diff options
Diffstat (limited to 'compiler/rustc_parse/src')
-rw-r--r-- | compiler/rustc_parse/src/errors.rs | 55 | ||||
-rw-r--r-- | compiler/rustc_parse/src/lexer/diagnostics.rs | 7 | ||||
-rw-r--r-- | compiler/rustc_parse/src/lexer/mod.rs | 14 | ||||
-rw-r--r-- | compiler/rustc_parse/src/lexer/tokentrees.rs | 80 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/diagnostics.rs | 211 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/expr.rs | 239 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/item.rs | 106 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/mod.rs | 35 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/nonterminal.rs | 37 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/pat.rs | 37 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/path.rs | 38 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/stmt.rs | 60 | ||||
-rw-r--r-- | compiler/rustc_parse/src/parser/ty.rs | 18 |
13 files changed, 672 insertions, 265 deletions
diff --git a/compiler/rustc_parse/src/errors.rs b/compiler/rustc_parse/src/errors.rs index 7c75e440a..8ab1ec298 100644 --- a/compiler/rustc_parse/src/errors.rs +++ b/compiler/rustc_parse/src/errors.rs @@ -431,6 +431,17 @@ pub(crate) struct ExpectedElseBlock { } #[derive(Diagnostic)] +#[diag(parse_expected_struct_field)] +pub(crate) struct ExpectedStructField { + #[primary_span] + #[label] + pub span: Span, + pub token: Token, + #[label(parse_ident_label)] + pub ident_span: Span, +} + +#[derive(Diagnostic)] #[diag(parse_outer_attribute_not_allowed_on_if_else)] pub(crate) struct OuterAttributeNotAllowedOnIfElse { #[primary_span] @@ -510,6 +521,14 @@ pub(crate) struct CatchAfterTry { } #[derive(Diagnostic)] +#[diag(parse_gen_fn)] +#[help] +pub(crate) struct GenFn { + #[primary_span] + pub span: Span, +} + +#[derive(Diagnostic)] #[diag(parse_comma_after_base_struct)] #[note] pub(crate) struct CommaAfterBaseStruct { @@ -1359,6 +1378,34 @@ pub(crate) struct FnPtrWithGenericsSugg { pub for_param_list_exists: bool, } +pub(crate) struct FnTraitMissingParen { + pub span: Span, + pub machine_applicable: bool, +} + +impl AddToDiagnostic for FnTraitMissingParen { + fn add_to_diagnostic_with<F>(self, diag: &mut rustc_errors::Diagnostic, _: F) + where + F: Fn( + &mut rustc_errors::Diagnostic, + rustc_errors::SubdiagnosticMessage, + ) -> rustc_errors::SubdiagnosticMessage, + { + diag.span_label(self.span, crate::fluent_generated::parse_fn_trait_missing_paren); + let applicability = if self.machine_applicable { + Applicability::MachineApplicable + } else { + Applicability::MaybeIncorrect + }; + diag.span_suggestion_short( + self.span.shrink_to_hi(), + crate::fluent_generated::parse_add_paren, + "()", + applicability, + ); + } +} + #[derive(Diagnostic)] #[diag(parse_unexpected_if_with_if)] pub(crate) struct UnexpectedIfWithIf( @@ -1533,6 +1580,14 @@ pub(crate) enum AmbiguousMissingKwForItemSub { } #[derive(Diagnostic)] +#[diag(parse_missing_fn_params)] +pub(crate) struct MissingFnParams { + #[primary_span] + #[suggestion(code = "()", applicability = "machine-applicable", style = "short")] + pub span: Span, +} + +#[derive(Diagnostic)] #[diag(parse_missing_trait_in_trait_impl)] pub(crate) struct MissingTraitInTraitImpl { #[primary_span] diff --git a/compiler/rustc_parse/src/lexer/diagnostics.rs b/compiler/rustc_parse/src/lexer/diagnostics.rs index b50bb47f2..b1bd4ac75 100644 --- a/compiler/rustc_parse/src/lexer/diagnostics.rs +++ b/compiler/rustc_parse/src/lexer/diagnostics.rs @@ -111,9 +111,10 @@ pub fn report_suspicious_mismatch_block( // If there is no suspicious span, give the last properly closed block may help if let Some(parent) = diag_info.matching_block_spans.last() && diag_info.open_braces.last().is_none() - && diag_info.empty_block_spans.iter().all(|&sp| sp != parent.0.to(parent.1)) { - err.span_label(parent.0, "this opening brace..."); - err.span_label(parent.1, "...matches this closing brace"); + && diag_info.empty_block_spans.iter().all(|&sp| sp != parent.0.to(parent.1)) + { + err.span_label(parent.0, "this opening brace..."); + err.span_label(parent.1, "...matches this closing brace"); } } } diff --git a/compiler/rustc_parse/src/lexer/mod.rs b/compiler/rustc_parse/src/lexer/mod.rs index a375a1d69..f2eed5c9b 100644 --- a/compiler/rustc_parse/src/lexer/mod.rs +++ b/compiler/rustc_parse/src/lexer/mod.rs @@ -64,10 +64,10 @@ pub(crate) fn parse_token_trees<'a>( override_span, nbsp_is_whitespace: false, }; - let (token_trees, unmatched_delims) = + let (stream, res, unmatched_delims) = tokentrees::TokenTreesReader::parse_all_token_trees(string_reader); - match token_trees { - Ok(stream) if unmatched_delims.is_empty() => Ok(stream), + match res { + Ok(()) if unmatched_delims.is_empty() => Ok(stream), _ => { // Return error if there are unmatched delimiters or unclosed delimiters. // We emit delimiter mismatch errors first, then emit the unclosing delimiter mismatch @@ -79,9 +79,11 @@ pub(crate) fn parse_token_trees<'a>( err.buffer(&mut buffer); } } - if let Err(err) = token_trees { - // Add unclosing delimiter error - err.buffer(&mut buffer); + if let Err(errs) = res { + // Add unclosing delimiter or diff marker errors + for err in errs { + err.buffer(&mut buffer); + } } Err(buffer) } diff --git a/compiler/rustc_parse/src/lexer/tokentrees.rs b/compiler/rustc_parse/src/lexer/tokentrees.rs index 07910113d..31d91fe80 100644 --- a/compiler/rustc_parse/src/lexer/tokentrees.rs +++ b/compiler/rustc_parse/src/lexer/tokentrees.rs @@ -5,7 +5,7 @@ use super::{StringReader, UnmatchedDelim}; use rustc_ast::token::{self, Delimiter, Token}; use rustc_ast::tokenstream::{DelimSpan, Spacing, TokenStream, TokenTree}; use rustc_ast_pretty::pprust::token_to_string; -use rustc_errors::{PErr, PResult}; +use rustc_errors::PErr; pub(super) struct TokenTreesReader<'a> { string_reader: StringReader<'a>, @@ -18,36 +18,42 @@ pub(super) struct TokenTreesReader<'a> { impl<'a> TokenTreesReader<'a> { pub(super) fn parse_all_token_trees( string_reader: StringReader<'a>, - ) -> (PResult<'a, TokenStream>, Vec<UnmatchedDelim>) { + ) -> (TokenStream, Result<(), Vec<PErr<'a>>>, Vec<UnmatchedDelim>) { let mut tt_reader = TokenTreesReader { string_reader, token: Token::dummy(), diag_info: TokenTreeDiagInfo::default(), }; - let res = tt_reader.parse_token_trees(/* is_delimited */ false); - (res, tt_reader.diag_info.unmatched_delims) + let (stream, res) = tt_reader.parse_token_trees(/* is_delimited */ false); + (stream, res, tt_reader.diag_info.unmatched_delims) } // Parse a stream of tokens into a list of `TokenTree`s. - fn parse_token_trees(&mut self, is_delimited: bool) -> PResult<'a, TokenStream> { + fn parse_token_trees( + &mut self, + is_delimited: bool, + ) -> (TokenStream, Result<(), Vec<PErr<'a>>>) { self.token = self.string_reader.next_token().0; let mut buf = Vec::new(); loop { match self.token.kind { - token::OpenDelim(delim) => buf.push(self.parse_token_tree_open_delim(delim)?), + token::OpenDelim(delim) => { + buf.push(match self.parse_token_tree_open_delim(delim) { + Ok(val) => val, + Err(errs) => return (TokenStream::new(buf), Err(errs)), + }) + } token::CloseDelim(delim) => { - return if is_delimited { - Ok(TokenStream::new(buf)) - } else { - Err(self.close_delim_err(delim)) - }; + return ( + TokenStream::new(buf), + if is_delimited { Ok(()) } else { Err(vec![self.close_delim_err(delim)]) }, + ); } token::Eof => { - return if is_delimited { - Err(self.eof_err()) - } else { - Ok(TokenStream::new(buf)) - }; + return ( + TokenStream::new(buf), + if is_delimited { Err(vec![self.eof_err()]) } else { Ok(()) }, + ); } _ => { // Get the next normal token. This might require getting multiple adjacent @@ -55,16 +61,14 @@ impl<'a> TokenTreesReader<'a> { let (this_spacing, next_tok) = loop { let (next_tok, is_next_tok_preceded_by_whitespace) = self.string_reader.next_token(); - if !is_next_tok_preceded_by_whitespace { - if let Some(glued) = self.token.glue(&next_tok) { - self.token = glued; - } else { - let this_spacing = - if next_tok.is_op() { Spacing::Joint } else { Spacing::Alone }; - break (this_spacing, next_tok); - } - } else { + if is_next_tok_preceded_by_whitespace { break (Spacing::Alone, next_tok); + } else if let Some(glued) = self.token.glue(&next_tok) { + self.token = glued; + } else { + let this_spacing = + if next_tok.is_punct() { Spacing::Joint } else { Spacing::Alone }; + break (this_spacing, next_tok); } }; let this_tok = std::mem::replace(&mut self.token, next_tok); @@ -99,7 +103,10 @@ impl<'a> TokenTreesReader<'a> { err } - fn parse_token_tree_open_delim(&mut self, open_delim: Delimiter) -> PResult<'a, TokenTree> { + fn parse_token_tree_open_delim( + &mut self, + open_delim: Delimiter, + ) -> Result<TokenTree, Vec<PErr<'a>>> { // The span for beginning of the delimited section let pre_span = self.token.span; @@ -108,7 +115,26 @@ impl<'a> TokenTreesReader<'a> { // Parse the token trees within the delimiters. // We stop at any delimiter so we can try to recover if the user // uses an incorrect delimiter. - let tts = self.parse_token_trees(/* is_delimited */ true)?; + let (tts, res) = self.parse_token_trees(/* is_delimited */ true); + if let Err(mut errs) = res { + // If there are unclosed delims, see if there are diff markers and if so, point them + // out instead of complaining about the unclosed delims. + let mut parser = crate::stream_to_parser(self.string_reader.sess, tts, None); + let mut diff_errs = vec![]; + while parser.token != token::Eof { + if let Err(diff_err) = parser.err_diff_marker() { + diff_errs.push(diff_err); + } + parser.bump(); + } + if !diff_errs.is_empty() { + errs.iter_mut().for_each(|err| { + err.delay_as_bug(); + }); + return Err(diff_errs); + } + return Err(errs); + } // Expand to cover the entire delimited token tree let delim_span = DelimSpan::from_pair(pre_span, self.token.span); diff --git a/compiler/rustc_parse/src/parser/diagnostics.rs b/compiler/rustc_parse/src/parser/diagnostics.rs index 06b1b1523..2a8eb6edd 100644 --- a/compiler/rustc_parse/src/parser/diagnostics.rs +++ b/compiler/rustc_parse/src/parser/diagnostics.rs @@ -39,7 +39,7 @@ use rustc_errors::{ use rustc_session::errors::ExprParenthesesNeeded; use rustc_span::source_map::Spanned; use rustc_span::symbol::{kw, sym, Ident}; -use rustc_span::{Span, SpanSnippetError, Symbol, DUMMY_SP}; +use rustc_span::{BytePos, Span, SpanSnippetError, Symbol, DUMMY_SP}; use std::mem::take; use std::ops::{Deref, DerefMut}; use thin_vec::{thin_vec, ThinVec}; @@ -314,11 +314,10 @@ impl<'a> Parser<'a> { // which uses `Symbol::to_ident_string()` and "helpfully" adds an implicit `r#` let ident_name = ident.name.to_string(); - Some(SuggEscapeIdentifier { - span: ident.span.shrink_to_lo(), - ident_name - }) - } else { None }; + Some(SuggEscapeIdentifier { span: ident.span.shrink_to_lo(), ident_name }) + } else { + None + }; let suggest_remove_comma = if self.token == token::Comma && self.look_ahead(1, |t| t.is_ident()) { @@ -375,9 +374,11 @@ impl<'a> Parser<'a> { // and current token should be Ident with the item name (i.e. the function name) // if there is a `<` after the fn name, then don't show a suggestion, show help - if !self.look_ahead(1, |t| *t == token::Lt) && - let Ok(snippet) = self.sess.source_map().span_to_snippet(generic.span) { - err.multipart_suggestion_verbose( + if !self.look_ahead(1, |t| *t == token::Lt) + && let Ok(snippet) = + self.sess.source_map().span_to_snippet(generic.span) + { + err.multipart_suggestion_verbose( format!("place the generic parameter name after the {ident_name} name"), vec![ (self.token.span.shrink_to_hi(), snippet), @@ -385,11 +386,11 @@ impl<'a> Parser<'a> { ], Applicability::MaybeIncorrect, ); - } else { - err.help(format!( - "place the generic parameter name after the {ident_name} name" - )); - } + } else { + err.help(format!( + "place the generic parameter name after the {ident_name} name" + )); + } } } Err(err) => { @@ -402,7 +403,9 @@ impl<'a> Parser<'a> { } } - if let Some(recovered_ident) = recovered_ident && recover { + if let Some(recovered_ident) = recovered_ident + && recover + { err.emit(); Ok(recovered_ident) } else { @@ -501,8 +504,10 @@ impl<'a> Parser<'a> { // Special-case "expected `;`" errors if expected.contains(&TokenType::Token(token::Semi)) { - if self.prev_token == token::Question && self.maybe_recover_from_ternary_operator() { - return Ok(true); + // If the user is trying to write a ternary expression, recover it and + // return an Err to prevent a cascade of irrelevant diagnostics + if self.prev_token == token::Question && let Err(e) = self.maybe_recover_from_ternary_operator() { + return Err(e); } if self.token.span == DUMMY_SP || self.prev_token.span == DUMMY_SP { @@ -617,19 +622,19 @@ impl<'a> Parser<'a> { } if let TokenKind::Ident(prev, _) = &self.prev_token.kind - && let TokenKind::Ident(cur, _) = &self.token.kind + && let TokenKind::Ident(cur, _) = &self.token.kind { - let concat = Symbol::intern(&format!("{prev}{cur}")); - let ident = Ident::new(concat, DUMMY_SP); - if ident.is_used_keyword() || ident.is_reserved() || ident.is_raw_guess() { - let span = self.prev_token.span.to(self.token.span); - err.span_suggestion_verbose( - span, - format!("consider removing the space to spell keyword `{concat}`"), - concat, - Applicability::MachineApplicable, - ); - } + let concat = Symbol::intern(&format!("{prev}{cur}")); + let ident = Ident::new(concat, DUMMY_SP); + if ident.is_used_keyword() || ident.is_reserved() || ident.is_raw_guess() { + let span = self.prev_token.span.to(self.token.span); + err.span_suggestion_verbose( + span, + format!("consider removing the space to spell keyword `{concat}`"), + concat, + Applicability::MachineApplicable, + ); + } } // `pub` may be used for an item or `pub(crate)` @@ -645,6 +650,26 @@ impl<'a> Parser<'a> { ); } + if let token::DocComment(kind, style, _) = self.token.kind { + // We have something like `expr //!val` where the user likely meant `expr // !val` + let pos = self.token.span.lo() + BytePos(2); + let span = self.token.span.with_lo(pos).with_hi(pos); + err.span_suggestion_verbose( + span, + format!( + "add a space before {} to write a regular comment", + match (kind, style) { + (token::CommentKind::Line, ast::AttrStyle::Inner) => "`!`", + (token::CommentKind::Block, ast::AttrStyle::Inner) => "`!`", + (token::CommentKind::Line, ast::AttrStyle::Outer) => "the last `/`", + (token::CommentKind::Block, ast::AttrStyle::Outer) => "the last `*`", + }, + ), + " ".to_string(), + Applicability::MachineApplicable, + ); + } + // Add suggestion for a missing closing angle bracket if '>' is included in expected_tokens // there are unclosed angle brackets if self.unmatched_angle_bracket_count > 0 @@ -827,6 +852,65 @@ impl<'a> Parser<'a> { None } + pub(super) fn recover_closure_body( + &mut self, + mut err: DiagnosticBuilder<'a, ErrorGuaranteed>, + before: token::Token, + prev: token::Token, + token: token::Token, + lo: Span, + decl_hi: Span, + ) -> PResult<'a, P<Expr>> { + err.span_label(lo.to(decl_hi), "while parsing the body of this closure"); + match before.kind { + token::OpenDelim(Delimiter::Brace) + if !matches!(token.kind, token::OpenDelim(Delimiter::Brace)) => + { + // `{ || () }` should have been `|| { () }` + err.multipart_suggestion( + "you might have meant to open the body of the closure, instead of enclosing \ + the closure in a block", + vec![ + (before.span, String::new()), + (prev.span.shrink_to_hi(), " {".to_string()), + ], + Applicability::MaybeIncorrect, + ); + err.emit(); + self.eat_to_tokens(&[&token::CloseDelim(Delimiter::Brace)]); + } + token::OpenDelim(Delimiter::Parenthesis) + if !matches!