https://bugs.winehq.org/show_bug.cgi?id=41290
--- Comment #2 from Bruno Jesus 00cpxxx@gmail.com --- The problem is that nBlockAlign is 193 and cbSrcLength is 192, so there is less data than a block so it returns an error. This off by one happens because quartz/mpegsplit.c always add +1 to nBlockAlign apparently related to padding, which is fixed for some reason instead of checking if the MP3 really has padding.
The patch introduced a check that seems to affect mp3 headers negatively, I'll have to dig further.