[FFmpeg-devel] [PATCH v5 3/3] fate/flac: Add test of 32 bps encoding/decoding

[Martijn van Beurden]

Op do 15 sep. 2022 om 21:52 schreef Martijn van Beurden <mvanb1 at gmail.com>:

> Op do 15 sep. 2022 om 20:20 schreef Andreas Rheinhardt <
> andreas.rheinhardt at outlook.com>:
>
>> (In any case, I don't get why there are two samples: The flac decoder is
>> intra-only, so if you made the samples so that one of the frames had a
>> wasted bit, you could test all the codepaths that two samples test. And
>> it would even have the advantage that it tests what happens when it
>> swiches from >0 wasted bits to 0 wasted bits to >0 wasted bits and back.)
>>
>
> It seemed easier for anyone triggering a problem with wasted bits to
> figure where the problem came from. A problem in the second file but not in
> the first is quite certain to be related to the wasted bits handling. But
> perhaps the probability of this happening is so low that it is not worth
> increasing the size of the fate sample collection. I'll create a new file
> and ask for a replacement.
>

Files have been replaced, there is now one file for this test with varying
levels of wasted bits. At the end it also has a differing number of wasted
bits between the left and right channel. In other decoders this has been a
problem in the past. It now changes 0-0, 1-1, 0-0, 1-1, 3-2, 2-3, where
each pair is the number of wasted bits for the left and right channel. I'll
send new patches in a moment.