(token.kind, token::OpenDelim(Delimiter::Brace)) => + { + // We are within a function call or tuple, we can emit the error + // and recover. + self.eat_to_tokens(&[&token::CloseDelim(Delimiter::Parenthesis), &token::Comma]); + + err.multipart_suggestion_verbose( + "you might have meant to open the body of the closure", + vec![ + (prev.span.shrink_to_hi(), " {".to_string()), + (self.token.span.shrink_to_lo(), "}".to_string()), + ], + Applicability::MaybeIncorrect, + ); + err.emit(); + } + _ if !matches!(token.kind, token::OpenDelim(Delimiter::Brace)) => { + // We don't have a heuristic to correctly identify where the block + // should be closed. + err.multipart_suggestion_verbose( + "you might have meant to open the body of the closure", + vec![(prev.span.shrink_to_hi(), " {".to_string())], + Applicability::HasPlaceholders, + ); + return Err(err); + } + _ => return Err(err), + } + Ok(self.mk_expr_err(lo.to(self.token.span))) + } + /// Eats and discards tokens until one of `kets` is encountered. Respects token trees, /// passes through any errors encountered. Used for error recovery. pub(super) fn eat_to_tokens(&mut self, kets: &[&TokenKind]) { @@ -1025,8 +1109,7 @@ impl<'a> Parser<'a> { .emit(); match self.parse_expr() { Ok(_) => { - *expr = - self.mk_expr_err(expr.span.to(self.prev_token.span)); + *expr = self.mk_expr_err(expr.span.to(self.prev_token.span)); return Ok(()); } Err(err) => { @@ -1218,7 +1301,9 @@ impl<'a> Parser<'a> { return if token::ModSep == self.token.kind { // We have some certainty that this was a bad turbofish at this point. // `foo< bar >::` - if let ExprKind::Binary(o, ..) = inner_op.kind && o.node == BinOpKind::Lt { + if let ExprKind::Binary(o, ..) = inner_op.kind + && o.node == BinOpKind::Lt + { err.suggest_turbofish = Some(op.span.shrink_to_lo()); } else { err.help_turbofish = Some(()); @@ -1248,7 +1333,9 @@ impl<'a> Parser<'a> { } else if token::OpenDelim(Delimiter::Parenthesis) == self.token.kind { // We have high certainty that this was a bad turbofish at this point. // `foo< bar >(` - if let ExprKind::Binary(o, ..) = inner_op.kind && o.node == BinOpKind::Lt { + if let ExprKind::Binary(o, ..) = inner_op.kind + && o.node == BinOpKind::Lt + { err.suggest_turbofish = Some(op.span.shrink_to_lo()); } else { err.help_turbofish = Some(()); @@ -1343,10 +1430,10 @@ impl<'a> Parser<'a> { /// Rust has no ternary operator (`cond ? then : else`). Parse it and try /// to recover from it if `then` and `else` are valid expressions. Returns - /// whether it was a ternary operator. - pub(super) fn maybe_recover_from_ternary_operator(&mut self) -> bool { + /// an err if this appears to be a ternary expression. + pub(super) fn maybe_recover_from_ternary_operator(&mut self) -> PResult<'a, ()> { if self.prev_token != token::Question { - return false; + return PResult::Ok(()); } let lo = self.prev_token.span.lo(); @@ -1364,20 +1451,18 @@ impl<'a> Parser<'a> { if self.eat_noexpect(&token::Colon) { match self.parse_expr() { Ok(_) => { - self.sess.emit_err(TernaryOperator { span: self.token.span.with_lo(lo) }); - return true; + return Err(self + .sess + .create_err(TernaryOperator { span: self.token.span.with_lo(lo) })); } Err(err) => { err.cancel(); - self.restore_snapshot(snapshot); } }; } - } else { - self.restore_snapshot(snapshot); - }; - - false + } + self.restore_snapshot(snapshot); + Ok(()) } pub(super) fn maybe_recover_from_bad_type_plus(&mut self, ty: &Ty) -> PResult<'a, ()> { @@ -1826,19 +1911,21 @@ impl<'a> Parser<'a> { let sm = self.sess.source_map(); let left = begin_par_sp; let right = self.prev_token.span; - let left_snippet = if let Ok(snip) = sm.span_to_prev_source(left) && - !snip.ends_with(' ') { - " ".to_string() - } else { - "".to_string() - }; + let left_snippet = if let Ok(snip) = sm.span_to_prev_source(left) + && !snip.ends_with(' ') + { + " ".to_string() + } else { + "".to_string() + }; - let right_snippet = if let Ok(snip) = sm.span_to_next_source(right) && - !snip.starts_with(' ') { - " ".to_string() - } else { - "".to_string() - }; + let right_snippet = if let Ok(snip) = sm.span_to_next_source(right) + && !snip.starts_with(' ') + { + " ".to_string() + } else { + "".to_string() + }; self.sess.emit_err(ParenthesesInForHead { span: vec![left, right], @@ -2721,8 +2808,15 @@ impl<'a> Parser<'a> { } pub fn recover_diff_marker(&mut self) { + if let Err(mut err) = self.err_diff_marker() { + err.emit(); + FatalError.raise(); + } + } + + pub fn err_diff_marker(&mut self) -> PResult<'a, ()> { let Some(start) = self.diff_marker(&TokenKind::BinOp(token::Shl), &TokenKind::Lt) else { - return; + return Ok(()); }; let mut spans = Vec::with_capacity(3); spans.push(start); @@ -2769,8 +2863,7 @@ impl<'a> Parser<'a> { "for an explanation on these markers from the `git` documentation, visit \ <https://git-scm.com/book/en/v2/Git-Tools-Advanced-Merging#_checking_out_conflicts>", ); - err.emit(); - FatalError.raise() + Err(err) } /// Parse and throw away a parenthesized comma separated diff --git a/compiler/rustc_parse/src/parser/expr.rs b/compiler/rustc_parse/src/parser/expr.rs index f4cee3a66..235b28b6e 100644 --- a/compiler/rustc_parse/src/parser/expr.rs +++ b/compiler/rustc_parse/src/parser/expr.rs @@ -9,7 +9,7 @@ use super::{ use crate::errors; use crate::maybe_recover_from_interpolated_ty_qpath; use ast::mut_visit::{noop_visit_expr, MutVisitor}; -use ast::{Path, PathSegment}; +use ast::{GenBlockKind, Path, PathSegment}; use core::mem; use rustc_ast::ptr::P; use rustc_ast::token::{self, Delimiter, Token, TokenKind}; @@ -32,10 +32,10 @@ use rustc_macros::Subdiagnostic; use rustc_session::errors::{report_lit_error, ExprParenthesesNeeded}; use rustc_session::lint::builtin::BREAK_WITH_LABEL_AND_LOOP; use rustc_session::lint::BuiltinLintDiagnostics; -use rustc_span::source_map::{self, Span, Spanned}; +use rustc_span::source_map::{self, Spanned}; use rustc_span::symbol::kw::PathRoot; use rustc_span::symbol::{kw, sym, Ident, Symbol}; -use rustc_span::{BytePos, Pos}; +use rustc_span::{BytePos, Pos, Span}; use thin_vec::{thin_vec, ThinVec}; /// Possibly accepts an `token::Interpolated` expression (a pre-parsed expression @@ -1007,8 +1007,9 @@ impl<'a> Parser<'a> { let span = self.token.span; let sm = self.sess.source_map(); let (span, actual) = match (&self.token.kind, self.subparser_name) { - (token::Eof, Some(_)) if let Ok(actual) = sm.span_to_snippet(sm.next_point(span)) => - (span.shrink_to_hi(), actual.into()), + (token::Eof, Some(_)) if let Ok(actual) = sm.span_to_snippet(sm.next_point(span)) => { + (span.shrink_to_hi(), actual.into()) + } _ => (span, actual), }; self.sess.emit_err(errors::UnexpectedTokenAfterDot { span, actual }); @@ -1440,14 +1441,20 @@ impl<'a> Parser<'a> { } else if this.token.uninterpolated_span().at_least_rust_2018() { // `Span:.at_least_rust_2018()` is somewhat expensive; don't get it repeatedly. if this.check_keyword(kw::Async) { - if this.is_async_block() { + if this.is_gen_block(kw::Async) { // Check for `async {` and `async move {`. - this.parse_async_block() + this.parse_gen_block() } else { this.parse_expr_closure() } } else if this.eat_keyword(kw::Await) { this.recover_incorrect_await_syntax(lo, this.prev_token.span) + } else if this.token.uninterpolated_span().at_least_rust_2024() { + if this.is_gen_block(kw::Gen) { + this.parse_gen_block() + } else { + this.parse_expr_lit() + } } else { this.parse_expr_lit() } @@ -1550,10 +1557,7 @@ impl<'a> Parser<'a> { self.sess.emit_err(errors::MacroInvocationWithQualifiedPath(path.span)); } let lo = path.span; - let mac = P(MacCall { - path, - args: self.parse_delim_args()?, - }); + let mac = P(MacCall { path, args: self.parse_delim_args()? }); (lo.to(self.prev_token.span), ExprKind::MacCall(mac)) } else if self.check(&token::OpenDelim(Delimiter::Brace)) && let Some(expr) = self.maybe_parse_struct_expr(&qself, &path) @@ -1592,7 +1596,7 @@ impl<'a> Parser<'a> { } else if !ate_colon && self.may_recover() && (matches!