diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-19 00:47:55 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-19 00:47:55 +0000 |
commit | 26a029d407be480d791972afb5975cf62c9360a6 (patch) | |
tree | f435a8308119effd964b339f76abb83a57c29483 /intl/encoding_glue/src/lib.rs | |
parent | Initial commit. (diff) | |
download | firefox-26a029d407be480d791972afb5975cf62c9360a6.tar.xz firefox-26a029d407be480d791972afb5975cf62c9360a6.zip |
Adding upstream version 124.0.1.upstream/124.0.1
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'intl/encoding_glue/src/lib.rs')
-rw-r--r-- | intl/encoding_glue/src/lib.rs | 609 |
1 files changed, 609 insertions, 0 deletions
diff --git a/intl/encoding_glue/src/lib.rs b/intl/encoding_glue/src/lib.rs new file mode 100644 index 0000000000..277a7cde70 --- /dev/null +++ b/intl/encoding_glue/src/lib.rs @@ -0,0 +1,609 @@ +// Copyright 2015-2016 Mozilla Foundation. See the COPYRIGHT +// file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// https://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +// Adapted from third_party/rust/encoding_rs/src/lib.rs, so the +// "top-level directory" in the above notice refers to +// third_party/rust/encoding_rs/. + +extern crate encoding_rs; +extern crate nserror; +extern crate nsstring; +extern crate xmldecl; + +use encoding_rs::*; +use nserror::*; +use nsstring::*; +use std::slice; + +/// Takes `Option<usize>`, the destination string and a value +/// to return on failure and tries to start a bulk write of the +/// destination string with the capacity given by the `usize` +/// wrapped in the first argument. Returns the bulk write +/// handle. +macro_rules! try_start_bulk_write { + ($needed:expr, + $dst:ident, + $ret:expr) => {{ + let needed = match $needed { + Some(needed) => needed, + None => { + return $ret; + } + }; + match unsafe { $dst.bulk_write(needed, 0, false) } { + Err(_) => { + return $ret; + } + Ok(handle) => handle, + } + }}; +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nsstring( + encoding: *mut *const Encoding, + src: *const u8, + src_len: usize, + dst: *mut nsAString, +) -> nsresult { + let (rv, enc) = decode_to_nsstring(&**encoding, slice::from_raw_parts(src, src_len), &mut *dst); + *encoding = enc as *const Encoding; + rv +} + +pub fn decode_to_nsstring( + encoding: &'static Encoding, + src: &[u8], + dst: &mut nsAString, +) -> (nsresult, &'static Encoding) { + if let Some((enc, bom_length)) = Encoding::for_bom(src) { + return ( + decode_to_nsstring_without_bom_handling(enc, &src[bom_length..], dst), + enc, + ); + } + ( + decode_to_nsstring_without_bom_handling(encoding, src, dst), + encoding, + ) +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nsstring_with_bom_removal( + encoding: *const Encoding, + src: *const u8, + src_len: usize, + dst: *mut nsAString, +) -> nsresult { + decode_to_nsstring_with_bom_removal(&*encoding, slice::from_raw_parts(src, src_len), &mut *dst) +} + +pub fn decode_to_nsstring_with_bom_removal( + encoding: &'static Encoding, + src: &[u8], + dst: &mut nsAString, +) -> nsresult { + let without_bom = if encoding == UTF_8 && src.starts_with(b"\xEF\xBB\xBF") { + &src[3..] + } else if (encoding == UTF_16LE && src.starts_with(b"\xFF\xFE")) + || (encoding == UTF_16BE && src.starts_with(b"\xFE\xFF")) + { + &src[2..] + } else { + src + }; + decode_to_nsstring_without_bom_handling(encoding, without_bom, dst) +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nsstring_without_bom_handling( + encoding: *const Encoding, + src: *const u8, + src_len: usize, + dst: *mut nsAString, +) -> nsresult { + decode_to_nsstring_without_bom_handling( + &*encoding, + slice::from_raw_parts(src, src_len), + &mut *dst, + ) +} + +pub fn decode_to_nsstring_without_bom_handling( + encoding: &'static Encoding, + src: &[u8], + dst: &mut nsAString, +) -> nsresult { + let mut decoder = encoding.new_decoder_without_bom_handling(); + let mut handle = try_start_bulk_write!