(&session, index_url)? { + self.delete(&session, &id)?; + } else { + eprintln!("not currently logged in to `{}`", index_url); + } + Ok(()) + } +} + +fn main() { + let op = match OnePasswordKeychain::new() { + Ok(op) => op, + Err(e) => { + eprintln!("error: {}", e); + std::process::exit(1); + } + }; + cargo_credential::main(op); +} diff --git a/crates/credential/cargo-credential-gnome-secret/Cargo.toml b/crates/credential/cargo-credential-gnome-secret/Cargo.toml new file mode 100644 index 0000000..12e25cf --- /dev/null +++ b/crates/credential/cargo-credential-gnome-secret/Cargo.toml @@ -0,0 +1,13 @@ +[package] +name = "cargo-credential-gnome-secret" +version = "0.2.0" +edition = "2021" +license = "MIT OR Apache-2.0" +repository = "https://github.com/rust-lang/cargo" +description = "A Cargo credential process that stores tokens with GNOME libsecret." + +[dependencies] +cargo-credential = { version = "0.2.0", path = "../cargo-credential" } + +[build-dependencies] +pkg-config = "0.3.19" diff --git a/crates/credential/cargo-credential-gnome-secret/build.rs b/crates/credential/cargo-credential-gnome-secret/build.rs new file mode 100644 index 0000000..9283535 --- /dev/null +++ b/crates/credential/cargo-credential-gnome-secret/build.rs @@ -0,0 +1,3 @@ +fn main() { + pkg_config::probe_library("libsecret-1").unwrap(); +} diff --git a/crates/credential/cargo-credential-gnome-secret/src/main.rs b/crates/credential/cargo-credential-gnome-secret/src/main.rs new file mode 100644 index 0000000..40972b0 --- /dev/null +++ b/crates/credential/cargo-credential-gnome-secret/src/main.rs @@ -0,0 +1,194 @@ +//! Cargo registry gnome libsecret credential process. + +use cargo_credential::{Credential, Error}; +use std::ffi::{CStr, CString}; +use std::os::raw::{c_char, c_int}; +use std::ptr::{null, null_mut}; + +#[allow(non_camel_case_types)] +type gchar = c_char; + +#[allow(non_camel_case_types)] +type gboolean = c_int; + +type GQuark = u32; + +#[repr(C)] +struct GError { + domain: GQuark, + code: c_int, + message: *mut gchar, +} + +#[repr(C)] +struct GCancellable { + _private: [u8; 0], +} + +#[repr(C)] +struct SecretSchema { + name: *const gchar, + flags: SecretSchemaFlags, + attributes: [SecretSchemaAttribute; 32], +} + +#[repr(C)] +#[derive(Copy, Clone)] +struct SecretSchemaAttribute { + name: *const gchar, + attr_type: SecretSchemaAttributeType, +} + +#[repr(C)] +enum SecretSchemaFlags { + None = 0, +} + +#[repr(C)] +#[derive(Copy, Clone)] +enum SecretSchemaAttributeType { + String = 0, +} + +extern "C" { + fn secret_password_store_sync( + schema: *const SecretSchema, + collection: *const gchar, + label: *const gchar, + password: *const gchar, + cancellable: *mut GCancellable, + error: *mut *mut GError, + ... + ) -> gboolean; + fn secret_password_clear_sync( + schema: *const SecretSchema, + cancellable: *mut GCancellable, + error: *mut *mut GError, + ... + ) -> gboolean; + fn secret_password_lookup_sync( + schema: *const SecretSchema, + cancellable: *mut GCancellable, + error: *mut *mut GError, + ... + ) -> *mut gchar; +} + +struct GnomeSecret; + +fn label(index_url: &str) -> CString { + CString::new(format!("cargo-registry:{}", index_url)).unwrap() +} + +fn schema() -> SecretSchema { + let mut attributes = [SecretSchemaAttribute { + name: null(), + attr_type: SecretSchemaAttributeType::String, + }; 32]; + attributes[0] = SecretSchemaAttribute { + name: b"url\0".as_ptr() as *const gchar, + attr_type: SecretSchemaAttributeType::String, + }; + SecretSchema { + name: b"org.rust-lang.cargo.registry\0".as_ptr() as *const gchar, + flags: SecretSchemaFlags::None, + attributes, + } +} + +impl Credential for GnomeSecret { + fn name(&self) -> &'static str { + env!("CARGO_PKG_NAME") + } + + fn get(&self, index_url: &str) -> Result<String, Error> { + let mut error: *mut GError = null_mut(); + let attr_url = CString::new("url").unwrap(); + let index_url_c = CString::new(index_url).unwrap(); + let schema = schema(); + unsafe { + let token_c = secret_password_lookup_sync( + &schema, + null_mut(), + &mut error, + attr_url.as_ptr(), + index_url_c.as_ptr(), + null() as *const gchar, + ); + if !error.is_null() { + return Err(format!( + "failed to get token: {}", + CStr::from_ptr((*error).message).to_str()? + ) + .into()); + } + if token_c.is_null() { + return Err(format!("cannot find token for {}", index_url).into()); + } + let token = CStr::from_ptr(token_c) + .to_str() + .map_err(|e| format!("expected utf8 token: {}", e))? + .to_string(); + Ok(token) + } + } + + fn store(&self, index_url: &str, token: &str, name: Option<&str>) -> Result<(), Error> { + let label = label(name.unwrap_or(index_url)); + let token = CString::new(token).unwrap(); + let mut error: *mut GError = null_mut(); + let attr_url = CString::new("url").unwrap(); + let index_url_c = CString::new(index_url).unwrap(); + let schema = schema(); + unsafe { + secret_password_store_sync( + &schema, + b"default\0".as_ptr() as *const gchar, + label.as_ptr(), + token.as_ptr(), + null_mut(), + &mut error, + attr_url.as_ptr(), + index_url_c.as_ptr(), + null() as *const gchar, + ); + if !error.is_null() { + return Err(format!( + "failed to store token: {}", + CStr::from_ptr((*error).message).to_str()? + ) + .into()); + } + } + Ok(()) + } + + fn erase(&self, index_url: &str) -> Result<(), Error> { + let schema = schema(); + let mut error: *mut GError = null_mut(); + let attr_url = CString::new("url").unwrap(); + let index_url_c = CString::new(index_url).unwrap(); + unsafe { + secret_password_clear_sync( + &schema, + null_mut(), + &mut error, + attr_url.as_ptr(), + index_url_c.as_ptr(), + null() as *const gchar, + ); + if !error.is_null() { + return Err(format!( + "failed to erase token: {}", + CStr::from_ptr((*error).message).to_str()? + ) + .into()); + } + } + Ok(()) + } +} + +fn main() { + cargo_credential::main(GnomeSecret); +} diff --git a/crates/credential/cargo-credential-macos-keychain/Cargo.toml b/crates/credential/cargo-credential-macos-keychain/Cargo.toml new file mode 100644 index 0000000..c2c22a4 --- /dev/null +++ b/crates/credential/cargo-credential-macos-keychain/Cargo.toml @@ -0,0 +1,11 @@ +[package] +name = "cargo-credential-macos-keychain" +version = "0.2.0" +edition = "2021" +license = "MIT OR Apache-2.0" +repository = "https://github.com/rust-lang/cargo" +description = "A Cargo credential process that stores tokens in a macOS keychain." + +[dependencies] +cargo-credential = { version = "0.2.0", path = "../cargo-credential" } +security-framework = "2.0.0" diff --git a/crates/credential/cargo-credential-macos-keychain/src/main.rs b/crates/credential/cargo-credential-macos-keychain/src/main.rs new file mode 100644 index 0000000..3fef3f9 --- /dev/null +++ b/crates/credential/cargo-credential-macos-keychain/src/main.rs @@ -0,0 +1,50 @@ +//! Cargo registry macos keychain credential process. + +use cargo_credential::{Credential, Error}; +use security_framework::os::macos::keychain::SecKeychain; + +struct MacKeychain; + +/// The account name is not used. +const ACCOUNT: &'static str = ""; + +fn registry(registry_name: &str) -> String { + format!("cargo-registry:{}", registry_name) +} + +impl Credential for MacKeychain { + fn name(&self) -> &'static str { + env!("CARGO_PKG_NAME") + } + + fn get(&self, index_url: &str) -> Result<String, Error> { + let keychain = SecKeychain::default().unwrap(); + let service_name = registry(index_url); + let (pass, _item) = keychain.find_generic_password(&service_name, ACCOUNT)?; + String::from_utf8(pass.as_ref().to_vec()) + .map_err(|_| "failed to convert token to UTF8".into()) + } + + fn store(&self, index_url: &str, token: &str, name: Option<&str>) -> Result<(), Error> { + let keychain = SecKeychain::default().unwrap(); + let service_name = registry(name.unwrap_or(index_url)); + if let Ok((_pass, mut item)) = keychain.find_generic_password(&service_name, ACCOUNT) { + item.set_password(token.as_bytes())?; + } else { + keychain.add_generic_password(&service_name, ACCOUNT, token.as_bytes())?; + } + Ok(()) + } + + fn erase(&self, index_url: &str) -> Result<(), Error> { + let keychain = SecKeychain::default().unwrap(); + let service_name = registry(index_url); + let (_pass, item) = keychain.find_generic_password(&service_name, ACCOUNT)?; + item.delete(); + Ok(()) + } +} + +fn main() { + cargo_credential::main(MacKeychain); +} diff --git a/crates/credential/cargo-credential-wincred/Cargo.toml b/crates/credential/cargo-credential-wincred/Cargo.toml new file mode 100644 index 0000000..83c38e8 --- /dev/null +++ b/crates/credential/cargo-credential-wincred/Cargo.toml @@ -0,0 +1,11 @@ +[package] +name = "cargo-credential-wincred" +version = "0.2.0" +edition = "2021" +license = "MIT OR Apache-2.0" +repository = "https://github.com/rust-lang/cargo" +description = "A Cargo credential process that stores tokens with Windows Credential Manager." + +[dependencies] +cargo-credential = { version = "0.2.0", path = "../cargo-credential" } +windows-sys = { version = "0.45", features = ["Win32_Foundation", "Win32_Security_Credentials"] } diff --git a/crates/credential/cargo-credential-wincred/src/main.rs b/crates/credential/cargo-credential-wincred/src/main.rs new file mode 100644 index 0000000..8ae48f3 --- /dev/null +++ b/crates/credential/cargo-credential-wincred/src/main.rs @@ -0,0 +1,111 @@ +//! Cargo registry windows credential process. + +use cargo_credential::{Credential, Error}; +use std::ffi::OsStr; +use std::os::windows::ffi::OsStrExt; + +use windows_sys::core::PWSTR; +use windows_sys::Win32::Foundation::ERROR_NOT_FOUND; +use windows_sys::Win32::Foundation::FILETIME; +use windows_sys::Win32::Foundation::TRUE; +use windows_sys::Win32::Security::Credentials::CredDeleteW; +use windows_sys::Win32::Security::Credentials::CredReadW; +use windows_sys::Win32::Security::Credentials::CredWriteW; +use windows_sys::Win32::Security::Credentials::CREDENTIALW; +use windows_sys::Win32::Security::Credentials::CRED_PERSIST_LOCAL_MACHINE; +use windows_sys::Win32::Security::Credentials::CRED_TYPE_GENERIC; + +struct WindowsCredential; + +/// Converts a string to a nul-terminated wide UTF-16 byte sequence. +fn wstr(s: &str) -> Vec<u16> { + let mut wide: Vec<u16> = OsStr::new(s).encode_wide().collect(); + if wide.iter().any(|b| *b == 0) { + panic!("nul byte in wide string"); + } + wide.push(0); + wide +} + +fn target_name(registry_name: &str) -> Vec<u16> { + wstr(&format!("cargo-registry:{}", registry_name)) +} + +impl Credential for WindowsCredential { + fn name(&self) -> &'static str { + env!("CARGO_PKG_NAME") + } + + fn get(&self, index_url: &str) -> Result<String, Error> { + let target_name = target_name(index_url); + let p_credential: *mut CREDENTIALW = std::ptr::null_mut() as *mut _; + unsafe { + if CredReadW( + target_name.as_ptr(), + CRED_TYPE_GENERIC, + 0, + p_credential as *mut _ as *mut _, + ) != TRUE + { + return Err( + format!("failed to fetch token: {}", std::io::Error::last_os_error()).into(), + ); + } + let bytes = std::slice::from_raw_parts( + (*p_credential).CredentialBlob, + (*p_credential).CredentialBlobSize as usize, + ); + String::from_utf8(bytes.to_vec()).map_err(|_| "failed to convert token to UTF8".into()) + } + } + + fn store(&self, index_url: &str, token: &str, name: Option<&str>) -> Result<(), Error> { + let token = token.as_bytes(); + let target_name = target_name(index_url); + let comment = match name { + Some(name) => wstr(&format!("Cargo registry token for {}", name)), + None => wstr("Cargo registry token"), + }; + let mut credential = CREDENTIALW { + Flags: 0, + Type: CRED_TYPE_GENERIC, + TargetName: target_name.as_ptr() as PWSTR, + Comment: comment.as_ptr() as PWSTR, + LastWritten: FILETIME { + dwLowDateTime: 0, + dwHighDateTime: 0, + }, + CredentialBlobSize: token.len() as u32, + CredentialBlob: token.as_ptr() as *mut u8, + Persist: CRED_PERSIST_LOCAL_MACHINE, + AttributeCount: 0, + Attributes: std::ptr::null_mut(), + TargetAlias: std::ptr::null_mut(), + UserName: std::ptr::null_mut(), + }; + let result = unsafe { CredWriteW(&mut credential, 0) }; + if result != TRUE { + let err = std::io::Error::last_os_error(); + return Err(format!("failed to store token: {}", err).into()); + } + Ok(()) + } + + fn erase(&self, index_url: &str) -> Result<(), Error> { + let target_name = target_name(index_url); + let result = unsafe { CredDeleteW(target_name.as_ptr(), CRED_TYPE_GENERIC, 0) }; + if result != TRUE { + let err = std::io::Error::last_os_error(); + if err.raw_os_error() == Some(ERROR_NOT_FOUND as i32) { + eprintln!("not currently logged in to `{}`", index_url); + return Ok(()); + } + return Err(format!("failed to remove token: {}", err).into()); + } + Ok(()) + } +} + +fn main() { + cargo_credential::main(WindowsCredential); +} diff --git a/crates/credential/cargo-credential/Cargo.toml b/crates/credential/cargo-credential/Cargo.toml new file mode 100644 index 0000000..2addaf5 --- /dev/null +++ b/crates/credential/cargo-credential/Cargo.toml @@ -0,0 +1,9 @@ +[package] +name = "cargo-credential" +version = "0.2.0" +edition = "2021" +license = "MIT OR Apache-2.0" +repository = "https://github.com/rust-lang/cargo" +description = "A library to assist writing Cargo credential helpers." + +[dependencies] diff --git a/crates/credential/cargo-credential/README.md b/crates/credential/cargo-credential/README.md new file mode 100644 index 0000000..1f75e59 --- /dev/null +++ b/crates/credential/cargo-credential/README.md @@ -0,0 +1,41 @@ +# cargo-credential + +This package is a library to assist writing a Cargo credential helper, which +provides an interface to store tokens for authorizing access to a registry +such as https://crates.io/. + +Documentation about credential processes may be found at +https://doc.rust-lang.org/nightly/cargo/reference/unstable.html#credential-process + +Example implementations may be found at +https://github.com/rust-lang/cargo/tree/master/crates/credential + +## Usage + +Create a Cargo project with this as a dependency: + +```toml +# Add this to your Cargo.toml: + +[dependencies] +cargo-credential = "0.1" +``` + +And then include a `main.rs` binary which implements the `Credential` trait, and calls +the `main` function which will call the appropriate method of the trait: + +```rust +// src/main.rs + +use cargo_credential::{Credential, Error}; + +struct MyCredential; + +impl Credential for MyCredential { + /// implement trait methods here... +} + +fn main() { + cargo_credential::main(MyCredential); +} +``` diff --git a/crates/credential/cargo-credential/src/lib.rs b/crates/credential/cargo-credential/src/lib.rs new file mode 100644 index 0000000..3baf42d --- /dev/null +++ b/crates/credential/cargo-credential/src/lib.rs @@ -0,0 +1,86 @@ +//! Helper library for writing Cargo credential processes. +//! +//! A credential process should have a `struct` that implements the `Credential` trait. +//! The `main` function should be called with an instance of that struct, such as: +//! +//! ```rust,ignore +//! fn main() { +//! cargo_credential::main(MyCredential); +//! } +//! ``` +//! +//! This will determine the action to perform (get/store/erase) by looking at +//! the CLI arguments for the first argument that does not start with `-`. It +//! will then call the corresponding method of the trait to perform the +//! requested action. + +pub type Error = Box<dyn std::error::Error>; + +pub trait Credential { + /// Returns the name of this credential process. + fn name(&self) -> &'static str; + + /// Retrieves a token for the given registry. + fn get(&self, index_url: &str) -> Result<String, Error>; + + /// Stores the given token for the given registry. + fn store(&self, index_url: &str, token: &str, name: Option<&str>) -> Result<(), Error>; + + /// Removes the token for the given registry. + /// + /// If the user is not logged in, this should print a message to stderr if + /// possible indicating that the user is not currently logged in, and + /// return `Ok`. + fn erase(&self, index_url: &str) -> Result<(), Error>; +} + +/// Runs the credential interaction by processing the command-line and +/// environment variables. +pub fn main(credential: impl Credential) { + let name = credential.name(); + if let Err(e) = doit(credential) { + eprintln!("{} error: {}", name, e); + std::process::exit(1); + } +} + +fn env(name: &str) -> Result<String, Error> { + std::env::var(name).map_err(|_| format!("environment variable `{}` is not set", name).into()) +} + +fn doit(credential: impl Credential) -> Result<(), Error> { + let which = std::env::args() + .skip(1) + .skip_while(|arg| arg.starts_with('-')) + .next() + .ok_or_else(|| "first argument must be the {action}")?; + let index_url = env("CARGO_REGISTRY_INDEX_URL")?; + let name = std::env::var("CARGO_REGISTRY_NAME_OPT").ok(); + let result = match which.as_ref() { + "get" => credential.get(&index_url).and_then(|token| { + println!("{}", token); + Ok(()) + }), + "store" => { + read_token().and_then(|token| credential.store(&index_url, &token, name.as_deref())) + } + "erase" => credential.erase(&index_url), + _ => { + return Err(format!( + "unexpected command-line argument `{}`, expected get/store/erase", + which + ) + .into()) + } + }; + result.map_err(|e| format!("failed to `{}` token: {}", which, e).into()) +} + +fn read_token() -> Result<String, Error> { + let mut buffer = String::new(); + std::io::stdin().read_line(&mut buffer)?; + if buffer.ends_with('\n') { + buffer.pop(); + } + Ok(buffer) +} diff --git a/crates/home/CHANGELOG.md b/crates/home/CHANGELOG.md new file mode 100644 index 0000000..7674667 --- /dev/null +++ b/crates/home/CHANGELOG.md @@ -0,0 +1,46 @@ +# Changelog +All notable changes to this project will be documented in this file. + +The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), +and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). + +<!-- ## [Unreleased] --> + +## [0.5.4] - 2022-10-10 +- Add `_with_env` variants of functions to support in-process threaded tests for + rustup. + +## [0.5.3] - 2020-01-07 + +Use Rust 1.36.0 as minimum Rust version. + +## [0.5.2] - 2020-01-05 + +*YANKED since it cannot be built on Rust 1.36.0* + +### Changed +- Check for emptiness of `CARGO_HOME` and `RUSTUP_HOME` environment variables. +- Windows: Use `SHGetFolderPath` to replace `GetUserProfileDirectory` syscall. + * Remove `scopeguard` dependency. + +## [0.5.1] - 2019-10-12 +### Changed +- Disable unnecessary features for `scopeguard`. Thanks @mati865. + +## [0.5.0] - 2019-08-21 +### Added +- Add `home_dir` implementation for Windows UWP platforms. + +### Fixed +- Fix `rustup_home` implementation when `RUSTUP_HOME` is an absolute directory. +- Fix `cargo_home` implementation when `CARGO_HOME` is an absolute directory. + +### Removed +- Remove support for `multirust` folder used in old version of `rustup`. + +[Unreleased]: https://github.com/brson/home/compare/v0.5.4...HEAD +[0.5.4]: https://github.com/brson/home/compare/v0.5.3...v0.5.4 +[0.5.3]: https://github.com/brson/home/compare/v0.5.2...v0.5.3 +[0.5.2]: https://github.com/brson/home/compare/v0.5.1...v0.5.2 +[0.5.1]: https://github.com/brson/home/compare/v0.5.0...v0.5.1 +[0.5.0]: https://github.com/brson/home/compare/0.4.2...v0.5.0 diff --git a/crates/home/Cargo.toml b/crates/home/Cargo.toml new file mode 100644 index 0000000..18459df --- /dev/null +++ b/crates/home/Cargo.toml @@ -0,0 +1,20 @@ +[package] +name = "home" +version = "0.5.4" # also update `html_root_url` in `src/lib.rs` +authors = [ "Brian Anderson <andersrb@gmail.com>" ] +documentation = "https://docs.rs/home" +edition = "2018" +include = [ + "/src", + "/Cargo.toml", + "/CHANGELOG", + "/LICENSE-*", + "/README.md", +] +license = "MIT OR Apache-2.0" +readme = "README.md" +repository = "https://github.com/brson/home" +description = "Shared definitions of home directories" + +[target.'cfg(windows)'.dependencies] +windows-sys = { version = "0.45.0", features = ["Win32_Foundation", "Win32_UI_Shell"] } diff --git a/crates/home/LICENSE-APACHE b/crates/home/LICENSE-APACHE new file mode 120000 index 0000000..1cd601d --- /dev/null +++ b/crates/home/LICENSE-APACHE @@ -0,0 +1 @@ +../../LICENSE-APACHE
\ No newline at end of file diff --git a/crates/home/LICENSE-MIT b/crates/home/LICENSE-MIT new file mode 120000 index 0000000..b2cfbdc --- /dev/null +++ b/crates/home/LICENSE-MIT @@ -0,0 +1 @@ +../../LICENSE-MIT
\ No newline at end of file diff --git a/crates/home/README.md b/crates/home/README.md new file mode 100644 index 0000000..db2ba92 --- /dev/null +++ b/crates/home/README.md @@ -0,0 +1,27 @@ +[](https://docs.rs/home) +[](https://crates.io/crates/home) + +Canonical definitions of `home_dir`, `cargo_home`, and `rustup_home`. + +This provides the definition of `home_dir` used by Cargo and rustup, +as well functions to find the correct value of `CARGO_HOME` and +`RUSTUP_HOME`. + +The definition of `home_dir` provided by the standard library is +incorrect because it considers the `HOME` environment variable on +Windows. This causes surprising situations where a Rust program will +behave differently depending on whether it is run under a Unix +emulation environment like Cygwin or MinGW. Neither Cargo nor rustup +use the standard libraries definition - they use the definition here. + +This crate further provides two functions, `cargo_home` and +`rustup_home`, which are the canonical way to determine the location +that Cargo and rustup store their data. + +See [rust-lang/rust#43321]. + +[rust-lang/rust#43321]: https://github.com/rust-lang/rust/issues/43321 + +## License + +MIT OR Apache-2.0 diff --git a/crates/home/src/env.rs b/crates/home/src/env.rs new file mode 100644 index 0000000..e47273b --- /dev/null +++ b/crates/home/src/env.rs @@ -0,0 +1,106 @@ +//! Lower-level utilities for mocking the process environment. + +use std::{ + ffi::OsString, + io, + path::{Path, PathBuf}, +}; + +/// Permits parameterizing the home functions via the _from variants - used for +/// in-process unit testing by rustup. +pub trait Env { + /// Return the path to the the users home dir, or None if any error occurs: + /// see home_inner. + fn home_dir(&self) -> Option<PathBuf>; + /// Return the current working directory. + fn current_dir(&self) -> io::Result<PathBuf>; + /// Get an environment variable, as per std::env::var_os. + fn var_os(&self, key: &str) -> Option<OsString>; +} + +/// Implements Env for the OS context, both Unix style and Windows. +/// +/// This is trait permits in-process testing by providing a control point to +/// allow in-process divergence on what is normally process wide state. +/// +/// Implementations should be provided by whatever testing framework the caller +/// is using. Code that is not performing in-process threaded testing requiring +/// isolated rustup/cargo directories does not need this trait or the _from +/// functions. +pub struct OsEnv; +impl Env for OsEnv { + fn home_dir(&self) -> Option<PathBuf> { + crate::home_dir_inner() + } + fn current_dir(&self) -> io::Result<PathBuf> { + std::env::current_dir() + } + fn var_os(&self, key: &str) -> Option<OsString> { + std::env::var_os(key) + } +} + +pub const OS_ENV: OsEnv = OsEnv {}; + +/// Returns the path of the current user's home directory from [`Env::home_dir`]. +pub fn home_dir_with_env(env: &dyn Env) -> Option<PathBuf> { + env.home_dir() +} + +/// Variant of cargo_home where the environment source is parameterized. This is +/// specifically to support in-process testing scenarios as environment +/// variables and user home metadata are normally process global state. See the +/// [`Env`] trait. +pub fn cargo_home_with_env(env: &dyn Env) -> io::Result<PathBuf> { + let cwd = env.current_dir()?; + cargo_home_with_cwd_env(env, &cwd) +} + +/// Variant of cargo_home_with_cwd where the environment source is +/// parameterized. This is specifically to support in-process testing scenarios +/// as environment variables and user home metadata are normally process global +/// state. See the OsEnv trait. +pub fn cargo_home_with_cwd_env(env: &dyn Env, cwd: &Path) -> io::Result<PathBuf> { + match env.var_os("CARGO_HOME").filter(|h| !h.is_empty()) { + Some(home) => { + let home = PathBuf::from(home); + if home.is_absolute() { + Ok(home) + } else { + Ok(cwd.join(&home)) + } + } + _ => home_dir_with_env(env) + .map(|p| p.join(".cargo")) + .ok_or_else(|| io::Error::new(io::ErrorKind::Other, "could not find cargo home dir")), + } +} + +/// Variant of cargo_home_with_cwd where the environment source is +/// parameterized. This is specifically to support in-process testing scenarios +/// as environment variables and user home metadata are normally process global +/// state. See the OsEnv trait. +pub fn rustup_home_with_env(env: &dyn Env) -> io::Result<PathBuf> { + let cwd = env.current_dir()?; + rustup_home_with_cwd_env(env, &cwd) +} + +/// Variant of cargo_home_with_cwd where the environment source is +/// parameterized. This is specifically to support in-process testing scenarios +/// as environment variables and user home metadata are normally process global +/// state. See the OsEnv trait. +pub fn rustup_home_with_cwd_env(env: &dyn Env, cwd: &Path) -> io::Result<PathBuf> { + match env.var_os("RUSTUP_HOME").filter(|h| !h.is_empty()) { + Some(home) => { + let home = PathBuf::from(home); + if home.is_absolute() { + Ok(home) + } else { + Ok(cwd.join(&home)) + } + } + _ => home_dir_with_env(env) + .map(|d| d.join(".rustup")) + .ok_or_else(|| io::Error::new(io::ErrorKind::Other, "could not find rustup home dir")), + } +} diff --git a/crates/home/src/lib.rs b/crates/home/src/lib.rs new file mode 100644 index 0000000..306026e --- /dev/null +++ b/crates/home/src/lib.rs @@ -0,0 +1,149 @@ +//! Canonical definitions of `home_dir`, `cargo_home`, and `rustup_home`. +//! +//! This provides the definition of `home_dir` used by Cargo and +//! rustup, as well functions to find the correct value of +//! `CARGO_HOME` and `RUSTUP_HOME`. +//! +//! See also the [`dirs`](https://docs.rs/dirs) crate. +//! +//! _Note that as of 2019/08/06 it appears that cargo uses this crate. And +//! rustup has used this crate since 2019/08/21._ +//! +//! The definition of `home_dir` provided by the standard library is +//! incorrect because it considers the `HOME` environment variable on +//! Windows. This causes surprising situations where a Rust program +//! will behave differently depending on whether it is run under a +//! Unix emulation environment like Cygwin or MinGW. Neither Cargo nor +//! rustup use the standard libraries definition - they use the +//! definition here. +//! +//! This crate further provides two functions, `cargo_home` and +//! `rustup_home`, which are the canonical way to determine the +//! location that Cargo and rustup store their data. +//! +//! See also this [discussion]. +//! +//! [discussion]: https://github.com/rust-lang/rust/pull/46799#issuecomment-361156935 + +#![doc(html_root_url = "https://docs.rs/home/0.5.4")] +#![deny(rust_2018_idioms)] + +pub mod env; + +#[cfg(target_os = "windows")] +mod windows; + +use std::io; +use std::path::{Path, PathBuf}; + +/// Returns the path of the current user's home directory if known. +/// +/// # Unix +/// +/// Returns the value of the `HOME` environment variable if it is set +/// and not equal to the empty string. Otherwise, it tries to determine the +/// home directory by invoking the `getpwuid_r` function on the UID of the +/// current user. +/// +/// # Windows +/// +/// Returns the value of the `USERPROFILE` environment variable if it +/// is set and not equal to the empty string. If both do not exist, +/// [`SHGetFolderPathW`][msdn] is used to return the appropriate path. +/// +/// [msdn]: https://docs.microsoft.com/en-us/windows/win32/api/shlobj_core/nf-shlobj_core-shgetfolderpathw +/// +/// # Examples +/// +/// ``` +/// match home::home_dir() { +/// Some(path) => println!