Repository navigation
Panic in RW2 decoder (decode_panasonic) on newer Panasonic compression variants #59
Description
Activity
Reproduction with a modern camera file
Confirmed with a Panasonic DC-S5M2X (Lumix S5 II X, 2023) RW2 file: 6000×4000 px, 29,769,728 bytes total.
Minimal repro:
fn main() { let path = std::env::args().nth(1).unwrap(); match rawloader::decode_file(&path) { Ok(img) => println!("decoded OK: {}x{} cpp={}", img.width, img.height, img.cpp), Err(e) => println!("decode error (no panic): {}", e), } }
with
rawloader = "=0.37.2":thread '<unnamed>' panicked at rawloader-0.37.2/src/decoders/rw2.rs:94:21: range start index 26673152 out of range for slice of length 25587216 thread '<unnamed>' panicked at rawloader-0.37.2/src/decoders/rw2.rs:94:21: range start index 26869760 out of range for slice of length 25587216 decode error (no panic): RawLoaderError: "Caught a panic while decoding. Please file a bug with a sample file at https://github.com/pedrocr/rawloader/issues/new"(Multiple rayon worker threads panic on successive row groups before the internal
catch_unwindturns it into anErr.)Raw section is 25,587,216 bytes ≈ 1.07 bytes/pixel for a 24 MP frame — confirming this is the newer ~8-bit/pixel Panasonic compression, which
decode_panasonic's 9/8-packing offset math can't address. The decoder selection inRw2Decoder::imagehas no branch for it, so it falls through to the panicking path.Environment: macOS 26 arm64, rustc stable, rawloader 0.37.2 (also verified identical code on
master).
Summary
Decoding Panasonic RW2 files from newer cameras (S5 II / S1R II / G9 II generation) panics in
decode_panasonicwith an out-of-bounds slice index.Panic
Reproducible on both 0.37.2 (crates.io) and current
master— the code at the panic site is identical:Analysis
The failing file's raw section is ~1.0 byte/pixel (24,175,632 bytes for a ~23.6 MP image), while
decode_panasonicassumes the classic 9/8 bytes-per-pixel encoding. The branch selection inRw2Decoder::imagedoesn't catch this variant:src.len() >= width*height*2→ falsesrc.len() >= width*height*3/2→ falsedecode_panasonic(src, width, height, false, …), which computes block offsets from the assumed 9/8 packing and overruns the buffer near the bottom rows.Newer Panasonic bodies write RW2 with a different compression (~8-bit/pixel), which this decoder path was never written for.
Expected behavior
Return an
Err(unsupported encoding) instead of panicking — a panic on a rayon worker thread aborts the whole decode and is unfriendly to downstream FFI consumers. At minimum a bounds check before the slice; ideally support for the newer compression.Environment
master)catch_unwindonly because rayon re-raises on the join)