( + decoder.max_utf16_buffer_length(src.len()), + dst, + NS_ERROR_OUT_OF_MEMORY + ); + let (result, read, written, had_errors) = + decoder.decode_to_utf16(src, handle.as_mut_slice(), true); + debug_assert_eq!(result, CoderResult::InputEmpty); + debug_assert_eq!(read, src.len()); + debug_assert!(written <= handle.as_mut_slice().len()); + let _ = handle.finish(written, true); + if had_errors { + return NS_OK_HAD_REPLACEMENTS; + } + NS_OK +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nsstring_without_bom_handling_and_without_replacement( + encoding: *const Encoding, + src: *const u8, + src_len: usize, + dst: *mut nsAString, +) -> nsresult { + decode_to_nsstring_without_bom_handling_and_without_replacement( + &*encoding, + slice::from_raw_parts(src, src_len), + &mut *dst, + ) +} + +pub fn decode_to_nsstring_without_bom_handling_and_without_replacement( + encoding: &'static Encoding, + src: &[u8], + dst: &mut nsAString, +) -> nsresult { + let mut decoder = encoding.new_decoder_without_bom_handling(); + let mut handle = try_start_bulk_write!( + decoder.max_utf16_buffer_length(src.len()), + dst, + NS_ERROR_OUT_OF_MEMORY + ); + let (result, read, written) = + decoder.decode_to_utf16_without_replacement(src, handle.as_mut_slice(), true); + match result { + DecoderResult::InputEmpty => { + debug_assert_eq!(read, src.len()); + debug_assert!(written <= handle.as_mut_slice().len()); + let _ = handle.finish(written, true); + NS_OK + } + DecoderResult::Malformed(_, _) => { + // Let handle's drop() run + NS_ERROR_UDEC_ILLEGALINPUT + } + DecoderResult::OutputFull => unreachable!(), + } +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_encode_from_utf16( + encoding: *mut *const Encoding, + src: *const u16, + src_len: usize, + dst: *mut nsACString, +) -> nsresult { + let (rv, enc) = encode_from_utf16(&**encoding, slice::from_raw_parts(src, src_len), &mut *dst); + *encoding = enc as *const Encoding; + rv +} + +pub fn encode_from_utf16( + encoding: &'static Encoding, + src: &[u16], + dst: &mut nsACString, +) -> (nsresult, &'static Encoding) { + let output_encoding = encoding.output_encoding(); + let mut encoder = output_encoding.new_encoder(); + let mut handle = try_start_bulk_write!( + encoder.max_buffer_length_from_utf16_if_no_unmappables(src.len()), + dst, + (NS_ERROR_OUT_OF_MEMORY, output_encoding) + ); + + let mut total_read = 0; + let mut total_written = 0; + let mut total_had_errors = false; + loop { + let (result, read, written, had_errors) = encoder.encode_from_utf16( + &src[total_read..], + &mut (handle.as_mut_slice())[total_written..], + true, + ); + total_read += read; + total_written += written; + total_had_errors |= had_errors; + match result { + CoderResult::InputEmpty => { + debug_assert_eq!(total_read, src.len()); + debug_assert!