mp4a Sound Description version 2 (QuickTime) fails open

[alexantoshuk]

Summary

Opening some QuickTime .mov files fails while parsing an AAC mp4a sample entry with Sound Description version = 2. The error is:

mp4a box contains a box with a larger size than it

The file is valid (plays in VLC / QuickTime). The video track is fine; parse dies on the audio stsd entry and aborts the whole moov, so video-only consumers cannot open the file either.

Repro

1. Download: https://download.blender.org/peach/bigbuckbunny_movies/big_buck_bunny_1080p_h264.mov.zip
2. Unzip → big_buck_bunny_1080p_h264.mov
3. Mp4::read_bytes / open the file with re_mp4 0.5.1

Root cause (observed)

In this file the audio mp4a has version == 2 (QTFF SoundDescriptionV2). After the ISO-style fields, the next u32 is sizeOfStructOnly (here 72); child atoms (e.g. wave → esds) start at that offset from the atom start.

Mp4aBox::read_box only skips the version 1 16-byte extension. For version 2 it continues box-walking from the wrong cursor, treats sizeOfStructOnly / the float64 sample rate as a child box header, then hits a misaligned size and returns InvalidData("mp4a box contains a box with a larger size than it").

There is also a related gap: on BoxType::WaveBox the reader does not skip/recurse into wave to find nested esds (comment mentions the layout, but the branch is empty).

Expected

• Parse QT Sound Description version 2 (sizeOfStructOnly + skip to extensions), matching QTFF / common players.
• Prefer finding esds inside wave when present.
• Ideally, a failure on one sample entry / audio track should not hard-fail the entire file when other tracks are usable (nice-to-have).

Environment

• re_mp4 0.5.1
• File: Blender Foundation Big Buck Bunny 1080p H.264 .mov (link above)