Arthur Taylor
Arthur Taylor
I have found one deficiency in the code related to pre-skip. The calculation of how many samples to skip doesn't take into account non-zero initial granule positions. ``` /* **...
Looking at the files, yes `toshiba-ffmpeg.ogg` would appear to have an incorrect prekip value in it's header, an thus be a malformed file. Both `toshiba-ffmpeg.ogg` and `toshiba-opusenc.ogg` have the same...
Looking at ffmpeg sources, yes it has a bug where the preskip amount is written at encode relative to the input sample rate. Worse, fmpeg always decodes Opus at 48kHz...
> Interestingly, libav also have this bug? It would appear so! Unlike ffmpeg, libav uses libopus, but then falls victim to a gotcha where the encoder returns the lookahead (preskip)...
Update: Patch accepted by ffmpeg upstream for preskip.
Humm. The seek test checks for bit-exact decoding after a seek. So, while -0.000935563 does equal -0.000935563, my hunch is that this is probably a deception and the values do...
Since #457 libsndfile should at least be consistent and only report the size of the first part now.
Might have been fixed inadvertently by #457. Need to test again.
Just my two cents, but I prefer K&R or other one-true-brace like styles over Allman braces. It's just a feel thing, but I dislike the whitespace gaps having the opening...
Should also point out there already exists a multitude of commenting formats in the project. I started using: ``` /* ** This is the fist line of a multiline comment...