("{}", path.display()), +/// None => println!("Impossible to get your home dir!"), +/// } +/// ``` +pub fn home_dir() -> Option<PathBuf> { + env::home_dir_with_env(&env::OS_ENV) +} + +#[cfg(windows)] +use windows::home_dir_inner; + +#[cfg(any(unix, target_os = "redox"))] +fn home_dir_inner() -> Option<PathBuf> { + #[allow(deprecated)] + std::env::home_dir() +} + +/// Returns the storage directory used by Cargo, often knowns as +/// `.cargo` or `CARGO_HOME`. +/// +/// It returns one of the following values, in this order of +/// preference: +/// +/// - The value of the `CARGO_HOME` environment variable, if it is +/// an absolute path. +/// - The value of the current working directory joined with the value +/// of the `CARGO_HOME` environment variable, if `CARGO_HOME` is a +/// relative directory. +/// - The `.cargo` directory in the user's home directory, as reported +/// by the `home_dir` function. +/// +/// # Errors +/// +/// This function fails if it fails to retrieve the current directory, +/// or if the home directory cannot be determined. +/// +/// # Examples +/// +/// ``` +/// match home::cargo_home() { +/// Ok(path) => println!("{}", path.display()), +/// Err(err) => eprintln!("Cannot get your cargo home dir: {:?}", err), +/// } +/// ``` +pub fn cargo_home() -> io::Result<PathBuf> { + env::cargo_home_with_env(&env::OS_ENV) +} + +/// Returns the storage directory used by Cargo within `cwd`. +/// For more details, see [`cargo_home`](fn.cargo_home.html). +pub fn cargo_home_with_cwd(cwd: &Path) -> io::Result<PathBuf> { + env::cargo_home_with_cwd_env(&env::OS_ENV, cwd) +} + +/// Returns the storage directory used by rustup, often knowns as +/// `.rustup` or `RUSTUP_HOME`. +/// +/// It returns one of the following values, in this order of +/// preference: +/// +/// - The value of the `RUSTUP_HOME` environment variable, if it is +/// an absolute path. +/// - The value of the current working directory joined with the value +/// of the `RUSTUP_HOME` environment variable, if `RUSTUP_HOME` is a +/// relative directory. +/// - The `.rustup` directory in the user's home directory, as reported +/// by the `home_dir` function. +/// +/// # Errors +/// +/// This function fails if it fails to retrieve the current directory, +/// or if the home directory cannot be determined. +/// +/// # Examples +/// +/// ``` +/// match home::rustup_home() { +/// Ok(path) => println!("{}", path.display()), +/// Err(err) => eprintln!("Cannot get your rustup home dir: {:?}", err), +/// } +/// ``` +pub fn rustup_home() -> io::Result<PathBuf> { + env::rustup_home_with_env(&env::OS_ENV) +} + +/// Returns the storage directory used by rustup within `cwd`. +/// For more details, see [`rustup_home`](fn.rustup_home.html). +pub fn rustup_home_with_cwd(cwd: &Path) -> io::Result<PathBuf> { + env::rustup_home_with_cwd_env(&env::OS_ENV, cwd) +} diff --git a/crates/home/src/windows.rs b/crates/home/src/windows.rs new file mode 100644 index 0000000..a35dc9c --- /dev/null +++ b/crates/home/src/windows.rs @@ -0,0 +1,66 @@ +use std::env; +use std::ffi::OsString; +use std::os::windows::ffi::OsStringExt; +use std::path::PathBuf; + +use windows_sys::Win32::Foundation::{MAX_PATH, S_OK}; +use windows_sys::Win32::UI::Shell::{SHGetFolderPathW, CSIDL_PROFILE}; + +pub fn home_dir_inner() -> Option<PathBuf> { + env::var_os("USERPROFILE") + .filter(|s| !s.is_empty()) + .map(PathBuf::from) + .or_else(home_dir_crt) +} + +#[cfg(not(target_vendor = "uwp"))] +fn home_dir_crt() -> Option<PathBuf> { + unsafe { + let mut path: Vec<u16> = Vec::with_capacity(MAX_PATH as usize); + match SHGetFolderPathW(0, CSIDL_PROFILE as i32, 0, 0, path.as_mut_ptr()) { + S_OK => { + let len = wcslen(path.as_ptr()); + path.set_len(len); + let s = OsString::from_wide(&path); + Some(PathBuf::from(s)) + } + _ => None, + } + } +} + +#[cfg(target_vendor = "uwp")] +fn home_dir_crt() -> Option<PathBuf> { + None +} + +extern "C" { + fn wcslen(buf: *const u16) -> usize; +} + +#[cfg(not(target_vendor = "uwp"))] +#[cfg(test)] +mod tests { + use super::home_dir_inner; + use std::env; + use std::ops::Deref; + use std::path::{Path, PathBuf}; + + #[test] + fn test_with_without() { + let olduserprofile = env::var_os("USERPROFILE").unwrap(); + + env::remove_var("HOME"); + env::remove_var("USERPROFILE"); + + assert_eq!(home_dir_inner(), Some(PathBuf::from(olduserprofile))); + + let home = Path::new(r"C:\Users\foo tar baz"); + + env::set_var("HOME", home.as_os_str()); + assert_ne!(home_dir_inner().as_ref().map(Deref::deref), Some(home)); + + env::set_var("USERPROFILE", home.as_os_str()); + assert_eq!(home_dir_inner().as_ref().map(Deref::deref), Some(home)); + } +} diff --git a/crates/mdman/Cargo.lock b/crates/mdman/Cargo.lock new file mode 100644 index 0000000..51fe47a --- /dev/null +++ b/crates/mdman/Cargo.lock @@ -0,0 +1,459 @@ +# This file is automatically @generated by Cargo. +# It is not intended for manual editing. +version = 3 + +[[package]] +name = "anyhow" +version = "1.0.32" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "6b602bfe940d21c130f3895acd65221e8a61270debe89d628b9cb4e3ccb8569b" + +[[package]] +name = "bitflags" +version = "1.3.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a" + +[[package]] +name = "block-buffer" +version = "0.7.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "c0940dc441f31689269e10ac70eb1002a3a1d3ad1390e030043662eb7fe4688b" +dependencies = [ + "block-padding", + "byte-tools", + "byteorder", + "generic-array", +] + +[[package]] +name = "block-padding" +version = "0.1.5" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "fa79dedbb091f449f1f39e53edf88d5dbe95f895dae6135a8d7b881fb5af73f5" +dependencies = [ + "byte-tools", +] + +[[package]] +name = "byte-tools" +version = "0.3.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "e3b5ca7a04898ad4bcd41c90c5285445ff5b791899bb1b0abdd2a2aa791211d7" + +[[package]] +name = "byteorder" +version = "1.3.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "08c48aae112d48ed9f069b33538ea9e3e90aa263cfa3d1c24309612b1f7472de" + +[[package]] +name = "cfg-if" +version = "0.1.10" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "4785bdd1c96b2a846b2bd7cc02e86b6b3dbf14e7e53446c4f54c92a361040822" + +[[package]] +name = "ctor" +version = "0.1.15" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "39858aa5bac06462d4dd4b9164848eb81ffc4aa5c479746393598fd193afa227" +dependencies = [ + "quote", + "syn", +] + +[[package]] +name = "diff" +version = "0.1.13" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "56254986775e3233ffa9c4d7d3faaf6d36a2c09d30b20687e9f88bc8bafc16c8" + +[[package]] +name = "digest" +version = "0.8.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "f3d0c8c8752312f9713efd397ff63acb9f85585afbf179282e720e7704954dd5" +dependencies = [ + "generic-array", +] + +[[package]] +name = "fake-simd" +version = "0.1.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "e88a8acf291dafb59c2d96e8f59828f3838bb1a70398823ade51a84de6a6deed" + +[[package]] +name = "form_urlencoded" +version = "1.0.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "5fc25a87fa4fd2094bffb06925852034d90a17f0d1e05197d4956d3555752191" +dependencies = [ + "matches", + "percent-encoding", +] + +[[package]] +name = "generic-array" +version = "0.12.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "c68f0274ae0e023facc3c97b2e00f076be70e254bc851d972503b328db79b2ec" +dependencies = [ + "typenum", +] + +[[package]] +name = "handlebars" +version = "3.3.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "86dbc8a0746b08f363d2e00da48e6c9ceb75c198ac692d2715fcbb5bee74c87d" +dependencies = [ + "log", + "pest", + "pest_derive", + "quick-error", + "serde", + "serde_json", + "walkdir", +] + +[[package]] +name = "idna" +version = "0.2.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "02e2673c30ee86b5b96a9cb52ad15718aa1f966f5ab9ad54a8b95d5ca33120a9" +dependencies = [ + "matches", + "unicode-bidi", + "unicode-normalization", +] + +[[package]] +name = "itoa" +version = "0.4.6" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "dc6f3ad7b9d11a0c00842ff8de1b60ee58661048eb8049ed33c73594f359d7e6" + +[[package]] +name = "log" +version = "0.4.11" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "4fabed175da42fed1fa0746b0ea71f412aa9d35e76e95e59b192c64b9dc2bf8b" +dependencies = [ + "cfg-if", +] + +[[package]] +name = "maplit" +version = "1.0.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "3e2e65a1a2e43cfcb47a895c4c8b10d1f4a61097f9f254f183aee60cad9c651d" + +[[package]] +name = "matches" +version = "0.1.8" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "7ffc5c5338469d4d3ea17d269fa8ea3512ad247247c30bd2df69e68309ed0a08" + +[[package]] +name = "mdman" +version = "0.1.0" +dependencies = [ + "anyhow", + "handlebars", + "pretty_assertions", + "pulldown-cmark", + "same-file", + "serde_json", + "url", +] + +[[package]] +name = "memchr" +version = "2.5.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "2dffe52ecf27772e601905b7522cb4ef790d2cc203488bbd0e2fe85fcb74566d" + +[[package]] +name = "opaque-debug" +version = "0.2.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "2839e79665f131bdb5782e51f2c6c9599c133c6098982a54c794358bf432529c" + +[[package]] +name = "output_vt100" +version = "0.1.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "53cdc5b785b7a58c5aad8216b3dfa114df64b0b06ae6e1501cef91df2fbdf8f9" +dependencies = [ + "winapi", +] + +[[package]] +name = "percent-encoding" +version = "2.1.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "d4fd5641d01c8f18a23da7b6fe29298ff4b55afcccdf78973b24cf3175fee32e" + +[[package]] +name = "pest" +version = "2.1.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "10f4872ae94d7b90ae48754df22fd42ad52ce740b8f370b03da4835417403e53" +dependencies = [ + "ucd-trie", +] + +[[package]] +name = "pest_derive" +version = "2.1.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "833d1ae558dc601e9a60366421196a8d94bc0ac980476d0b67e1d0988d72b2d0" +dependencies = [ + "pest", + "pest_generator", +] + +[[package]] +name = "pest_generator" +version = "2.1.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "99b8db626e31e5b81787b9783425769681b347011cc59471e33ea46d2ea0cf55" +dependencies = [ + "pest", + "pest_meta", + "proc-macro2", + "quote", + "syn", +] + +[[package]] +name = "pest_meta" +version = "2.1.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "54be6e404f5317079812fc8f9f5279de376d8856929e21c184ecf6bbd692a11d" +dependencies = [ + "maplit", + "pest", + "sha-1", +] + +[[package]] +name = "pretty_assertions" +version = "1.3.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "a25e9bcb20aa780fd0bb16b72403a9064d6b3f22f026946029acb941a50af755" +dependencies = [ + "ctor", + "diff", + "output_vt100", + "yansi", +] + +[[package]] +name = "proc-macro2" +version = "1.0.19" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "04f5f085b5d71e2188cb8271e5da0161ad52c3f227a661a3c135fdf28e258b12" +dependencies = [ + "unicode-xid", +] + +[[package]] +name = "pulldown-cmark" +version = "0.9.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "2d9cc634bc78768157b5cbfe988ffcd1dcba95cd2b2f03a88316c08c6d00ed63" +dependencies = [ + "bitflags", + "memchr", + "unicase", +] + +[[package]] +name = "quick-error" +version = "1.2.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "a1d01941d82fa2ab50be1e79e6714289dd7cde78eba4c074bc5a4374f650dfe0" + +[[package]] +name = "quote" +version = "1.0.7" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "aa563d17ecb180e500da1cfd2b028310ac758de548efdd203e18f283af693f37" +dependencies = [ + "proc-macro2", +] + +[[package]] +name = "ryu" +version = "1.0.5" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "71d301d4193d031abdd79ff7e3dd721168a9572ef3fe51a1517aba235bd8f86e" + +[[package]] +name = "same-file" +version = "1.0.6" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502" +dependencies = [ + "winapi-util", +] + +[[package]] +name = "serde" +version = "1.0.114" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "5317f7588f0a5078ee60ef675ef96735a1442132dc645eb1d12c018620ed8cd3" + +[[package]] +name = "serde_json" +version = "1.0.57" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "164eacbdb13512ec2745fb09d51fd5b22b0d65ed294a1dcf7285a360c80a675c" +dependencies = [ + "itoa", + "ryu", + "serde", +] + +[[package]] +name = "sha-1" +version = "0.8.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "f7d94d0bede923b3cea61f3f1ff57ff8cdfd77b400fb8f9998949e0cf04163df" +dependencies = [ + "block-buffer", + "digest", + "fake-simd", + "opaque-debug", +] + +[[package]] +name = "syn" +version = "1.0.36" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "4cdb98bcb1f9d81d07b536179c269ea15999b5d14ea958196413869445bb5250" +dependencies = [ + "proc-macro2", + "quote", + "unicode-xid", +] + +[[package]] +name = "tinyvec" +version = "0.3.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "53953d2d3a5ad81d9f844a32f14ebb121f50b650cd59d0ee2a07cf13c617efed" + +[[package]] +name = "typenum" +version = "1.12.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "373c8a200f9e67a0c95e62a4f52fbf80c23b4381c05a17845531982fa99e6b33" + +[[package]] +name = "ucd-trie" +version = "0.1.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "56dee185309b50d1f11bfedef0fe6d036842e3fb77413abef29f8f8d1c5d4c1c" + +[[package]] +name = "unicase" +version = "2.6.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "50f37be617794602aabbeee0be4f259dc1778fabe05e2d67ee8f79326d5cb4f6" +dependencies = [ + "version_check", +] + +[[package]] +name = "unicode-bidi" +version = "0.3.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "49f2bd0c6468a8230e1db229cff8029217cf623c767ea5d60bfbd42729ea54d5" +dependencies = [ + "matches", +] + +[[package]] +name = "unicode-normalization" +version = "0.1.13" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "6fb19cf769fa8c6a80a162df694621ebeb4dafb606470b2b2fce0be40a98a977" +dependencies = [ + "tinyvec", +] + +[[package]] +name = "unicode-xid" +version = "0.2.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "f7fe0bb3479651439c9112f72b6c505038574c9fbb575ed1bf3b797fa39dd564" + +[[package]] +name = "url" +version = "2.2.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "a507c383b2d33b5fc35d1861e77e6b383d158b2da5e14fe51b83dfedf6fd578c" +dependencies = [ + "form_urlencoded", + "idna", + "matches", + "percent-encoding", +] + +[[package]] +name = "version_check" +version = "0.9.2" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "b5a972e5669d67ba988ce3dc826706fb0a8b01471c088cb0b6110b805cc36aed" + +[[package]] +name = "walkdir" +version = "2.3.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "777182bc735b6424e1a57516d35ed72cb8019d85c8c9bf536dccb3445c1a2f7d" +dependencies = [ + "same-file", + "winapi", + "winapi-util", +] + +[[package]] +name = "winapi" +version = "0.3.9" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "5c839a674fcd7a98952e593242ea400abe93992746761e38641405d28b00f419" +dependencies = [ + "winapi-i686-pc-windows-gnu", + "winapi-x86_64-pc-windows-gnu", +] + +[[package]] +name = "winapi-i686-pc-windows-gnu" +version = "0.4.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6" + +[[package]] +name = "winapi-util" +version = "0.1.5" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "70ec6ce85bb158151cae5e5c87f95a8e97d2c0c4b001223f33a334e3ce5de178" +dependencies = [ + "winapi", +] + +[[package]] +name = "winapi-x86_64-pc-windows-gnu" +version = "0.4.0" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f" + +[[package]] +name = "yansi" +version = "0.5.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "09041cd90cf85f7f8b2df60c646f853b7f535ce68f85244eb6731cf89fa498ec" diff --git a/crates/mdman/Cargo.toml b/crates/mdman/Cargo.toml new file mode 100644 index 0000000..92cdf2e --- /dev/null +++ b/crates/mdman/Cargo.toml @@ -0,0 +1,17 @@ +[package] +name = "mdman" +version = "0.1.0" +edition = "2021" +license = "MIT OR Apache-2.0" +description = "Creates a man page page from markdown." + +[dependencies] +anyhow = "1.0.31" +handlebars = { version = "3.2.1", features = ["dir_source"] } +pulldown-cmark = { version = "0.9.2", default-features = false } +same-file = "1.0.6" +serde_json = "1.0.56" +url = "2.2.2" + +[dev-dependencies] +pretty_assertions = "1.3.0" diff --git a/crates/mdman/README.md b/crates/mdman/README.md new file mode 100644 index 0000000..e28b596 --- /dev/null +++ b/crates/mdman/README.md @@ -0,0 +1,7 @@ +# mdman + +mdman is a small utility for creating man pages from markdown text files. + +## Usage + +See the [man page](doc/out/mdman.md) generated by this tool. diff --git a/crates/mdman/build-man.sh b/crates/mdman/build-man.sh new file mode 100755 index 0000000..9286b17 --- /dev/null +++ b/crates/mdman/build-man.sh @@ -0,0 +1,7 @@ +#!/bin/bash + +set -e + +cargo run -- -t md -o doc/out doc/*.md +cargo run -- -t txt -o doc/out doc/*.md +cargo run -- -t man -o doc/out doc/*.md diff --git a/crates/mdman/doc/mdman.md b/crates/mdman/doc/mdman.md new file mode 100644 index 0000000..2025c13 --- /dev/null +++ b/crates/mdman/doc/mdman.md @@ -0,0 +1,95 @@ +# mdman(1) + +## NAME + +mdman - Converts markdown to a man page + +## SYNOPSIS + +`mdman` [_options_] `-t` _type_ `-o` _outdir_ _sources..._ + +## DESCRIPTION + +Converts a markdown file to a man page. + +The source file is first processed as a +[handlebars](https://handlebarsjs.com/) template. Then, it is processed as +markdown into the target format. This supports different output formats, +such as troff or plain text. + +Every man page should start with a level-1 header with the man name and +section, such as `# mdman(1)`. + +The handlebars template has several special tags to assist with generating the +man page: + +{{{{raw}}}} +- Every block of command-line options must be wrapped between `{{#options}}` + and `{{/options}}` tags. This tells the processor where the options start + and end. +- Each option must be expressed with a `{{#option}}` block. The parameters to + the the block are a sequence of strings indicating the option. For example, + ```{{#option "`-p` _spec_..." "`--package` _spec_..."}}``` is an option that + has two different forms. The text within the string is processed as markdown. + It is recommended to use formatting similar to this example. + + The content of the `{{#option}}` block should contain a detailed description + of the option. + + Use the `{{/option}}` tag to end the option block. +- References to other man pages should use the `{{man name section}}` + expression. For example, `{{man "mdman" 1}}` will generate a reference to + the `mdman(1)` man page. For non-troff output, the `--man` option will tell + `mdman` how to create links to the man page. If there is no matching `--man` + option, then it links to a file named _name_`.md` in the same directory. +- Variables can be set with `{{*set name="value"}}`. These variables can + then be referenced with `{{name}}` expressions. +- Partial templates should be placed in a directory named `includes` + next to the source file. Templates can be included with an expression like + `{{> template-name}}`. +- Other helpers include: + - `{{lower value}}` Converts the given value to lowercase. +{{{{/raw}}}} + +## OPTIONS + +{{#options}} + +{{#option "`-t` _type_"}} +Specifies the output type. The following output types are supported: +- `man` — A troff-style man page. Outputs with a numbered extension (like + `.1`) matching the man page section. +- `md` — A markdown file, after all handlebars processing has been finished. + Outputs with the `.md` extension. +- `txt` — A text file, rendered for situations where a man page viewer isn't + available. Outputs with the `.txt` extension. +{{/option}} + +{{#option "`-o` _outdir_"}} +Specifies the directory where to save the output. +{{/option}} + +{{#option "`--url` _base_url_"}} +Specifies a base URL to use for relative URLs within the document. Any +relative URL will be joined with this URL. +{{/option}} + +{{#option "`--man` _name_`:`_section_`=`_url_"}} +Specifies a URL to use for the given man page. When the `\{{man name +section}}` expression is used, the given URL will be inserted as a link. This +may be specified multiple times. If a man page reference does not have a +matching `--man` entry, then a relative link to a file named _name_`.md` will +be used. +{{/option}} + +{{#option "_sources..._"}} +The source input filename, may be specified multiple times. +{{/option}} + +{{/options}} + +## EXAMPLES + +1. Convert the given documents to man pages: + + mdman -t man -o doc doc/mdman.md diff --git a/crates/mdman/doc/out/mdman.1 b/crates/mdman/doc/out/mdman.1 new file mode 100644 index 0000000..0718d6d --- /dev/null +++ b/crates/mdman/doc/out/mdman.1 @@ -0,0 +1,124 @@ +'\" t +.TH "MDMAN" "1" +.nh +.ad l +.ss \n[.ss] 0 +.SH "NAME" +mdman \- Converts markdown to a man page +.SH "SYNOPSIS" +\fBmdman\fR [\fIoptions\fR] \fB\-t\fR \fItype\fR \fB\-o\fR \fIoutdir\fR \fIsources...\fR +.SH "DESCRIPTION" +Converts a markdown file to a man page. +.sp +The source file is first processed as a +\fIhandlebars\fR <https://handlebarsjs.com/> template. Then, it is processed as +markdown into the target format. This supports different output formats, +such as troff or plain text. +.sp +Every man page should start with a level\-1 header with the man name and +section, such as \fB# mdman(1)\fR\&. +.sp +The handlebars template has several special tags to assist with generating the +man page: +.sp +.RS 4 +\h'-04'\(bu\h'+02'Every block of command\-line options must be wrapped between \fB{{#options}}\fR +and \fB{{/options}}\fR tags. This tells the processor where the options start +and end. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Each option must be expressed with a \fB{{#option}}\fR block. The parameters to +the the block are a sequence of strings indicating the option. For example, +\fB{{#option "`\-p` _spec_..." "`\-\-package` _spec_..."}}\fR is an option that +has two different forms. The text within the string is processed as markdown. +It is recommended to use formatting similar to this example. +.sp +The content of the \fB{{#option}}\fR block should contain a detailed description +of the option. +.sp +Use the \fB{{/option}}\fR tag to end the option block. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'References to other man pages should use the \fB{{man name section}}\fR +expression. For example, \fB{{man "mdman" 1}}\fR will generate a reference to +the \fBmdman(1)\fR man page. For non\-troff output, the \fB\-\-man\fR option will tell +\fBmdman\fR how to create links to the man page. If there is no matching \fB\-\-man\fR +option, then it links to a file named \fIname\fR\fB\&.md\fR in the same directory. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Variables can be set with \fB{{*set name="value"}}\fR\&. These variables can +then be referenced with \fB{{name}}\fR expressions. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Partial templates should be placed in a directory named \fBincludes\fR +next to the source file. Templates can be included with an expression like +\fB{{> template\-name}}\fR\&. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Other helpers include: +.sp +.RS 4 +\h'-04'\(bu\h'+02'\fB{{lower value}}\fR Converts the given value to lowercase. +.RE +.RE +.SH "OPTIONS" +.sp +\fB\-t\fR \fItype\fR +.RS 4 +Specifies the output type. The following output types are supported: +.sp +.RS 4 +\h'-04'\(bu\h'+02'\fBman\fR \[em] A troff\-style man page. Outputs with a numbered extension (like +\fB\&.1\fR) matching the man page section. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'\fBmd\fR \[em] A markdown file, after all handlebars processing has been finished. +Outputs with the \fB\&.md\fR extension. +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'\fBtxt\fR \[em] A text file, rendered for situations where a man page viewer isn't +available. Outputs with the \fB\&.txt\fR extension. +.RE +.RE +.sp +\fB\-o\fR \fIoutdir\fR +.RS 4 +Specifies the directory where to save the output. +.RE +.sp +\fB\-\-url\fR \fIbase_url\fR +.RS 4 +Specifies a base URL to use for relative URLs within the document. Any +relative URL will be joined with this URL. +.RE +.sp +\fB\-\-man\fR \fIname\fR\fB:\fR\fIsection\fR\fB=\fR\fIurl\fR +.RS 4 +Specifies a URL to use for the given man page. When the \fB{{man name section}}\fR expression is used, the given URL will be inserted as a link. This +may be specified multiple times. If a man page reference does not have a +matching \fB\-\-man\fR entry, then a relative link to a file named \fIname\fR\fB\&.md\fR will +be used. +.RE +.sp +\fIsources...