libchromaprint1 reports errors with some mp3 files even when no error exists

[eharris]

System: Debian bookworm amd64
Version: libchromaprint1 1.5.1-2+b1

When using libchromaprint1 via fpcalc, fpcalc sometimes reports the following error and returns rc = 3 when there doesn't seem to be anything wrong with the input file:

ERROR: Error decoding audio frame (End of file)

This causes picard to consider the fingerprint calculation to have failed.

mp3val and mp3diags report no problems with the source files that cause this error.
ffmpeg also reports no errors when directly decoding the same file, even with -loglevel debug.

This appears to be caused by fpcalc -> libchromaprint1 -> FFmpegAudioReader::Read() -> avcodec_receive_frame()

returning a negative value in some cases when reaching the end of the data stream.

[phw]

I think this should be closed:

1. The decoding errors reported are an ffmpeg thing and would need to be solved there (and it is not unlikely they already are with newer versions).
2. Status code 3 is explicitly returned for non-critical issues that still allowed the fingerprint to be calculated. Picard ignores those errors since version 2.11 PICARD-2813

[MufaroDKaseke]

System: Debian bookworm amd64
Version: libchromaprint1 1.5.1-2+b1

When using libchromaprint1 via fpcalc, fpcalc sometimes reports the following error and returns rc = 3 when there doesn't seem to be anything wrong with the input file:

ERROR: Error decoding audio frame (End of file)

This causes picard to consider the fingerprint calculation to have failed.

mp3val and mp3diags report no problems with the source files that cause this error.
ffmpeg also reports no errors when directly decoding the same file, even with -loglevel debug.

This appears to be caused by fpcalc -> libchromaprint1 -> FFmpegAudioReader::Read() -> avcodec_receive_frame()

returning a negative value in some cases when reaching the end of the data stream.

This error had me lose a hackathon yesterday, couldn't find a way around. It sucks🤤