Switch from encoding to encoding_rs.

See https://github.com/rustsec/advisory-db/issues/1605.
This commit is contained in:
Ethiraric 2024-03-23 16:34:35 +01:00
parent 6052436852
commit eddea11e01
3 changed files with 155 additions and 19 deletions

View file

@ -1,5 +1,27 @@
# Changelog # Changelog
## Upcoming
### Breaking changes
- The `encoding` library has been replaced with `encoding_rs`. If you use the
`trap` of `YamlDecoder`, this change will make your code not compile.
An additional enum `YamlDecoderTrap` has been added to abstract the
underlying library and avoid breaking changes in the future. This
additionally lifts the `encoding` dependency on _your_ project if you were
using that feature.
- The `encoding::types::DecoderTrap` has been replaced with `YamlDecoderTrap`.
- The signature of the function for `YamlDecoderTrap::Call` has changed:
```rs
// Before, with `encoding::types::DecoderTrap::Call`
fn(_: &mut encoding::RawDecoder, _: &[u8], _: &mut encoding::StringWriter) -> bool;
// Now, with `YamlDecoderTrap::Call`
fn(_: u8, _: u8, _: &[u8], _: &mut String) -> ControlFlow<Cow<'static str>>;
```
Please refer to the `YamlDecoderTrapFn` documentation for more details.
**Features**:
**Development**:
## v0.7.0 ## v0.7.0
**Features**: **Features**:

View file

@ -15,7 +15,7 @@ edition = "2021"
[dependencies] [dependencies]
arraydeque = "0.5.1" arraydeque = "0.5.1"
encoding = "0.2" encoding_rs = "0.8.33"
hashlink = "0.8" hashlink = "0.8"
[dev-dependencies] [dev-dependencies]