\fR +.RS 4 +The source input filename, may be specified multiple times. +.RE +.SH "EXAMPLES" +.sp +.RS 4 +\h'-04' 1.\h'+01'Convert the given documents to man pages: +.sp +.RS 4 +.nf +mdman \-t man \-o doc doc/mdman.md +.fi +.RE +.RE diff --git a/crates/mdman/doc/out/mdman.md b/crates/mdman/doc/out/mdman.md new file mode 100644 index 0000000..d0dd345 --- /dev/null +++ b/crates/mdman/doc/out/mdman.md @@ -0,0 +1,95 @@ +# mdman(1) + +## NAME + +mdman - Converts markdown to a man page + +## SYNOPSIS + +`mdman` [_options_] `-t` _type_ `-o` _outdir_ _sources..._ + +## DESCRIPTION + +Converts a markdown file to a man page. + +The source file is first processed as a +[handlebars](https://handlebarsjs.com/) template. Then, it is processed as +markdown into the target format. This supports different output formats, +such as troff or plain text. + +Every man page should start with a level-1 header with the man name and +section, such as `# mdman(1)`. + +The handlebars template has several special tags to assist with generating the +man page: + +- Every block of command-line options must be wrapped between `{{#options}}` + and `{{/options}}` tags. This tells the processor where the options start + and end. +- Each option must be expressed with a `{{#option}}` block. The parameters to + the the block are a sequence of strings indicating the option. For example, + ```{{#option "`-p` _spec_..." "`--package` _spec_..."}}``` is an option that + has two different forms. The text within the string is processed as markdown. + It is recommended to use formatting similar to this example. + + The content of the `{{#option}}` block should contain a detailed description + of the option. + + Use the `{{/option}}` tag to end the option block. +- References to other man pages should use the `{{man name section}}` + expression. For example, `{{man "mdman" 1}}` will generate a reference to + the `mdman(1)` man page. For non-troff output, the `--man` option will tell + `mdman` how to create links to the man page. If there is no matching `--man` + option, then it links to a file named _name_`.md` in the same directory. +- Variables can be set with `{{*set name="value"}}`. These variables can + then be referenced with `{{name}}` expressions. +- Partial templates should be placed in a directory named `includes` + next to the source file. Templates can be included with an expression like + `{{> template-name}}`. +- Other helpers include: + - `{{lower value}}` Converts the given value to lowercase. + + +## OPTIONS + +<dl> + +<dt class="option-term" id="option-mdman--t"><a class="option-anchor" href="#option-mdman--t"></a><code>-t</code> <em>type</em></dt> +<dd class="option-desc">Specifies the output type. The following output types are supported:</p> +<ul> +<li><code>man</code> — A troff-style man page. Outputs with a numbered extension (like +<code>.1</code>) matching the man page section.</li> +<li><code>md</code> — A markdown file, after all handlebars processing has been finished. +Outputs with the <code>.md</code> extension.</li> +<li><code>txt</code> — A text file, rendered for situations where a man page viewer isn't +available. Outputs with the <code>.txt</code> extension.</li> +</ul></dd> + + +<dt class="option-term" id="option-mdman--o"><a class="option-anchor" href="#option-mdman--o"></a><code>-o</code> <em>outdir</em></dt> +<dd class="option-desc">Specifies the directory where to save the output.</dd> + + +<dt class="option-term" id="option-mdman---url"><a class="option-anchor" href="#option-mdman---url"></a><code>--url</code> <em>base_url</em></dt> +<dd class="option-desc">Specifies a base URL to use for relative URLs within the document. Any +relative URL will be joined with this URL.</dd> + + +<dt class="option-term" id="option-mdman---man"><a class="option-anchor" href="#option-mdman---man"></a><code>--man</code> <em>name</em><code>:</code><em>section</em><code>=</code><em>url</em></dt> +<dd class="option-desc">Specifies a URL to use for the given man page. When the <code>{{man name section}}</code> expression is used, the given URL will be inserted as a link. This +may be specified multiple times. If a man page reference does not have a +matching <code>--man</code> entry, then a relative link to a file named <em>name</em><code>.md</code> will +be used.</dd> + + +<dt class="option-term" id="option-mdman-sources..."><a class="option-anchor" href="#option-mdman-sources..."></a><em>sources...</em></dt> +<dd class="option-desc">The source input filename, may be specified multiple times.</dd> + + +</dl> + +## EXAMPLES + +1. Convert the given documents to man pages: + + mdman -t man -o doc doc/mdman.md diff --git a/crates/mdman/doc/out/mdman.txt b/crates/mdman/doc/out/mdman.txt new file mode 100644 index 0000000..83fa7de --- /dev/null +++ b/crates/mdman/doc/out/mdman.txt @@ -0,0 +1,91 @@ +MDMAN(1) + +NAME + mdman - Converts markdown to a man page + +SYNOPSIS + mdman [options] -t type -o outdir sources... + +DESCRIPTION + Converts a markdown file to a man page. + + The source file is first processed as a handlebars + <https://handlebarsjs.com/> template. Then, it is processed as markdown + into the target format. This supports different output formats, such as + troff or plain text. + + Every man page should start with a level-1 header with the man name and + section, such as # mdman(1). + + The handlebars template has several special tags to assist with + generating the man page: + + o Every block of command-line options must be wrapped between + {{#options}} and {{/options}} tags. This tells the processor where + the options start and end. + + o Each option must be expressed with a {{#option}} block. The + parameters to the the block are a sequence of strings indicating the + option. For example, {{#option "`-p` _spec_..." "`--package` + _spec_..."}} is an option that has two different forms. The text + within the string is processed as markdown. It is recommended to use + formatting similar to this example. + + The content of the {{#option}} block should contain a detailed + description of the option. + + Use the {{/option}} tag to end the option block. + + o References to other man pages should use the {{man name section}} + expression. For example, {{man "mdman" 1}} will generate a reference + to the mdman(1) man page. For non-troff output, the --man option will + tell mdman how to create links to the man page. If there is no + matching --man option, then it links to a file named name.md in the + same directory. + + o Variables can be set with {{*set name="value"}}. These variables can + then be referenced with {{name}} expressions. + + o Partial templates should be placed in a directory named includes next + to the source file. Templates can be included with an expression like + {{> template-name}}. + + o Other helpers include: + + o {{lower value}} Converts the given value to lowercase. + +OPTIONS + -t type + Specifies the output type. The following output types are supported: + + o man — A troff-style man page. Outputs with a numbered extension + (like .1) matching the man page section. + + o md — A markdown file, after all handlebars processing has been + finished. Outputs with the .md extension. + + o txt — A text file, rendered for situations where a man page + viewer isn't available. Outputs with the .txt extension. + + -o outdir + Specifies the directory where to save the output. + + --url base_url + Specifies a base URL to use for relative URLs within the document. + Any relative URL will be joined with this URL. + + --man name:section=url + Specifies a URL to use for the given man page. When the {{man name + section}} expression is used, the given URL will be inserted as a + link. This may be specified multiple times. If a man page reference + does not have a matching --man entry, then a relative link to a file + named name.md will be used. + + sources... + The source input filename, may be specified multiple times. + +EXAMPLES + 1. Convert the given documents to man pages: + + mdman -t man -o doc doc/mdman.md + diff --git a/crates/mdman/src/format.rs b/crates/mdman/src/format.rs new file mode 100644 index 0000000..7bc9781 --- /dev/null +++ b/crates/mdman/src/format.rs @@ -0,0 +1,20 @@ +use anyhow::Error; + +pub mod man; +pub mod md; +pub mod text; + +pub trait Formatter { + /// Renders the given markdown to the formatter's output. + fn render(&self, input: &str) -> Result<String, Error>; + /// Renders the start of a block of options (triggered by `{{#options}}`). + fn render_options_start(&self) -> &'static str; + /// Renders the end of a block of options (triggered by `{{/options}}`). + fn render_options_end(&self) -> &'static str; + /// Renders an option (triggered by `{{#option}}`). + fn render_option(&self, params: &[&str], block: &str, man_name: &str) -> Result<String, Error>; + /// Converts a man page reference into markdown that is appropriate for this format. + /// + /// Triggered by `{{man name section}}`. + fn linkify_man_to_md(&self, name: &str, section: u8) -> Result<String, Error>; +} diff --git a/crates/mdman/src/format/man.rs b/crates/mdman/src/format/man.rs new file mode 100644 index 0000000..9767fdd --- /dev/null +++ b/crates/mdman/src/format/man.rs @@ -0,0 +1,436 @@ +//! Man-page formatter. + +use crate::util::{header_text, parse_name_and_section}; +use crate::EventIter; +use anyhow::{bail, Error}; +use pulldown_cmark::{Alignment, Event, HeadingLevel, LinkType, Tag}; +use std::fmt::Write; +use url::Url; + +pub struct ManFormatter { + url: Option<Url>, +} + +impl ManFormatter { + pub fn new(url: Option<Url>) -> ManFormatter { + ManFormatter { url } + } +} + +impl super::Formatter for ManFormatter { + fn render(&self, input: &str) -> Result<String, Error> { + ManRenderer::render(input, self.url.clone()) + } + + fn render_options_start(&self) -> &'static str { + // Tell pulldown_cmark to ignore this. + // This will be stripped out later. + "<![CDATA[" + } + + fn render_options_end(&self) -> &'static str { + "]]>" + } + + fn render_option( + &self, + params: &[&str], + block: &str, + _man_name: &str, + ) -> Result<String, Error> { + let rendered_options = params + .iter() + .map(|param| { + let r = self.render(param)?; + Ok(r.trim().trim_start_matches(".sp").to_string()) + }) + .collect::<Result<Vec<_>, Error>>()?; + let rendered_block = self.render(block)?; + let rendered_block = rendered_block.trim().trim_start_matches(".sp").trim(); + // .RS = move left margin to right 4. + // .RE = move margin back one level. + Ok(format!( + "\n.sp\n{}\n.RS 4\n{}\n.RE\n", + rendered_options.join(", "), + rendered_block + )) + } + + fn linkify_man_to_md(&self, name: &str, section: u8) -> Result<String, Error> { + Ok(format!("`{}`({})", name, section)) + } +} + +#[derive(Copy, Clone)] +enum Font { + Bold, + Italic, +} + +impl Font { + fn str_from_stack(font_stack: &[Font]) -> &'static str { + let has_bold = font_stack.iter().any(|font| matches!(font, Font::Bold)); + let has_italic = font_stack.iter().any(|font| matches!(font, Font::Italic)); + match (has_bold, has_italic) { + (false, false) => "\\fR", // roman (normal) + (false, true) => "\\fI", // italic + (true, false) => "\\fB", // bold + (true, true) => "\\f(BI", // bold italic + } + } +} + +struct ManRenderer<'e> { + output: String, + parser: EventIter<'e>, + font_stack: Vec<Font>, +} + +impl<'e> ManRenderer<'e> { + fn render(input: &str, url: Option<Url>) -> Result<String, Error> { + let parser = crate::md_parser(input, url); + let output = String::with_capacity(input.len() * 3 / 2); + let mut mr = ManRenderer { + parser, + output, + font_stack: Vec::new(), + }; + mr.push_man()?; + Ok(mr.output) + } + + fn push_man(&mut self) -> Result<(), Error> { + // If this is true, this is inside a cdata block used for hiding + // content from pulldown_cmark. + let mut in_cdata = false; + // The current list stack. None if unordered, Some if ordered with the + // given number as the current index. + let mut list: Vec<Option<u64>> = Vec::new(); + // Used in some cases where spacing isn't desired. + let mut suppress_paragraph = false; + let mut table_cell_index = 0; + + while let Some((event, range)) = self.parser.next() { + let this_suppress_paragraph = suppress_paragraph; + suppress_paragraph = false; + match event { + Event::Start(tag) => { + match tag { + Tag::Paragraph => { + if !this_suppress_paragraph { + self.flush(); + self.output.push_str(".sp\n"); + } + } + Tag::Heading(level, ..) => { + if level == HeadingLevel::H1 { + self.push_top_header()?; + } else if level == HeadingLevel::H2 { + // Section header + let text = header_text(&mut self.parser)?; + self.flush(); + write!(self.output, ".SH \"{}\"\n", text)?; + suppress_paragraph = true; + } else { + // Subsection header + let text = header_text(&mut self.parser)?; + self.flush(); + write!(self.output, ".SS \"{}\"\n", text)?; + suppress_paragraph = true; + } + } + Tag::BlockQuote => { + self.flush(); + // .RS = move left margin over 3 + // .ll = shrink line length + self.output.push_str(".RS 3\n.ll -5\n.sp\n"); + suppress_paragraph = true; + } + Tag::CodeBlock(_kind) => { + // space down, indent 4, no-fill mode + self.flush(); + self.output.push_str(".sp\n.RS 4\n.nf\n"); + } + Tag::List(start) => list.push(start), + Tag::Item => { + // Note: This uses explicit movement instead of .IP + // because the spacing on .IP looks weird to me. + // space down, indent 4 + self.flush(); + self.output.push_str(".sp\n.RS 4\n"); + match list.last_mut().expect("item must have list start") { + // Ordered list. + Some(n) => { + // move left 4, output the list index number, move right 1. + write!(self.output, "\\h'-04' {}.\\h'+01'", n)?; + *n += 1; + } + // Unordered list. + None => self.output.push_str("\\h'-04'\\(bu\\h'+02'"), + } + suppress_paragraph = true; + } + Tag::FootnoteDefinition(_label) => unimplemented!(), + Tag::Table(alignment) => { + // Table start + // allbox = draw a box around all the cells + // tab(:) = Use `:` to separate cell data (instead of tab) + // ; = end of options + self.output.push_str( + "\n.TS\n\ + allbox tab(:);\n", + ); + let alignments: Vec<_> = alignment + .iter() + .map(|a| match a { + Alignment::Left | Alignment::None => "lt", + Alignment::Center => "ct", + Alignment::Right => "rt", + }) + .collect(); + self.output.push_str(&alignments.join(" ")); + self.output.push_str(".\n"); + table_cell_index = 0; + } + Tag::TableHead => { + table_cell_index = 0; + } + Tag::TableRow => { + table_cell_index = 0; + self.output.push('\n'); + } + Tag::TableCell => { + if table_cell_index != 0 { + // Separator between columns. + self.output.push(':'); + } + // Start a text block. + self.output.push_str("T{\n"); + table_cell_index += 1 + } + Tag::Emphasis => self.push_font(Font::Italic), + Tag::Strong => self.push_font(Font::Bold), + // Strikethrough isn't usually supported for TTY. + Tag::Strikethrough => self.output.push_str("~~"), + Tag::Link(link_type, dest_url, _title) => { + if dest_url.starts_with('#') { + // In a man page, page-relative anchors don't + // have much meaning. + continue; + } + match link_type { + LinkType::Autolink | LinkType::Email => { + // The text is a copy of the URL, which is not needed. + match self.parser.next() { + Some((Event::Text(_), _range)) => {} + _ => bail!("expected text after autolink"), + } + } + LinkType::Inline + | LinkType::Reference + | LinkType::Collapsed + | LinkType::Shortcut => { + self.push_font(Font::Italic); + } + // This is currently unused. This is only + // emitted with a broken link callback, but I + // felt it is too annoying to escape `[` in + // option descriptions. + LinkType::ReferenceUnknown + | LinkType::CollapsedUnknown + | LinkType::ShortcutUnknown => { + bail!( + "link with missing reference `{}` located at offset {}", + dest_url, + range.start + ); + } + } + } + Tag::Image(_link_type, _dest_url, _title) => { + bail!("images are not currently supported") + } + } + } + Event::End(tag) => { + match &tag { + Tag::Paragraph => self.flush(), + Tag::Heading(..) => {} + Tag::BlockQuote => { + self.flush(); + // restore left margin, restore line length + self.output.push_str(".br\n.RE\n.ll\n"); + } + Tag::CodeBlock(_kind) => { + self.flush(); + // Restore fill mode, move margin back one level. + self.output.push_str(".fi\n.RE\n"); + } + Tag::List(_) => { + list.pop(); + } + Tag::Item => { + self.flush(); + // Move margin back one level. + self.output.push_str(".RE\n"); + } + Tag::FootnoteDefinition(_label) => {} + Tag::Table(_) => { + // Table end + // I don't know why, but the .sp is needed to provide + // space with the following content. + self.output.push_str("\n.TE\n.sp\n"); + } + Tag::TableHead => {} + Tag::TableRow => {} + Tag::TableCell => { + // End text block. + self.output.push_str("\nT}"); + } + Tag::Emphasis | Tag::Strong => self.pop_font(), + Tag::Strikethrough => self.output.push_str("~~"), + Tag::Link(link_type, dest_url, _title) => { + if dest_url.starts_with('#') { + continue; + } + match link_type { + LinkType::Autolink | LinkType::Email => {} + LinkType::Inline + | LinkType::Reference + | LinkType::Collapsed + | LinkType::Shortcut => { + self.pop_font(); + self.output.push(' '); + } + _ => { + panic!("unexpected tag {:?}", tag); + } + } + write!(self.output, "<{}>", escape(&dest_url)?)?; + } + Tag::Image(_link_type, _dest_url, _title) => {} + } + } + Event::Text(t) => { + self.output.push_str(&escape(&t)?); + } + Event::Code(t) => { + self.push_font(Font::Bold); + self.output.push_str(&escape(&t)?); + self.pop_font(); + } + Event::Html(t) => { + if t.starts_with("<![CDATA[") { + // CDATA is a special marker used for handling options. + in_cdata = true; + } else if in_cdata { + if t.trim().ends_with("]]>") { + in_cdata = false; + } else if !t.trim().is_empty() { + self.output.push_str(&t); + } + } else { + self.output.push_str(&escape(&t)?); + } + } + Event::FootnoteReference(_t) => {} + Event::SoftBreak => self.output.push('\n'), + Event::HardBreak => { + self.flush(); + self.output.push_str(".br\n"); + } + Event::Rule => { + self.flush(); + // \l' **length** ' Draw horizontal line (default underscore). + // \n(.lu Gets value from register "lu" (current line length) + self.output.push_str("\\l'\\n(.lu'\n"); + } + Event::TaskListMarker(_b) => unimplemented!(), + } + } + Ok(()) + } + + fn flush(&mut self) { + if !self.output.ends_with('\n') { + self.output.push('\n'); + } + } + + /// Switch to the given font. + /// + /// Because the troff sequence `\fP` for switching to the "previous" font + /// doesn't support nesting, this needs to emulate it here. This is needed + /// for situations like **hi _there_**. + fn push_font(&mut self, font: Font) { + self.font_stack.push(font); + self.output.push_str(Font::str_from_stack(&self.font_stack)); + } + + fn pop_font(&mut self) { + self.font_stack.pop(); + self.output.push_str(Font::str_from_stack(&self.font_stack)); + } + + /// Parse and render the first top-level header of the document. + fn push_top_header(&mut self) -> Result<(), Error> { + // This enables the tbl preprocessor for tables. + // This seems to be enabled by default on every modern system I could + // find, but it doesn't seem to hurt to enable this. + self.output.push_str("'\\\" t\n"); + // Extract the name of the man page. + let text = header_text(&mut self.parser)?; + let (name, section) = parse_name_and_section(&text)?; + // .TH = Table header + // .nh = disable hyphenation + // .ad l = Left-adjust mode (disable justified). + // .ss sets sentence_space_size to 0 (prevents double spaces after . + // if . is last on the line) + write!( + self.output, + ".TH \"{}\" \"{}\"\n\ + .nh\n\ + .ad l\n\ + .ss \\n[.ss] 0\n", + escape(&name.to_uppercase())?, + section + )?; + Ok(()) + } +} + +fn escape(s: &str) -> Result<String, Error> { + // Note: Possible source on output escape sequences: https://man7.org/linux/man-pages/man7/groff_char.7.html. + // Otherwise, use generic escaping in the form `\[u1EE7]` or `\[u1F994]`. + + let mut replaced = s + .replace('\\', "\\(rs") + .replace('-', "\\-") + .replace('\u{00A0}', "\\ ") // non-breaking space (non-stretchable) + .replace('–', "\\[en]") // \u{2013} en-dash + .replace('—', "\\[em]") // \u{2014} em-dash + .replace('‘', "\\[oq]") // \u{2018} left single quote + .replace('’', "\\[cq]") // \u{2019} right single quote or apostrophe + .replace('“', "\\[lq]") // \u{201C} left double quote + .replace('”', "\\[rq]") // \u{201D} right double quote + .replace('…', "\\[u2026]") // \u{2026} ellipsis + .replace('│', "|") // \u{2502} box drawing light vertical (could use \[br]) + .replace('├', "|") // \u{251C} box drawings light vertical and right + .replace('└', "`") // \u{2514} box drawings light up and right + .replace('─', "\\-") // \u{2500} box drawing light horizontal + ; + if replaced.starts_with('.') { + replaced = format!("\\&.{}", &replaced[1..]); + } + + if let Some(ch) = replaced.chars().find(|ch| { + !matches!(ch, '\n' | ' ' | '!'..='/' | '0'..='9' + | ':'..='@' | 'A'..='Z' | '['..='`' | 'a'..='z' | '{'..='~') + }) { + bail!( + "character {:?} is not allowed (update the translation table if needed)", + ch + ); + } + Ok(replaced) +} diff --git a/crates/mdman/src/format/md.rs b/crates/mdman/src/format/md.rs new file mode 100644 index 0000000..0e1c498 --- /dev/null +++ b/crates/mdman/src/format/md.rs @@ -0,0 +1,112 @@ +//! Markdown formatter. + +use crate::util::unwrap; +use crate::ManMap; +use anyhow::{bail, format_err, Error}; +use std::fmt::Write; + +pub struct MdFormatter { + man_map: ManMap, +} + +impl MdFormatter { + pub fn new(man_map: ManMap) -> MdFormatter { + MdFormatter { man_map } + } +} + +impl MdFormatter { + fn render_html(&self, input: &str) -> Result<String, Error> { + let parser = crate::md_parser(input, None); + let mut html_output: String = String::with_capacity(input.len() * 3 / 2); + pulldown_cmark::html::push_html(&mut html_output, parser.map(|(e, _r)| e)); + Ok(html_output) + } +} + +impl super::Formatter for MdFormatter { + fn render(&self, input: &str) -> Result<String, Error> { + Ok(input.replace("\r\n", "\n")) + } + + fn render_options_start(&self) -> &'static str { + "<dl>" + } + + fn render_options_end(&self) -> &'static str { + "</dl>" + } + + fn render_option(&self, params: &[&str], block: &str, man_name: &str) -> Result<String, Error> { + let mut result = String::new(); + fn unwrap_p(t: &str) -> &str { + unwrap(t, "<p>", "</p>") + } + + for param in params { + let rendered = self.render_html(param)?; + let no_p = unwrap_p(&rendered); + // split out first term to use as the id. + let first = no_p + .split_whitespace() + .next() + .ok_or_else(|| format_err!("did not expect option `{}` to be empty", param))?; + let no_tags = trim_tags(first); + if no_tags.is_empty() { + bail!("unexpected empty option with no tags `{}`", param); + } + let id = format!("option-{}-{}", man_name, no_tags); + write!( + result, + "<dt class=\"option-term\" id=\"{ID}\">\ + <a class=\"option-anchor\" href=\"#{ID}\"></a>{OPTION}</dt>\n", + ID = id, + OPTION = no_p + )?; + } + let rendered_block = self.render_html(block)?; + write!( + result, + "<dd class=\"option-desc\">{}</dd>\n", + unwrap_p(&rendered_block) + )?; + Ok(result) + } + + fn linkify_man_to_md(&self, name: &str, section: u8) -> Result<String, Error> { + let s = match self.man_map.get(&(name.to_string(), section)) { + Some(link) => format!("[{}({})]({})", name, section, link), + None => format!("[{}({})]({}.html)", name, section, name), + }; + Ok(s) + } +} + +fn trim_tags(s: &str) -> String { + // This is a hack. It removes all HTML tags. + let mut in_tag = false; + let mut in_char_ref = false; + s.chars() + .filter(|&ch| match ch { + '<' if in_tag => panic!("unexpected nested tag"), + '&' if in_char_ref => panic!("unexpected nested char ref"), + '<' => { + in_tag = true; + false + } + '&' => { + in_char_ref = true; + false + } + '>' if in_tag => { + in_tag = false; + false + } + ';' if in_char_ref => { + in_char_ref = false; + false + } + _ => !in_tag && !in_char_ref, + }) + .collect() +} diff --git a/crates/mdman/src/format/text.rs b/crates/mdman/src/format/text.rs new file mode 100644 index 0000000..ae07985 --- /dev/null +++ b/crates/mdman/src/format/text.rs @@ -0,0 +1,605 @@ +//! Text formatter. + +use crate::util::{header_text, unwrap}; +use crate::EventIter; +use anyhow::{bail, Error}; +use pulldown_cmark::{Alignment, Event, HeadingLevel, LinkType, Tag}; +use std::fmt::Write; +use std::mem; +use url::Url; + +pub struct TextFormatter { + url: Option<Url>, +} + +impl TextFormatter { + pub fn new(url: Option<Url>) -> TextFormatter { + TextFormatter { url } + } +} + +impl super::Formatter for TextFormatter { + fn render(&self, input: &str) -> Result<String, Error> { + TextRenderer::render(input, self.url.clone(), 0) + } + + fn render_options_start(&self) -> &'static str { + // Tell pulldown_cmark to ignore this. + // This will be stripped out later. + "<![CDATA[" + } + + fn render_options_end(&self) -> &'static str { + "]]>" + } + + fn render_option( + &self, + params: &[&str], + block: &str, + _man_name: &str, + ) -> Result<String, Error> { + let rendered_options = params + .iter() + .map(|param| TextRenderer::render(param, self.url.clone(), 0)) + .collect::<Result<Vec<_>, Error>>()?; + let trimmed: Vec<_> = rendered_options.iter().map(|o| o.trim()).collect(); + // Wrap in HTML tags, they will be stripped out during rendering. + Ok(format!( + "<dt>{}</dt>\n<dd>{}</dd>\n<br>\n", + trimmed.join(", "), + block + )) + } + + fn linkify_man_to_md(&self, name: &str, section: u8) -> Result<String, Error> { + Ok(format!