(total_written <= handle.as_mut_slice().len()); + let _ = handle.finish(total_written, true); + if total_had_errors { + return (NS_OK_HAD_REPLACEMENTS, output_encoding); + } + return (NS_OK, output_encoding); + } + CoderResult::OutputFull => { + if let Some(needed) = checked_add( + total_written, + encoder.max_buffer_length_from_utf16_if_no_unmappables(src.len() - total_read), + ) { + if unsafe { + handle + .restart_bulk_write(needed, total_written, false) + .is_ok() + } { + continue; + } + } + return (NS_ERROR_OUT_OF_MEMORY, output_encoding); + } + } + } +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nscstring( + encoding: *mut *const Encoding, + src: *const nsACString, + dst: *mut nsACString, +) -> nsresult { + debug_assert_ne!(src as usize, dst as usize); + let (rv, enc) = decode_to_nscstring(&**encoding, &*src, &mut *dst); + *encoding = enc as *const Encoding; + rv +} + +pub fn decode_to_nscstring( + encoding: &'static Encoding, + src: &nsACString, + dst: &mut nsACString, +) -> (nsresult, &'static Encoding) { + if let Some((enc, bom_length)) = Encoding::for_bom(src) { + return ( + decode_from_slice_to_nscstring_without_bom_handling(enc, &src[bom_length..], dst, 0), + enc, + ); + } + ( + decode_to_nscstring_without_bom_handling(encoding, src, dst), + encoding, + ) +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nscstring_with_bom_removal( + encoding: *const Encoding, + src: *const nsACString, + dst: *mut nsACString, +) -> nsresult { + debug_assert_ne!(src as usize, dst as usize); + decode_to_nscstring_with_bom_removal(&*encoding, &*src, &mut *dst) +} + +pub fn decode_to_nscstring_with_bom_removal( + encoding: &'static Encoding, + src: &nsACString, + dst: &mut nsACString, +) -> nsresult { + let without_bom = if encoding == UTF_8 && src.starts_with(b"\xEF\xBB\xBF") { + &src[3..] + } else if (encoding == UTF_16LE && src.starts_with(b"\xFF\xFE")) + || (encoding == UTF_16BE && src.starts_with(b"\xFE\xFF")) + { + &src[2..] + } else { + return decode_to_nscstring_without_bom_handling(encoding, src, dst); + }; + decode_from_slice_to_nscstring_without_bom_handling(encoding, without_bom, dst, 0) +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nscstring_without_bom_handling( + encoding: *const Encoding, + src: *const nsACString, + dst: *mut nsACString, +) -> nsresult { + debug_assert_ne!(src as usize, dst as usize); + decode_to_nscstring_without_bom_handling(&*encoding, &*src, &mut *dst) +} + +pub fn decode_to_nscstring_without_bom_handling( + encoding: &'static Encoding, + src: &nsACString, + dst: &mut nsACString, +) -> nsresult { + let bytes = &src[..]; + let valid_up_to = if encoding == UTF_8 { + Encoding::utf8_valid_up_to(bytes) + } else if encoding.is_ascii_compatible() { + Encoding::ascii_valid_up_to(bytes) + } else if encoding == ISO_2022_JP { + Encoding::iso_2022_jp_ascii_valid_up_to(bytes) + } else { + return decode_from_slice_to_nscstring_without_bom_handling(encoding, src, dst, 0); + }; + if valid_up_to == bytes.len() { + if dst.fallible_assign(src).is_err() { + return NS_ERROR_OUT_OF_MEMORY; + } + return NS_OK; + } + decode_from_slice_to_nscstring_without_bom_handling(encoding, src, dst, valid_up_to) +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_from_slice_to_nscstring_without_bom_handling( + encoding: *const Encoding, + src: *const u8, + src_len: usize, + dst: *mut nsACString, + already_validated: usize, +) -> nsresult { + decode_from_slice_to_nscstring_without_bom_handling( + &*encoding, + slice::from_raw_parts(src, src_len), + &mut *dst, + already_validated, + ) +} + +fn decode_from_slice_to_nscstring_without_bom_handling( + encoding: &'static Encoding, + src: &[u8], + dst: &mut nsACString, + already_validated: usize, +) -> nsresult { + let bytes = src; + let mut decoder = encoding.new_decoder_without_bom_handling(); + let mut handle = try_start_bulk_write!