[Bf-committers] [BF-committers] Updating FFMPEG
Peter Schlaile
peter at schlaile.de
Sat Aug 8 15:26:57 CEST 2009
Hi Nicolas,
> I've looked into theses, and all are patches targeted to be reviewed
> and merged in ffmpeg upstream.
> No patch seems required to build against distro ffmpeg shared version.
the problem is, that those patches are rejected for one or the other
reason by ffmpeg folks.
That affects explicitly:
* ffmpeg_accel_bgra32.patch
I was told, that this special optimisation wasn't generic enough.
Still wondering, what they ment...
* ffmpeg_work_around_ac3_sse2_crash.patch
Disables SSE2 in a certain case, haven't submitted upstream, since
I'm sure, that I'll be told, that we have to fix our compilers.
* ffmpeg_work_around_dv_crash.patch
Haven't tried to submit. (Don't know, if I have the nerves. Any
volunteers? This could actually be accepted after a little battle on
the ffmpeg mailing list...)
Unaffected:
* swscale_stride_bugfix.patch
Was surprisingly accepted after some discussions...
If you have the time, to
a) make those patches appear in ffmpeg upstream
or
b) set up the suggested build environment, that will work with all our
build systems even with custom patches by our own *and* without
messing up the ffmpeg installation of our users
go ahead, be my guest.
And yes: those patches are *requirements* for Blender Sequencer, otherwise
you'll get *really* slow playback (the color conversion has to be done
with MMX-optimisations) and/or nasty crashes in DV and AC3 decoding (other
patches). The reason, why the MMX-optimisation only affects Blender is,
that no other video application uses BGRA32 for image buffers.
Can I expect you to be the new ffmpeg maintainer, that makes a) or b)
happen? It's not, that I'm very fond of the job and I'm getting really
angry, if people start "voting" about technical issues, that simply
aren't resolved yet. (reminds me of the way, the german government handles
internet related issues...)
Cheers,
Peter
P.S.: Nathan: It would have been nice, if you had first asked me
directly before starting a large discussion on the list...
----
Peter Schlaile
More information about the Bf-committers
mailing list