("`{}`({})", name, section)) + } +} + +struct TextRenderer<'e> { + output: String, + indent: usize, + /// The current line being written. Once a line break is encountered (such + /// as starting a new paragraph), this will be written to `output` via + /// `flush`. + line: String, + /// The current word being written. Once a break is encountered (such as a + /// space) this will be written to `line` via `flush_word`. + word: String, + parser: EventIter<'e>, + /// The base URL used for relative URLs. + url: Option<Url>, + table: Table, +} + +impl<'e> TextRenderer<'e> { + fn render(input: &str, url: Option<Url>, indent: usize) -> Result<String, Error> { + let parser = crate::md_parser(input, url.clone()); + let output = String::with_capacity(input.len() * 3 / 2); + let mut mr = TextRenderer { + output, + indent, + line: String::new(), + word: String::new(), + parser, + url, + table: Table::new(), + }; + mr.push_md()?; + Ok(mr.output) + } + + fn push_md(&mut self) -> Result<(), Error> { + // If this is true, this is inside a cdata block used for hiding + // content from pulldown_cmark. + let mut in_cdata = false; + // The current list stack. None if unordered, Some if ordered with the + // given number as the current index. + let mut list: Vec<Option<u64>> = Vec::new(); + // Used in some cases where spacing isn't desired. + let mut suppress_paragraph = false; + // Whether or not word-wrapping is enabled. + let mut wrap_text = true; + + while let Some((event, range)) = self.parser.next() { + let this_suppress_paragraph = suppress_paragraph; + // Always reset suppression, even if the next event isn't a + // paragraph. This is in essence, a 1-token lookahead where the + // suppression is only enabled if the next event is a paragraph. + suppress_paragraph = false; + match event { + Event::Start(tag) => { + match tag { + Tag::Paragraph => { + if !this_suppress_paragraph { + self.flush(); + } + } + Tag::Heading(level, ..) => { + self.flush(); + if level == HeadingLevel::H1 { + let text = header_text(&mut self.parser)?; + self.push_to_line(&text.to_uppercase()); + self.hard_break(); + self.hard_break(); + } else if level == HeadingLevel::H2 { + let text = header_text(&mut self.parser)?; + self.push_to_line(&text.to_uppercase()); + self.flush(); + self.indent = 7; + } else { + let text = header_text(&mut self.parser)?; + self.push_indent((level as usize - 2) * 3); + self.push_to_line(&text); + self.flush(); + self.indent = (level as usize - 1) * 3 + 1; + } + } + Tag::BlockQuote => { + self.indent += 3; + } + Tag::CodeBlock(_kind) => { + self.flush(); + wrap_text = false; + self.indent += 4; + } + Tag::List(start) => list.push(start), + Tag::Item => { + self.flush(); + match list.last_mut().expect("item must have list start") { + // Ordered list. + Some(n) => { + self.push_indent(self.indent); + write!(self.line, "{}.", n)?; + *n += 1; + } + // Unordered list. + None => { + self.push_indent(self.indent); + self.push_to_line("o ") + } + } + self.indent += 3; + suppress_paragraph = true; + } + Tag::FootnoteDefinition(_label) => unimplemented!(), + Tag::Table(alignment) => { + assert!(self.table.alignment.is_empty()); + self.flush(); + self.table.alignment.extend(alignment); + let table = self.table.process(&mut self.parser, self.indent)?; + self.output.push_str(&table); + self.hard_break(); + self.table = Table::new(); + } + Tag::TableHead | Tag::TableRow | Tag::TableCell => { + bail!("unexpected table element") + } + Tag::Emphasis => {} + Tag::Strong => {} + // Strikethrough isn't usually supported for TTY. + Tag::Strikethrough => self.word.push_str("~~"), + Tag::Link(link_type, dest_url, _title) => { + if dest_url.starts_with('#') { + // In a man page, page-relative anchors don't + // have much meaning. + continue; + } + match link_type { + LinkType::Autolink | LinkType::Email => { + // The text is a copy of the URL, which is not needed. + match self.parser.next() { + Some((Event::Text(_), _range)) => {} + _ => bail!("expected text after autolink"), + } + } + LinkType::Inline + | LinkType::Reference + | LinkType::Collapsed + | LinkType::Shortcut => {} + // This is currently unused. This is only + // emitted with a broken link callback, but I + // felt it is too annoying to escape `[` in + // option descriptions. + LinkType::ReferenceUnknown + | LinkType::CollapsedUnknown + | LinkType::ShortcutUnknown => { + bail!( + "link with missing reference `{}` located at offset {}", + dest_url, + range.start + ); + } + } + } + Tag::Image(_link_type, _dest_url, _title) => { + bail!("images are not currently supported") + } + } + } + Event::End(tag) => match &tag { + Tag::Paragraph => { + self.flush(); + self.hard_break(); + } + Tag::Heading(..) => {} + Tag::BlockQuote => { + self.indent -= 3; + } + Tag::CodeBlock(_kind) => { + self.hard_break(); + wrap_text = true; + self.indent -= 4; + } + Tag::List(_) => { + list.pop(); + } + Tag::Item => { + self.flush(); + self.indent -= 3; + self.hard_break(); + } + Tag::FootnoteDefinition(_label) => {} + Tag::Table(_) => {} + Tag::TableHead => {} + Tag::TableRow => {} + Tag::TableCell => {} + Tag::Emphasis => {} + Tag::Strong => {} + Tag::Strikethrough => self.word.push_str("~~"), + Tag::Link(link_type, dest_url, _title) => { + if dest_url.starts_with('#') { + continue; + } + match link_type { + LinkType::Autolink | LinkType::Email => {} + LinkType::Inline + | LinkType::Reference + | LinkType::Collapsed + | LinkType::Shortcut => self.flush_word(), + _ => { + panic!("unexpected tag {:?}", tag); + } + } + self.flush_word(); + write!(self.word, "<{}>", dest_url)?; + } + Tag::Image(_link_type, _dest_url, _title) => {} + }, + Event::Text(t) | Event::Code(t) => { + if wrap_text { + let chunks = split_chunks(&t); + for chunk in chunks { + if chunk == " " { + self.flush_word(); + } else { + self.word.push_str(chunk); + } + } + } else { + for line in t.lines() { + self.push_indent(self.indent); + self.push_to_line(line); + self.flush(); + } + } + } + Event::Html(t) => { + if t.starts_with("<![CDATA[") { + // CDATA is a special marker used for handling options. + in_cdata = true; + self.flush(); + } else if in_cdata { + if t.trim().ends_with("]]>") { + in_cdata = false; + } else { + let trimmed = t.trim(); + if trimmed.is_empty() { + continue; + } + if trimmed == "<br>" { + self.hard_break(); + } else if trimmed.starts_with("<dt>") { + let opts = unwrap(trimmed, "<dt>", "</dt>"); + self.push_indent(self.indent); + self.push_to_line(opts); + self.flush(); + } else if trimmed.starts_with("<dd>") { + let mut def = String::new(); + while let Some((Event::Html(t), _range)) = self.parser.next() { + if t.starts_with("</dd>") { + break; + } + def.push_str(&t); + } + let rendered = + TextRenderer::render(&def, self.url.clone(), self.indent + 4)?; + self.push_to_line(rendered.trim_end()); + self.flush(); + } else { + self.push_to_line(&t); + self.flush(); + } + } + } else { + self.push_to_line(&t); + self.flush(); + } + } + Event::FootnoteReference(_t) => {} + Event::SoftBreak => self.flush_word(), + Event::HardBreak => self.flush(), + Event::Rule => { + self.flush(); + self.push_indent(self.indent); + self.push_to_line(&"_".repeat(79 - self.indent * 2)); + self.flush(); + } + Event::TaskListMarker(_b) => unimplemented!(), + } + } + Ok(()) + } + + fn flush(&mut self) { + self.flush_word(); + if !self.line.is_empty() { + self.output.push_str(&self.line); + self.output.push('\n'); + self.line.clear(); + } + } + + fn hard_break(&mut self) { + self.flush(); + if !self.output.ends_with("\n\n") { + self.output.push('\n'); + } + } + + fn flush_word(&mut self) { + if self.word.is_empty() { + return; + } + if self.line.len() + self.word.len() >= 79 { + self.output.push_str(&self.line); + self.output.push('\n'); + self.line.clear(); + } + if self.line.is_empty() { + self.push_indent(self.indent); + self.line.push_str(&self.word); + } else { + self.line.push(' '); + self.line.push_str(&self.word); + } + self.word.clear(); + } + + fn push_indent(&mut self, indent: usize) { + for _ in 0..indent { + self.line.push(' '); + } + } + + fn push_to_line(&mut self, text: &str) { + self.flush_word(); + self.line.push_str(text); + } +} + +/// Splits the text on whitespace. +/// +/// Consecutive whitespace is collapsed to a single ' ', and is included as a +/// separate element in the result. +fn split_chunks(text: &str) -> Vec<&str> { + let mut result = Vec::new(); + let mut start = 0; + while start < text.len() { + match text[start..].find(' ') { + Some(i) => { + if i != 0 { + result.push(&text[start..start + i]); + } + result.push(" "); + // Skip past whitespace. + match text[start + i..].find(|c| c != ' ') { + Some(n) => { + start = start + i + n; + } + None => { + break; + } + } + } + None => { + result.push(&text[start..]); + break; + } + } + } + result +} + +struct Table { + alignment: Vec<Alignment>, + rows: Vec<Vec<String>>, + row: Vec<String>, + cell: String, +} + +impl Table { + fn new() -> Table { + Table { + alignment: Vec::new(), + rows: Vec::new(), + row: Vec::new(), + cell: String::new(), + } + } + + /// Processes table events and generates a text table. + fn process(&mut self, parser: &mut EventIter<'_>, indent: usize) -> Result<String, Error> { + while let Some((event, _range)) = parser.next() { + match event { + Event::Start(tag) => match tag { + Tag::TableHead + | Tag::TableRow + | Tag::TableCell + | Tag::Emphasis + | Tag::Strong => {} + Tag::Strikethrough => self.cell.push_str("~~"), + // Links not yet supported, they usually won't fit. + Tag::Link(_, _, _) => {} + _ => bail!("unexpected tag in table: {:?}", tag), + }, + Event::End(tag) => match tag { + Tag::Table(_) => return self.render(indent), + Tag::TableCell => { + let cell = mem::replace(&mut self.cell, String::new()); + self.row.push(cell); + } + Tag::TableHead | Tag::TableRow => { + let row = mem::replace(&mut self.row, Vec::new()); + self.rows.push(row); + } + Tag::Strikethrough => self.cell.push_str("~~"), + _ => {} + }, + Event::Text(t) | Event::Code(t) => { + self.cell.push_str(&t); + } + Event::Html(t) => bail!("html unsupported in tables: {:?}", t), + _ => bail!("unexpected event in table: {:?}", event), + } + } + bail!("table end not reached"); + } + + fn render(&self, indent: usize) -> Result<String, Error> { + // This is an extremely primitive layout routine. + // First compute the potential maximum width of each cell. + // 2 for 1 space margin on left and right. + let width_acc = vec![2; self.alignment.len()]; + let mut col_widths = self + .rows + .iter() + .map(|row| row.iter().map(|cell| cell.len())) + .fold(width_acc, |mut acc, row| { + acc.iter_mut() + .zip(row) + // +3 for left/right margin and | symbol + .for_each(|(a, b)| *a = (*a).max(b + 3)); + acc + }); + // Shrink each column until it fits the total width, proportional to + // the columns total percent width. + let max_width = 78 - indent; + // Include total len for | characters, and +1 for final |. + let total_width = col_widths.iter().sum::<usize>() + col_widths.len() + 1; + if total_width > max_width { + let to_shrink = total_width - max_width; + // Compute percentage widths, and shrink each column based on its + // total percentage. + for width in &mut col_widths { + let percent = *width as f64 / total_width as f64; + *width -= (to_shrink as f64 * percent).ceil() as usize; + } + } + // Start rendering. + let mut result = String::new(); + + // Draw the horizontal line separating each row. + let mut row_line = String::new(); + row_line.push_str(&" ".repeat(indent)); + row_line.push('+'); + let lines = col_widths + .iter() + .map(|width| "-".repeat(*width)) + .collect::<Vec<_>>(); + row_line.push_str(&lines.join("+")); + row_line.push('+'); + row_line.push('\n'); + + // Draw top of the table. + result.push_str(&row_line); + // Draw each row. + for row in &self.rows { + // Word-wrap and fill each column as needed. + let filled = fill_row(row, &col_widths, &self.alignment); + // Need to transpose the cells across rows for cells that span + // multiple rows. + let height = filled.iter().map(|c| c.len()).max().unwrap(); + for row_i in 0..height { + result.push_str(&" ".repeat(indent)); + result.push('|'); + for filled_row in &filled { + let cell = &filled_row[row_i]; + result.push_str(cell); + result.push('|'); + } + result.push('\n'); + } + result.push_str(&row_line); + } + Ok(result) + } +} + +/// Formats a row, filling cells with spaces and word-wrapping text. +/// +/// Returns a vec of cells, where each cell is split into multiple lines. +fn fill_row(row: &[String], col_widths: &[usize], alignment: &[Alignment]) -> Vec<Vec<String>> { + let mut cell_lines = row + .iter() + .zip(col_widths) + .zip(alignment) + .map(|((cell, width), alignment)| fill_cell(cell, *width - 2, *alignment)) + .collect::<Vec<_>>(); + // Fill each cell to match the maximum vertical height of the tallest cell. + let max_lines = cell_lines.iter().map(|cell| cell.len()).max().unwrap(); + for (cell, width) in cell_lines.iter_mut().zip(col_widths) { + if cell.len() < max_lines { + cell.extend(std::iter::repeat(" ".repeat(*width)).take(max_lines - cell.len())); + } + } + cell_lines +} + +/// Formats a cell. Word-wraps based on width, and adjusts based on alignment. +/// +/// Returns a vec of lines for the cell. +fn fill_cell(text: &str, width: usize, alignment: Alignment) -> Vec<String> { + let fill_width = |text: &str| match alignment { + Alignment::None | Alignment::Left => format!(" {:<width$} ", text, width = width), + Alignment::Center => format!(" {:^width$} ", text, width = width), + Alignment::Right => format!(" {:>width$} ", text, width = width), + }; + if text.len() < width { + // No wrapping necessary, just format. + vec![fill_width(text)] + } else { + // Word-wrap the cell. + let mut result = Vec::new(); + let mut line = String::new(); + for word in text.split_whitespace() { + if line.len() + word.len() >= width { + // todo: word.len() > width + result.push(fill_width(&line)); + line.clear(); + } + if line.is_empty() { + line.push_str(word); + } else { + line.push(' '); + line.push_str(&word); + } + } + if !line.is_empty() { + result.push(fill_width(&line)); + } + + result + } +} diff --git a/crates/mdman/src/hbs.rs b/crates/mdman/src/hbs.rs new file mode 100644 index 0000000..81ad7ee --- /dev/null +++ b/crates/mdman/src/hbs.rs @@ -0,0 +1,215 @@ +//! Handlebars template processing. + +use crate::format::Formatter; +use anyhow::Error; +use handlebars::{ + handlebars_helper, Context, Decorator, Handlebars, Helper, HelperDef, HelperResult, Output, + RenderContext, RenderError, Renderable, +}; +use std::collections::HashMap; +use std::path::Path; + +type FormatterRef<'a> = &'a (dyn Formatter + Send + Sync); + +/// Processes the handlebars template at the given file. +pub fn expand(file: &Path, formatter: FormatterRef) -> Result<String, Error> { + let mut handlebars = Handlebars::new(); + handlebars.set_strict_mode(true); + handlebars.register_helper("lower", Box::new(lower)); + handlebars.register_helper("options", Box::new(OptionsHelper { formatter })); + handlebars.register_helper("option", Box::new(OptionHelper { formatter })); + handlebars.register_helper("man", Box::new(ManLinkHelper { formatter })); + handlebars.register_decorator("set", Box::new(set_decorator)); + handlebars.register_template_file("template", file)?; + let includes = file.parent().unwrap().join("includes"); + handlebars.register_templates_directory(".md", includes)?; + let man_name = file + .file_stem() + .expect("expected filename") + .to_str() + .expect("utf8 filename") + .to_string(); + let data = HashMap::from([("man_name", man_name)]); + let expanded = handlebars.render("template", &data)?; + Ok(expanded) +} + +/// Helper for `{{#options}}` block. +struct OptionsHelper<'a> { + formatter: FormatterRef<'a>, +} + +impl HelperDef for OptionsHelper<'_> { + fn call<'reg: 'rc, 'rc>( + &self, + h: &Helper<'reg, 'rc>, + r: &'reg Handlebars<'reg>, + ctx: &'rc Context, + rc: &mut RenderContext<'reg, 'rc>, + out: &mut dyn Output, + ) -> HelperResult { + if in_options(rc) { + return Err(RenderError::new("options blocks cannot be nested")); + } + // Prevent nested {{#options}}. + set_in_context(rc, "__MDMAN_IN_OPTIONS", serde_json::Value::Bool(true)); + let s = self.formatter.render_options_start(); + out.write(&s)?; + let t = match h.template() { + Some(t) => t, + None => return Err(RenderError::new("options block must not be empty")), + }; + let block = t.renders(r, ctx, rc)?; + out.write(&block)?; + + let s = self.formatter.render_options_end(); + out.write(&s)?; + remove_from_context(rc, "__MDMAN_IN_OPTIONS"); + Ok(()) + } +} + +/// Whether or not the context is currently inside a `{{#options}}` block. +fn in_options(rc: &RenderContext<'_, '_>) -> bool { + rc.context() + .map_or(false, |ctx| ctx.data().get("__MDMAN_IN_OPTIONS").is_some()) +} + +/// Helper for `{{#option}}` block. +struct OptionHelper<'a> { + formatter: FormatterRef<'a>, +} + +impl HelperDef for OptionHelper<'_> { + fn call<'reg: 'rc, 'rc>( + &self, + h: &Helper<'reg, 'rc>, + r: &'reg Handlebars<'reg>, + ctx: &'rc Context, + rc: &mut RenderContext<'reg, 'rc>, + out: &mut dyn Output, + ) -> HelperResult { + if !in_options(rc) { + return Err(RenderError::new("option must be in options block")); + } + let params = h.params(); + if params.is_empty() { + return Err(RenderError::new( + "option block must have at least one param", + )); + } + // Convert params to strings. + let params = params + .iter() + .map(|param| { + param + .value() + .as_str() + .ok_or_else(|| RenderError::new("option params must be strings")) + }) + .collect::<Result<Vec<&str>, RenderError>>()?; + let t = match h.template() { + Some(t) => t, + None => return Err(RenderError::new("option block must not be empty")), + }; + // Render the block. + let block = t.renders(r, ctx, rc)?; + + // Get the name of this page. + let man_name = ctx + .data() + .get("man_name") + .expect("expected man_name in context") + .as_str() + .expect("expect man_name str"); + + // Ask the formatter to convert this option to its format. + let option = self + .formatter + .render_option(¶ms, &block, man_name) + .map_err(|e| RenderError::new(format!("option render failed: {}", e)))?; + out.write(&option)?; + Ok(()) + } +} + +/// Helper for `{{man name section}}` expression. +struct ManLinkHelper<'a> { + formatter: FormatterRef<'a>, +} + +impl HelperDef for ManLinkHelper<'_> { + fn call<'reg: 'rc, 'rc>( + &self, + h: &Helper<'reg, 'rc>, + _r: &'reg Handlebars<'reg>, + _ctx: &'rc Context, + _rc: &mut RenderContext<'reg, 'rc>, + out: &mut dyn Output, + ) -> HelperResult { + let params = h.params(); + if params.len() != 2 { + return Err(RenderError::new("{{man}} must have two arguments")); + } + let name = params[0] + .value() + .as_str() + .ok_or_else(|| RenderError::new("man link name must be a string"))?; + let section = params[1] + .value() + .as_u64() + .ok_or_else(|| RenderError::new("man link section must be an integer"))?; + let section = + u8::try_from(section).map_err(|_e| RenderError::new("section number too large"))?; + let link = self + .formatter + .linkify_man_to_md(name, section) + .map_err(|e| RenderError::new(format!("failed to linkify man: {}", e)))?; + out.write(&link)?; + Ok(()) + } +} + +/// `{{*set var=value}}` decorator. +/// +/// This sets a variable to a value within the template context. +fn set_decorator( + d: &Decorator, + _: &Handlebars, + _ctx: &Context, + rc: &mut RenderContext, +) -> Result<(), RenderError> { + let data_to_set = d.hash(); + for (k, v) in data_to_set { + set_in_context(rc, k, v.value().clone()); + } + Ok(()) +} + +/// Sets a variable to a value within the context. +fn set_in_context(rc: &mut RenderContext, key: &str, value: serde_json::Value) { + let mut ctx = match rc.context() { + Some(c) => (*c).clone(), + None => Context::wraps(serde_json::Value::Object(serde_json::Map::new())).unwrap(), + }; + if let serde_json::Value::Object(m) = ctx.data_mut() { + m.insert(key.to_string(), value); + rc.set_context(ctx); + } else { + panic!("expected object in context"); + } +} + +/// Removes a variable from the context. +fn remove_from_context(rc: &mut RenderContext, key: &str) { + let ctx = rc.context().expect("cannot remove from null context"); + let mut ctx = (*ctx).clone(); + if let serde_json::Value::Object(m) = ctx.data_mut() { + m.remove(key); + rc.set_context(ctx); + } else { + panic!("expected object in context"); + } +} + +handlebars_helper!(lower: |s: str| s.to_lowercase()); diff --git a/crates/mdman/src/lib.rs b/crates/mdman/src/lib.rs new file mode 100644 index 0000000..01c3c8d --- /dev/null +++ b/crates/mdman/src/lib.rs @@ -0,0 +1,122 @@ +//! mdman markdown to man converter. + +use anyhow::{bail, Context, Error}; +use pulldown_cmark::{CowStr, Event, LinkType, Options, Parser, Tag}; +use std::collections::HashMap; +use std::fs; +use std::io::{self, BufRead}; +use std::ops::Range; +use std::path::Path; +use url::Url; + +mod format; +mod hbs; +mod util; + +use format::Formatter; + +/// Mapping of `(name, section)` of a man page to a URL. +pub type ManMap = HashMap<(String, u8), String>; + +/// A man section. +pub type Section = u8; + +/// The output formats supported by mdman. +#[derive(Copy, Clone)] +pub enum Format { + Man, + Md, + Text, +} + +impl Format { + /// The filename extension for the format. + pub fn extension(&self, section: Section) -> String { + match self { + Format::Man => section.to_string(), + Format::Md => "md".to_string(), + Format::Text => "txt".to_string(), + } + } +} + +/// Converts the handlebars markdown file at the given path into the given +/// format, returning the translated result. +pub fn convert( + file: &Path, + format: Format, + url: Option<Url>, + man_map: ManMap, +) -> Result<String, Error> { + let formatter: Box<dyn Formatter + Send + Sync> = match format { + Format::Man => Box::new(format::man::ManFormatter::new(url)), + Format::Md => Box::new(format::md::MdFormatter::new(man_map)), + Format::Text => Box::new(format::text::TextFormatter::new(url)), + }; + let expanded = hbs::expand(file, &*formatter)?; + // pulldown-cmark can behave a little differently with Windows newlines, + // just normalize it. + let expanded = expanded.replace("\r\n", "\n"); + formatter.render(&expanded) +} + +/// Pulldown-cmark iterator yielding an `(event, range)` tuple. +type EventIter<'a> = Box<dyn Iterator<Item = (Event<'a>, Range<usize>)> + 'a>; + +/// Creates a new markdown parser with the given input. +pub(crate) fn md_parser(input: &str, url: Option<Url>) -> EventIter { + let mut options = Options::empty(); + options.insert(Options::ENABLE_TABLES); + options.insert(Options::ENABLE_FOOTNOTES); + options.insert(Options::ENABLE_STRIKETHROUGH); + options.insert(Options::ENABLE_SMART_PUNCTUATION); + let parser = Parser::new_ext(input, options); + let parser = parser.into_offset_iter(); + // Translate all links to include the base url. + let parser = parser.map(move |(event, range)| match event { + Event::Start(Tag::Link(lt, dest_url, title)) if !matches!(lt, LinkType::Email) => ( + Event::Start(Tag::Link(lt, join_url(url.as_ref(), dest_url), title)), + range, + ), + Event::End(Tag::Link(lt, dest_url, title)) if !matches!(lt, LinkType::Email) => ( + Event::End(Tag::Link(lt, join_url(url.as_ref(), dest_url), title)), + range, + ), + _ => (event, range), + }); + Box::new(parser) +} + +fn join_url<'a>(base: Option<&Url>, dest: CowStr<'a>) -> CowStr<'a> { + match base { + Some(base_url) => { + // Absolute URL or page-relative anchor doesn't need to be translated. + if dest.contains(':') || dest.starts_with('#') { + dest + } else { + let joined = base_url.join(&dest).unwrap_or_else(|e| { + panic!("failed to join URL `{}` to `{}`: {}", dest, base_url, e) + }); + String::from(joined).into() + } + } + None => dest, + } +} + +pub fn extract_section(file: &Path) -> Result<Section, Error> { + let f = fs::File::open(file).with_context(|| format!("could not open `{}`", file.display()))?; + let mut f = io::BufReader::new(f); + let mut line = String::new(); + f.read_line(&mut line)?; + if !line.starts_with("# ") { + bail!("expected input file to start with # header"); + } + let (_name, section) = util::parse_name_and_section(&line[2..].trim()).with_context(|| { + format!( + "expected input file to have header with the format `# command-name(1)`, found: `{}`", + line + ) + })?; + Ok(section) +} diff --git a/crates/mdman/src/main.rs b/crates/mdman/src/main.rs new file mode 100644 index 0000000..2bdf96d --- /dev/null +++ b/crates/mdman/src/main.rs @@ -0,0 +1,133 @@ +use anyhow::{bail, format_err, Context, Error}; +use mdman::{Format, ManMap}; +use std::collections::HashMap; +use std::path::{Path, PathBuf}; +use url::Url; + +/// Command-line options. +struct Options { + format: Format, + output_dir: PathBuf, + sources: Vec<PathBuf>, + url: Option<Url>, + man_map: ManMap, +} + +fn main() { + if let Err(e) = run() { + eprintln!("error: {}", e); + for cause in e.chain().skip(1) { + eprintln!("\nCaused by:"); + for line in cause.to_string().lines() { + if line.is_empty() { + eprintln!(); + } else { + eprintln!(" {}", line); + } + } + } + std::process::exit(1); + } +} + +fn run() -> Result<(), Error> { + let opts = process_args()?; + if !opts.output_dir.exists() { + std::fs::create_dir_all(&opts.output_dir).with_context(|| { + format!( + "failed to create output directory {}", + opts.output_dir.display() + ) + })?; + } + for source in &opts.sources { + let section = mdman::extract_section(source)?; + let filename = + Path::new(source.file_name().unwrap()).with_extension(opts.format.extension(section)); + let out_path = opts.output_dir.join(filename); + if same_file::is_same_file(source, &out_path).unwrap_or(false) { + bail!("cannot output to the same file as the source"); + } + println!("Converting {} -> {}", source.display(), out_path.display()); + let result = mdman::convert(&source, opts.format, opts.url.clone(), opts.man_map.clone()) + .with_context(|| format!("failed to translate {}", source.display()))?; + + std::fs::write(out_path, result)?; + } + Ok(()) +} + +fn process_args() -> Result<Options, Error> { + let mut format = None; + let mut output = None; + let mut url = None; + let mut man_map: ManMap = HashMap::new(); + let mut sources = Vec::new(); + let mut args = std::env::args().skip(1); + while let Some(arg) = args.next() { + match arg.as_str() { + "-t" => { + format = match args.next().as_deref() { + Some("man") => Some(Format::Man), + Some("md") => Some(Format::Md), + Some("txt") => Some(Format::Text), + Some(s) => bail!("unknown output format: {}", s), + None => bail!("-t requires a value (man, md, txt)"), + }; + } + "-o" => { + output = match args.next() { + Some(s) => Some(PathBuf::from(s)), + None => bail!