View file

@ -2,8 +2,11 @@
#![allow(clippy::module_name_repetitions)] #![allow(clippy::module_name_repetitions)]
use std::borrow::Cow;
use std::ops::ControlFlow;
use std::{collections::BTreeMap, convert::TryFrom, mem, ops::Index}; use std::{collections::BTreeMap, convert::TryFrom, mem, ops::Index};
use encoding_rs::{Decoder, DecoderResult, Encoding};
use hashlink::LinkedHashMap; use hashlink::LinkedHashMap;
use crate::parser::{Event, MarkedEventReceiver, Parser, Tag}; use crate::parser::{Event, MarkedEventReceiver, Parser, Tag};
@ -238,11 +241,51 @@ impl YamlLoader {
} }
} }
/// The signature of the function to call when using [`YAMLDecodingTrap::Call`].
///
/// The arguments are as follows:
/// * `malformation_length`: The length of the sequence the decoder failed to decode.
/// * `bytes_read_after_malformation`: The number of lookahead bytes the decoder consumed after
/// the malformation.
/// * `input_at_malformation`: What the input buffer is at the malformation.
/// This is the buffer starting at the malformation. The first `malformation_length` bytes are
/// the problematic sequence. The following `bytes_read_after_malformation` are already stored
/// in the decoder and will not be re-fed.
/// * `output`: The output string.
///
/// The function must modify `output` as it feels is best. For instance, one could recreate the
/// behavior of [`YAMLDecodingTrap::Ignore`] with an empty function, [`YAMLDecodingTrap::Replace`]
/// by pushing a `\u{FFFD}` into `output` and [`YAMLDecodingTrap::Strict`] by returning
/// [`ControlFlow::Break`].
///
/// # Returns
/// The function must return [`ControlFlow::Continue`] if decoding may continue or
/// [`ControlFlow::Break`] if decoding must be aborted. An optional error string may be supplied.
pub type YAMLDecodingTrapFn = fn(
malformation_length: u8,
bytes_read_after_malformation: u8,
input_at_malformation: &[u8],
output: &mut String,
) -> ControlFlow<Cow<'static, str>>;
/// The behavior [`YamlDecoder`] must have when an decoding error occurs.
#[derive(Copy, Clone, PartialEq, Eq)]
pub enum YAMLDecodingTrap {
/// Ignore the offending bytes, remove them from the output.
Ignore,
/// Error out.
Strict,
/// Replace them with the Unicode REPLACEMENT CHARACTER.
Replace,
/// Call the user-supplied function upon decoding malformation.
Call(YAMLDecodingTrapFn),
}
/// `YamlDecoder` is a `YamlLoader` builder that allows you to supply your own encoding error trap. /// `YamlDecoder` is a `YamlLoader` builder that allows you to supply your own encoding error trap.
/// For example, to read a YAML file while ignoring Unicode decoding errors you can set the /// For example, to read a YAML file while ignoring Unicode decoding errors you can set the
/// `encoding_trap` to `encoding::DecoderTrap::Ignore`. /// `encoding_trap` to `encoding::DecoderTrap::Ignore`.
/// ```rust /// ```rust
/// use yaml_rust2::yaml::YamlDecoder; /// use yaml_rust2::yaml::{YamlDecoder, YAMLDecodingTrap};
/// ///
/// let string = b"--- /// let string = b"---
/// a\xa9: 1 /// a\xa9: 1
@ -250,13 +293,13 @@ impl YamlLoader {
/// c: [1, 2] /// c: [1, 2]
/// "; /// ";
/// let out = YamlDecoder::read(string as &[u8]) /// let out = YamlDecoder::read(string as &[u8])
/// .encoding_trap(encoding::DecoderTrap::Ignore) /// .encoding_trap(YAMLDecodingTrap::Ignore)
/// .decode() /// .decode()
/// .unwrap(); /// .unwrap();
/// ``` /// ```
pub struct YamlDecoder<T: std::io::Read> { pub struct YamlDecoder<T: std::io::Read> {
source: T, source: T,
trap: encoding::types::DecoderTrap, trap: YAMLDecodingTrap,
} }
impl<T: std::io::Read> YamlDecoder<T> { impl<T: std::io::Read> YamlDecoder<T> {
@ -264,12 +307,12 @@ impl<T: std::io::Read> YamlDecoder<T> {
pub fn read(source: T) -> YamlDecoder<T> { pub fn read(source: T) -> YamlDecoder<T> {
YamlDecoder { YamlDecoder {
source, source,
trap: encoding::DecoderTrap::Strict, trap: YAMLDecodingTrap::Strict,
} }
} }
/// Set the behavior of the decoder when the encoding is invalid. /// Set the behavior of the decoder when the encoding is invalid.
pub fn encoding_trap(&mut self, trap: encoding::types::DecoderTrap) -> &mut Self { pub fn encoding_trap(&mut self, trap: YAMLDecodingTrap) -> &mut Self {
self.trap = trap; self.trap = trap;
self self
} }
@ -282,13 +325,84 @@ impl<T: std::io::Read> YamlDecoder<T> {
let mut buffer = Vec::new(); let mut buffer = Vec::new();
self.source.read_to_end(&mut buffer)?; self.source.read_to_end(&mut buffer)?;
