Panic when calling loudness_range about ebur128

[sdroege]

Thanks for reporting this. The problem is in

    combined.sort_unstable_by(|a, b| a.partial_cmp(b).unwrap());

Apparently the history contains NaNs sometimes? Can you provide a testcase for this, even if it only triggers rarely?

[sdroege]

@overdrivenpotato Any updates here?

[overdrivenpotato]

I've shifted off that particular project but for some added context:

This was triggered by feeding a simple non-NaN 96kHz FLAC input into FFMPEG, which was then decoded, resampled to 96kHz (same rate, no-op?), and then fed into ebur128. I would imagine the data coming out of the FFMPEG FLAC decoder/resampler is constant as the FLAC input was always identical, and the sample rate was always the same between I/O. The spurious nature of the panic is puzzling.

[sdroege]

Indeed. If you can somehow produce a testcase for this, even if it only triggers the problem sometimes, that would be really great. I don't see how any of these values can end up as NaN so without a testcase there's not much I can do.

Would that be possible for you? Otherwise I guess we can close this issue until someone can reproduce it :)