gif

• EncodingError and DecodingError are now #[non_exhaustive]
• Modified several error paths to return a new variant of EncodingError instead of boxing them into
  an io::Error, several return Result types are adjusted accordingly.
• The Decoded enum no longer communicates data from decoded sub-blocks such as the repetition count.
  It now only contains the meta information on framing and extension data. This ensures we are yielding
  less often for performance.
• last_ext was renamed to last_extension_sub_block for clarity.
• The Decoder will now collect XMP and ICC metadata subblocks, making them available after decoding.

• Fix interpretation of LZW stream when multiple intermediate reset codes are used.
• Allow extension blocks of unknown type with the allow_unknown_blocks option.

• Simplified internal decoder size

Bugfixes:

• Fixed writing palettes with non-power-of-two sizes

Features:

• Added Iterator interface for the Decoder
• Added reading number of loop repetitions Decoder::repeat.
• Added skip_frame_decoding option to read Frame.buffer as LZW data. It works together with write_lzw_pre_encoded_frame for quick rewriting of GIF files.
• Added support pre-allocated Vecs in from_palette_pixels
• Added ability to recover the io::Reader after decoding.
• Added support for decompressing Frame.buffer with LZW data, which enables fully parallel GIF re-encoding (see examples/parallel.rs),

Optimization:

• Less buffering, copying, and lower peak memory usage.

Other changes:

• Removed unsafe {} code.
• Enforced memory limit also on metadata extension blocks, and added out-of-memory checks where possible.
• EncodingFormatError enum is public.
• Removed defunct skip_extensions
• Added validation of frame dimensions. The buffer must be large enough for all pixels, and if the width or height is 0, the buffer must be empty.

Features:

• Add compression of pre-compressed frame data, via Encoder::write_lzw_pre_encoded_frame.
• The color_quant dependency is now optional. Turning it off disables some
  interfaces that would internally build quantization tables. The generic
  implementation of creating such tables can be prohibitively costly compared
  to specialized algorithms in some use cases.

Optimization:

• Avoid some allocations in by replacing flat_map argument with arrays

sha256sum: 80792593675e051cf94a4b111980da2ba60d4a83e43e0048c5693baab3977045 target/package/gif-0.12.0.crate

Bufixes:

• Fix decoding confusing superfluous image data from previous frames with
  current frame data.
• Bump minimum required version of weezl.

Features:

• Add Encoder::{get_ref, get_mut, into_inner} to access underlying stream.

Bugfixes:

• Fix panic while decoding some images, has no precise cause in the file.
• Warn about set_extensions being unimplemented...

Features:

• Added StreamingDecoder::version to query the precise version of the
  standard used for encoding the file. This is merely a hint.
• Added DecodeOptions::allow_unknown_blocks to skip over unknown or
  unspecified block kinds.

Optimization:

• Frame::from_rgba now recognizes when less than 256 colors are being used,
  dynamically skipping the quantization phase.
• Encoding image chunks is faster and simpler

• Fix panic when LZW code size is invalid
• Added option to omit check for lzw end code

sha256sum:
5a668f699973d0f573d15749b7002a9ac9e1f9c6b220e7b165601334c173d8de target/package/gif-0.11.2.crate

• Frames out-of-bounds of the screen descriptor are again accepted by default.
• Added DecodeOptions::check_frame_consistency to turn this validation on.