("-o requires a value"), + }; + } + "--url" => { + url = match args.next() { + Some(s) => { + let url = Url::parse(&s) + .with_context(|| format!("could not convert `{}` to a url", s))?; + if !url.path().ends_with('/') { + bail!("url `{}` should end with a /", url); + } + Some(url) + } + None => bail!("--url requires a value"), + } + } + "--man" => { + let man = args + .next() + .ok_or_else(|| format_err!("--man requires a value"))?; + let parts: Vec<_> = man.splitn(2, '=').collect(); + let key_parts: Vec<_> = parts[0].splitn(2, ':').collect(); + if parts.len() != 2 || key_parts.len() != 2 { + bail!("--man expected value with form name:1=link"); + } + let section: u8 = key_parts[1].parse().with_context(|| { + format!("expected unsigned integer for section, got `{}`", parts[1]) + })?; + man_map.insert((key_parts[0].to_string(), section), parts[1].to_string()); + } + s => { + sources.push(PathBuf::from(s)); + } + } + } + if format.is_none() { + bail!("-t must be specified (man, md, txt)"); + } + if output.is_none() { + bail!("-o must be specified (output directory)"); + } + if sources.is_empty() { + bail!("at least one source must be specified"); + } + let opts = Options { + format: format.unwrap(), + output_dir: output.unwrap(), + sources, + url, + man_map, + }; + Ok(opts) +} diff --git a/crates/mdman/src/util.rs b/crates/mdman/src/util.rs new file mode 100644 index 0000000..a4c71ad --- /dev/null +++ b/crates/mdman/src/util.rs @@ -0,0 +1,44 @@ +///! General utilities. +use crate::EventIter; +use anyhow::{bail, format_err, Context, Error}; +use pulldown_cmark::{CowStr, Event, Tag}; + +/// Splits the text `foo(1)` into "foo" and `1`. +pub fn parse_name_and_section(text: &str) -> Result<(&str, u8), Error> { + let mut i = text.split_terminator(&['(', ')'][..]); + let name = i + .next() + .ok_or_else(|| format_err!("man reference must have a name"))?; + let section = i + .next() + .ok_or_else(|| format_err!("man reference must have a section such as mycommand(1)"))?; + if let Some(s) = i.next() { + bail!( + "man reference must have the form mycommand(1), got extra part `{}`", + s + ); + } + let section: u8 = section + .parse() + .with_context(|| format!("section must be a number, got {}", section))?; + Ok((name, section)) +} + +/// Extracts the text from a header after Tag::Heading has been received. +pub fn header_text<'e>(parser: &mut EventIter<'e>) -> Result<CowStr<'e>, Error> { + let text = match parser.next() { + Some((Event::Text(t), _range)) => t, + e => bail!("expected plain text in man header, got {:?}", e), + }; + match parser.next() { + Some((Event::End(Tag::Heading(..)), _range)) => { + return Ok(text); + } + e => bail!("expected plain text in man header, got {:?}", e), + } +} + +/// Removes tags from the front and back of a string. +pub fn unwrap<'t>(text: &'t str, front: &str, back: &str) -> &'t str { + text.trim().trim_start_matches(front).trim_end_matches(back) +} diff --git a/crates/mdman/tests/compare.rs b/crates/mdman/tests/compare.rs new file mode 100644 index 0000000..3e679d1 --- /dev/null +++ b/crates/mdman/tests/compare.rs @@ -0,0 +1,48 @@ +//! Compares input to expected output. +//! +//! Use the MDMAN_BLESS environment variable to automatically update the +//! expected output. + +use mdman::{Format, ManMap}; +use pretty_assertions::assert_eq; +use std::path::PathBuf; +use url::Url; + +fn run(name: &str) { + let input = PathBuf::from(format!("tests/compare/{}.md", name)); + let url = Some(Url::parse("https://example.org/").unwrap()); + let mut map = ManMap::new(); + map.insert( + ("other-cmd".to_string(), 1), + "https://example.org/commands/other-cmd.html".to_string(), + ); + + for &format in &[Format::Man, Format::Md, Format::Text] { + let section = mdman::extract_section(&input).unwrap(); + let result = mdman::convert(&input, format, url.clone(), map.clone()).unwrap(); + let expected_path = format!( + "tests/compare/expected/{}.{}", + name, + format.extension(section) + ); + if std::env::var("MDMAN_BLESS").is_ok() { + std::fs::write(&expected_path, result).unwrap(); + } else { + let expected = std::fs::read_to_string(&expected_path).unwrap(); + // Fix if Windows checked out with autocrlf. + let expected = expected.replace("\r\n", "\n"); + assert_eq!(expected, result); + } + } +} + +macro_rules! test( ($name:ident) => ( + #[test] + fn $name() { run(stringify!($name)); } +) ); + +test!(formatting); +test!(links); +test!(options); +test!(tables); +test!(vars); diff --git a/crates/mdman/tests/compare/expected/formatting.1 b/crates/mdman/tests/compare/expected/formatting.1 new file mode 100644 index 0000000..840734c --- /dev/null +++ b/crates/mdman/tests/compare/expected/formatting.1 @@ -0,0 +1,118 @@ +'\" t +.TH "FORMATTING" "1" +.nh +.ad l +.ss \n[.ss] 0 +.sp +This is \fBnested \f(BIformatting\fB \fBtext\fB\fR\&. +.SH "SECOND HEADING" +Some text at second level. +.SS "Third heading" +Some text at third level. +.SS "Fourth heading" +Some text at fourth level. +.SH "Quotes and blocks." +Here are some quotes and blocks. +.RS 3 +.ll -5 +.sp +This is a block quote. Ambidextrously koala apart that prudent blindly alas +far amid dear goodness turgid so exact inside oh and alas much fanciful that +dark on spoon\-fed adequately insolent walking crud. +.br +.RE +.ll +.sp +.RS 4 +.nf +This is a code block. Groundhog watchfully sudden firefly some self\-consciously hotly jeepers satanic after that this parrot this at virtuous +some mocking the leaned jeez nightingale as much mallard so because jeez +turned dear crud grizzly strenuously. + + Indented and should be unmodified. +.fi +.RE +.sp +.RS 4 +.nf +This is an indented code block. Egregiously yikes animatedly since outside beseechingly a badger hey shakily giraffe a one wow one this +goodness regarding reindeer so astride before. + + Doubly indented +.fi +.RE +.SH "Lists" +.sp +.RS 4 +\h'-04' 1.\h'+01'Ordered list +.sp +.RS 4 +\h'-04'\(bu\h'+02'Unordered list +.sp +With a second paragraph inside it +.sp +.RS 4 +\h'-04' 1.\h'+01'Inner ordered list +.RE +.sp +.RS 4 +\h'-04' 2.\h'+01'Another +.RE +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Eggs +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Milk +.sp +.RS 4 +\h'-04' 5.\h'+01'Don\[cq]t start at one. +.RE +.sp +.RS 4 +\h'-04' 6.\h'+01'tamarind +.RE +.RE +.RE +.sp +.RS 4 +\h'-04' 2.\h'+01'Second element +.RE +.sp +.RS 4 +\h'-04' 3.\h'+01'Third element +.RE +.SH "Breaks" +This has a +.br +hard break in it +and a soft one. +.SH "Horizontal rule" +This should contain a line: +\l'\n(.lu' +.sp +Nice! +.SH "Strange characters" +Handles escaping for characters +.sp +\&.dot at the start of a line. +.sp +\(rsfBnot really troff +.sp +Various characters \(rs \- \[en] \[em] \- | | ` +.sp +.RS 4 +.nf +tree +`\-\- example + |\-\- salamander + | |\-\- honey + | `\-\- some + |\-\- fancifully + `\-\- trout +.fi +.RE +.sp +\ \ \ \ non\-breaking space. diff --git a/crates/mdman/tests/compare/expected/formatting.md b/crates/mdman/tests/compare/expected/formatting.md new file mode 100644 index 0000000..3b9f5b8 --- /dev/null +++ b/crates/mdman/tests/compare/expected/formatting.md @@ -0,0 +1,95 @@ +# formatting(1) + +This is **nested _formatting_ `text`**. + +## SECOND HEADING + +Some text at second level. + +### Third heading + +Some text at third level. + +#### Fourth heading + +Some text at fourth level. + +## Quotes and blocks. + +Here are some quotes and blocks. + +> This is a block quote. Ambidextrously koala apart that prudent blindly alas +> far amid dear goodness turgid so exact inside oh and alas much fanciful that +> dark on spoon-fed adequately insolent walking crud. + +``` +This is a code block. Groundhog watchfully sudden firefly some self-consciously hotly jeepers satanic after that this parrot this at virtuous +some mocking the leaned jeez nightingale as much mallard so because jeez +turned dear crud grizzly strenuously. + + Indented and should be unmodified. +``` + + This is an indented code block. Egregiously yikes animatedly since outside beseechingly a badger hey shakily giraffe a one wow one this + goodness regarding reindeer so astride before. + + Doubly indented + +## Lists + +1. Ordered list + + * Unordered list + + With a second paragraph inside it + + 1. Inner ordered list + + 1. Another + + * Eggs + + * Milk + + 5. Don't start at one. + 6. tamarind + +1. Second element + +1. Third element + +## Breaks + +This has a\ +hard break in it +and a soft one. + +## Horizontal rule + +This should contain a line: + +--- + +Nice! + +## Strange characters + +Handles escaping for characters + +.dot at the start of a line. + +\fBnot really troff + +Various characters \ - – — ─ │ ├ └ + +``` +tree +└── example + ├── salamander + │ ├── honey + │ └── some + ├── fancifully + └── trout +``` + + non-breaking space. diff --git a/crates/mdman/tests/compare/expected/formatting.txt b/crates/mdman/tests/compare/expected/formatting.txt new file mode 100644 index 0000000..b5258c4 --- /dev/null +++ b/crates/mdman/tests/compare/expected/formatting.txt @@ -0,0 +1,84 @@ +FORMATTING(1) + +This is nested formatting text. + +SECOND HEADING + Some text at second level. + + Third heading + Some text at third level. + + Fourth heading + Some text at fourth level. + +QUOTES AND BLOCKS. + Here are some quotes and blocks. + + This is a block quote. Ambidextrously koala apart that prudent + blindly alas far amid dear goodness turgid so exact inside oh and + alas much fanciful that dark on spoon-fed adequately insolent walking + crud. + + This is a code block. Groundhog watchfully sudden firefly some self-consciously hotly jeepers satanic after that this parrot this at virtuous + some mocking the leaned jeez nightingale as much mallard so because jeez + turned dear crud grizzly strenuously. + + Indented and should be unmodified. + + This is an indented code block. Egregiously yikes animatedly since outside beseechingly a badger hey shakily giraffe a one wow one this + goodness regarding reindeer so astride before. + + Doubly indented + +LISTS + 1. Ordered list + + o Unordered list + + With a second paragraph inside it + + 1. Inner ordered list + + 2. Another + + o Eggs + + o Milk + + 5. Don’t start at one. + + 6. tamarind + + 2. Second element + + 3. Third element + +BREAKS + This has a + hard break in it and a soft one. + +HORIZONTAL RULE + This should contain a line: + + _________________________________________________________________ + Nice! + +STRANGE CHARACTERS + Handles escaping for characters + + .dot at the start of a line. + + \fBnot really troff + + Various characters \ - – — ─ │ ├ └ + + tree + └── example + ├── salamander + │ ├── honey + │ └── some + ├── fancifully + └── trout + + non-breaking space. + diff --git a/crates/mdman/tests/compare/expected/links.1 b/crates/mdman/tests/compare/expected/links.1 new file mode 100644 index 0000000..e56cef7 --- /dev/null +++ b/crates/mdman/tests/compare/expected/links.1 @@ -0,0 +1,45 @@ +'\" t +.TH "LINKS" "1" +.nh +.ad l +.ss \n[.ss] 0 +.SH "NAME" +links \- Test of different link kinds +.SH "DESCRIPTION" +Inline link: \fIinline link\fR <https://example.com/inline> +.sp +Reference link: \fIthis is a link\fR <https://example.com/bar> +.sp +Collapsed: \fIcollapsed\fR <https://example.com/collapsed> +.sp +Shortcut: \fIshortcut\fR <https://example.com/shortcut> +.sp +Autolink: <https://example.com/auto> +.sp +Email: <foo@example.com> +.sp +Relative link: \fIrelative link\fR <https://example.org/foo/bar.html> +.sp +Collapsed unknown: [collapsed unknown][] +.sp +Reference unknown: [foo][unknown] +.sp +Shortcut unknown: [shortcut unknown] +.sp +\fBother\-cmd\fR(1) +.sp +\fBlocal\-cmd\fR(1) +.sp +\fISome link\fR <https://example.org/foo.html> +.sp +\fB\-\-include\fR +.RS 4 +Testing an \fIincluded link\fR <https://example.org/included_link.html>\&. +.RE +.SH "OPTIONS" +.sp +\fB\-\-foo\-bar\fR +.RS 4 +Example \fIlink\fR <https://example.org/bar.html>\&. +See \fBother\-cmd\fR(1), \fBlocal\-cmd\fR(1) +.RE diff --git a/crates/mdman/tests/compare/expected/links.md b/crates/mdman/tests/compare/expected/links.md new file mode 100644 index 0000000..11afcf3 --- /dev/null +++ b/crates/mdman/tests/compare/expected/links.md @@ -0,0 +1,56 @@ +# links(1) + +## NAME + +links - Test of different link kinds + +## DESCRIPTION + +Inline link: [inline link](https://example.com/inline) + +Reference link: [this is a link][bar] + +Collapsed: [collapsed][] + +Shortcut: [shortcut] + +Autolink: <https://example.com/auto> + +Email: <foo@example.com> + +Relative link: [relative link](foo/bar.html) + +Collapsed unknown: [collapsed unknown][] + +Reference unknown: [foo][unknown] + +Shortcut unknown: [shortcut unknown] + +[other-cmd(1)](https://example.org/commands/other-cmd.html) + +[local-cmd(1)](local-cmd.html) + +[Some link](foo.html) + +<dl> +<dt class="option-term" id="option-links---include"><a class="option-anchor" href="#option-links---include"></a><code>--include</code></dt> +<dd class="option-desc">Testing an <a href="included_link.html">included link</a>.</dd> + +</dl> + + +## OPTIONS + +<dl> + +<dt class="option-term" id="option-links---foo-bar"><a class="option-anchor" href="#option-links---foo-bar"></a><code>--foo-bar</code></dt> +<dd class="option-desc">Example <a href="bar.html">link</a>. +See <a href="https://example.org/commands/other-cmd.html">other-cmd(1)</a>, <a href="local-cmd.html">local-cmd(1)</a></dd> + + +</dl> + + +[bar]: https://example.com/bar +[collapsed]: https://example.com/collapsed +[shortcut]: https://example.com/shortcut diff --git a/crates/mdman/tests/compare/expected/links.txt b/crates/mdman/tests/compare/expected/links.txt new file mode 100644 index 0000000..7748c3d --- /dev/null +++ b/crates/mdman/tests/compare/expected/links.txt @@ -0,0 +1,40 @@ +LINKS(1) + +NAME + links - Test of different link kinds + +DESCRIPTION + Inline link: inline link <https://example.com/inline> + + Reference link: this is a link <https://example.com/bar> + + Collapsed: collapsed <https://example.com/collapsed> + + Shortcut: shortcut <https://example.com/shortcut> + + Autolink: <https://example.com/auto> + + Email: <foo@example.com> + + Relative link: relative link <https://example.org/foo/bar.html> + + Collapsed unknown: [collapsed unknown][] + + Reference unknown: [foo][unknown] + + Shortcut unknown: [shortcut unknown] + + other-cmd(1) + + local-cmd(1) + + Some link <https://example.org/foo.html> + + --include + Testing an included link <https://example.org/included_link.html>. + +OPTIONS + --foo-bar + Example link <https://example.org/bar.html>. See other-cmd(1), + local-cmd(1) + diff --git a/crates/mdman/tests/compare/expected/options.1 b/crates/mdman/tests/compare/expected/options.1 new file mode 100644 index 0000000..d362421 --- /dev/null +++ b/crates/mdman/tests/compare/expected/options.1 @@ -0,0 +1,94 @@ +'\" t +.TH "MY\-COMMAND" "1" +.nh +.ad l +.ss \n[.ss] 0 +.SH "NAME" +my\-command \- A brief description +.SH "SYNOPSIS" +\fBmy\-command\fR [\fB\-\-abc\fR | \fB\-\-xyz\fR] \fIname\fR +.br +\fBmy\-command\fR [\fB\-f\fR \fIfile\fR] +.br +\fBmy\-command\fR (\fB\-m\fR | \fB\-M\fR) [\fIoldbranch\fR] \fInewbranch\fR +.br +\fBmy\-command\fR (\fB\-d\fR | \fB\-D\fR) [\fB\-r\fR] \fIbranchname\fR\[u2026] +.SH "DESCRIPTION" +A description of the command. +.sp +.RS 4 +\h'-04'\(bu\h'+02'One +.sp +.RS 4 +\h'-04'\(bu\h'+02'Sub one +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Sub two +.RE +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Two +.RE +.sp +.RS 4 +\h'-04'\(bu\h'+02'Three +.RE +.SH "OPTIONS" +.SS "Command options" +.sp +\fB\-\-foo\-bar\fR +.RS 4 +Demo \fIemphasis\fR, \fBstrong\fR, ~~strike~~ +.RE +.sp +\fB\-p\fR \fIspec\fR, +\fB\-\-package\fR \fIspec\fR +.RS 4 +This has multiple flags. +.RE +.sp +\fInamed\-arg\[u2026]\fR +.RS 4 +A named argument. +.RE +.SS "Common Options" +.sp +\fB@\fR\fIfilename\fR +.RS 4 +Load from filename. +.RE +.sp +\fB\-\-foo\fR [\fIbar\fR] +.RS 4 +Flag with optional value. +.RE +.sp +\fB\-\-foo\fR[\fB=\fR\fIbar\fR] +.RS 4 +Alternate syntax for optional value (with required = for disambiguation). +.RE +.SH "EXAMPLES" +.sp +.RS 4 +\h'-04' 1.\h'+01'An example +.sp +.RS 4 +.nf +my\-command \-\-abc +.fi +.RE +.RE +.sp +.RS 4 +\h'-04' 2.\h'+01'Another example +.sp +.RS 4 +.nf +my\-command \-\-xyz +.fi +.RE +.RE +.SH "SEE ALSO" +\fBother\-command\fR(1) \fBabc\fR(7) diff --git a/crates/mdman/tests/compare/expected/options.md b/crates/mdman/tests/compare/expected/options.md new file mode 100644 index 0000000..19b0b44 --- /dev/null +++ b/crates/mdman/tests/compare/expected/options.md @@ -0,0 +1,77 @@ +# my-command(1) + +## NAME + +my-command - A brief description + +## SYNOPSIS + +`my-command` [`--abc` | `--xyz`] _name_\ +`my-command` [`-f` _file_]\ +`my-command` (`-m` | `-M`) [_oldbranch_] _newbranch_\ +`my-command` (`-d` | `-D`) [`-r`] _branchname_... + +## DESCRIPTION + +A description of the command. + +* One + * Sub one + * Sub two +* Two +* Three + + +## OPTIONS + +### Command options + +<dl> + +<dt class="option-term" id="option-options---foo-bar"><a class="option-anchor" href="#option-options---foo-bar"></a><code>--foo-bar</code></dt> +<dd class="option-desc">Demo <em>emphasis</em>, <strong>strong</strong>, <del>strike</del></dd> + + +<dt class="option-term" id="option-options--p"><a class="option-anchor" href="#option-options--p"></a><code>-p</code> <em>spec</em></dt> +<dt class="option-term" id="option-options---package"><a class="option-anchor" href="#option-options---package"></a><code>--package</code> <em>spec</em></dt> +<dd class="option-desc">This has multiple flags.</dd> + + +<dt class="option-term" id="option-options-named-arg…"><a class="option-anchor" href="#option-options-named-arg…"></a><em>named-arg…</em></dt> +<dd class="option-desc">A named argument.</dd> + + +</dl> + +### Common Options + +<dl> +<dt class="option-term" id="option-options-@filename"><a class="option-anchor" href="#option-options-@filename"></a><code>@</code><em>filename</em></dt> +<dd class="option-desc">Load from filename.</dd> + + +<dt class="option-term" id="option-options---foo"><a class="option-anchor" href="#option-options---foo"></a><code>--foo</code> [<em>bar</em>]</dt> +<dd class="option-desc">Flag with optional value.</dd> + + +<dt class="option-term" id="option-options---foo[=bar]"><a class="option-anchor" href="#option-options---foo[=bar]"></a><code>--foo</code>[<code>=</code><em>bar</em>]</dt> +<dd class="option-desc">Alternate syntax for optional value (with required = for disambiguation).</dd> + + +</dl> + + +## EXAMPLES + +1. An example + + ``` + my-command --abc + ``` + +1. Another example + + my-command --xyz + +## SEE ALSO +[other-command(1)](other-command.html) [abc(7)](abc.html) diff --git a/crates/mdman/tests/compare/expected/options.txt b/crates/mdman/tests/compare/expected/options.txt new file mode 100644 index 0000000..9bfdec6 --- /dev/null +++ b/crates/mdman/tests/compare/expected/options.txt @@ -0,0 +1,57 @@ +MY-COMMAND(1) + +NAME + my-command - A brief description + +SYNOPSIS + my-command [--abc | --xyz] name + my-command [-f file] + my-command (-m | -M) [oldbranch] newbranch + my-command (-d | -D) [-r] branchname… + +DESCRIPTION + A description of the command. + + o One + o Sub one + + o Sub two + + o Two + + o Three + +OPTIONS + Command options + --foo-bar + Demo emphasis, strong, ~~strike~~ + + -p spec, --package spec + This has multiple flags. + + named-arg… + A named argument. + + Common Options + @filename + Load from filename. + + --foo [bar] + Flag with optional value. + + --foo[=bar] + Alternate syntax for optional value (with required = for + disambiguation). + +EXAMPLES + 1. An example + + my-command --abc + + 2. Another example + + my-command --xyz + +SEE ALSO + other-command(1) abc(7) + diff --git a/crates/mdman/tests/compare/expected/tables.1 b/crates/mdman/tests/compare/expected/tables.1 new file mode 100644 index 0000000..7175a3e --- /dev/null +++ b/crates/mdman/tests/compare/expected/tables.1 @@ -0,0 +1,108 @@ +'\" t +.TH "TABLES" "1" +.nh +.ad l +.ss \n[.ss] 0 +.SH "DESCRIPTION" +Testing tables. + +.TS +allbox tab(:); +lt. +T{ +Single col +T} +T{ +Hi! :) +T} +.TE +.sp + +.TS +allbox tab(:); +lt lt lt. +T{ +Header content +T}:T{ +With \fBformat\fR \fItext\fR +T}:T{ +Another column +T} +T{ +Some data +T}:T{ +More data +T}:T{ + +T} +T{ +Extra long amount of text within a column +T}:T{ +hi +T}:T{ +there +T} +.TE +.sp + +.TS +allbox tab(:); +lt ct rt. +T{ +Left aligned +T}:T{ +Center aligned +T}:T{ +Right aligned +T} +T{ +abc +T}:T{ +def +T}:T{ +ghi +T} +.TE +.sp + +.TS +allbox tab(:); +lt ct rt. +T{ +Left aligned +T}:T{ +Center aligned +T}:T{ +Right aligned +T} +T{ +X +T}:T{ +X +T}:T{ +X +T} +T{ +Extra long text 123456789012 with mixed widths. +T}:T{ +Extra long text 123456789012 with mixed widths. +T}:T{ +Extra long text 123456789012 with mixed widths. +T} +.TE +.sp + +.TS +allbox tab(:); +lt. +T{ +Link check +T} +T{ +\fIfoo\fR <https://example.com/> +T} +T{ +<https://example.com/> +T} +.TE +.sp diff --git a/crates/mdman/tests/compare/expected/tables.md b/crates/mdman/tests/compare/expected/tables.md new file mode 100644 index 0000000..831132c --- /dev/null +++ b/crates/mdman/tests/compare/expected/tables.md @@ -0,0 +1,35 @@ +# tables(1) + +## DESCRIPTION + +Testing tables. + +| Single col | +-------------- +| Hi! :) | + + +Header content | With `format` *text* | Another column +---------------|----------------------|---------------- +Some data | More data | +Extra long amount of text within a column | hi | there + + +Left aligned | Center aligned | Right aligned +-------------|:--------------:|--------------: +abc | def | ghi + + +Left aligned | Center aligned | Right aligned +-------------|:--------------:|--------------: +X | X | X +Extra long text 123456789012 with mixed widths. | Extra long text 123456789012 with mixed widths. | Extra long text 123456789012 with mixed widths. + + +| Link check | +-------------- +| [foo] | +| <https://example.com/> | + + +[foo]: https://example.com/ diff --git a/crates/mdman/tests/compare/expected/tables.txt b/crates/mdman/tests/compare/expected/tables.txt new file mode 100644 index 0000000..fed53f9 --- /dev/null +++ b/crates/mdman/tests/compare/expected/tables.txt @@ -0,0 +1,45 @@ +TABLES(1) + +DESCRIPTION + Testing tables. + + +-------------+ + | Single col | + +-------------+ + | Hi! :) | + +-------------+ + + +-------------------------------------+----------------+--------------+ + | Header content | With format | Another | + | | text | column | + +-------------------------------------+----------------+--------------+ + | Some data | More data | | + +-------------------------------------+----------------+--------------+ + | Extra long amount of text within a | hi | there | + | column | | | + +-------------------------------------+----------------+--------------+ + + +---------------+-----------------+----------------+ + | Left aligned | Center aligned | Right aligned | + +---------------+-----------------+----------------+ + | abc | def | ghi | + +---------------+-----------------+----------------+ + + +-----------------------+-----------------------+-----------------------+ + | Left aligned | Center aligned | Right aligned | + +-----------------------+-----------------------+-----------------------+ + | X | X | X | + +-----------------------+-----------------------+-----------------------+ + | Extra long text | Extra long text | Extra long text | + | 123456789012 with | 123456789012 with | 123456789012 with | + | mixed widths. | mixed widths. | mixed widths. | + +-----------------------+-----------------------+-----------------------+ + + +-----------------------+ + | Link check | + +-----------------------+ + | foo | + +-----------------------+ + | https://example.com/ | + +-----------------------+ + diff --git a/crates/mdman/tests/compare/expected/vars.7 b/crates/mdman/tests/compare/expected/vars.7 new file mode 100644 index 0000000..0ee33ad --- /dev/null +++ b/crates/mdman/tests/compare/expected/vars.7 @@ -0,0 +1,9 @@ +'\" t +.TH "VARS" "7" +.nh +.ad l +.ss \n[.ss] 0 +.sp +Bar +.sp +bar diff --git a/crates/mdman/tests/compare/expected/vars.md b/crates/mdman/tests/compare/expected/vars.md new file mode 100644 index 0000000..2493aca --- /dev/null +++ b/crates/mdman/tests/compare/expected/vars.md @@ -0,0 +1,7 @@ +# vars(7) + + + +Bar + +bar diff --git a/crates/mdman/tests/compare/expected/vars.txt b/crates/mdman/tests/compare/expected/vars.txt new file mode 100644 index 0000000..11d34ca --- /dev/null +++ b/crates/mdman/tests/compare/expected/vars.txt @@ -0,0 +1,6 @@ +VARS(7) + +Bar + +bar + diff --git a/crates/mdman/tests/compare/formatting.md b/crates/mdman/tests/compare/formatting.md new file mode 100644 index 0000000..3b9f5b8 --- /dev/null +++ b/crates/mdman/tests/compare/formatting.md @@ -0,0 +1,95 @@ +# formatting(1) + +This is **nested _formatting_ `text`**. + +## SECOND HEADING + +Some text at second level. + +### Third heading + +Some text at third level. + +#### Fourth heading + +Some text at fourth level. + +## Quotes and blocks. + +Here are some quotes and blocks. + +> This is a block quote. Ambidextrously koala apart that prudent blindly alas +> far amid dear goodness turgid so exact inside oh and alas much fanciful that +> dark on spoon-fed adequately insolent walking crud. + +``` +This is a code block. Groundhog watchfully sudden firefly some self-consciously hotly jeepers satanic after that this parrot this at virtuous +some mocking the leaned jeez nightingale as much mallard so because jeez +turned dear crud grizzly strenuously. + + Indented and should be unmodified. +``` + + This is an indented code block. Egregiously yikes animatedly since outside beseechingly a badger hey shakily giraffe a one wow one this + goodness regarding reindeer so astride before. + + Doubly indented + +## Lists + +1. Ordered list + + * Unordered list + + With a second paragraph inside it + + 1. Inner ordered list + + 1. Another + + * Eggs + + * Milk + + 5. Don't start at one. + 6. tamarind + +1. Second element + +1. Third element + +## Breaks + +This has a\ +hard break in it +and a soft one. + +## Horizontal