[FFmpeg-devel] [RFC]Use ptrdiff_t for linesize and stride
nfxjfg at googlemail.com
Sat Jan 13 05:06:43 EET 2018
On Sat, 13 Jan 2018 03:47:31 +0100
Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
> 2018-01-13 3:19 GMT+01:00 wm4 <nfxjfg at googlemail.com>:
> > On Sat, 13 Jan 2018 02:46:51 +0100
> > Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
> >> Hi!
> >> Seeing 6e80079a it appears that the unstable period is not over yet.
> >> Vittorio posted a patch that introduces a temporary type to avoid
> >> breaking API, but that may not be relevant if the api currently
> >> unstable.
> >> Attached patch (inspired by Vittorio's RFC) passes fate with
> >> --disable-asm and does not introduce new warnings with gcc 6.3.
> >> Do we generally need this change?
> >> Do we generally want it?
> >> If yes, I will likely need help for the asm changes.
> >> I would like to avoid a temporary type and I would like to avoid a
> >> future version bump only because this change suddenly becomes
> >> necessary.
> >> Please comment, Carl Eugen
> > Doing that now would break EVERY API user.
> As opposed to causing trouble only to those users who happen
> to run libavcodec on a new Intel cpu? We seem to disagree
> which is worse...
> Anyway, since we had no release since the last bump, we
> could simply bump again if this change has any advantage.
What does this have to do with new Intel CPUs?
Also you can't be serious about breaking the API this badly. Also,
we're in an ABI unstable period - we never have API unstable periods.
The suggested change does not only change the ABI, but it breaks the
API as well, because the change is not source compatible.
More information about the ffmpeg-devel