Created attachment 197764 [details] [review]
speexenc: fix calculation of filler data size

Is this still valid after the GstAudioEncoder port?

The new code seems to do the flush-with-0 at every buffer. Is that intended ?
If so, then it handles the filler well, so my patch is unneeded.
Current code also writes the buffer twice, but only when the input isn't a multiple of bytes, so I guess not very often anyway.

It performs it now sort-of at every buffer (intended), because AudioEncoder is configured to ensure that there won't be some mod leftover (except when performing flushing/draining).

If by "writing the buffer twice" is meant first some 0-filling and then a subsequent copy, that should not be big deal as the flushing/draining should only happen (insignificantly) rarely.

So closing.