(self.token.kind, token::CloseDelim(_) | token::Comma) - || self.token.is_op()) + || self.token.is_punct()) { let (lit, _) = self.recover_unclosed_char(label_.ident, Parser::mk_token_lit_char, |self_| { @@ -1771,7 +1775,9 @@ impl<'a> Parser<'a> { fn parse_expr_break(&mut self) -> PResult<'a, P<Expr>> { let lo = self.prev_token.span; let mut label = self.eat_label(); - let kind = if self.token == token::Colon && let Some(label) = label.take() { + let kind = if self.token == token::Colon + && let Some(label) = label.take() + { // The value expression can be a labeled loop, see issue #86948, e.g.: // `loop { break 'label: loop { break 'label 42; }; }` let lexpr = self.parse_expr_labeled(label, true)?; @@ -1848,7 +1854,7 @@ impl<'a> Parser<'a> { let lo = self.prev_token.span; let kind = ExprKind::Yield(self.parse_expr_opt()?); let span = lo.to(self.prev_token.span); - self.sess.gated_spans.gate(sym::generators, span); + self.sess.gated_spans.gate(sym::yield_expr, span); let expr = self.mk_expr(span, kind); self.maybe_recover_from_bad_qpath(expr) } @@ -2209,6 +2215,7 @@ impl<'a> Parser<'a> { fn parse_expr_closure(&mut self) -> PResult<'a, P<Expr>> { let lo = self.token.span; + let before = self.prev_token.clone(); let binder = if self.check_keyword(kw::For) { let lo = self.token.span; let lifetime_defs = self.parse_late_bound_lifetime_defs()?; @@ -2239,7 +2246,12 @@ impl<'a> Parser<'a> { FnRetTy::Default(_) => { let restrictions = self.restrictions - Restrictions::STMT_EXPR - Restrictions::ALLOW_LET; - self.parse_expr_res(restrictions, None)? + let prev = self.prev_token.clone(); + let token = self.token.clone(); + match self.parse_expr_res(restrictions, None) { + Ok(expr) => expr, + Err(err) => self.recover_closure_body(err, before, prev, token, lo, decl_hi)?, + } } _ => { // If an explicit return type is given, require a block to appear (RFC 968). @@ -2291,13 +2303,14 @@ impl<'a> Parser<'a> { /// Parses an optional `move` prefix to a closure-like construct. fn parse_capture_clause(&mut self) -> PResult<'a, CaptureBy> { if self.eat_keyword(kw::Move) { + let move_kw_span = self.prev_token.span; // Check for `move async` and recover if self.check_keyword(kw::Async) { let move_async_span = self.token.span.with_lo(self.prev_token.span.data().lo); Err(errors::AsyncMoveOrderIncorrect { span: move_async_span } .into_diagnostic(&self.sess.span_diagnostic)) } else { - Ok(CaptureBy::Value) + Ok(CaptureBy::Value { move_kw: move_kw_span }) } } else { Ok(CaptureBy::Ref) @@ -2371,16 +2384,18 @@ impl<'a> Parser<'a> { let mut recover_block_from_condition = |this: &mut Self| { let block = match &mut cond.kind { ExprKind::Binary(Spanned { span: binop_span, .. }, _, right) - if let ExprKind::Block(_, None) = right.kind => { - self.sess.emit_err(errors::IfExpressionMissingThenBlock { - if_span: lo, - missing_then_block_sub: - errors::IfExpressionMissingThenBlockSub::UnfinishedCondition(cond_span.shrink_to_lo().to(*binop_span)), - let_else_sub: None, - - }); - std::mem::replace(right, this.mk_expr_err(binop_span.shrink_to_hi())) - }, + if let ExprKind::Block(_, None) = right.kind => + { + self.sess.emit_err(errors::IfExpressionMissingThenBlock { + if_span: lo, + missing_then_block_sub: + errors::IfExpressionMissingThenBlockSub::UnfinishedCondition( + cond_span.shrink_to_lo().to(*binop_span), + ), + let_else_sub: None, + }); + std::mem::replace(right, this.mk_expr_err(binop_span.shrink_to_hi())) + } ExprKind::Block(_, None) => { self.sess.emit_err(errors::IfExpressionMissingCondition { if_span: lo.shrink_to_hi(), @@ -2426,10 +2441,26 @@ impl<'a> Parser<'a> { self.error_on_extra_if(&cond)?; // Parse block, which will always fail, but we can add a nice note to the error self.parse_block().map_err(|mut err| { - err.span_note( - cond_span, - "the `if` expression is missing a block after this condition", - ); + if self.prev_token == token::Semi + && self.token == token::AndAnd + && let maybe_let = self.look_ahead(1, |t| t.clone()) + && maybe_let.is_keyword(kw::Let) + { + err.span_suggestion( + self.prev_token.span, + "consider removing this semicolon to parse the `let` as part of the same chain", + "", + Applicability::MachineApplicable, + ).span_note( + self.token.span.to(maybe_let.span), + "you likely meant to continue parsing the let-chain starting here", + ); + } else { + err.span_note( + cond_span, + "the `if` expression is missing a block after this condition", + ); + } err })? } @@ -2459,10 +2490,16 @@ impl<'a> Parser<'a> { /// Parses a `let $pat = $expr` pseudo-expression. fn parse_expr_let(&mut self, restrictions: Restrictions) -> PResult<'a, P<Expr>> { let is_recovered = if !restrictions.contains(Restrictions::ALLOW_LET) { - Some(self.sess.emit_err(errors::ExpectedExpressionFoundLet { + let err = errors::ExpectedExpressionFoundLet { span: self.token.span, reason: ForbiddenLetReason::OtherForbidden, - })) + }; + if self.prev_token.kind == token::BinOp(token::Or) { + // This was part of a closure, the that part of the parser recover. + return Err(err.into_diagnostic(&self.sess.span_diagnostic)); + } else { + Some(self.sess.emit_err(err)) + } } else { None }; @@ -2557,13 +2594,16 @@ impl<'a> Parser<'a> { } fn error_on_extra_if(&mut self, cond: &P<Expr>) -> PResult<'a, ()> { - if let ExprKind::Binary(Spanned { span: binop_span, node: binop}, _, right) = &cond.kind && - let BinOpKind::And = binop && - let ExprKind::If(cond, ..) = &right.kind { - Err(self.sess.create_err(errors::UnexpectedIfWithIf(binop_span.shrink_to_hi().to(cond.span.shrink_to_lo())))) - } else { - Ok(()) - } + if let ExprKind::Binary(Spanned { span: binop_span, node: binop }, _, right) = &cond.kind + && let BinOpKind::And = binop + && let ExprKind::If(cond, ..) = &right.kind + { + Err(self.sess.create_err(errors::UnexpectedIfWithIf( + binop_span.shrink_to_hi().to(cond.span.shrink_to_lo()), + ))) + } else { + Ok(()) + } } /// Parses `for <src_pat> in <src_expr> <src_loop_block>` (`for` token already eaten). @@ -2834,7 +2874,7 @@ impl<'a> Parser<'a> { )?; let guard = if this.eat_keyword(kw::If) { let if_span = this.prev_token.span; - let mut cond = this.parse_expr_res(Restrictions::ALLOW_LET, None)?; + let mut cond = this.parse_match_guard_condition()?; CondChecker { parser: this, forbid_let_reason: None }.visit_expr(&mut cond); @@ -2860,9 +2900,9 @@ impl<'a> Parser<'a> { { err.span_suggestion( this.token.span, - "try using a fat arrow here", + "use a fat arrow to start a match arm", "=>", - Applicability::MaybeIncorrect, + Applicability::MachineApplicable, ); err.emit(); this.bump(); @@ -2911,9 +2951,9 @@ impl<'a> Parser<'a> { .or_else(|mut err| { if this.token == token::FatArrow { if let Ok(expr_lines) = sm.span_to_lines(expr.span) - && let Ok(arm_start_lines) = sm.span_to_lines(arm_start_span) - && arm_start_lines.lines[0].end_col == expr_lines.lines[0].end_col - && expr_lines.lines.len() == 2 + && let Ok(arm_start_lines) = sm.span_to_lines(arm_start_span) + && arm_start_lines.lines[0].end_col == expr_lines.lines[0].end_col + && expr_lines.lines.len() == 2 { // We check whether there's any trailing code in the parse span, // if there isn't, we very likely have the following: @@ -2979,6 +3019,33 @@ impl<'a> Parser<'a> { }) } + fn parse_match_guard_condition(&mut