rule + +This should contain a line: + +--- + +Nice! + +## Strange characters + +Handles escaping for characters + +.dot at the start of a line. + +\fBnot really troff + +Various characters \ - – — ─ │ ├ └ + +``` +tree +└── example + ├── salamander + │ ├── honey + │ └── some + ├── fancifully + └── trout +``` + + non-breaking space. diff --git a/crates/mdman/tests/compare/includes/links-include.md b/crates/mdman/tests/compare/includes/links-include.md new file mode 100644 index 0000000..7373360 --- /dev/null +++ b/crates/mdman/tests/compare/includes/links-include.md @@ -0,0 +1,7 @@ +[Some link](foo.html) + +{{#options}} +{{#option "`--include`"}} +Testing an [included link](included_link.html). +{{/option}} +{{/options}} diff --git a/crates/mdman/tests/compare/includes/options-common.md b/crates/mdman/tests/compare/includes/options-common.md new file mode 100644 index 0000000..07404e3 --- /dev/null +++ b/crates/mdman/tests/compare/includes/options-common.md @@ -0,0 +1,14 @@ +{{#options}} +{{#option "`@`_filename_"}} +Load from filename. +{{/option}} + +{{#option "`--foo` [_bar_]"}} +Flag with optional value. +{{/option}} + +{{#option "`--foo`[`=`_bar_]"}} +Alternate syntax for optional value (with required = for disambiguation). +{{/option}} + +{{/options}} diff --git a/crates/mdman/tests/compare/links.md b/crates/mdman/tests/compare/links.md new file mode 100644 index 0000000..949f374 --- /dev/null +++ b/crates/mdman/tests/compare/links.md @@ -0,0 +1,49 @@ +# links(1) + +## NAME + +links - Test of different link kinds + +## DESCRIPTION + +Inline link: [inline link](https://example.com/inline) + +Reference link: [this is a link][bar] + +Collapsed: [collapsed][] + +Shortcut: [shortcut] + +Autolink: <https://example.com/auto> + +Email: <foo@example.com> + +Relative link: [relative link](foo/bar.html) + +Collapsed unknown: [collapsed unknown][] + +Reference unknown: [foo][unknown] + +Shortcut unknown: [shortcut unknown] + +{{man "other-cmd" 1}} + +{{man "local-cmd" 1}} + +{{> links-include}} + +## OPTIONS + +{{#options}} + +{{#option "`--foo-bar`"}} +Example [link](bar.html). +See {{man "other-cmd" 1}}, {{man "local-cmd" 1}} +{{/option}} + +{{/options}} + + +[bar]: https://example.com/bar +[collapsed]: https://example.com/collapsed +[shortcut]: https://example.com/shortcut diff --git a/crates/mdman/tests/compare/options.md b/crates/mdman/tests/compare/options.md new file mode 100644 index 0000000..51415b0 --- /dev/null +++ b/crates/mdman/tests/compare/options.md @@ -0,0 +1,62 @@ +# my-command(1) + +## NAME + +my-command - A brief description + +## SYNOPSIS + +`my-command` [`--abc` | `--xyz`] _name_\ +`my-command` [`-f` _file_]\ +`my-command` (`-m` | `-M`) [_oldbranch_] _newbranch_\ +`my-command` (`-d` | `-D`) [`-r`] _branchname_... + +## DESCRIPTION + +A description of the command. + +* One + * Sub one + * Sub two +* Two +* Three + + +## OPTIONS + +### Command options + +{{#options}} + +{{#option "`--foo-bar`"}} +Demo *emphasis*, **strong**, ~~strike~~ +{{/option}} + +{{#option "`-p` _spec_" "`--package` _spec_"}} +This has multiple flags. +{{/option}} + +{{#option "_named-arg..._"}} +A named argument. +{{/option}} + +{{/options}} + +### Common Options + +{{> options-common}} + +## EXAMPLES + +1. An example + + ``` + my-command --abc + ``` + +1. Another example + + my-command --xyz + +## SEE ALSO +{{man "other-command" 1}} {{man "abc" 7}} diff --git a/crates/mdman/tests/compare/tables.md b/crates/mdman/tests/compare/tables.md new file mode 100644 index 0000000..831132c --- /dev/null +++ b/crates/mdman/tests/compare/tables.md @@ -0,0 +1,35 @@ +# tables(1) + +## DESCRIPTION + +Testing tables. + +| Single col | +-------------- +| Hi! :) | + + +Header content | With `format` *text* | Another column +---------------|----------------------|---------------- +Some data | More data | +Extra long amount of text within a column | hi | there + + +Left aligned | Center aligned | Right aligned +-------------|:--------------:|--------------: +abc | def | ghi + + +Left aligned | Center aligned | Right aligned +-------------|:--------------:|--------------: +X | X | X +Extra long text 123456789012 with mixed widths. | Extra long text 123456789012 with mixed widths. | Extra long text 123456789012 with mixed widths. + + +| Link check | +-------------- +| [foo] | +| <https://example.com/> | + + +[foo]: https://example.com/ diff --git a/crates/mdman/tests/compare/vars.md b/crates/mdman/tests/compare/vars.md new file mode 100644 index 0000000..d41b765 --- /dev/null +++ b/crates/mdman/tests/compare/vars.md @@ -0,0 +1,7 @@ +# vars(7) + +{{*set foo="Bar"}} + +{{foo}} + +{{lower foo}} diff --git a/crates/mdman/tests/invalid.rs b/crates/mdman/tests/invalid.rs new file mode 100644 index 0000000..cc81d06 --- /dev/null +++ b/crates/mdman/tests/invalid.rs @@ -0,0 +1,34 @@ +//! Tests for errors and invalid input. + +use mdman::{Format, ManMap}; +use pretty_assertions::assert_eq; +use std::path::PathBuf; + +fn run(name: &str, expected_error: &str) { + let input = PathBuf::from(format!("tests/invalid/{}", name)); + match mdman::convert(&input, Format::Man, None, ManMap::new()) { + Ok(_) => { + panic!("expected {} to fail", name); + } + Err(e) => { + assert_eq!(expected_error, e.to_string()); + } + } +} + +macro_rules! test( ($name:ident, $file_name:expr, $error:expr) => ( + #[test] + fn $name() { run($file_name, $error); } +) ); + +test!( + nested, + "nested.md", + "Error rendering \"template\" line 4, col 1: options blocks cannot be nested" +); + +test!( + not_inside_options, + "not-inside-options.md", + "Error rendering \"template\" line 3, col 1: option must be in options block" +); diff --git a/crates/mdman/tests/invalid/nested.md b/crates/mdman/tests/invalid/nested.md new file mode 100644 index 0000000..6a33e6d --- /dev/null +++ b/crates/mdman/tests/invalid/nested.md @@ -0,0 +1,6 @@ +# nested(1) + +{{#options}} +{{#options}} +{{/options}} +{{/options}} diff --git a/crates/mdman/tests/invalid/not-inside-options.md b/crates/mdman/tests/invalid/not-inside-options.md new file mode 100644 index 0000000..b6c816f --- /dev/null +++ b/crates/mdman/tests/invalid/not-inside-options.md @@ -0,0 +1,5 @@ +# not-inside-options(1) + +{{#option "`-o`"}} +Testing without options block. +{{/option}} diff --git a/crates/resolver-tests/Cargo.toml b/crates/resolver-tests/Cargo.toml new file mode 100644 index 0000000..cc50ad3 --- /dev/null +++ b/crates/resolver-tests/Cargo.toml @@ -0,0 +1,12 @@ +[package] +name = "resolver-tests" +version = "0.1.0" +edition = "2018" + +[dependencies] +cargo = { path = "../.." } +cargo-util = { path = "../cargo-util" } +is-terminal = "0.4.0" +lazy_static = "1.3.0" +proptest = "0.9.1" +varisat = "0.2.1" diff --git a/crates/resolver-tests/src/lib.rs b/crates/resolver-tests/src/lib.rs new file mode 100644 index 0000000..3ffb6c5 --- /dev/null +++ b/crates/resolver-tests/src/lib.rs @@ -0,0 +1,991 @@ +#![allow(clippy::all)] + +use std::cell::RefCell; +use std::cmp::PartialEq; +use std::cmp::{max, min}; +use std::collections::{BTreeMap, BTreeSet, HashMap, HashSet}; +use std::fmt; +use std::fmt::Write; +use std::rc::Rc; +use std::task::Poll; +use std::time::Instant; + +use cargo::core::dependency::DepKind; +use cargo::core::resolver::{self, ResolveOpts, VersionPreferences}; +use cargo::core::source::{GitReference, QueryKind, SourceId}; +use cargo::core::Resolve; +use cargo::core::{Dependency, PackageId, Registry, Summary}; +use cargo::util::{CargoResult, Config, Graph, IntoUrl}; + +use proptest::collection::{btree_map, vec}; +use proptest::prelude::*; +use proptest::sample::Index; +use proptest::string::string_regex; +use varisat::{self, ExtendFormula}; + +pub fn resolve(deps: Vec<Dependency>, registry: &[Summary]) -> CargoResult<Vec<PackageId>> { + resolve_with_config(deps, registry, &Config::default().unwrap()) +} + +pub fn resolve_and_validated( + deps: Vec<Dependency>, + registry: &[Summary], + sat_resolve: Option<SatResolve>, +) -> CargoResult<Vec<PackageId>> { + let resolve = resolve_with_config_raw(deps.clone(), registry, &Config::default().unwrap()); + + match resolve { + Err(e) => { + let sat_resolve = sat_resolve.unwrap_or_else(|| SatResolve::new(registry)); + if sat_resolve.sat_resolve(&deps) { + panic!( + "the resolve err but the sat_resolve thinks this will work:\n{}", + sat_resolve.use_packages().unwrap() + ); + } + Err(e) + } + Ok(resolve) => { + let mut stack = vec![pkg_id("root")]; + let mut used = HashSet::new(); + let mut links = HashSet::new(); + while let Some(p) = stack.pop() { + assert!(resolve.contains(&p)); + if used.insert(p) { + // in the tests all `links` crates end in `-sys` + if p.name().ends_with("-sys") { + assert!(links.insert(p.name())); + } + stack.extend(resolve.deps(p).map(|(dp, deps)| { + for d in deps { + assert!(d.matches_id(dp)); + } + dp + })); + } + } + let out = resolve.sort(); + assert_eq!(out.len(), used.len()); + + let mut pub_deps: HashMap<PackageId, HashSet<_>> = HashMap::new(); + for &p in out.iter() { + // make the list of `p` public dependencies + let mut self_pub_dep = HashSet::new(); + self_pub_dep.insert(p); + for (dp, deps) in resolve.deps(p) { + if deps.iter().any(|d| d.is_public()) { + self_pub_dep.extend(pub_deps[&dp].iter().cloned()) + } + } + pub_deps.insert(p, self_pub_dep); + + // check if `p` has a public dependencies conflicts + let seen_dep: BTreeSet<_> = resolve + .deps(p) + .flat_map(|(dp, _)| pub_deps[&dp].iter().cloned()) + .collect(); + let seen_dep: Vec<_> = seen_dep.iter().collect(); + for a in seen_dep.windows(2) { + if a[0].name() == a[1].name() { + panic!( + "the package {:?} can publicly see {:?} and {:?}", + p, a[0], a[1] + ) + } + } + } + let sat_resolve = sat_resolve.unwrap_or_else(|| SatResolve::new(registry)); + if !sat_resolve.sat_is_valid_solution(&out) { + panic!( + "the sat_resolve err but the resolve thinks this will work:\n{:?}", + resolve + ); + } + Ok(out) + } + } +} + +pub fn resolve_with_config( + deps: Vec<Dependency>, + registry: &[Summary], + config: &Config, +) -> CargoResult<Vec<PackageId>> { + let resolve = resolve_with_config_raw(deps, registry, config)?; + Ok(resolve.sort()) +} + +pub fn resolve_with_config_raw( + deps: Vec<Dependency>, + registry: &[Summary], + config: &Config, +) -> CargoResult<Resolve> { + struct MyRegistry<'a> { + list: &'a [Summary], + used: HashSet<PackageId>, + } + impl<'a> Registry for MyRegistry<'a> { + fn query( + &mut self, + dep: &Dependency, + kind: QueryKind, + f: &mut dyn FnMut(Summary), + ) -> Poll<CargoResult<()>> { + for summary in self.list.iter() { + let matched = match kind { + QueryKind::Exact => dep.matches(summary), + QueryKind::Fuzzy => true, + }; + if matched { + self.used.insert(summary.package_id()); + f(summary.clone()); + } + } + Poll::Ready(Ok(())) + } + + fn describe_source(&self, _src: SourceId) -> String { + String::new() + } + + fn is_replaced(&self, _src: SourceId) -> bool { + false + } + + fn block_until_ready(&mut self) -> CargoResult<()> { + Ok(()) + } + } + impl<'a> Drop for MyRegistry<'a> { + fn drop(&mut self) { + if std::thread::panicking() && self.list.len() != self.used.len() { + // we found a case that causes a panic and did not use all of the input. + // lets print the part of the input that was used for minimization. + println!( + "{:?}", + PrettyPrintRegistry( + self.list + .iter() + .filter(|s| { self.used.contains(&s.package_id()) }) + .cloned() + .collect() + ) + ); + } + } + } + let mut registry = MyRegistry { + list: registry, + used: HashSet::new(), + }; + let summary = Summary::new( + config, + pkg_id("root"), + deps, + &BTreeMap::new(), + None::<&String>, + ) + .unwrap(); + let opts = ResolveOpts::everything(); + let start = Instant::now(); + let resolve = resolver::resolve( + &[(summary, opts)], + &[], + &mut registry, + &VersionPreferences::default(), + Some(config), + true, + ); + + // The largest test in our suite takes less then 30 sec. + // So lets fail the test if we have ben running for two long. + assert!(start.elapsed().as_secs() < 60); + resolve +} + +const fn num_bits<T>() -> usize { + std::mem::size_of::<T>() * 8 +} + +fn log_bits(x: usize) -> usize { + if x == 0 { + return 0; + } + assert!(x > 0); + (num_bits::<usize>() as u32 - x.leading_zeros()) as usize +} + +fn sat_at_most_one(solver: &mut impl varisat::ExtendFormula, vars: &[varisat::Var]) { + if vars.len() <= 1 { + return; + } else if vars.len() == 2 { + solver.add_clause(&[vars[0].negative(), vars[1].negative()]); + return; + } else if vars.len() == 3 { + solver.add_clause(&[vars[0].negative(), vars[1].negative()]); + solver.add_clause(&[vars[0].negative(), vars[2].negative()]); + solver.add_clause(&[vars[1].negative(), vars[2].negative()]); + return; + } + // use the "Binary Encoding" from + // https://www.it.uu.se/research/group/astra/ModRef10/papers/Alan%20M.%20Frisch%20and%20Paul%20A.%20Giannoros.%20SAT%20Encodings%20of%20the%20At-Most-k%20Constraint%20-%20ModRef%202010.pdf + let bits: Vec<varisat::Var> = solver.new_var_iter(log_bits(vars.len())).collect(); + for (i, p) in vars.iter().enumerate() { + for b in 0..bits.len() { + solver.add_clause(&[p.negative(), bits[b].lit(((1 << b) & i) > 0)]); + } + } +} + +fn sat_at_most_one_by_key<K: std::hash::Hash + Eq>( + cnf: &mut impl varisat::ExtendFormula, + data: impl Iterator<Item = (K, varisat::Var)>, +) -> HashMap<K, Vec<varisat::Var>> { + // no two packages with the same links set + let mut by_keys: HashMap<K, Vec<varisat::Var>> = HashMap::new(); + for (p, v) in data { + by_keys.entry(p).or_default().push(v) + } + for key in by_keys.values() { + sat_at_most_one(cnf, key); + } + by_keys +} + +/// Resolution can be reduced to the SAT problem. So this is an alternative implementation +/// of the resolver that uses a SAT library for the hard work. This is intended to be easy to read, +/// as compared to the real resolver. +/// +/// For the subset of functionality that are currently made by `registry_strategy` this will, +/// find a valid resolution if one exists. The big thing that the real resolver does, +/// that this one does not do is work with features and optional dependencies. +/// +/// The SAT library dose not optimize for the newer version, +/// so the selected packages may not match the real resolver. +#[derive(Clone)] +pub struct SatResolve(Rc<RefCell<SatResolveInner>>); +struct SatResolveInner { + solver: varisat::Solver<'static>, + var_for_is_packages_used: HashMap<PackageId, varisat::Var>, + by_name: HashMap<&'static str, Vec<PackageId>>, +} + +impl SatResolve { + pub fn new(registry: &[Summary]) -> Self { + let mut cnf = varisat::CnfFormula::new(); + let var_for_is_packages_used: HashMap<PackageId, varisat::Var> = registry + .iter() + .map(|s| (s.package_id(), cnf.new_var())) + .collect(); + + // no two packages with the same links set + sat_at_most_one_by_key( + &mut cnf, + registry + .iter() + .map(|s| (s.links(), var_for_is_packages_used[&s.package_id()])) + .filter(|(l, _)| l.is_some()), + ); + + // no two semver compatible versions of the same package + let by_activations_keys = sat_at_most_one_by_key( + &mut cnf, + var_for_is_packages_used + .iter() + .map(|(p, &v)| (p.as_activations_key(), v)), + ); + + let mut by_name: HashMap<&'static str, Vec<PackageId>> = HashMap::new(); + + for p in registry.iter() { + by_name + .entry(p.name().as_str()) + .or_default() + .push(p.package_id()) + } + + let empty_vec = vec![]; + + let mut graph: Graph<PackageId, ()> = Graph::new(); + + let mut version_selected_for: HashMap< + PackageId, + HashMap<Dependency, HashMap<_, varisat::Var>>, + > = HashMap::new(); + // active packages need each of there `deps` to be satisfied + for p in registry.iter() { + graph.add(p.package_id()); + for dep in p.dependencies() { + // This can more easily be written as: + // !is_active(p) or one of the things that match dep is_active + // All the complexity, from here to the end, is to support public and private dependencies! + let mut by_key: HashMap<_, Vec<varisat::Lit>> = HashMap::new(); + for &m in by_name + .get(dep.package_name().as_str()) + .unwrap_or(&empty_vec) + .iter() + .filter(|&p| dep.matches_id(*p)) + { + graph.link(p.package_id(), m); + by_key + .entry(m.as_activations_key()) + .or_default() + .push(var_for_is_packages_used[&m].positive()); + } + let keys: HashMap<_, _> = by_key.keys().map(|&k| (k, cnf.new_var())).collect(); + + // if `p` is active then we need to select one of the keys + let matches: Vec<_> = keys + .values() + .map(|v| v.positive()) + .chain(Some(var_for_is_packages_used[&p.package_id()].negative())) + .collect(); + cnf.add_clause(&matches); + + // if a key is active then we need to select one of the versions + for (key, vars) in by_key.iter() { + let mut matches = vars.clone(); + matches.push(keys[key].negative()); + cnf.add_clause(&matches); + } + + version_selected_for + .entry(p.package_id()) + .or_default() + .insert(dep.clone(), keys); + } + } + + let topological_order = graph.sort(); + + // we already ensure there is only one version for each `activations_key` so we can think of + // `publicly_exports` as being in terms of a set of `activations_key`s + let mut publicly_exports: HashMap<_, HashMap<_, varisat::Var>> = HashMap::new(); + + for &key in by_activations_keys.keys() { + // everything publicly depends on itself + let var = publicly_exports + .entry(key) + .or_default() + .entry(key) + .or_insert_with(|| cnf.new_var()); + cnf.add_clause(&[var.positive()]); + } + + // if a `dep` is public then `p` `publicly_exports` all the things that the selected version `publicly_exports` + for &p in topological_order.iter() { + if let Some(deps) = version_selected_for.get(&p) { + let mut p_exports = publicly_exports.remove(&p.as_activations_key()).unwrap(); + for (_, versions) in deps.iter().filter(|(d, _)| d.is_public()) { + for (ver, sel) in versions { + for (&export_pid, &export_var) in publicly_exports[ver].iter() { + let our_var = + p_exports.entry(export_pid).or_insert_with(|| cnf.new_var()); + cnf.add_clause(&[ + sel.negative(), + export_var.negative(), + our_var.positive(), + ]); + } + } + } + publicly_exports.insert(p.as_activations_key(), p_exports); + } + } + + // we already ensure there is only one version for each `activations_key` so we can think of + // `can_see` as being in terms of a set of `activations_key`s + // and if `p` `publicly_exports` `export` then it `can_see` `export` + let mut can_see: HashMap<_, HashMap<_, varisat::Var>> = HashMap::new(); + + // if `p` has a `dep` that selected `ver` then it `can_see` all the things that the selected version `publicly_exports` + for (&p, deps) in version_selected_for.iter() { + let p_can_see = can_see.entry(p).or_default(); + for (_, versions) in deps.iter() { + for (&ver, sel) in versions { + for (&export_pid, &export_var) in publicly_exports[&ver].iter() { + let our_var = p_can_see.entry(export_pid).or_insert_with(|| cnf.new_var()); + cnf.add_clause(&[ + sel.negative(), + export_var.negative(), + our_var.positive(), + ]); + } + } + } + } + + // a package `can_see` only one version by each name + for (_, see) in can_see.iter() { + sat_at_most_one_by_key(&mut cnf, see.iter().map(|((name, _, _), &v)| (name, v))); + } + let mut solver = varisat::Solver::new(); + solver.add_formula(&cnf); + + // We dont need to `solve` now. We know that "use nothing" will satisfy all the clauses so far. + // But things run faster if we let it spend some time figuring out how the constraints interact before we add assumptions. + solver + .solve() + .expect("docs say it can't error in default config"); + SatResolve(Rc::new(RefCell::new(SatResolveInner { + solver, + var_for_is_packages_used, + by_name, + }))) + } + pub fn sat_resolve(&self, deps: &[Dependency]) -> bool { + let mut s = self.0.borrow_mut(); + let mut assumption = vec![]; + let mut this_call = None; + + // the starting `deps` need to be satisfied + for dep in deps.iter() { + let empty_vec = vec![]; + let matches: Vec<varisat::Lit> = s + .by_name + .get(dep.package_name().as_str()) + .unwrap_or(&empty_vec) + .iter() + .filter(|&p| dep.matches_id(*p)) + .map(|p| s.var_for_is_packages_used[p].positive()) + .collect(); + if matches.is_empty() { + return false; + } else if matches.len() == 1 { + assumption.extend_from_slice(&matches) + } else { + if this_call.is_none() { + let new_var = s.solver.new_var(); + this_call = Some(new_var); + assumption.push(new_var.positive()); + } + let mut matches = matches; + matches.push(this_call.unwrap().negative()); + s.solver.add_clause(&matches); + } + } + + s.solver.assume(&assumption); + + s.solver + .solve() + .expect("docs say it can't error in default config") + } + pub fn sat_is_valid_solution(&self, pids: &[PackageId]) -> bool { + let mut s = self.0.borrow_mut(); + for p in pids { + if p.name().as_str() != "root" && !s.var_for_is_packages_used.contains_key(p) { + return false; + } + } + let assumption: Vec<_> = s + .var_for_is_packages_used + .iter() + .map(|(p, v)| v.lit(pids.contains(p))) + .collect(); + + s.solver.assume(&assumption); + + s.solver + .solve() + .expect("docs say it can't error in default config") + } + fn use_packages(&self) -> Option<String> { + self.0.borrow().solver.model().map(|lits| { + let lits: HashSet<_> = lits + .iter() + .filter(|l| l.is_positive()) + .map(|l| l.var()) + .collect(); + let mut out = String::new(); + out.push_str("used:\n"); + for (p, v) in self.0.borrow().var_for_is_packages_used.iter() { + if lits.contains(v) { + writeln!(&mut out, " {}", p).unwrap(); + } + } + out + }) + } +} + +pub trait ToDep { + fn to_dep(self) -> Dependency; +} + +impl ToDep for &'static str { + fn to_dep(self) -> Dependency { + Dependency::parse(self, Some("1.0.0"), registry_loc()).unwrap() + } +} + +impl ToDep for Dependency { + fn to_dep(self) -> Dependency { + self + } +} + +pub trait ToPkgId { + fn to_pkgid(&self) -> PackageId; +} + +impl ToPkgId for PackageId { + fn to_pkgid(&self) -> PackageId { + *self + } +} + +impl<'a> ToPkgId for &'a str { + fn to_pkgid(&self) -> PackageId { + PackageId::new(*self, "1.0.0", registry_loc()).unwrap() + } +} + +impl<T: AsRef<str>, U: AsRef<str>> ToPkgId for (T, U) { + fn to_pkgid(&self) -> PackageId { + let (name, vers) = self; + PackageId::new(name.as_ref(), vers.as_ref(), registry_loc()).unwrap() + } +} + +#[macro_export] +macro_rules! pkg { + ($pkgid:expr => [$($deps:expr),+ $(,)* ]) => ({ + let d: Vec<Dependency> = vec![$($deps.to_dep()),+]; + $crate::pkg_dep($pkgid, d) + }); + + ($pkgid:expr) => ({ + $crate::pkg($pkgid) + }) +} + +fn registry_loc() -> SourceId { + lazy_static::lazy_static! { + static ref EXAMPLE_DOT_COM: SourceId = + SourceId::for_registry(&"https://example.com".into_url().unwrap()).unwrap(); + } + *EXAMPLE_DOT_COM +} + +pub fn pkg<T: ToPkgId>(name: T) -> Summary { + pkg_dep(name, Vec::new()) +} + +pub fn pkg_dep<T: ToPkgId>(name: T, dep: Vec<Dependency>) -> Summary { + let pkgid = name.to_pkgid(); + let link = if pkgid.name().ends_with("-sys") { + Some(pkgid.name().as_str()) + } else { + None + }; + Summary::new( + &Config::default().unwrap(), + name.to_pkgid(), + dep, + &BTreeMap::new(), + link, + ) + .unwrap() +} + +pub fn pkg_id(name: &str) -> PackageId { + PackageId::new(name, "1.0.0", registry_loc()).unwrap() +} + +fn pkg_id_loc(name: &str, loc: &str) -> PackageId { + let remote = loc.into_url(); + let master = GitReference::Branch("master".to_string()); + let source_id = SourceId::for_git(&remote.unwrap(), master).unwrap(); + + PackageId::new(name, "1.0.0", source_id).unwrap() +} + +pub fn pkg_loc(name: &str, loc: &str) -> Summary { + let link = if name.ends_with("-sys") { + Some(name) + } else { + None + }; + Summary::new( + &Config::default().unwrap(), + pkg_id_loc(name, loc), + Vec::new(), + &BTreeMap::new(), + link, + ) + .unwrap() +} + +pub fn remove_dep(sum: &Summary, ind: usize) -> Summary { + let mut deps = sum.dependencies().to_vec(); + deps.remove(ind); + // note: more things will need to be copied over in the future, but it works for now. + Summary::new( + &Config::default().unwrap(), + sum.package_id(), + deps, + &BTreeMap::new(), + sum.links().map(|a| a.as_str()), + ) + .unwrap() +} + +pub fn dep(name: &str) -> Dependency { + dep_req(name, "*") +} +pub fn dep_req(name: &str, req: &str) -> Dependency { + Dependency::parse(name, Some(req), registry_loc()).unwrap() +} +pub fn dep_req_kind(name: &str, req: &str, kind: DepKind, public: bool) -> Dependency { + let mut dep = dep_req(name, req); + dep.set_kind(kind); + dep.set_public(public); + dep +} + +pub fn dep_loc(name: &str, location: &str) -> Dependency { + let url = location.into_url().unwrap(); + let master = GitReference::Branch("master".to_string()); + let source_id = SourceId::for_git(&url, master).unwrap(); + Dependency::parse(name, Some("1.0.0"), source_id).unwrap() +} +pub fn dep_kind(name: &str, kind: DepKind) -> Dependency { + dep(name).set_kind(kind).clone() +} + +pub fn registry(pkgs: Vec<Summary>) -> Vec<Summary> { + pkgs +} + +pub fn names<P: ToPkgId>(names: &[P]) -> Vec<PackageId> { + names.iter().map(|name| name.to_pkgid()).collect() +} + +pub fn loc_names(names: &[(&'static str, &'static str)]) -> Vec<PackageId> { + names + .iter() + .map(|&(name, loc)| pkg_id_loc(name, loc)) + .collect() +} + +/// By default `Summary` and `Dependency` have a very verbose `Debug` representation. +/// This replaces with a representation that uses constructors from this file. +/// +/// If `registry_strategy` is improved to modify more fields +/// then this needs to update to display the corresponding constructor. +pub struct PrettyPrintRegistry(pub Vec<Summary>); + +impl fmt::Debug for PrettyPrintRegistry { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "vec![")?; + for s in &self.0 { + if s.dependencies().is_empty() { + write!(f, "pkg!((\"{}\", \"{}\")),", s.name(), s.version())?; + } else { + write!(f, "pkg!((\"{}\", \"{}\") => [", s.name(), s.version())?; + for d in s.dependencies() { + if d.kind() == DepKind::Normal + && &d.version_req().to_string() == "*" + && !d.is_public() + { + write!(f, "dep(\"{}\"),", d.name_in_toml())?; + } else if d.kind() == DepKind::Normal && !d.is_public() { + write!