- Remove manual gzip header parsing - libdeflate handles all format details
- Rename linearize_chunked_content to build_input_buffer and free chunks as we copy
- Add output chunking to split large decompressed data into 1MB chunks
- Add comment explaining libdeflate's whole-buffer requirement
- Use better initial size heuristic based on compression ratio
Co-authored-by: nyh <584227+nyh@users.noreply.github.com>
- Check if total_decompressed >= length_limit before allocating output buffer
- Prevents allocating a zero-sized buffer when limit is already reached
- Ensures clear error message when limit is exceeded
Co-authored-by: nyh <584227+nyh@users.noreply.github.com>
- Removed unused get_gzip_member_size function
- Rely on libdeflate_gzip_decompress to tell us how many input bytes were consumed
- Added check for zero bytes consumed to detect invalid state
- Simplified the logic by removing unnecessary header size tracking
Co-authored-by: nyh <584227+nyh@users.noreply.github.com>
- Created utils/gzip.hh header with ungzip function declaration
- Created utils/gzip.cc implementation using libdeflate
- Updated utils/CMakeLists.txt to include gzip.cc and link libdeflate
- Created comprehensive test suite in test/boost/gzip_test.cc
- Added gzip_test to test/boost/CMakeLists.txt
The implementation:
- Uses libdeflate for high-performance gzip decompression
- Handles chunked_content input/output (vector of temporary_buffer)
- Supports concatenated gzip files
- Validates gzip headers and detects invalid/truncated/corrupted data
- Enforces size limits to prevent memory exhaustion
- Runs in async context to avoid blocking the reactor
Co-authored-by: nyh <584227+nyh@users.noreply.github.com>
