MP3: Mismatch with delta 0.0004 on MySpace Dragon Hoard corpus

[Shnatsel]

Following the fix for #68, there is a single mismatch above the 0.00002 threshold on the MySpace Dragon Hoard corpus. The max delta is 0.00042961, the file is attached:
std_9dc496596e1558362a28966aacdab41d.mp3.gz

Tested vs ffmpeg version 4.2.4-1ubuntu0.1 built with gcc 9 (Ubuntu 9.3.0-10ubuntu2)

Tested on commit 16497ee

[pdeljanov]

This was an interesting find, because there was a small bug in Symphonia, however the proper fix in my opinion would deviate from both ffmpeg and mpg123. For now I've chosen to just do as they do but print a different message instead.

[Shnatsel]

Wow! Thanks for the quick fix!

I'm surprised this was actually actionable 😁

I'll re-run symphonia-check on my other MP3s shortly.

[pdeljanov]

No problem. On a side note, I think I actually a bug in ffmpeg regarding this issue. This may cause you to get failures when you test more MP3s.

If you see count1 overrun, malformed bitstream or big_values overrun, malformed bitstream and get a failure, check if it also fails with mpg123. If it doesn't then I think we're good.