self) -> PResult<'a, P<Expr>> { + self.parse_expr_res(Restrictions::ALLOW_LET | Restrictions::IN_IF_GUARD, None).map_err( + |mut err| { + if self.prev_token == token::OpenDelim(Delimiter::Brace) { + let sugg_sp = self.prev_token.span.shrink_to_lo(); + // Consume everything within the braces, let's avoid further parse + // errors. + self.recover_stmt_(SemiColonMode::Ignore, BlockMode::Ignore); + let msg = "you might have meant to start a match arm after the match guard"; + if self.eat(&token::CloseDelim(Delimiter::Brace)) { + let applicability = if self.token.kind != token::FatArrow { + // We have high confidence that we indeed didn't have a struct + // literal in the match guard, but rather we had some operation + // that ended in a path, immediately followed by a block that was + // meant to be the match arm. + Applicability::MachineApplicable + } else { + Applicability::MaybeIncorrect + }; + err.span_suggestion_verbose(sugg_sp, msg, "=> ".to_string(), applicability); + } + } + err + }, + ) + } + pub(crate) fn is_builtin(&self) -> bool { self.token.is_keyword(kw::Builtin) && self.look_ahead(1, |t| *t == token::Pound) } @@ -3015,18 +3082,24 @@ impl<'a> Parser<'a> { && self.token.uninterpolated_span().at_least_rust_2018() } - /// Parses an `async move? {...}` expression. - fn parse_async_block(&mut self) -> PResult<'a, P<Expr>> { + /// Parses an `async move? {...}` or `gen move? {...}` expression. + fn parse_gen_block(&mut self) -> PResult<'a, P<Expr>> { let lo = self.token.span; - self.expect_keyword(kw::Async)?; + let kind = if self.eat_keyword(kw::Async) { + GenBlockKind::Async + } else { + assert!(self.eat_keyword(kw::Gen)); + self.sess.gated_spans.gate(sym::gen_blocks, lo.to(self.token.span)); + GenBlockKind::Gen + }; let capture_clause = self.parse_capture_clause()?; let (attrs, body) = self.parse_inner_attrs_and_block()?; - let kind = ExprKind::Async(capture_clause, body); + let kind = ExprKind::Gen(capture_clause, body, kind); Ok(self.mk_expr_with_attrs(lo.to(self.prev_token.span), kind, attrs)) } - fn is_async_block(&self) -> bool { - self.token.is_keyword(kw::Async) + fn is_gen_block(&self, kw: Symbol) -> bool { + self.token.is_keyword(kw) && (( // `async move {` self.is_keyword_ahead(1, &[kw::Move]) @@ -3049,9 +3122,10 @@ impl<'a> Parser<'a> { || self.look_ahead(2, |t| t == &token::Colon) && ( // `{ ident: token, ` cannot start a block. - self.look_ahead(4, |t| t == &token::Comma) || - // `{ ident: ` cannot start a block unless it's a type ascription `ident: Type`. - self.look_ahead(3, |t| !t.can_begin_type()) + self.look_ahead(4, |t| t == &token::Comma) + // `{ ident: ` cannot start a block unless it's a type ascription + // `ident: Type`. + || self.look_ahead(3, |t| !t.can_begin_type()) ) ) } @@ -3091,6 +3165,7 @@ impl<'a> Parser<'a> { let mut fields = ThinVec::new(); let mut base = ast::StructRest::None; let mut recover_async = false; + let in_if_guard = self.restrictions.contains(Restrictions::IN_IF_GUARD); let mut async_block_err = |e: &mut Diagnostic, span: Span| { recover_async = true; @@ -3128,6 +3203,26 @@ impl<'a> Parser<'a> { e.span_label(pth.span, "while parsing this struct"); } + if let Some((ident, _)) = self.token.ident() + && !self.token.is_reserved_ident() + && self.look_ahead(1, |t| { + AssocOp::from_token(&t).is_some() + || matches!(t.kind, token::OpenDelim(_)) + || t.kind == token::Dot + }) + { + // Looks like they tried to write a shorthand, complex expression. + e.span_suggestion_verbose( + self.token.span.shrink_to_lo(), + "try naming a field", + &format!("{ident}: ",), + Applicability::MaybeIncorrect, + ); + } + if in_if_guard && close_delim == Delimiter::Brace { + return Err(e); + } + if !recover { return Err(e); } @@ -3173,19 +3268,6 @@ impl<'a> Parser<'a> { ",", Applicability::MachineApplicable, ); - } else if is_shorthand - && (AssocOp::from_token(&self.token).is_some() - || matches!(&self.token.kind, token::OpenDelim(_)) - || self.token.kind == token::Dot) - { - // Looks like they tried to write a shorthand, complex expression. - let ident = parsed_field.expect("is_shorthand implies Some").ident; - e.span_suggestion( - ident.span.shrink_to_lo(), - "try naming a field", - &format!("{ident}: "), - Applicability::HasPlaceholders, - ); } } if !recover { @@ -3288,6 +3370,24 @@ impl<'a> Parser<'a> { // Check if a colon exists one ahead. This means we're parsing a fieldname. let is_shorthand = !this.look_ahead(1, |t| t == &token::Colon || t == &token::Eq); + // Proactively check whether parsing the field will be incorrect. + let is_wrong = this.token.is_ident() + && !this.token.is_reserved_ident() + && !this.look_ahead(1, |t| { + t == &token::Colon + || t == &token::Eq + || t == &token::Comma + || t == &token::CloseDelim(Delimiter::Brace) + || t == &token::CloseDelim(Delimiter::Parenthesis) + }); + if is_wrong { + return Err(errors::ExpectedStructField { + span: this.look_ahead(1, |t| t.span), + ident_span: this.token.span, + token: this.look_ahead(1, |t| t.clone()), + } + .into_diagnostic(&self.sess.span_diagnostic)); + } let (ident, expr) = if is_shorthand { // Mimic `x: x` for the `x` field shorthand. let ident = this.parse_ident_common(false)?; @@ -3508,8 +3608,7 @@ impl MutVisitor for CondChecker<'_> { noop_visit_expr(e, self); self.forbid_let_reason = forbid_let_reason; } - ExprKind::Cast(ref mut op, _) - | ExprKind::Type(ref mut op, _) => { + ExprKind::Cast(ref mut op, _) | ExprKind::Type(ref mut op, _) => { let forbid_let_reason = self.forbid_let_reason; self.forbid_let_reason = Some(OtherForbidden); self.visit_expr(op); @@ -3526,7 +3625,7 @@ impl MutVisitor for CondChecker<'_> { | ExprKind::Match(_, _) | ExprKind::Closure(_) | ExprKind::Block(_, _) - | ExprKind::Async(_, _) + | ExprKind::Gen(_, _, _) | ExprKind::TryBlock(_) | ExprKind::Underscore | ExprKind::Path(_, _) diff --git a/compiler/rustc_parse/src/parser/item.rs b/compiler/rustc_parse/src/parser/item.rs index aad4edaba..801860c21 100644 --- a/compiler/rustc_parse/src/parser/item.rs +++ b/compiler/rustc_parse/src/parser/item.rs @@ -22,9 +22,9 @@ use rustc_errors::{ }; use rustc_span::edit_distance::edit_distance; use rustc_span::edition::Edition; -use rustc_span::source_map::{self, Span}; +use rustc_span::source_map; use rustc_span::symbol::{kw, sym, Ident, Symbol}; -use rustc_span::DUMMY_SP; +use rustc_span::{Span, DUMMY_SP}; use std::fmt::Write; use std::mem; use thin_vec::{thin_vec, ThinVec}; @@ -122,7 +122,9 @@ impl<'a> Parser<'a> { ) -> PResult<'a, Option<Item>> { // Don't use `maybe_whole` so that we have precise control // over when we bump the parser - if let token::Interpolated(nt) = &self.token.kind && let token::NtItem(item) = &**nt { + if let token::Interpolated(nt) = &self.token.kind + && let token::NtItem(item) = &**nt + { let mut item = item.clone(); self.bump(); @@ -623,11 +625,10 @@ impl<'a> Parser<'a> { // `impl<T: Default> impl Default for Wrapper<T>` // ^^^^^ let extra_impl_kw = ty_first.span.until(bound.span()); - self.sess - .emit_err(errors::ExtraImplKeywordInTraitImpl { - extra_impl_kw, - impl_trait_span: ty_first.span - }); + self.sess.emit_err(errors::ExtraImplKeywordInTraitImpl { + extra_impl_kw, + impl_trait_span: ty_first.span, + }); } else { self.sess.emit_err(errors::ExpectedTraitInTraitImplFoundType { span: ty_first.span, @@ -813,7 +814,12 @@ impl<'a> Parser<'a> { fn parse_item_trait(&mut self, attrs: &mut AttrVec, lo: Span) -> PResult<'a, ItemInfo> { let unsafety = self.parse_unsafety(Case::Sensitive); // Parse optional `auto` prefix. - let is_auto = if self.eat_keyword(kw::Auto) { IsAuto::Yes } else { IsAuto::No }; + let is_auto = if self.eat_keyword(kw::Auto) { + self.sess.gated_spans.gate(sym::auto_traits, self.prev_token.span); + IsAuto::Yes + } else { + IsAuto::No + }; self.expect_keyword(kw::Trait)?; let ident = self.parse_ident()?; @@ -1301,7 +1307,9 @@ impl<'a> Parser<'a> { // Provide a nice error message if the user placed a where-clause before the item body. // Users may be tempted to write such code if they are still used to the deprecated // where-clause location on type aliases and associated types. See also #89122. - if before_where_clause.has_where_token && let Some(expr) = &expr { + if before_where_clause.has_where_token + && let Some(expr) = &expr + { self.sess.emit_err(errors::WhereClauseBeforeConstBody { span: before_where_clause.span, name: ident.span, @@ -1944,7 +1952,8 @@ impl<'a> Parser<'a> { let mut err = self.expected_ident_found_err(); if self.eat_keyword_noexpect(kw::Let) && let removal_span = self.prev_token.span.until(self.token.span) - && let Ok(ident) = self.parse_ident_common(false) + && let Ok(ident) = self + .parse_ident_common(false) // Cancel this error, we don't need it. .map_err(|err| err.cancel()) && self.token.kind == TokenKind::Colon @@ -2269,6 +2278,18 @@ impl<'a> Parser<'a> { err.span_label(ident.span, "while parsing this `fn`"); err.emit(); } else { + // check for typo'd Fn* trait bounds such as + // fn foo<F>() where F: FnOnce -> () {} + if self.token.kind == token::RArrow { + let machine_applicable = [sym::FnOnce, sym::FnMut, sym::Fn] + .into_iter() + .any(|s| self.prev_token.is_ident_named(s)); + + err.subdiagnostic(errors::FnTraitMissingParen { + span: self.prev_token.span, + machine_applicable, + }); + } return Err(err); } } @@ -2288,9 +2309,9 @@ impl<'a> Parser<'a> { // `pub` is added in case users got confused with the ordering like `async pub fn`, // only if it wasn't preceded by `default` as `default pub` is invalid. let quals: &[Symbol] = if check_pub { - &[kw::Pub, kw::Const, kw::Async, kw::Unsafe, kw::Extern] + &[kw::Pub, kw::Gen, kw::Const, kw::Async, kw::Unsafe, kw::Extern] } else { - &[kw::Const, kw::Async, kw::Unsafe, kw::Extern] + &[kw::Gen, kw::Const, kw::Async, kw::Unsafe, kw::Extern] }; self.check_keyword_case(kw::Fn, case) // Definitely an `fn`. // `$qual fn` or `$qual $qual`: @@ -2344,6 +2365,9 @@ impl<'a> Parser<'a> { let async_start_sp = self.token.span; let asyncness = self.parse_asyncness(case); + let _gen_start_sp = self.token.span; + let genness = self.parse_genness(case); + let unsafe_start_sp = self.token.span; let unsafety = self.parse_unsafety(case); @@ -2359,6 +2383,10 @@ impl<'a> Parser<'a> { } } + if let Gen::Yes { span, .. } = genness { + self.sess.emit_err(errors::GenFn { span }); + } + if !self.eat_keyword_case(kw::Fn, case) { // It is possible for `expect_one_of` to recover given the contents of // `self.expected_tokens`, therefore, do not use `self.unexpected()` which doesn't @@ -2373,22 +2401,39 @@ impl<'a> Parser<'a> { Misplaced(Span), } + // We may be able to recover + let mut recover_constness = constness; + let mut recover_asyncness = asyncness; + let mut recover_unsafety = unsafety; // This will allow the machine fix to directly place the keyword in the correct place or to indicate // that the keyword is already present and the second instance should be removed. let wrong_kw = if self.check_keyword(kw::Const) { match constness { Const::Yes(sp) => Some(WrongKw::Duplicated(sp)), - Const::No => Some(WrongKw::Misplaced(async_start_sp)), + Const::No => { + recover_constness = Const::Yes(self.token.span); + Some(WrongKw::Misplaced(async_start_sp)) + } } } else if self.check_keyword(kw::Async) { match asyncness { Async::Yes { span, .. } => Some(WrongKw::Duplicated(span)), - Async::No => Some(WrongKw::Misplaced(unsafe_start_sp)), + Async::No => { + recover_asyncness = Async::Yes { + span: self.token.span, + closure_id: DUMMY_NODE_ID, + return_impl_trait_id: DUMMY_NODE_ID, + }; + Some(WrongKw::Misplaced(unsafe_start_sp)) + } } } else if self.check_keyword(kw::Unsafe) { match unsafety { Unsafe::Yes(sp) => Some(WrongKw::Duplicated(sp)), - Unsafe::No => Some(WrongKw::Misplaced(ext_start_sp)), + Unsafe::No => { + recover_unsafety = Unsafe::Yes(self.token.span); + Some(WrongKw::Misplaced(ext_start_sp)) + } } } else { None @@ -2458,6 +2503,23 @@ impl<'a> Parser<'a> { } } } + + if wrong_kw.is_some() + && self.may_recover() + && self.look_ahead(1, |tok| tok.is_keyword_case(kw::Fn, case)) + { + // Advance past the misplaced keyword and `fn` + self.bump(); + self.bump(); + err.emit(); + return Ok(FnHeader { + constness: recover_constness, + unsafety: recover_unsafety, + asyncness: recover_asyncness, + ext, + }); + } + return Err(err); } } @@ -2483,11 +2545,23 @@ impl<'a> Parser<'a> { pub(super) fn parse_fn_params(&mut self, req_name: ReqName) -> PResult<'a, ThinVec<Param>> { let mut first_param = true; // Parse the arguments, starting out with `self` being allowed... + if self.token.kind != TokenKind::OpenDelim(Delimiter::Parenthesis) + // might be typo'd trait impl, handled elsewhere + && !self.token.is_keyword(kw::For) + { + // recover from missing argument list, e.g. `fn main -> () {}` + self.sess + .emit_err(errors::MissingFnParams { span: self.prev_token.span.shrink_to_hi() }); + return Ok(ThinVec::new()); + } + let (mut params, _) = self.parse_paren_comma_seq(|p| { p.recover_diff_marker(); + let snapshot = p.create_snapshot_for_diagnostic(); let param = p.parse_param_general(req_name, first_param).or_else(|mut e| { e.emit(); let lo = p.prev_token.span; + p.restore_snapshot(snapshot); // Skip every token until next possible arg or end. p.eat_to_tokens(&[&token::Comma, &token::CloseDelim(Delimiter::Parenthesis)]); // Create a placeholder argument for proper arg count (issue #34264). diff --git a/compiler/rustc_parse/src/parser/mod.rs b/compiler/rustc_parse/src/parser/mod.rs index e84d8f5b3..1a7ae4069 100644 --- a/compiler/rustc_parse/src/parser/mod.rs +++ b/compiler/rustc_parse/src/parser/mod.rs @@ -11,6 +11,7 @@ mod stmt; mod ty; use crate::lexer::UnmatchedDelim; +use ast::Gen; pub use attr_wrapper::AttrWrapper; pub use diagnostics::AttemptLocalParseRecovery; pub(crate) use expr::ForbiddenLetReason; @@ -35,8 +36,8 @@ use rustc_errors::{ Applicability, DiagnosticBuilder, ErrorGuaranteed, FatalError, IntoDiagnostic, MultiSpan, }; use rustc_session::parse::ParseSess; -use rustc_span::source_map::{Span, DUMMY_SP}; use rustc_span::symbol::{kw, sym, Ident, Symbol}; +use rustc_span::{Span, DUMMY_SP}; use std::ops::Range; use std::{mem, slice}; use thin_vec::ThinVec; @@ -52,6 +53,7 @@ bitflags::bitflags! { const NO_STRUCT_LITERAL = 1 << 1; const CONST_EXPR = 1 << 2; const ALLOW_LET = 1 << 3; + const IN_IF_GUARD = 1 << 4; } } @@ -158,8 +160,9 @@ pub struct Parser<'a> { /// appropriately. /// /// See the comments in the `parse_path_segment` function for more details. - unmatched_angle_bracket_count: u32, - max_angle_bracket_count: u32, + unmatched_angle_bracket_count: u16, + max_angle_bracket_count: u16, + angle_bracket_nesting: u16, last_unexpected_token_span: Option<Span>, /// If present, this `Parser` is not parsing Rust code but rather a macro call. @@ -393,6 +396,7 @@ impl<'a> Parser<'a> { break_last_token: false, unmatched_angle_bracket_count: 0, max_angle_bracket_count: 0, + angle_bracket_nesting: 0, last_unexpected_token_span: None, subparser_name, capture_state: CaptureState { @@ -557,8 +561,9 @@ impl<'a> Parser<'a> { } if case == Case::Insensitive - && let Some((ident, /* is_raw */ false)) = self.token.ident() - && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() { + && let Some((ident, /* is_raw */ false)) = self.token.ident() + && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() + { true } else { false @@ -586,12 +591,10 @@ impl<'a> Parser<'a> { } if case == Case::Insensitive - && let Some((ident, /* is_raw */ false)) = self.token.ident() - && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() { - self.sess.emit_err(errors::KwBadCase { - span: ident.span, - kw: kw.as_str() - }); + && let Some((ident, /* is_raw */ false)) = self.token.ident() + && ident.as_str().to_lowercase() == kw.as_str().to_lowercase() + { + self.sess.emit_err(errors::KwBadCase { span: ident.span, kw: kw.as_str() }); self.bump(); return true; } @@ -1126,6 +1129,16 @@ impl<'a> Parser<'a> { } } + /// Parses genness: `gen` or nothing. + fn parse_genness(&mut self, case: Case) -> Gen { + if self.token.span.at_least_rust_2024() && self.eat_keyword_case(kw::Gen, case) { + let span = self.prev_token.uninterpolated_span(); + Gen::Yes { span, closure_id: DUMMY_NODE_ID, return_impl_trait_id: DUMMY_NODE_ID } + } else { + Gen::No + } + } + /// Parses unsafety: `unsafe` or nothing. fn parse_unsafety(&mut self, case: Case) -> Unsafe { if self.eat_keyword_case(kw::Unsafe, case) { diff --git a/compiler/rustc_parse/src/parser/nonterminal.rs b/compiler/rustc_parse/src/parser/nonterminal.rs index ff059a7e8..025b0615a 100644 --- a/compiler/rustc_parse/src/parser/nonterminal.rs +++ b/compiler/rustc_parse/src/parser/nonterminal.rs @@ -115,7 +115,7 @@ impl<'a> Parser<'a> { Some(item) => NtItem(item), None => { return Err(UnexpectedNonterminal::Item(self.token.span) - .into_diagnostic(&self.sess.span_diagnostic)); + .into_diagnostic(&self.sess.span_diagnostic)); } }, NonterminalKind::Block => { @@ -127,7 +127,7 @@ impl<'a> Parser<'a> { Some(s) => NtStmt(P(s)), None => { return Err(UnexpectedNonterminal::Statement(self.token.span) - .into_diagnostic(&self.sess.span_diagnostic)); + .into_diagnostic(&self.sess.span_diagnostic)); } }, NonterminalKind::PatParam { .. } | NonterminalKind::PatWithOr => { @@ -146,19 +146,15 @@ impl<'a> Parser<'a> { NonterminalKind::Expr => NtExpr(self.parse_expr_force_collect()?), NonterminalKind::Literal => { // The `:literal` matcher does not support attributes - NtLiteral( - self.collect_tokens_no_attrs(|this| this.parse_literal_maybe_minus())?, - ) + NtLiteral(self.collect_tokens_no_attrs(|this| this.parse_literal_maybe_minus())?) } - NonterminalKind::Ty => NtTy( - self.collect_tokens_no_attrs(|this| this.parse_ty_no_question_mark_recover())?, - ), + NonterminalKind::Ty => { + NtTy(self.collect_tokens_no_attrs(|this| this.parse_ty_no_question_mark_recover())?) + } // this could be handled like a token, since it is one - NonterminalKind::Ident - if let Some((ident, is_raw)) = get_macro_ident(&self.token) => - { + NonterminalKind::Ident if let Some((ident, is_raw)) = get_macro_ident(&self.token) => { self.bump(); NtIdent(ident, is_raw) } @@ -166,15 +162,17 @@ impl<'a> Parser<'a> { return Err(UnexpectedNonterminal::Ident { span: self.token.span, token: self.token.clone(), - }.into_diagnostic(&self.sess.span_diagnostic)); + } + .into_diagnostic(&self.sess.span_diagnostic)); + } + NonterminalKind::Path => { + NtPath(P(self.collect_tokens_no_attrs(|this| this.parse_path(PathStyle::Type))?)) } - NonterminalKind::Path => NtPath( - P(self.collect_tokens_no_attrs(|this| this.parse_path(PathStyle::Type))?), - ), NonterminalKind::Meta => NtMeta(P(self.parse_attr_item(true)?)), - NonterminalKind::Vis => NtVis( - P(self.collect_tokens_no_attrs(|this| this.parse_visibility(FollowedByType::Yes))?), - ), + NonterminalKind::Vis => { + NtVis(P(self + .collect_tokens_no_attrs(|this| this.parse_visibility(FollowedByType::Yes))?)) + } NonterminalKind::Lifetime => { if self.check_lifetime() { NtLifetime(self.expect_lifetime().ident) @@ -182,7 +180,8 @@ impl<'a> Parser<'a> { return Err(UnexpectedNonterminal::Lifetime { span: self.token.span, token: self.token.clone(), - }.into_diagnostic(&self.sess.span_diagnostic)); + } + .into_diagnostic(&self.sess.span_diagnostic)); } } }; diff --git a/compiler/rustc_parse/src/parser/pat.rs b/compiler/rustc_parse/src/parser/pat.rs index 3e4e92789..0a4c7c17d 100644 --- a/compiler/rustc_parse/src/parser/pat.rs +++ b/compiler/rustc_parse/src/parser/pat.rs @@ -20,8 +20,9 @@ use rustc_ast::{ use rustc_ast_pretty::pprust; use rustc_errors::{Applicability, DiagnosticBuilder, ErrorGuaranteed, IntoDiagnostic, PResult}; use rustc_session::errors::ExprParenthesesNeeded; -use rustc_span::source_map::{respan, Span, Spanned}; +use rustc_span::source_map::{respan, Spanned}; use rustc_span::symbol::{kw, sym, Ident}; +use rustc_span::Span; use thin_vec::{thin_vec, ThinVec}; #[derive(PartialEq, Copy, Clone)] @@ -967,11 +968,12 @@ impl<'a> Parser<'a> { // check that a comma comes after every field if !ate_comma { - let err = ExpectedCommaAfterPatternField { span: self.token.span } + let mut err = ExpectedCommaAfterPatternField { span: self.token.span } .into_diagnostic(&self.sess.span_diagnostic); if let Some(mut delayed) = delayed_err { delayed.emit(); } + self.recover_misplaced_pattern_modifiers(&fields, &mut err); return Err(err); } ate_comma = false; @@ -1109,6 +1111,37 @@ impl<'a> Parser<'a> { Ok((fields, etc)) } + /// If the user writes `S { ref field: name }` instead of `S { field: ref name }`, we suggest + /// the correct code. + fn recover_misplaced_pattern_modifiers( + &self, + fields: &ThinVec<PatField>, + err: &mut DiagnosticBuilder<'a, ErrorGuaranteed>, + ) { + if let Some(last) = fields.iter().last() + && last.is_shorthand + && let PatKind::Ident(binding, ident, None) = last.pat.kind + && binding != BindingAnnotation::NONE + && self.token == token::Colon + // We found `ref mut? ident:`, try to parse a `name,` or `name }`. + && let Some(name_span) = self.look_ahead(1, |t| t.is_ident().then(|| t.span)) + && self.look_ahead(2, |t| { + t == &token::Comma || t == &token::CloseDelim(Delimiter::Brace) + }) + { + let span = last.pat.span.with_hi(ident.span.lo()); + // We have `S { ref field: name }` instead of `S { field: ref name }` + err.multipart_suggestion( + "the pattern modifiers belong after the `:`", + vec![ + (span, String::new()), + (name_span.shrink_to_lo(), binding.prefix_str().to_string()), + ], + Applicability::MachineApplicable, + ); + } + } + /// Recover on `...` or `_` as if it were `..` to avoid further errors. /// See issue #46718. fn recover_bad_dot_dot(&self) { diff --git a/compiler/rustc_parse/src/parser/path.rs b/compiler/rustc_parse/src/parser/path.rs index 445516c03..8626dbe40 100644 --- a/compiler/rustc_parse/src/parser/path.rs +++ b/compiler/rustc_parse/src/parser/path.rs @@ -10,8 +10,8 @@ use rustc_ast::{ Path, PathSegment, QSelf, }; use rustc_errors::{Applicability, IntoDiagnostic, PResult}; -use rustc_span::source_map::{BytePos, Span}; use rustc_span::symbol::{kw, sym, Ident}; +use rustc_span::{BytePos, Span}; use std::mem; use thin_vec::ThinVec; use tracing::debug; @@ -487,10 +487,24 @@ impl<'a> Parser<'a> { // Take a snapshot before attempting to parse - we can restore this later. let snapshot = is_first_invocation.then(|| self.clone()); + self.angle_bracket_nesting += 1; debug!("parse_generic_args_with_leading_angle_bracket_recovery: (snapshotting)"); match self.parse_angle_args(ty_generics) { - Ok(args) => Ok(args), + Ok(args) => { + self.angle_bracket_nesting -= 1; + Ok(args) + } + Err(mut e) if self.angle_bracket_nesting > 10 => { + self.angle_bracket_nesting -= 1; + // When encountering severely malformed code where there are several levels of + // nested unclosed angle args (`f::<f::<f::<f::<...