( + f, + "dep_req(\"{}\", \"{}\"),", + d.name_in_toml(), + d.version_req() + )?; + } else { + write!( + f, + "dep_req_kind(\"{}\", \"{}\", {}, {}),", + d.name_in_toml(), + d.version_req(), + match d.kind() { + DepKind::Development => "DepKind::Development", + DepKind::Build => "DepKind::Build", + DepKind::Normal => "DepKind::Normal", + }, + d.is_public() + )?; + } + } + write!(f, "]),")?; + } + } + write!(f, "]") + } +} + +#[test] +fn meta_test_deep_pretty_print_registry() { + assert_eq!( + &format!( + "{:?}", + PrettyPrintRegistry(vec![ + pkg!(("foo", "1.0.1") => [dep_req("bar", "1")]), + pkg!(("foo", "1.0.0") => [dep_req("bar", "2")]), + pkg!(("foo", "2.0.0") => [dep_req("bar", "*")]), + pkg!(("bar", "1.0.0") => [dep_req("baz", "=1.0.2"), + dep_req("other", "1")]), + pkg!(("bar", "2.0.0") => [dep_req("baz", "=1.0.1")]), + pkg!(("baz", "1.0.2") => [dep_req("other", "2")]), + pkg!(("baz", "1.0.1")), + pkg!(("cat", "1.0.2") => [dep_req_kind("other", "2", DepKind::Build, false)]), + pkg!(("cat", "1.0.3") => [dep_req_kind("other", "2", DepKind::Development, false)]), + pkg!(("dep_req", "1.0.0")), + pkg!(("dep_req", "2.0.0")), + ]) + ), + "vec![pkg!((\"foo\", \"1.0.1\") => [dep_req(\"bar\", \"^1\"),]),\ + pkg!((\"foo\", \"1.0.0\") => [dep_req(\"bar\", \"^2\"),]),\ + pkg!((\"foo\", \"2.0.0\") => [dep(\"bar\"),]),\ + pkg!((\"bar\", \"1.0.0\") => [dep_req(\"baz\", \"=1.0.2\"),dep_req(\"other\", \"^1\"),]),\ + pkg!((\"bar\", \"2.0.0\") => [dep_req(\"baz\", \"=1.0.1\"),]),\ + pkg!((\"baz\", \"1.0.2\") => [dep_req(\"other\", \"^2\"),]),\ + pkg!((\"baz\", \"1.0.1\")),\ + pkg!((\"cat\", \"1.0.2\") => [dep_req_kind(\"other\", \"^2\", DepKind::Build, false),]),\ + pkg!((\"cat\", \"1.0.3\") => [dep_req_kind(\"other\", \"^2\", DepKind::Development, false),]),\ + pkg!((\"dep_req\", \"1.0.0\")),\ + pkg!((\"dep_req\", \"2.0.0\")),]" + ) +} + +/// This generates a random registry index. +/// Unlike vec((Name, Ver, vec((Name, VerRq), ..), ..) +/// This strategy has a high probability of having valid dependencies +pub fn registry_strategy( + max_crates: usize, + max_versions: usize, + shrinkage: usize, +) -> impl Strategy<Value = PrettyPrintRegistry> { + let name = string_regex("[A-Za-z][A-Za-z0-9_-]*(-sys)?").unwrap(); + + let raw_version = ..max_versions.pow(3); + let version_from_raw = move |r: usize| { + let major = ((r / max_versions) / max_versions) % max_versions; + let minor = (r / max_versions) % max_versions; + let patch = r % max_versions; + format!("{}.{}.{}", major, minor, patch) + }; + + // If this is false then the crate will depend on the nonexistent "bad" + // instead of the complex set we generated for it. + let allow_deps = prop::bool::weighted(0.99); + + let list_of_versions = + btree_map(raw_version, allow_deps, 1..=max_versions).prop_map(move |ver| { + ver.into_iter() + .map(|a| (version_from_raw(a.0), a.1)) + .collect::<Vec<_>>() + }); + + let list_of_crates_with_versions = + btree_map(name, list_of_versions, 1..=max_crates).prop_map(|mut vers| { + // root is the name of the thing being compiled + // so it would be confusing to have it in the index + vers.remove("root"); + // bad is a name reserved for a dep that won't work + vers.remove("bad"); + vers + }); + + // each version of each crate can depend on each crate smaller then it. + // In theory shrinkage should be 2, but in practice we get better trees with a larger value. + let max_deps = max_versions * (max_crates * (max_crates - 1)) / shrinkage; + + let raw_version_range = (any::<Index>(), any::<Index>()); + let raw_dependency = ( + any::<Index>(), + any::<Index>(), + raw_version_range, + 0..=1, + Just(false), + // TODO: ^ this needs to be set back to `any::<bool>()` and work before public & private dependencies can stabilize + ); + + fn order_index(a: Index, b: Index, size: usize) -> (usize, usize) { + let (a, b) = (a.index(size), b.index(size)); + (min(a, b), max(a, b)) + } + + let list_of_raw_dependency = vec(raw_dependency, ..=max_deps); + + // By default a package depends only on other packages that have a smaller name, + // this helps make sure that all things in the resulting index are DAGs. + // If this is true then the DAG is maintained with grater instead. + let reverse_alphabetical = any::<bool>().no_shrink(); + + ( + list_of_crates_with_versions, + list_of_raw_dependency, + reverse_alphabetical, + ) + .prop_map( + |(crate_vers_by_name, raw_dependencies, reverse_alphabetical)| { + let list_of_pkgid: Vec<_> = crate_vers_by_name + .iter() + .flat_map(|(name, vers)| vers.iter().map(move |x| ((name.as_str(), &x.0), x.1))) + .collect(); + let len_all_pkgid = list_of_pkgid.len(); + let mut dependency_by_pkgid = vec![vec![]; len_all_pkgid]; + for (a, b, (c, d), k, p) in raw_dependencies { + let (a, b) = order_index(a, b, len_all_pkgid); + let (a, b) = if reverse_alphabetical { (b, a) } else { (a, b) }; + let ((dep_name, _), _) = list_of_pkgid[a]; + if (list_of_pkgid[b].0).0 == dep_name { + continue; + } + let s = &crate_vers_by_name[dep_name]; + let s_last_index = s.len() - 1; + let (c, d) = order_index(c, d, s.len()); + + dependency_by_pkgid[b].push(dep_req_kind( + dep_name, + &if c == 0 && d == s_last_index { + "*".to_string() + } else if c == 0 { + format!("<={}", s[d].0) + } else if d == s_last_index { + format!(">={}", s[c].0) + } else if c == d { + format!("={}", s[c].0) + } else { + format!(">={}, <={}", s[c].0, s[d].0) + }, + match k { + 0 => DepKind::Normal, + 1 => DepKind::Build, + // => DepKind::Development, // Development has no impact so don't gen + _ => panic!("bad index for DepKind"), + }, + p && k == 0, + )) + } + + let mut out: Vec<Summary> = list_of_pkgid + .into_iter() + .zip(dependency_by_pkgid.into_iter()) + .map(|(((name, ver), allow_deps), deps)| { + pkg_dep( + (name, ver).to_pkgid(), + if !allow_deps { + vec![dep_req("bad", "*")] + } else { + let mut deps = deps; + deps.sort_by_key(|d| d.name_in_toml()); + deps.dedup_by_key(|d| d.name_in_toml()); + deps + }, + ) + }) + .collect(); + + if reverse_alphabetical { + // make sure the complicated cases are at the end + out.reverse(); + } + + PrettyPrintRegistry(out) + }, + ) +} + +/// This test is to test the generator to ensure +/// that it makes registries with large dependency trees +#[test] +fn meta_test_deep_trees_from_strategy() { + use proptest::strategy::ValueTree; + use proptest::test_runner::TestRunner; + + let mut dis = [0; 21]; + + let strategy = registry_strategy(50, 20, 60); + let mut test_runner = TestRunner::deterministic(); + for _ in 0..128 { + let PrettyPrintRegistry(input) = strategy + .new_tree(&mut TestRunner::new_with_rng( + Default::default(), + test_runner.new_rng(), + )) + .unwrap() + .current(); + let reg = registry(input.clone()); + for this in input.iter().rev().take(10) { + let res = resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + ); + dis[res + .as_ref() + .map(|x| min(x.len(), dis.len()) - 1) + .unwrap_or(0)] += 1; + if dis.iter().all(|&x| x > 0) { + return; + } + } + } + + panic!( + "In 1280 tries we did not see a wide enough distribution of dependency trees! dis: {:?}", + dis + ); +} + +/// This test is to test the generator to ensure +/// that it makes registries that include multiple versions of the same library +#[test] +fn meta_test_multiple_versions_strategy() { + use proptest::strategy::ValueTree; + use proptest::test_runner::TestRunner; + + let mut dis = [0; 10]; + + let strategy = registry_strategy(50, 20, 60); + let mut test_runner = TestRunner::deterministic(); + for _ in 0..128 { + let PrettyPrintRegistry(input) = strategy + .new_tree(&mut TestRunner::new_with_rng( + Default::default(), + test_runner.new_rng(), + )) + .unwrap() + .current(); + let reg = registry(input.clone()); + for this in input.iter().rev().take(10) { + let res = resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + ); + if let Ok(mut res) = res { + let res_len = res.len(); + res.sort_by_key(|s| s.name()); + res.dedup_by_key(|s| s.name()); + dis[min(res_len - res.len(), dis.len() - 1)] += 1; + } + if dis.iter().all(|&x| x > 0) { + return; + } + } + } + panic!( + "In 1280 tries we did not see a wide enough distribution of multiple versions of the same library! dis: {:?}", + dis + ); +} + +/// Assert `xs` contains `elems` +#[track_caller] +pub fn assert_contains<A: PartialEq>(xs: &[A], elems: &[A]) { + for elem in elems { + assert!(xs.contains(elem)); + } +} + +#[track_caller] +pub fn assert_same<A: PartialEq>(a: &[A], b: &[A]) { + assert_eq!(a.len(), b.len()); + assert_contains(b, a); +} diff --git a/crates/resolver-tests/tests/resolve.rs b/crates/resolver-tests/tests/resolve.rs new file mode 100644 index 0000000..9d8f25a --- /dev/null +++ b/crates/resolver-tests/tests/resolve.rs @@ -0,0 +1,1504 @@ +use cargo::core::dependency::DepKind; +use cargo::core::Dependency; +use cargo::util::Config; +use cargo_util::is_ci; + +use resolver_tests::{ + assert_contains, assert_same, dep, dep_kind, dep_loc, dep_req, dep_req_kind, loc_names, names, + pkg, pkg_id, pkg_loc, registry, registry_strategy, remove_dep, resolve, resolve_and_validated, + resolve_with_config, PrettyPrintRegistry, SatResolve, ToDep, ToPkgId, +}; + +use proptest::prelude::*; + +// NOTE: proptest is a form of fuzz testing. It generates random input and makes sure that +// certain universal truths are upheld. Therefore, it can pass when there is a problem, +// but if it fails then there really is something wrong. When testing something as +// complicated as the resolver, the problems can be very subtle and hard to generate. +// We have had a history of these tests only failing on PRs long after a bug is introduced. +// If you have one of these test fail please report it on #6258, +// and if you did not change the resolver then feel free to retry without concern. +proptest! { + #![proptest_config(ProptestConfig { + max_shrink_iters: + if is_ci() || !is_terminal::IsTerminal::is_terminal(&std::io::stderr()){ + // This attempts to make sure that CI will fail fast, + 0 + } else { + // but that local builds will give a small clear test case. + u32::MAX + }, + result_cache: prop::test_runner::basic_result_cache, + .. ProptestConfig::default() + })] + + /// NOTE: if you think this test has failed spuriously see the note at the top of this macro. + #[test] + fn prop_passes_validation( + PrettyPrintRegistry(input) in registry_strategy(50, 20, 60) + ) { + let reg = registry(input.clone()); + let sat_resolve = SatResolve::new(®); + // there is only a small chance that any one + // crate will be interesting. + // So we try some of the most complicated. + for this in input.iter().rev().take(20) { + let _ = resolve_and_validated( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + Some(sat_resolve.clone()), + ); + } + } + + /// NOTE: if you think this test has failed spuriously see the note at the top of this macro. + #[test] + fn prop_minimum_version_errors_the_same( + PrettyPrintRegistry(input) in registry_strategy(50, 20, 60) + ) { + let mut config = Config::default().unwrap(); + config.nightly_features_allowed = true; + config + .configure( + 1, + false, + None, + false, + false, + false, + &None, + &["minimal-versions".to_string()], + &[], + ) + .unwrap(); + + let reg = registry(input.clone()); + // there is only a small chance that any one + // crate will be interesting. + // So we try some of the most complicated. + for this in input.iter().rev().take(10) { + // minimal-versions change what order the candidates + // are tried but not the existence of a solution + let res = resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + ); + + let mres = resolve_with_config( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + &config, + ); + + prop_assert_eq!( + res.is_ok(), + mres.is_ok(), + "minimal-versions and regular resolver disagree about whether `{} = \"={}\"` can resolve", + this.name(), + this.version() + ) + } + } + + /// NOTE: if you think this test has failed spuriously see the note at the top of this macro. + #[test] + fn prop_removing_a_dep_cant_break( + PrettyPrintRegistry(input) in registry_strategy(50, 20, 60), + indexes_to_remove in prop::collection::vec((any::<prop::sample::Index>(), any::<prop::sample::Index>()), ..10) + ) { + let reg = registry(input.clone()); + let mut removed_input = input.clone(); + for (summary_idx, dep_idx) in indexes_to_remove { + if !removed_input.is_empty() { + let summary_idx = summary_idx.index(removed_input.len()); + let deps = removed_input[summary_idx].dependencies(); + if !deps.is_empty() { + let new = remove_dep(&removed_input[summary_idx], dep_idx.index(deps.len())); + removed_input[summary_idx] = new; + } + } + } + let removed_reg = registry(removed_input); + // there is only a small chance that any one + // crate will be interesting. + // So we try some of the most complicated. + for this in input.iter().rev().take(10) { + if resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + ).is_ok() { + prop_assert!( + resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + &removed_reg, + ).is_ok(), + "full index worked for `{} = \"={}\"` but removing some deps broke it!", + this.name(), + this.version(), + ) + } + } + } + + /// NOTE: if you think this test has failed spuriously see the note at the top of this macro. + #[test] + fn prop_limited_independence_of_irrelevant_alternatives( + PrettyPrintRegistry(input) in registry_strategy(50, 20, 60), + indexes_to_unpublish in prop::collection::vec(any::<prop::sample::Index>(), ..10) + ) { + let reg = registry(input.clone()); + // there is only a small chance that any one + // crate will be interesting. + // So we try some of the most complicated. + for this in input.iter().rev().take(10) { + let res = resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + ®, + ); + + match res { + Ok(r) => { + // If resolution was successful, then unpublishing a version of a crate + // that was not selected should not change that. + let not_selected: Vec<_> = input + .iter() + .cloned() + .filter(|x| !r.contains(&x.package_id())) + .collect(); + if !not_selected.is_empty() { + let indexes_to_unpublish: Vec<_> = indexes_to_unpublish.iter().map(|x| x.get(¬_selected)).collect(); + + let new_reg = registry( + input + .iter() + .cloned() + .filter(|x| !indexes_to_unpublish.contains(&x)) + .collect(), + ); + + let res = resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + &new_reg, + ); + + // Note: that we can not assert that the two `res` are identical + // as the resolver does depend on irrelevant alternatives. + // It uses how constrained a dependency requirement is + // to determine what order to evaluate requirements. + + prop_assert!( + res.is_ok(), + "unpublishing {:?} stopped `{} = \"={}\"` from working", + indexes_to_unpublish.iter().map(|x| x.package_id()).collect::<Vec<_>>(), + this.name(), + this.version() + ) + } + } + + Err(_) => { + // If resolution was unsuccessful, then it should stay unsuccessful + // even if any version of a crate is unpublished. + let indexes_to_unpublish: Vec<_> = indexes_to_unpublish.iter().map(|x| x.get(&input)).collect(); + + let new_reg = registry( + input + .iter() + .cloned() + .filter(|x| !indexes_to_unpublish.contains(&x)) + .collect(), + ); + + let res = resolve( + vec![dep_req(&this.name(), &format!("={}", this.version()))], + &new_reg, + ); + + prop_assert!( + res.is_err(), + "full index did not work for `{} = \"={}\"` but unpublishing {:?} fixed it!", + this.name(), + this.version(), + indexes_to_unpublish.iter().map(|x| x.package_id()).collect::<Vec<_>>(), + ) + } + } + } + } +} + +#[test] +#[should_panic(expected = "pub dep")] // The error handling is not yet implemented. +fn pub_fail() { + let input = vec![ + pkg!(("a", "0.0.4")), + pkg!(("a", "0.0.5")), + pkg!(("e", "0.0.6") => [dep_req_kind("a", "<= 0.0.4", DepKind::Normal, true),]), + pkg!(("kB", "0.0.3") => [dep_req("a", ">= 0.0.5"),dep("e"),]), + ]; + let reg = registry(input); + assert!(resolve_and_validated(vec![dep("kB")], ®, None).is_err()); +} + +#[test] +fn basic_public_dependency() { + let reg = registry(vec![ + pkg!(("A", "0.1.0")), + pkg!(("A", "0.2.0")), + pkg!("B" => [dep_req_kind("A", "0.1", DepKind::Normal, true)]), + pkg!("C" => [dep("A"), dep("B")]), + ]); + + let res = resolve_and_validated(vec![dep("C")], ®, None).unwrap(); + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("C", "1.0.0"), + ("B", "1.0.0"), + ("A", "0.1.0"), + ]), + ); +} + +#[test] +fn public_dependency_filling_in() { + // The resolver has an optimization where if a candidate to resolve a dependency + // has already bean activated then we skip looking at the candidates dependencies. + // However, we have to be careful as the new path may make pub dependencies invalid. + + // Triggering this case requires dependencies to be resolved in a specific order. + // Fuzzing found this unintuitive case, that triggers this unfortunate order of operations: + // 1. `d`'s dep on `c` is resolved + // 2. `d`'s dep on `a` is resolved with `0.1.1` + // 3. `c`'s dep on `b` is resolved with `0.0.2` + // 4. `b`'s dep on `a` is resolved with `0.0.6` no pub dev conflict as `b` is private to `c` + // 5. `d`'s dep on `b` is resolved with `0.0.2` triggering the optimization. + // Do we notice that `d` has a pub dep conflict on `a`? Lets try it and see. + let reg = registry(vec![ + pkg!(("a", "0.0.6")), + pkg!(("a", "0.1.1")), + pkg!(("b", "0.0.0") => [dep("bad")]), + pkg!(("b", "0.0.1") => [dep("bad")]), + pkg!(("b", "0.0.2") => [dep_req_kind("a", "=0.0.6", DepKind::Normal, true)]), + pkg!("c" => [dep_req("b", ">=0.0.1")]), + pkg!("d" => [dep("c"), dep("a"), dep("b")]), + ]); + + let res = resolve_and_validated(vec![dep("d")], ®, None).unwrap(); + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("d", "1.0.0"), + ("c", "1.0.0"), + ("b", "0.0.2"), + ("a", "0.0.6"), + ]), + ); +} + +#[test] +fn public_dependency_filling_in_and_update() { + // The resolver has an optimization where if a candidate to resolve a dependency + // has already bean activated then we skip looking at the candidates dependencies. + // However, we have to be careful as the new path may make pub dependencies invalid. + + // Triggering this case requires dependencies to be resolved in a specific order. + // Fuzzing found this unintuitive case, that triggers this unfortunate order of operations: + // 1. `D`'s dep on `B` is resolved + // 2. `D`'s dep on `C` is resolved + // 3. `B`'s dep on `A` is resolved with `0.0.0` + // 4. `C`'s dep on `B` triggering the optimization. + // So did we add `A 0.0.0` to the deps `C` can see? + // Or are we going to resolve `C`'s dep on `A` with `0.0.2`? + // Lets try it and see. + let reg = registry(vec![ + pkg!(("A", "0.0.0")), + pkg!(("A", "0.0.2")), + pkg!("B" => [dep_req_kind("A", "=0.0.0", DepKind::Normal, true),]), + pkg!("C" => [dep("A"),dep("B")]), + pkg!("D" => [dep("B"),dep("C")]), + ]); + let res = resolve_and_validated(vec![dep("D")], ®, None).unwrap(); + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("D", "1.0.0"), + ("C", "1.0.0"), + ("B", "1.0.0"), + ("A", "0.0.0"), + ]), + ); +} + +#[test] +fn public_dependency_skipping() { + // When backtracking due to a failed dependency, if Cargo is + // trying to be clever and skip irrelevant dependencies, care must + // the effects of pub dep must be accounted for. + let input = vec![ + pkg!(("a", "0.2.0")), + pkg!(("a", "2.0.0")), + pkg!(("b", "0.0.0") => [dep("bad")]), + pkg!(("b", "0.2.1") => [dep_req_kind("a", "0.2.0", DepKind::Normal, true)]), + pkg!("c" => [dep("a"),dep("b")]), + ]; + let reg = registry(input); + + resolve_and_validated(vec![dep("c")], ®, None).unwrap(); +} + +#[test] +fn public_dependency_skipping_in_backtracking() { + // When backtracking due to a failed dependency, if Cargo is + // trying to be clever and skip irrelevant dependencies, care must + // the effects of pub dep must be accounted for. + let input = vec![ + pkg!(("A", "0.0.0") => [dep("bad")]), + pkg!(("A", "0.0.1") => [dep("bad")]), + pkg!(("A", "0.0.2") => [dep("bad")]), + pkg!(("A", "0.0.3") => [dep("bad")]), + pkg!(("A", "0.0.4")), + pkg!(("A", "0.0.5")), + pkg!("B" => [dep_req_kind("A", ">= 0.0.3", DepKind::Normal, true)]), + pkg!("C" => [dep_req("A", "<= 0.0.4"), dep("B")]), + ]; + let reg = registry(input); + + resolve_and_validated(vec![dep("C")], ®, None).unwrap(); +} + +#[test] +fn public_sat_topological_order() { + let input = vec![ + pkg!(("a", "0.0.1")), + pkg!(("a", "0.0.0")), + pkg!(("b", "0.0.1") => [dep_req_kind("a", "= 0.0.1", DepKind::Normal, true),]), + pkg!(("b", "0.0.0") => [dep("bad"),]), + pkg!("A" => [dep_req("a", "= 0.0.0"),dep_req_kind("b", "*", DepKind::Normal, true)]), + ]; + + let reg = registry(input); + assert!(resolve_and_validated(vec![dep("A")], ®, None).is_err()); +} + +#[test] +fn public_sat_unused_makes_things_pub() { + let input = vec![ + pkg!(("a", "0.0.1")), + pkg!(("a", "0.0.0")), + pkg!(("b", "8.0.1") => [dep_req_kind("a", "= 0.0.1", DepKind::Normal, true),]), + pkg!(("b", "8.0.0") => [dep_req("a", "= 0.0.1"),]), + pkg!("c" => [dep_req("b", "= 8.0.0"),dep_req("a", "= 0.0.0"),]), + ]; + let reg = registry(input); + + resolve_and_validated(vec![dep("c")], ®, None).unwrap(); +} + +#[test] +fn public_sat_unused_makes_things_pub_2() { + let input = vec![ + pkg!(("c", "0.0.2")), + pkg!(("c", "0.0.1")), + pkg!(("a-sys", "0.0.2")), + pkg!(("a-sys", "0.0.1") => [dep_req_kind("c", "= 0.0.1", DepKind::Normal, true),]), + pkg!("P" => [dep_req_kind("a-sys", "*", DepKind::Normal, true),dep_req("c", "= 0.0.1"),]), + pkg!("A" => [dep("P"),dep_req("c", "= 0.0.2"),]), + ]; + let reg = registry(input); + + resolve_and_validated(vec![dep("A")], ®, None).unwrap(); +} + +#[test] +#[should_panic(expected = "assertion failed: !name.is_empty()")] +fn test_dependency_with_empty_name() { + // Bug 5229, dependency-names must not be empty + "".to_dep(); +} + +#[test] +fn test_resolving_empty_dependency_list() { + let res = resolve(Vec::new(), ®istry(vec![])).unwrap(); + + assert_eq!(res, names(&["root"])); +} + +#[test] +fn test_resolving_only_package() { + let reg = registry(vec![pkg!("foo")]); + let res = resolve(vec![dep("foo")], ®).unwrap(); + assert_same(&res, &names(&["root", "foo"])); +} + +#[test] +fn test_resolving_one_dep() { + let reg = registry(vec![pkg!("foo"), pkg!("bar")]); + let res = resolve(vec![dep("foo")], ®).unwrap(); + assert_same(&res, &names(&["root", "foo"])); +} + +#[test] +fn test_resolving_multiple_deps() { + let reg = registry(vec![pkg!("foo"), pkg!("bar"), pkg!("baz")]); + let res = resolve(vec![dep("foo"), dep("baz")], ®).unwrap(); + assert_same(&res, &names(&["root", "foo", "baz"])); +} + +#[test] +fn test_resolving_transitive_deps() { + let reg = registry(vec![pkg!("foo"), pkg!("bar" => ["foo"])]); + let res = resolve(vec![dep("bar")], ®).unwrap(); + + assert_same(&res, &names(&["root", "foo", "bar"])); +} + +#[test] +fn test_resolving_common_transitive_deps() { + let reg = registry(vec![pkg!("foo" => ["bar"]), pkg!("bar")]); + let res = resolve(vec![dep("foo"), dep("bar")], ®).unwrap(); + + assert_same(&res, &names(&["root", "foo", "bar"])); +} + +#[test] +fn test_resolving_with_same_name() { + let list = vec![ + pkg_loc("foo", "https://first.example.com"), + pkg_loc("bar", "https://second.example.com"), + ]; + + let reg = registry(list); + let res = resolve( + vec![ + dep_loc("foo", "https://first.example.com"), + dep_loc("bar", "https://second.example.com"), + ], + ®, + ) + .unwrap(); + + let mut names = loc_names(&[ + ("foo", "https://first.example.com"), + ("bar", "https://second.example.com"), + ]); + + names.push(pkg_id("root")); + assert_same(&res, &names); +} + +#[test] +fn test_resolving_with_dev_deps() { + let reg = registry(vec![ + pkg!("foo" => ["bar", dep_kind("baz", DepKind::Development)]), + pkg!("baz" => ["bat", dep_kind("bam", DepKind::Development)]), + pkg!("bar"), + pkg!("bat"), + ]); + + let res = resolve( + vec![dep("foo"), dep_kind("baz", DepKind::Development)], + ®, + ) + .unwrap(); + + assert_same(&res, &names(&["root", "foo", "bar", "baz", "bat"])); +} + +#[test] +fn resolving_with_many_versions() { + let reg = registry(vec![pkg!(("foo", "1.0.1")), pkg!(("foo", "1.0.2"))]); + + let res = resolve(vec![dep("foo")], ®).unwrap(); + + assert_same(&res, &names(&[("root", "1.0.0"), ("foo", "1.0.2")])); +} + +#[test] +fn resolving_with_specific_version() { + let reg = registry(vec![pkg!(("foo", "1.0.1")), pkg!(("foo", "1.0.2"))]); + + let res = resolve(vec![dep_req("foo", "=1.0.1")], ®).unwrap(); + + assert_same(&res, &names(&[("root", "1.0.0"), ("foo", "1.0.1")])); +} + +#[test] +fn test_resolving_maximum_version_with_transitive_deps() { + let reg = registry(vec![ + pkg!(("util", "1.2.2")), + pkg!(("util", "1.0.0")), + pkg!(("util", "1.1.1")), + pkg!("foo" => [dep_req("util", "1.0.0")]), + pkg!("bar" => [dep_req("util", ">=1.0.1")]), + ]); + + let res = resolve(vec![dep_req("foo", "1.0.0"), dep_req("bar", "1.0.0")], ®).unwrap(); + + assert_contains( + &res, + &names(&[ + ("root", "1.0.0"), + ("foo", "1.0.0"), + ("bar", "1.0.0"), + ("util", "1.2.2"), + ]), + ); + assert!(!res.contains(&("util", "1.0.1").to_pkgid())); + assert!(!res.contains(&("util", "1.1.1").to_pkgid())); +} + +#[test] +fn test_resolving_minimum_version_with_transitive_deps() { + let reg = registry(vec![ + pkg!(("util", "1.2.2")), + pkg!(("util", "1.0.0")), + pkg!(("util", "1.1.1")), + pkg!("foo" => [dep_req("util", "1.0.0")]), + pkg!("bar" => [dep_req("util", ">=1.0.1")]), + ]); + + let mut config = Config::default().unwrap(); + // -Z minimal-versions + // When the minimal-versions config option is specified then the lowest + // possible version of a package should be selected. "util 1.0.0" can't be + // selected because of the requirements of "bar", so the minimum version + // must be 1.1.1. + config.nightly_features_allowed = true; + config + .configure( + 1, + false, + None, + false, + false, + false, + &None, + &["minimal-versions".to_string()], + &[], + ) + .unwrap(); + + let res = resolve_with_config( + vec![dep_req("foo", "1.0.0"), dep_req("bar", "1.0.0")], + ®, + &config, + ) + .unwrap(); + + assert_contains( + &res, + &names(&[ + ("root", "1.0.0"), + ("foo", "1.0.0"), + ("bar", "1.0.0"), + ("util", "1.1.1"), + ]), + ); + assert!