(Some(src.len()), dst, NS_ERROR_OUT_OF_MEMORY); + + if already_validated != 0 { + (handle.as_mut_slice())[..already_validated].copy_from_slice(&bytes[..already_validated]); + } + let mut total_read = already_validated; + let mut total_written = already_validated; + let mut total_had_errors = false; + loop { + let (result, read, written, had_errors) = decoder.decode_to_utf8( + &bytes[total_read..], + &mut (handle.as_mut_slice())[total_written..], + true, + ); + total_read += read; + total_written += written; + total_had_errors |= had_errors; + match result { + CoderResult::InputEmpty => { + debug_assert_eq!(total_read, bytes.len()); + let _ = handle.finish(total_written, true); + if total_had_errors { + return NS_OK_HAD_REPLACEMENTS; + } + return NS_OK; + } + CoderResult::OutputFull => { + // Allocate for the worst case. That is, we should come + // here at most once per invocation of this method. + if let Some(needed) = checked_add( + total_written, + decoder.max_utf8_buffer_length(bytes.len() - total_read), + ) { + if unsafe { + handle + .restart_bulk_write(needed, total_written, false) + .is_ok() + } { + continue; + } + } + return NS_ERROR_OUT_OF_MEMORY; + } + } + } +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_decode_to_nscstring_without_bom_handling_and_without_replacement( + encoding: *const Encoding, + src: *const nsACString, + dst: *mut nsACString, +) -> nsresult { + decode_to_nscstring_without_bom_handling_and_without_replacement(&*encoding, &*src, &mut *dst) +} + +pub fn decode_to_nscstring_without_bom_handling_and_without_replacement( + encoding: &'static Encoding, + src: &nsACString, + dst: &mut nsACString, +) -> nsresult { + let bytes = &src[..]; + if encoding == UTF_8 { + let valid_up_to = Encoding::utf8_valid_up_to(bytes); + if valid_up_to == bytes.len() { + if dst.fallible_assign(src).is_err() { + return NS_ERROR_OUT_OF_MEMORY; + } + return NS_OK; + } + return NS_ERROR_UDEC_ILLEGALINPUT; + } + let valid_up_to = if encoding.is_ascii_compatible() { + Encoding::ascii_valid_up_to(bytes) + } else if encoding == ISO_2022_JP { + Encoding::iso_2022_jp_ascii_valid_up_to(bytes) + } else { + 0 + }; + if valid_up_to == bytes.len() { + if dst.fallible_assign(src).is_err() { + return NS_ERROR_OUT_OF_MEMORY; + } + return NS_OK; + } + let mut decoder = encoding.new_decoder_without_bom_handling(); + let mut handle = try_start_bulk_write!( + checked_add( + valid_up_to, + decoder.max_utf8_buffer_length_without_replacement(bytes.len() - valid_up_to) + ), + dst, + NS_ERROR_OUT_OF_MEMORY + ); + let (result, read, written) = { + let dest = handle.as_mut_slice(); + dest[..valid_up_to].copy_from_slice(&bytes[..valid_up_to]); + decoder.decode_to_utf8_without_replacement( + &src[valid_up_to..], + &mut dest[valid_up_to..], + true, + ) + }; + match result { + DecoderResult::InputEmpty => { + debug_assert_eq!(valid_up_to + read, src.len()); + debug_assert!(valid_up_to + written <= handle.as_mut_slice().len()); + let _ = handle.finish(valid_up_to + written, true); + NS_OK + } + DecoderResult::Malformed(_, _) => { + // let handle's drop() run + NS_ERROR_UDEC_ILLEGALINPUT + } + DecoderResult::OutputFull => unreachable!