`), we avoid severe O(n^2) + // behavior by bailing out earlier (#117080). + e.emit(); + rustc_errors::FatalError.raise(); + } Err(e) if is_first_invocation && self.unmatched_angle_bracket_count > 0 => { + self.angle_bracket_nesting -= 1; + // Swap `self` with our backup of the parser state before attempting to parse // generic arguments. let snapshot = mem::replace(self, snapshot.unwrap()); @@ -520,8 +534,8 @@ impl<'a> Parser<'a> { // Make a span over ${unmatched angle bracket count} characters. // This is safe because `all_angle_brackets` ensures that there are only `<`s, // i.e. no multibyte characters, in this range. - let span = - lo.with_hi(lo.lo() + BytePos(snapshot.unmatched_angle_bracket_count)); + let span = lo + .with_hi(lo.lo() + BytePos(snapshot.unmatched_angle_bracket_count.into())); self.sess.emit_err(errors::UnmatchedAngle { span, plural: snapshot.unmatched_angle_bracket_count > 1, @@ -531,7 +545,10 @@ impl<'a> Parser<'a> { self.parse_angle_args(ty_generics) } } - Err(e) => Err(e), + Err(e) => { + self.angle_bracket_nesting -= 1; + Err(e) + } } } @@ -764,7 +781,8 @@ impl<'a> Parser<'a> { if let ast::TyKind::Slice(inner_ty) | ast::TyKind::Array(inner_ty, _) = &ty.kind && let ast::TyKind::Err = inner_ty.kind && let Some(snapshot) = snapshot - && let Some(expr) = self.recover_unbraced_const_arg_that_can_begin_ty(snapshot) + && let Some(expr) = + self.recover_unbraced_const_arg_that_can_begin_ty(snapshot) { return Ok(Some(self.dummy_const_arg_needs_braces( self.struct_span_err(expr.span, "invalid const generic expression"), @@ -776,12 +794,10 @@ impl<'a> Parser<'a> { } Err(err) => { if let Some(snapshot) = snapshot - && let Some(expr) = self.recover_unbraced_const_arg_that_can_begin_ty(snapshot) + && let Some(expr) = + self.recover_unbraced_const_arg_that_can_begin_ty(snapshot) { - return Ok(Some(self.dummy_const_arg_needs_braces( - err, - expr.span, - ))); + return Ok(Some(self.dummy_const_arg_needs_braces(err, expr.span))); } // Try to recover from possible `const` arg without braces. return self.recover_const_arg(start, err).map(Some); diff --git a/compiler/rustc_parse/src/parser/stmt.rs b/compiler/rustc_parse/src/parser/stmt.rs index 12c267351..aa939a71d 100644 --- a/compiler/rustc_parse/src/parser/stmt.rs +++ b/compiler/rustc_parse/src/parser/stmt.rs @@ -20,8 +20,8 @@ use rustc_ast::{AttrStyle, AttrVec, LocalKind, MacCall, MacCallStmt, MacStmtStyl use rustc_ast::{Block, BlockCheckMode, Expr, ExprKind, HasAttrs, Local, Stmt}; use rustc_ast::{StmtKind, DUMMY_NODE_ID}; use rustc_errors::{Applicability, DiagnosticBuilder, ErrorGuaranteed, PResult}; -use rustc_span::source_map::{BytePos, Span}; use rustc_span::symbol::{kw, sym, Ident}; +use rustc_span::{BytePos, Span}; use std::borrow::Cow; use std::mem; @@ -52,7 +52,9 @@ impl<'a> Parser<'a> { // Don't use `maybe_whole` so that we have precise control // over when we bump the parser - if let token::Interpolated(nt) = &self.token.kind && let token::NtStmt(stmt) = &**nt { + if let token::Interpolated(nt) = &self.token.kind + && let token::NtStmt(stmt) = &**nt + { let mut stmt = stmt.clone(); self.bump(); stmt.visit_attrs(|stmt_attrs| { @@ -227,8 +229,9 @@ impl<'a> Parser<'a> { /// Also error if the previous token was a doc comment. fn error_outer_attrs(&self, attrs: AttrWrapper) { if !attrs.is_empty() - && let attrs = attrs.take_for_recovery(self.sess) - && let attrs @ [.., last] = &*attrs { + && let attrs = attrs.take_for_recovery(self.sess) + && let attrs @ [.., last] = &*attrs + { if last.is_doc_comment() { self.sess.emit_err(errors::DocCommentDoesNotDocumentAnything { span: last.span, @@ -616,34 +619,19 @@ impl<'a> Parser<'a> { match &mut stmt.kind { // Expression without semicolon. StmtKind::Expr(expr) - if self.token != token::Eof && classify::expr_requires_semi_to_be_stmt(expr) => { + if self.token != token::Eof && classify::expr_requires_semi_to_be_stmt(expr) => + { // Just check for errors and recover; do not eat semicolon yet. // `expect_one_of` returns PResult<'a, bool /* recovered */> - let expect_result = self.expect_one_of(&[], &[token::Semi, token::CloseDelim(Delimiter::Brace)]); + let expect_result = + self.expect_one_of(&[], &[token::Semi, token::CloseDelim(Delimiter::Brace)]); let replace_with_err = 'break_recover: { match expect_result { - // Recover from parser, skip type error to avoid extra errors. + // Recover from parser, skip type error to avoid extra errors. Ok(true) => true, Err(mut e) => { - if let TokenKind::DocComment(..) = self.token.kind - && let Ok(snippet) = self.span_to_snippet(self.token.span) - { - let sp = self.token.span; - let marker = &snippet[..3]; - let (comment_marker, doc_comment_marker) = marker.split_at(2); - - e.span_suggestion( - sp.with_hi(sp.lo() + BytePos(marker.len() as u32)), - format!( - "add a space before `{doc_comment_marker}` to use a regular comment", - ), - format!("{comment_marker} {doc_comment_marker}"), - Applicability::MaybeIncorrect, - ); - } - if self.recover_colon_as_semi() { // recover_colon_as_semi has already emitted a nicer error. e.delay_as_bug(); @@ -654,14 +642,19 @@ impl<'a> Parser<'a> { } match &expr.kind { - ExprKind::Path(None, ast::Path { segments, .. }) if segments.len() == 1 => { + ExprKind::Path(None, ast::Path { segments, .. }) + if segments.len() == 1 => + { if self.token == token::Colon && self.look_ahead(1, |token| { - token.is_whole_block() || matches!( - token.kind, - token::Ident(kw::For | kw::Loop | kw::While, false) - | token::OpenDelim(Delimiter::Brace) - ) + token.is_whole_block() + || matches!( + token.kind, + token::Ident( + kw::For | kw::Loop | kw::While, + false + ) | token::OpenDelim(Delimiter::Brace) + ) }) { let snapshot = self.create_snapshot_for_diagnostic(); @@ -702,9 +695,8 @@ impl<'a> Parser<'a> { } true - } - Ok(false) => false + Ok(false) => false, } }; @@ -727,7 +719,9 @@ impl<'a> Parser<'a> { } eat_semi = false; } - StmtKind::Empty | StmtKind::Item(_) | StmtKind::Local(_) | StmtKind::Semi(_) => eat_semi = false, + StmtKind::Empty | StmtKind::Item(_) | StmtKind::Local(_) | StmtKind::Semi(_) => { + eat_semi = false + } } if add_semi_to_stmt || (eat_semi && self.eat(&token::Semi)) { diff --git a/compiler/rustc_parse/src/parser/ty.rs b/compiler/rustc_parse/src/parser/ty.rs index a25b0f1f8..be2cbaf30 100644 --- a/compiler/rustc_parse/src/parser/ty.rs +++ b/compiler/rustc_parse/src/parser/ty.rs @@ -18,9 +18,8 @@ use rustc_ast::{ TraitObjectSyntax, Ty, TyKind, }; use rustc_errors::{Applicability, PResult}; -use rustc_span::source_map::Span; use rustc_span::symbol::{kw, sym, Ident}; -use rustc_span::Symbol; +use rustc_span::{Span, Symbol}; use thin_vec::{thin_vec, ThinVec}; /// Any `?`, `!`, or `~const` modifiers that appear at the start of a bound. @@ -247,7 +246,7 @@ impl<'a> Parser<'a> { )?; FnRetTy::Ty(ty) } else { - FnRetTy::Default(self.token.span.shrink_to_lo()) + FnRetTy::Default(self.prev_token.span.shrink_to_hi()) }) } @@ -893,13 +892,15 @@ impl<'a> Parser<'a> { // to recover from errors, not make more). let path = if self.may_recover() { let (span, message, sugg, path, applicability) = match &ty.kind { - TyKind::Ptr(..) | TyKind::Ref(..) if let TyKind::Path(_, path) = &ty.peel_refs().kind => { + TyKind::Ptr(..) | TyKind::Ref(..) + if let TyKind::Path(_, path) = &ty.peel_refs().kind => + { ( ty.span.until(path.span), "consider removing the indirection", "", path, - Applicability::MaybeIncorrect + Applicability::MaybeIncorrect, ) } TyKind::ImplTrait(_, bounds) @@ -910,10 +911,10 @@ impl<'a> Parser<'a> { "use the trait bounds directly", "", &tr.trait_ref.path, - Applicability::MachineApplicable + Applicability::MachineApplicable, ) } - _ => return Err(err) + _ => return Err(err), }; err.span_suggestion_verbose(span, message, sugg, applicability); @@ -1027,7 +1028,8 @@ impl<'a> Parser<'a> { args.into_iter() .filter_map(|arg| { if let ast::AngleBracketedArg::Arg(generic_arg) = arg - && let ast::GenericArg::Lifetime(lifetime) = generic_arg { + && let ast::GenericArg::Lifetime(lifetime) = generic_arg + { Some(lifetime) } else { None |