(!res.contains(&("util", "1.2.2").to_pkgid())); + assert!(!res.contains(&("util", "1.0.0").to_pkgid())); +} + +#[test] +fn resolving_incompat_versions() { + let reg = registry(vec![ + pkg!(("foo", "1.0.1")), + pkg!(("foo", "1.0.2")), + pkg!("bar" => [dep_req("foo", "=1.0.2")]), + ]); + + assert!(resolve(vec![dep_req("foo", "=1.0.1"), dep("bar")], ®).is_err()); +} + +#[test] +fn resolving_wrong_case_from_registry() { + // In the future we may #5678 allow this to happen. + // For back compatibility reasons, we probably won't. + // But we may want to future prove ourselves by understanding it. + // This test documents the current behavior. + let reg = registry(vec![pkg!(("foo", "1.0.0")), pkg!("bar" => ["Foo"])]); + + assert!(resolve(vec![dep("bar")], ®).is_err()); +} + +#[test] +fn resolving_mis_hyphenated_from_registry() { + // In the future we may #2775 allow this to happen. + // For back compatibility reasons, we probably won't. + // But we may want to future prove ourselves by understanding it. + // This test documents the current behavior. + let reg = registry(vec![pkg!(("fo-o", "1.0.0")), pkg!("bar" => ["fo_o"])]); + + assert!(resolve(vec![dep("bar")], ®).is_err()); +} + +#[test] +fn resolving_backtrack() { + let reg = registry(vec![ + pkg!(("foo", "1.0.2") => [dep("bar")]), + pkg!(("foo", "1.0.1") => [dep("baz")]), + pkg!("bar" => [dep_req("foo", "=2.0.2")]), + pkg!("baz"), + ]); + + let res = resolve(vec![dep_req("foo", "^1")], ®).unwrap(); + + assert_contains( + &res, + &names(&[("root", "1.0.0"), ("foo", "1.0.1"), ("baz", "1.0.0")]), + ); +} + +#[test] +fn resolving_backtrack_features() { + // test for cargo/issues/4347 + let mut bad = dep("bar"); + bad.set_features(vec!["bad"]); + + let reg = registry(vec![ + pkg!(("foo", "1.0.2") => [bad]), + pkg!(("foo", "1.0.1") => [dep("bar")]), + pkg!("bar"), + ]); + + let res = resolve(vec![dep_req("foo", "^1")], ®).unwrap(); + + assert_contains( + &res, + &names(&[("root", "1.0.0"), ("foo", "1.0.1"), ("bar", "1.0.0")]), + ); +} + +#[test] +fn resolving_allows_multiple_compatible_versions() { + let reg = registry(vec![ + pkg!(("foo", "1.0.0")), + pkg!(("foo", "2.0.0")), + pkg!(("foo", "0.1.0")), + pkg!(("foo", "0.2.0")), + pkg!("bar" => ["d1", "d2", "d3", "d4"]), + pkg!("d1" => [dep_req("foo", "1")]), + pkg!("d2" => [dep_req("foo", "2")]), + pkg!("d3" => [dep_req("foo", "0.1")]), + pkg!("d4" => [dep_req("foo", "0.2")]), + ]); + + let res = resolve(vec![dep("bar")], ®).unwrap(); + + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("foo", "1.0.0"), + ("foo", "2.0.0"), + ("foo", "0.1.0"), + ("foo", "0.2.0"), + ("d1", "1.0.0"), + ("d2", "1.0.0"), + ("d3", "1.0.0"), + ("d4", "1.0.0"), + ("bar", "1.0.0"), + ]), + ); +} + +#[test] +fn resolving_with_deep_backtracking() { + let reg = registry(vec![ + pkg!(("foo", "1.0.1") => [dep_req("bar", "1")]), + pkg!(("foo", "1.0.0") => [dep_req("bar", "2")]), + pkg!(("bar", "1.0.0") => [dep_req("baz", "=1.0.2"), + dep_req("other", "1")]), + pkg!(("bar", "2.0.0") => [dep_req("baz", "=1.0.1")]), + pkg!(("baz", "1.0.2") => [dep_req("other", "2")]), + pkg!(("baz", "1.0.1")), + pkg!(("dep_req", "1.0.0")), + pkg!(("dep_req", "2.0.0")), + ]); + + let res = resolve(vec![dep_req("foo", "1")], ®).unwrap(); + + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("foo", "1.0.0"), + ("bar", "2.0.0"), + ("baz", "1.0.1"), + ]), + ); +} + +#[test] +fn resolving_with_sys_crates() { + // This is based on issues/4902 + // With `l` a normal library we get 2copies so everyone gets the newest compatible. + // But `l-sys` a library with a links attribute we make sure there is only one. + let reg = registry(vec![ + pkg!(("l-sys", "0.9.1")), + pkg!(("l-sys", "0.10.0")), + pkg!(("l", "0.9.1")), + pkg!(("l", "0.10.0")), + pkg!(("d", "1.0.0") => [dep_req("l-sys", ">=0.8.0, <=0.10.0"), dep_req("l", ">=0.8.0, <=0.10.0")]), + pkg!(("r", "1.0.0") => [dep_req("l-sys", "0.9"), dep_req("l", "0.9")]), + ]); + + let res = resolve(vec![dep_req("d", "1"), dep_req("r", "1")], ®).unwrap(); + + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("d", "1.0.0"), + ("r", "1.0.0"), + ("l-sys", "0.9.1"), + ("l", "0.9.1"), + ("l", "0.10.0"), + ]), + ); +} + +#[test] +fn resolving_with_constrained_sibling_backtrack_parent() { + // There is no point in considering all of the backtrack_trap{1,2} + // candidates since they can't change the result of failing to + // resolve 'constrained'. Cargo should (ideally) skip past them and resume + // resolution once the activation of the parent, 'bar', is rolled back. + // Note that the traps are slightly more constrained to make sure they + // get picked first. + let mut reglist = vec![ + pkg!(("foo", "1.0.0") => [dep_req("bar", "1.0"), + dep_req("constrained", "=1.0.0")]), + pkg!(("bar", "1.0.0") => [dep_req("backtrack_trap1", "1.0.2"), + dep_req("backtrack_trap2", "1.0.2"), + dep_req("constrained", "1.0.0")]), + pkg!(("constrained", "1.0.0")), + pkg!(("backtrack_trap1", "1.0.0")), + pkg!(("backtrack_trap2", "1.0.0")), + ]; + // Bump this to make the test harder - it adds more versions of bar that will + // fail to resolve, and more versions of the traps to consider. + const NUM_BARS_AND_TRAPS: usize = 50; // minimum 2 + for i in 1..NUM_BARS_AND_TRAPS { + let vsn = format!("1.0.{}", i); + reglist.push( + pkg!(("bar", vsn.clone()) => [dep_req("backtrack_trap1", "1.0.2"), + dep_req("backtrack_trap2", "1.0.2"), + dep_req("constrained", "1.0.1")]), + ); + reglist.push(pkg!(("backtrack_trap1", vsn.clone()))); + reglist.push(pkg!(("backtrack_trap2", vsn.clone()))); + reglist.push(pkg!(("constrained", vsn.clone()))); + } + let reg = registry(reglist); + + let res = resolve(vec![dep_req("foo", "1")], ®).unwrap(); + + assert_contains( + &res, + &names(&[ + ("root", "1.0.0"), + ("foo", "1.0.0"), + ("bar", "1.0.0"), + ("constrained", "1.0.0"), + ]), + ); +} + +#[test] +fn resolving_with_many_equivalent_backtracking() { + let mut reglist = Vec::new(); + + const DEPTH: usize = 200; + const BRANCHING_FACTOR: usize = 100; + + // Each level depends on the next but the last level does not exist. + // Without cashing we need to test every path to the last level O(BRANCHING_FACTOR ^ DEPTH) + // and this test will time out. With cashing we need to discover that none of these + // can be activated O(BRANCHING_FACTOR * DEPTH) + for l in 0..DEPTH { + let name = format!("level{}", l); + let next = format!("level{}", l + 1); + for i in 1..BRANCHING_FACTOR { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!((name.as_str(), vsn.as_str()) => [dep(next.as_str())])); + } + } + + let reg = registry(reglist.clone()); + + let res = resolve(vec![dep("level0")], ®); + + assert!(res.is_err()); + + // It is easy to write code that quickly returns an error. + // Lets make sure we can find a good answer if it is there. + reglist.push(pkg!(("level0", "1.0.0"))); + + let reg = registry(reglist.clone()); + + let res = resolve(vec![dep("level0")], ®).unwrap(); + + assert_contains(&res, &names(&[("root", "1.0.0"), ("level0", "1.0.0")])); + + // Make sure we have not special case no candidates. + reglist.push(pkg!(("constrained", "1.1.0"))); + reglist.push(pkg!(("constrained", "1.0.0"))); + reglist.push( + pkg!((format!("level{}", DEPTH).as_str(), "1.0.0") => [dep_req("constrained", "=1.0.0")]), + ); + + let reg = registry(reglist.clone()); + + let res = resolve(vec![dep("level0"), dep("constrained")], ®).unwrap(); + + assert_contains( + &res, + &names(&[ + ("root", "1.0.0"), + ("level0", "1.0.0"), + ("constrained", "1.1.0"), + ]), + ); + + let reg = registry(reglist.clone()); + + let res = resolve(vec![dep_req("level0", "1.0.1"), dep("constrained")], ®).unwrap(); + + assert_contains( + &res, + &names(&[ + ("root", "1.0.0"), + (format!("level{}", DEPTH).as_str(), "1.0.0"), + ("constrained", "1.0.0"), + ]), + ); + + let reg = registry(reglist); + + let res = resolve( + vec![dep_req("level0", "1.0.1"), dep_req("constrained", "1.1.0")], + ®, + ); + + assert!(res.is_err()); +} + +#[test] +fn resolving_with_deep_traps() { + let mut reglist = Vec::new(); + + const DEPTH: usize = 200; + const BRANCHING_FACTOR: usize = 100; + + // Each backtrack_trap depends on the next, and adds a backtrack frame. + // None of witch is going to help with `bad`. + for l in 0..DEPTH { + let name = format!("backtrack_trap{}", l); + let next = format!("backtrack_trap{}", l + 1); + for i in 1..BRANCHING_FACTOR { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!((name.as_str(), vsn.as_str()) => [dep(next.as_str())])); + } + } + { + let name = format!("backtrack_trap{}", DEPTH); + for i in 1..BRANCHING_FACTOR { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!((name.as_str(), vsn.as_str()))); + } + } + { + // slightly less constrained to make sure `cloaking` gets picked last. + for i in 1..(BRANCHING_FACTOR + 10) { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!(("cloaking", vsn.as_str()) => [dep_req("bad", "1.0.1")])); + } + } + + let reg = registry(reglist); + + let res = resolve(vec![dep("backtrack_trap0"), dep("cloaking")], ®); + + assert!(res.is_err()); +} + +#[test] +fn resolving_with_constrained_cousins_backtrack() { + let mut reglist = Vec::new(); + + const DEPTH: usize = 100; + const BRANCHING_FACTOR: usize = 50; + + // Each backtrack_trap depends on the next. + // The last depends on a specific ver of constrained. + for l in 0..DEPTH { + let name = format!("backtrack_trap{}", l); + let next = format!("backtrack_trap{}", l + 1); + for i in 1..BRANCHING_FACTOR { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!((name.as_str(), vsn.as_str()) => [dep(next.as_str())])); + } + } + { + let name = format!("backtrack_trap{}", DEPTH); + for i in 1..BRANCHING_FACTOR { + let vsn = format!("1.0.{}", i); + reglist.push( + pkg!((name.as_str(), vsn.as_str()) => [dep_req("constrained", ">=1.1.0, <=2.0.0")]), + ); + } + } + { + // slightly less constrained to make sure `constrained` gets picked last. + for i in 0..(BRANCHING_FACTOR + 10) { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!(("constrained", vsn.as_str()))); + } + reglist.push(pkg!(("constrained", "1.1.0"))); + reglist.push(pkg!(("constrained", "2.0.0"))); + reglist.push(pkg!(("constrained", "2.0.1"))); + } + reglist.push(pkg!(("cloaking", "1.0.0") => [dep_req("constrained", "~1.0.0")])); + + let reg = registry(reglist.clone()); + + // `backtrack_trap0 = "*"` is a lot of ways of saying `constrained = ">=1.1.0, <=2.0.0"` + // but `constrained= "2.0.1"` is already picked. + // Only then to try and solve `constrained= "~1.0.0"` which is incompatible. + let res = resolve( + vec![ + dep("backtrack_trap0"), + dep_req("constrained", "2.0.1"), + dep("cloaking"), + ], + ®, + ); + + assert!(res.is_err()); + + // Each level depends on the next but the last depends on incompatible deps. + // Let's make sure that we can cache that a dep has incompatible deps. + for l in 0..DEPTH { + let name = format!("level{}", l); + let next = format!("level{}", l + 1); + for i in 1..BRANCHING_FACTOR { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!((name.as_str(), vsn.as_str()) => [dep(next.as_str())])); + } + } + reglist.push( + pkg!((format!("level{}", DEPTH).as_str(), "1.0.0") => [dep("backtrack_trap0"), + dep("cloaking") + ]), + ); + + let reg = registry(reglist); + + let res = resolve(vec![dep("level0"), dep_req("constrained", "2.0.1")], ®); + + assert!(res.is_err()); + + let res = resolve(vec![dep("level0"), dep_req("constrained", "2.0.0")], ®).unwrap(); + + assert_contains( + &res, + &names(&[("constrained", "2.0.0"), ("cloaking", "1.0.0")]), + ); +} + +#[test] +fn resolving_with_constrained_sibling_backtrack_activation() { + // It makes sense to resolve most-constrained deps first, but + // with that logic the backtrack traps here come between the two + // attempted resolutions of 'constrained'. When backtracking, + // cargo should skip past them and resume resolution once the + // number of activations for 'constrained' changes. + let mut reglist = vec![ + pkg!(("foo", "1.0.0") => [dep_req("bar", "=1.0.0"), + dep_req("backtrack_trap1", "1.0"), + dep_req("backtrack_trap2", "1.0"), + dep_req("constrained", "<=1.0.60")]), + pkg!(("bar", "1.0.0") => [dep_req("constrained", ">=1.0.60")]), + ]; + // Bump these to make the test harder, but you'll also need to + // change the version constraints on `constrained` above. To correctly + // exercise Cargo, the relationship between the values is: + // NUM_CONSTRAINED - vsn < NUM_TRAPS < vsn + // to make sure the traps are resolved between `constrained`. + const NUM_TRAPS: usize = 45; // min 1 + const NUM_CONSTRAINED: usize = 100; // min 1 + for i in 0..NUM_TRAPS { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!(("backtrack_trap1", vsn.clone()))); + reglist.push(pkg!(("backtrack_trap2", vsn.clone()))); + } + for i in 0..NUM_CONSTRAINED { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!(("constrained", vsn.clone()))); + } + let reg = registry(reglist); + + let res = resolve(vec![dep_req("foo", "1")], ®).unwrap(); + + assert_contains( + &res, + &names(&[ + ("root", "1.0.0"), + ("foo", "1.0.0"), + ("bar", "1.0.0"), + ("constrained", "1.0.60"), + ]), + ); +} + +#[test] +fn resolving_with_public_constrained_sibling() { + // It makes sense to resolve most-constrained deps first, but + // with that logic the backtrack traps here come between the two + // attempted resolutions of 'constrained'. When backtracking, + // cargo should skip past them and resume resolution once the + // number of activations for 'constrained' changes. + let mut reglist = vec![ + pkg!(("foo", "1.0.0") => [dep_req("bar", "=1.0.0"), + dep_req("backtrack_trap1", "1.0"), + dep_req("backtrack_trap2", "1.0"), + dep_req("constrained", "<=60")]), + pkg!(("bar", "1.0.0") => [dep_req_kind("constrained", ">=60", DepKind::Normal, true)]), + ]; + // Bump these to make the test harder, but you'll also need to + // change the version constraints on `constrained` above. To correctly + // exercise Cargo, the relationship between the values is: + // NUM_CONSTRAINED - vsn < NUM_TRAPS < vsn + // to make sure the traps are resolved between `constrained`. + const NUM_TRAPS: usize = 45; // min 1 + const NUM_CONSTRAINED: usize = 100; // min 1 + for i in 0..NUM_TRAPS { + let vsn = format!("1.0.{}", i); + reglist.push(pkg!(("backtrack_trap1", vsn.clone()))); + reglist.push(pkg!(("backtrack_trap2", vsn.clone()))); + } + for i in 0..NUM_CONSTRAINED { + let vsn = format!("{}.0.0", i); + reglist.push(pkg!(("constrained", vsn.clone()))); + } + let reg = registry(reglist); + + let _ = resolve_and_validated(vec![dep_req("foo", "1")], ®, None); +} + +#[test] +fn resolving_with_constrained_sibling_transitive_dep_effects() { + // When backtracking due to a failed dependency, if Cargo is + // trying to be clever and skip irrelevant dependencies, care must + // be taken to not miss the transitive effects of alternatives. E.g. + // in the right-to-left resolution of the graph below, B may + // affect whether D is successfully resolved. + // + // A + // / | \ + // B C D + // | | + // C D + let reg = registry(vec![ + pkg!(("A", "1.0.0") => [dep_req("B", "1.0"), + dep_req("C", "1.0"), + dep_req("D", "1.0.100")]), + pkg!(("B", "1.0.0") => [dep_req("C", ">=1.0.0")]), + pkg!(("B", "1.0.1") => [dep_req("C", ">=1.0.1")]), + pkg!(("C", "1.0.0") => [dep_req("D", "1.0.0")]), + pkg!(("C", "1.0.1") => [dep_req("D", ">=1.0.1,<1.0.100")]), + pkg!(("C", "1.0.2") => [dep_req("D", ">=1.0.2,<1.0.100")]), + pkg!(("D", "1.0.0")), + pkg!(("D", "1.0.1")), + pkg!(("D", "1.0.2")), + pkg!(("D", "1.0.100")), + pkg!(("D", "1.0.101")), + pkg!(("D", "1.0.102")), + pkg!(("D", "1.0.103")), + pkg!(("D", "1.0.104")), + pkg!(("D", "1.0.105")), + ]); + + let res = resolve(vec![dep_req("A", "1")], ®).unwrap(); + + assert_same( + &res, + &names(&[ + ("root", "1.0.0"), + ("A", "1.0.0"), + ("B", "1.0.0"), + ("C", "1.0.0"), + ("D", "1.0.105"), + ]), + ); +} + +#[test] +fn incomplete_information_skipping() { + // When backtracking due to a failed dependency, if Cargo is + // trying to be clever and skip irrelevant dependencies, care must + // be taken to not miss the transitive effects of alternatives. + // Fuzzing discovered that for some reason cargo was skipping based + // on incomplete information in the following case: + // minimized bug found in: + // https://github.com/rust-lang/cargo/commit/003c29b0c71e5ea28fbe8e72c148c755c9f3f8d9 + let input = vec![ + pkg!(("a", "1.0.0")), + pkg!(("a", "1.1.0")), + pkg!("b" => [dep("a")]), + pkg!(("c", "1.0.0")), + pkg!(("c", "1.1.0")), + pkg!("d" => [dep_req("c", "=1.0")]), + pkg!(("e", "1.0.0")), + pkg!(("e", "1.1.0") => [dep_req("c", "1.1")]), + pkg!("to_yank"), + pkg!(("f", "1.0.0") => [ + dep("to_yank"), + dep("d"), + ]), + pkg!(("f", "1.1.0") => [dep("d")]), + pkg!("g" => [ + dep("b"), + dep("e"), + dep("f"), + ]), + ]; + let reg = registry(input.clone()); + + let res = resolve(vec![dep("g")], ®).unwrap(); + let package_to_yank = "to_yank".to_pkgid(); + // this package is not used in the resolution. + assert!(!res.contains(&package_to_yank)); + // so when we yank it + let new_reg = registry( + input + .iter() + .cloned() + .filter(|x| package_to_yank != x.package_id()) + .collect(), + ); + assert_eq!(input.len(), new_reg.len() + 1); + // it should still build + assert!(resolve(vec![dep("g")], &new_reg).is_ok()); +} + +#[test] +fn incomplete_information_skipping_2() { + // When backtracking due to a failed dependency, if Cargo is + // trying to be clever and skip irrelevant dependencies, care must + // be taken to not miss the transitive effects of alternatives. + // Fuzzing discovered that for some reason cargo was skipping based + // on incomplete information in the following case: + // https://github.com/rust-lang/cargo/commit/003c29b0c71e5ea28fbe8e72c148c755c9f3f8d9 + let input = vec![ + pkg!(("b", "3.8.10")), + pkg!(("b", "8.7.4")), + pkg!(("b", "9.4.6")), + pkg!(("c", "1.8.8")), + pkg!(("c", "10.2.5")), + pkg!(("d", "4.1.2") => [ + dep_req("bad", "=6.10.9"), + ]), + pkg!(("d", "5.5.6")), + pkg!(("d", "5.6.10")), + pkg!(("to_yank", "8.0.1")), + pkg!(("to_yank", "8.8.1")), + pkg!(("e", "4.7.8") => [ + dep_req("d", ">=5.5.6, <=5.6.10"), + dep_req("to_yank", "=8.0.1"), + ]), + pkg!(("e", "7.4.9") => [ + dep_req("bad", "=4.7.5"), + ]), + pkg!("f" => [ + dep_req("d", ">=4.1.2, <=5.5.6"), + ]), + pkg!("g" => [ + dep("bad"), + ]), + pkg!(("h", "3.8.3") => [ + dep("g"), + ]), + pkg!(("h", "6.8.3") => [ + dep("f"), + ]), + pkg!(("h", "8.1.9") => [ + dep_req("to_yank", "=8.8.1"), + ]), + pkg!("i" => [ + dep("b"), + dep("c"), + dep("e"), + dep("h"), + ]), + ]; + let reg = registry(input.clone()); + + let res = resolve(vec![dep("i")], ®).unwrap(); + let package_to_yank = ("to_yank", "8.8.1").to_pkgid(); + // this package is not used in the resolution. + assert!(!res.contains(&package_to_yank)); + // so when we yank it + let new_reg = registry( + input + .iter() + .cloned() + .filter(|x| package_to_yank != x.package_id()) + .collect(), + ); + assert_eq!(input.len(), new_reg.len() + 1); + // it should still build + assert!(resolve(vec![dep("i")], &new_reg).is_ok()); +} + +#[test] +fn incomplete_information_skipping_3() { + // When backtracking due to a failed dependency, if Cargo is + // trying to be clever and skip irrelevant dependencies, care must + // be taken to not miss the transitive effects of alternatives. + // Fuzzing discovered that for some reason cargo was skipping based + // on incomplete information in the following case: + // minimized bug found in: + // https://github.com/rust-lang/cargo/commit/003c29b0c71e5ea28fbe8e72c148c755c9f3f8d9 + let input = vec![ + pkg! {("to_yank", "3.0.3")}, + pkg! {("to_yank", "3.3.0")}, + pkg! {("to_yank", "3.3.1")}, + pkg! {("a", "3.3.0") => [ + dep_req("to_yank", "=3.0.3"), + ] }, + pkg! {("a", "3.3.2") => [ + dep_req("to_yank", "<=3.3.0"), + ] }, + pkg! {("b", "0.1.3") => [ + dep_req("a", "=3.3.0"), + ] }, + pkg! {("b", "2.0.2") => [ + dep_req("to_yank", "3.3.0"), + dep("a"), + ] }, + pkg! {("b", "2.3.3") => [ + dep_req("to_yank", "3.3.0"), + dep_req("a", "=3.3.0"), + ] }, + ]; + let reg = registry(input.clone()); + + let res = resolve(vec![dep("b")], ®).unwrap(); + let package_to_yank = ("to_yank", "3.0.3").to_pkgid(); + // this package is not used in the resolution. + assert!(!res.contains(&package_to_yank)); + // so when we yank it + let new_reg = registry( + input + .iter() + .cloned() + .filter(|x| package_to_yank != x.package_id()) + .collect(), + ); + assert_eq!(input.len(), new_reg.len() + 1); + // it should still build + assert!(resolve(vec![dep("b")], &new_reg).is_ok()); +} + +#[test] +fn resolving_but_no_exists() { + let reg = registry(vec![]); + + let res = resolve(vec![dep_req("foo", "1")], ®); + assert!(res.is_err()); + + assert_eq!( + res.err().unwrap().to_string(), + "no matching package named `foo` found\n\ + location searched: registry `https://example.com/`\n\ + required by package `root v1.0.0 (registry `https://example.com/`)`\ + " + ); +} + +#[test] +fn resolving_cycle() { + let reg = registry(vec![pkg!("foo" => ["foo"])]); + + let _ = resolve(vec![dep_req("foo", "1")], ®); +} + +#[test] +fn hard_equality() { + let reg = registry(vec![ + pkg!(("foo", "1.0.1")), + pkg!(("foo", "1.0.0")), + pkg!(("bar", "1.0.0") => [dep_req("foo", "1.0.0")]), + ]); + + let res = resolve(vec![dep_req("bar", "1"), dep_req("foo", "=1.0.0")], ®).unwrap(); + + assert_same( + &res, + &names(&[("root", "1.0.0"), ("foo", "1.0.0"), ("bar", "1.0.0")]), + ); +} + +#[test] +fn large_conflict_cache() { + let mut input = vec![ + pkg!(("last", "0.0.0") => [dep("bad")]), // just to make sure last is less constrained + ]; + let mut root_deps = vec![dep("last")]; + const NUM_VERSIONS: u8 = 20; + for name in 0..=NUM_VERSIONS { + // a large number of conflicts can easily be generated by a sys crate. + let sys_name = format!("{}-sys", (b'a' + name) as char); + let in_len = input.len(); + input.push(pkg!(("last", format!("{}.0.0", in_len)) => [dep_req(&sys_name, "=0.0.0")])); + root_deps.push(dep_req(&sys_name, ">= 0.0.1")); + + // a large number of conflicts can also easily be generated by a major release version. + let plane_name = format!("{}", (b'a' + name) as char); + let in_len = input.len(); + input.push(pkg!(("last", format!("{}.0.0", in_len)) => [dep_req(&plane_name, "=1.0.0")])); + root_deps.push(dep_req(&plane_name, ">= 1.0.1")); + + for i in 0..=NUM_VERSIONS { + input.push(pkg!((&sys_name, format!("{}.0.0", i)))); + input.push(pkg!((&plane_name, format!("1.0.{}", i)))); + } + } + let reg = registry(input); + let _ = resolve(root_deps, ®); +} + +#[test] +fn off_by_one_bug() { + let input = vec![ + pkg!(("A-sys", "0.0.1")), + pkg!(("A-sys", "0.0.4")), + pkg!(("A-sys", "0.0.6")), + pkg!(("A-sys", "0.0.7")), + pkg!(("NA", "0.0.0") => [dep_req("A-sys", "<= 0.0.5"),]), + pkg!(("NA", "0.0.1") => [dep_req("A-sys", ">= 0.0.6, <= 0.0.8"),]), + pkg!(("a", "0.0.1")), + pkg!(("a", "0.0.2")), + pkg!(("aa", "0.0.0") => [dep_req("A-sys", ">= 0.0.4, <= 0.0.6"),dep_req("NA", "<= 0.0.0"),]), + pkg!(("f", "0.0.3") => [dep("NA"),dep_req("a", "<= 0.0.2"),dep("aa"),]), + ]; + + let reg = registry(input); + let _ = resolve_and_validated(vec![dep("f")], ®, None); +} + +#[test] +fn conflict_store_bug() { + let input = vec![ + pkg!(("A", "0.0.3")), + pkg!(("A", "0.0.5")), + pkg!(("A", "0.0.9") => [dep("bad"),]), + pkg!(("A", "0.0.10") => [dep("bad"),]), + pkg!(("L-sys", "0.0.1") => [dep("bad"),]), + pkg!(("L-sys", "0.0.5")), + pkg!(("R", "0.0.4") => [ + dep_req("L-sys", "= 0.0.5"), + ]), + pkg!(("R", "0.0.6")), + pkg!(("a-sys", "0.0.5")), + pkg!(("a-sys", "0.0.11")), + pkg!(("c", "0.0.12") => [ + dep_req("R", ">= 0.0.3, <= 0.0.4"), + ]), + pkg!(("c", "0.0.13") => [ + dep_req("a-sys", ">= 0.0.8, <= 0.0.11"), + ]), + pkg!(("c0", "0.0.6") => [ + dep_req("L-sys", "<= 0.0.2"), + ]), + pkg!(("c0", "0.0.10") => [ + dep_req("A", ">= 0.0.9, <= 0.0.10"), + dep_req("a-sys", "= 0.0.5"), + ]), + pkg!("j" => [ + dep_req("A", ">= 0.0.3, <= 0.0.5"), + dep_req("R", ">=0.0.4, <= 0.0.6"), + dep_req("c", ">= 0.0.9"), + dep_req("c0", ">= 0.0.6"), + ]), + ]; + + let reg = registry(input); + let _ = resolve_and_validated(vec![dep("j")], ®, None); +} + +#[test] +fn conflict_store_more_then_one_match() { + let input = vec![ + pkg!(("A", "0.0.0")), + pkg!(("A", "0.0.1")), + pkg!(("A-sys", "0.0.0")), + pkg!(("A-sys", "0.0.1")), + pkg!(("A-sys", "0.0.2")), + pkg!(("A-sys", "0.0.3")), + pkg!(("A-sys", "0.0.12")), + pkg!(("A-sys", "0.0.16")), + pkg!(("B-sys", "0.0.0")), + pkg!(("B-sys", "0.0.1")), + pkg!(("B-sys", "0.0.2") => [dep_req("A-sys", "= 0.0.12"),]), + pkg!(("BA-sys", "0.0.0") => [dep_req("A-sys","= 0.0.16"),]), + pkg!(("BA-sys", "0.0.1") => [dep("bad"),]), + pkg!(("BA-sys", "0.0.2") => [dep("bad"),]), + pkg!("nA" => [ + dep("A"), + dep_req("A-sys", "<= 0.0.3"), + dep("B-sys"), + dep("BA-sys"), + ]), + ]; + let reg = registry(input); + let _ = resolve_and_validated(vec![dep("nA")], ®, None); +} + +#[test] +fn bad_lockfile_from_8249() { + let input = vec![ + pkg!(("a-sys", "0.2.0")), + pkg!(("a-sys", "0.1.0")), + pkg!(("b", "0.1.0") => [ + dep_req("a-sys", "0.1"), // should be optional: true, but not deeded for now + ]), + pkg!(("c", "1.0.0") => [ + dep_req("b", "=0.1.0"), + ]), + pkg!("foo" => [ + dep_req("a-sys", "=0.2.0"), + { + let mut b = dep_req("b", "=0.1.0"); + b.set_features(vec!["a-sys"]); + b + }, + dep_req("c", "=1.0.0"), + ]), + ]; + let reg = registry(input); + let _ = resolve_and_validated(vec![dep("foo")], ®, None); +} + +#[test] +fn cyclic_good_error_message() { + let input = vec![ + pkg!(("A", "0.0.0") => [dep("C")]), + pkg!(("B", "0.0.0") => [dep("C")]), + pkg!(("C", "0.0.0") => [dep("A")]), + ]; + let reg = registry(input); + let error = resolve(vec![dep("A"), dep("B")], ®).unwrap_err(); + println!("{}", error); + assert_eq!("\ +cyclic package dependency: package `A v0.0.0 (registry `https://example.com/`)` depends on itself. Cycle: +package `A v0.0.0 (registry `https://example.com/`)` + ... which satisfies dependency `A = \"*\"` of package `C v0.0.0 (registry `https://example.com/`)` + ... which satisfies dependency `C = \"*\"` of package `A v0.0.0 (registry `https://example.com/`)`\ +", error.to_string()); +} |