Format support matrix

All advertised formats use the same FFmpeg decoder and AudioTrack lifecycle. There is no platform-decoder fallback.

The library scanner indexes only the enabled extension set. Metadata for every indexed format comes from FFmpeg. The build also carries the ASF demuxer so the same metadata API can inspect WMA containers, but WMA remains unadvertised and unplayable because no WMA decoder is enabled.

Embedded chapter metadata in an M4B does not prevent indexing or playback. The current audiobook model presents one physical file as one playable item; it does not expose embedded chapter boundaries as separate chapter rows.

There is no runtime format probe: support is a build property checked against the native configure allowlists by unit tests. Database migration version 9 clears old framework-specific playback failures and probe rows; version 10 drops the obsolete probe table. Version 11 stores the unified FFmpeg metadata record and schedules one refresh of existing library rows.

All formats decode through the same 44.1 kHz stereo float32 internal path. Sources at other rates or layouts are converted by libswresample; application gain, ReplayGain, balance, fades, ducking, and crossfade retain float precision. The final AudioTrack output remains 44.1 kHz stereo PCM16 because that is the only normal primary format verified in the stock Y2 audio policy. This is not a claim of bit-perfect or high-rate hardware output.