// Decodes the input buffer using either UTF-8, UTF-16LE or UTF-16BE depending on the BOM codepoint. // Check if the `encoding` library can detect encoding from the BOM, otherwise use
// If the buffer doesn't start with a BOM codepoint, it will use a fallback encoding obtained by // `detect_utf16_endianness`.
// detect_utf16_endianness. let (encoding, _) =
let (res, _) = Encoding::for_bom(&buffer).unwrap_or_else(|| (detect_utf16_endianness(&buffer), 2));
encoding::types::decode(&buffer, self.trap, detect_utf16_endianness(&buffer)); let mut decoder = encoding.new_decoder();
let s = res.map_err(LoadError::Decode)?; let mut output = String::new();
YamlLoader::load_from_str(&s).map_err(LoadError::Scan)
// Decode the input buffer.
decode_loop(&buffer, &mut output, &mut decoder, self.trap)?;
YamlLoader::load_from_str(&output).map_err(LoadError::Scan)
}
}
/// Perform a loop of [`Decoder::decode_to_string`], reallocating `output` if needed.
fn decode_loop(
input: &[u8],
output: &mut String,
decoder: &mut Decoder,
trap: YAMLDecodingTrap,
) -> Result<(), LoadError> {
output.reserve(input.len());
let mut total_bytes_read = 0;
loop {
match decoder.decode_to_string_without_replacement(&input[total_bytes_read..], output, true)
{
// If the input is empty, we processed the whole input.
(DecoderResult::InputEmpty, _) => break Ok(()),
// If the output is full, we must reallocate.
(DecoderResult::OutputFull, bytes_read) => {
total_bytes_read += bytes_read;
// The output is already reserved to the size of the input. We slowly resize. Here,
// we're expecting that 10% of bytes will double in size when converting to UTF-8.
output.reserve(input.len() / 10);
}
(DecoderResult::Malformed(malformed_len, bytes_after_malformed), bytes_read) => {
total_bytes_read += bytes_read;
match trap {
// Ignore (skip over) malformed character.
YAMLDecodingTrap::Ignore => {}
// Replace them with the Unicode REPLACEMENT CHARACTER.
YAMLDecodingTrap::Replace => {
output.push('\u{FFFD}');
}
// Otherwise error, getting as much context as possible.
YAMLDecodingTrap::Strict => {
let malformed_len = malformed_len as usize;
let bytes_after_malformed = bytes_after_malformed as usize;
let byte_idx = total_bytes_read - (malformed_len + bytes_after_malformed);
let malformed_sequence = &input[byte_idx..byte_idx + malformed_len];
break Err(LoadError::Decode(Cow::Owned(format!(
"Invalid character sequence at {byte_idx}: {malformed_sequence:?}",
))));
}
YAMLDecodingTrap::Call(callback) => {
let byte_idx =
total_bytes_read - ((malformed_len + bytes_after_malformed) as usize);
let malformed_sequence =
&input[byte_idx..byte_idx + malformed_len as usize];
if let ControlFlow::Break(error) = callback(
malformed_len,
bytes_after_malformed,
&input[byte_idx..],
output,
) {
if error.is_empty() {
break Err(LoadError::Decode(Cow::Owned(format!(
"Invalid character sequence at {byte_idx}: {malformed_sequence:?}",
))));
}
break Err(LoadError::Decode(error));
}
}
}
}
}
} }
} }
@ -301,15 +415,15 @@ impl<T: std::io::Read> YamlDecoder<T> {
/// This allows the encoding to be deduced by the pattern of null (#x00) characters. /// This allows the encoding to be deduced by the pattern of null (#x00) characters.
// //
/// See spec at <https://yaml.org/spec/1.2/spec.html#id2771184> /// See spec at <https://yaml.org/spec/1.2/spec.html#id2771184>
fn detect_utf16_endianness(b: &[u8]) -> encoding::types::EncodingRef { fn detect_utf16_endianness(b: &[u8]) -> &'static Encoding {
if b.len() > 1 && (b[0] != b[1]) { if b.len() > 1 && (b[0] != b[1]) {
if b[0] == 0 { if b[0] == 0 {
return encoding::all::UTF_16BE; return encoding_rs::UTF_16BE;
} else if b[1] == 0 { } else if b[1] == 0 {
return encoding::all::UTF_16LE; return encoding_rs::UTF_16LE;
} }
} }
encoding::all::UTF_8 encoding_rs::UTF_8
} }
macro_rules! define_as ( macro_rules! define_as (
@ -550,7 +664,7 @@ impl Iterator for YamlIter {
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use super::{Yaml, YamlDecoder}; use super::{YAMLDecodingTrap, Yaml, YamlDecoder};
#[test] #[test]
fn test_read_bom() { fn test_read_bom() {
@ -623,7 +737,7 @@ b: 2.2
c: [1, 2] c: [1, 2]
"; ";
let out = YamlDecoder::read(s as &[u8]) let out = YamlDecoder::read(s as &[u8])
.encoding_trap(encoding::DecoderTrap::Ignore) .encoding_trap(YAMLDecodingTrap::Ignore)
.decode() .decode()
.unwrap(); .unwrap();
let doc = &out[0]; let doc = &out[0];