(), + } +} + +#[no_mangle] +pub unsafe extern "C" fn mozilla_encoding_encode_from_nscstring( + encoding: *mut *const Encoding, + src: *const nsACString, + dst: *mut nsACString, +) -> nsresult { + let (rv, enc) = encode_from_nscstring(&**encoding, &*src, &mut *dst); + *encoding = enc as *const Encoding; + rv +} + +pub fn encode_from_nscstring( + encoding: &'static Encoding, + src: &nsACString, + dst: &mut nsACString, +) -> (nsresult, &'static Encoding) { + let output_encoding = encoding.output_encoding(); + let bytes = &src[..]; + if output_encoding == UTF_8 { + let valid_up_to = Encoding::utf8_valid_up_to(bytes); + if valid_up_to == bytes.len() { + if dst.fallible_assign(src).is_err() { + return (NS_ERROR_OUT_OF_MEMORY, output_encoding); + } + return (NS_OK, output_encoding); + } + return (NS_ERROR_UDEC_ILLEGALINPUT, output_encoding); + } + let valid_up_to = if output_encoding == ISO_2022_JP { + Encoding::iso_2022_jp_ascii_valid_up_to(bytes) + } else { + debug_assert!(output_encoding.is_ascii_compatible()); + Encoding::ascii_valid_up_to(bytes) + }; + if valid_up_to == bytes.len() { + if dst.fallible_assign(src).is_err() { + return (NS_ERROR_OUT_OF_MEMORY, output_encoding); + } + return (NS_OK, output_encoding); + } + + // Encoder requires valid UTF-8. Using std instead of encoding_rs + // to avoid unsafe blocks. + let trail = if let Ok(trail) = ::std::str::from_utf8(&bytes[valid_up_to..]) { + trail + } else { + return (NS_ERROR_UDEC_ILLEGALINPUT, output_encoding); + }; + + let mut encoder = output_encoding.new_encoder(); + let mut handle = try_start_bulk_write!( + checked_add( + valid_up_to, + encoder.max_buffer_length_from_utf8_if_no_unmappables(trail.len()) + ), + dst, + (NS_ERROR_OUT_OF_MEMORY, output_encoding) + ); + + if valid_up_to != 0 { + // to_mut() shouldn't fail right after setting length. + (handle.as_mut_slice())[..valid_up_to].copy_from_slice(&bytes[..valid_up_to]); + } + + // `total_read` tracks `trail` only but `total_written` tracks the overall situation! + // This asymmetry is here, because trail is materialized as `str` without resorting + // to unsafe code here. + let mut total_read = 0; + let mut total_written = valid_up_to; + let mut total_had_errors = false; + loop { + let (result, read, written, had_errors) = encoder.encode_from_utf8( + &trail[total_read..], + &mut (handle.as_mut_slice())[total_written..], + true, + ); + total_read += read; + total_written += written; + total_had_errors |= had_errors; + match result { + CoderResult::InputEmpty => { + debug_assert_eq!(valid_up_to + total_read, src.len()); + debug_assert!(total_written <= handle.as_mut_slice().len()); + let _ = handle.finish(total_written, true); + if total_had_errors { + return (NS_OK_HAD_REPLACEMENTS, output_encoding); + } + return (NS_OK, output_encoding); + } + CoderResult::OutputFull => { + if let Some(needed) = checked_add( + total_written, + encoder.max_buffer_length_from_utf8_if_no_unmappables(trail.len() - total_read), + ) { + if unsafe { + handle + .restart_bulk_write(needed, total_written, false) + .is_ok() + } { + continue; + } + } + return (NS_ERROR_OUT_OF_MEMORY, output_encoding); + } + } + } +} + +#[inline(always)] +fn checked_add(num: usize, opt: Option<usize>) -> Option<usize> { + if let Some(n) = opt { + n.checked_add(num) + } else { + None + } +} + +// Declared in nsHtml5StreamParser.cpp +#[no_mangle] +pub unsafe extern "C" fn xmldecl_parse( + buf: *const u8, + buf_len: usize, +) -> *const encoding_rs::Encoding { + if let Some(encoding) = xmldecl::parse(std::slice::from_raw_parts(buf, buf_len)) { + encoding + } else { + std::ptr::null() + } +} |