[FFmpeg-devel] FFmpeg 3.5 / 4.0
nfxjfg at googlemail.com
Thu Apr 19 17:43:16 EEST 2018
On Thu, 19 Apr 2018 16:33:47 +0200
Nicolas George <george at nsup.org> wrote:
> James Almer (2018-04-19):
> > Had it been in 3.4 it would have mean a considerable ABI breakage as
> > well, at least without the eventual backwards compat change.
> Can you explain why you think that?
> When advising on these changes and reviewing patches, I was very careful
> that they do not introduce API nor ABI changes.
Your care is appreciated, but it still caused API changes and some
rather critical bugs.
> Apart from bugs in protocols that were not fixed immediately, the only
6 months later is "immediately"? Strange sense of time.
> change visible for applications is if they register a custom callback
> for a packet protocol and yet decide to return 0 to indicate EOF. This
> was never a documented practice, is logically absurd (0 is a valid
> packet size) and inconsistent with similar practices (UDP socket do not
> return 0 for EOF).
Applications could have relied on this behavior, though. Also not all
packet based I/O mechanisms need to be UDP or sockets.
Yes, we all know EOF behavior isn't well documented, which means we
should cope with whatever behavior applications expect.
More information about the ffmpeg-devel