[Ffmpeg-devel-irc] ffmpeg-devel.log.20120509

[burek]

[00:02] <Daemon404> -async causes a stream of infinite silence on this sample... awesome
[00:02] <Daemon404> -async 1*
[00:04] <ubitux> this?
[00:04] <Daemon404> random sample someone gave me
[00:04] <Daemon404> flac in avi
[00:04] <Daemon404> should be publically avail
[00:06] Action: Daemon404 checks
[02:58] <CIA-122> ffmpeg: 03Michael Niedermayer 07master * r9501b93241 10ffmpeg/libavutil/parseutils.c: 
[02:58] <CIA-122> ffmpeg: parseutils: fix ;/, error
[02:58] <CIA-122> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:58] <CIA-122> ffmpeg: 03Michael Niedermayer 07master * rf8b536a786 10ffmpeg/ (libavutil/parseutils.c tests/ref/fate/parseutils): 
[02:58] <CIA-122> ffmpeg: parseutils: use strings instead of integers for error codes.
[02:58] <CIA-122> ffmpeg: error values can differ between platforms.
[02:58] <CIA-122> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[09:07] <ubitux> it seems our make checkheaders is heavily broken
[09:08] <ubitux> i've fixed a decade of them, but there is still some left
[09:08] <ubitux> i think i'll add a fate instance for this
[09:44] <ubitux> ok done.
[11:31] <CIA-122> ffmpeg: 03Michael Niedermayer 07master * r2c83265290 10ffmpeg/libavutil/error.c: 
[11:31] <CIA-122> ffmpeg: error: Try to make av_strerror() closer to binary identical across platforms.
[11:31] <CIA-122> ffmpeg: This should fix the parseutils fate test.
[11:31] <CIA-122> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[11:59] <CIA-122> ffmpeg: 03Carl Eugen Hoyos 07master * reae77f4631 10ffmpeg/libavcodec/lagarith.c: Fix lagarith YUY2 output.
[15:40] <ubitux> michaelni: it won't invaluate your argument, but do you know git grep?
[15:43] <michaelni> hmm
[15:43] <michaelni> i thought i knew most git command, but indeed i didnt know about that one :)
[15:44] <michaelni> thanks, i guess this one will proof usefull
[15:48] <ubitux> :)
[17:12] <ubitux> does anyone know if there is some kind of "gcc assert"?
[17:13] <ubitux> __builtin_expect() is supposed to be what i want but i'm not able to make it work as i want
[17:13] <ubitux> for example, with: int f(int x) { return x / 2; }
[17:14] <ubitux> if we add a assert(x>0);, the branch for "unsigned" will be simpler
[17:14] <ubitux> the code becomes if (x<0) return x>>1; else [crap]
[17:14] <ubitux> (x > 0 sorry)
[17:15] <ubitux> i'm trying to make gcc understands it's always the case by some mean
[17:15] <ubitux> (without the assert since it adds a branch)
[17:15] <ubitux> doing int f(int x) { __builtin_expect(x > 0, 1); return x / 2; } doesn't help
[17:17] <kierank> there is a likely() function somewhere in ffmpeg iirc
[17:18] <kierank> if that's what you mean
[17:19] <ubitux> given the name it looks like it will just make the compiler re-organize the branches in the more optimized way
[17:19] <ubitux> not drop them
[17:20] <ubitux> ah and it seems to use __builtin_expect()
[17:21] <ubitux> (i may be blind but i don't see any usage)
[17:22] <iive> ubitux: if it always the case (x is unsigned)  then make it unsigned.
[17:22] <ubitux> well this is just as a test purpose
[17:23] <ubitux> it could be a function which assumes a limited range of values which would help some misc. optim
[17:23] <ubitux> which gcc could be able to detect if told so about the possible values
[17:24] <ubitux> the signed/unsigned was the most obvious one i had in mind
[18:24] <CIA-122> ffmpeg: 03Michael Niedermayer 07master * r1ca37e79d0 10ffmpeg/libavfilter/src_buffer.c: 
[18:25] <CIA-122> ffmpeg: src_buffer: ifdef->if CONFIG_*
[18:25] <CIA-122> ffmpeg: Reviewed-by: Stefano
[18:25] <CIA-122> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:25] <CIA-122> ffmpeg: 03Michael Niedermayer 07master * ra54867121d 10ffmpeg/ffmpeg.c: 
[18:25] <CIA-122> ffmpeg: ffmpeg: fix frame duration used for video sync.
[18:25] <CIA-122> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:28] <PapaSmurf007> hmm, can MPEG-2 have multiple MB types per frame?
[21:31] <kierank> it can mix frame/field if that's what you mean
[21:36] <PapaSmurf007> no, I'm asking if one MB can be encoded as a inter-predicted block and another be encoded as an intra-predicted block
[21:36] <PapaSmurf007> in the same frame
[21:37] <kierank> yes
[21:37] <kierank> in a p/b frame
[21:45] <Compn> there are so many ways to watermark a video its strange we havent seen more
[21:45] <Compn> big ugly brown dots are not cool, NOT COOL!
[23:48] <ubitux> michaelni: can we backport 162e4004643c929a8a799f4362c68bef5a53ba50 to 0.10?
[23:48] <ubitux> in 0.10 the filtering_audio example is broken because of this
[23:49] <ubitux> it's not that important since the header is deprecated now though
[23:56] <michaelni> ubitux, feel free to backport what you like, maybe ask saste/nicolas
[23:56] <michaelni> first though ...
[23:58] Action: ubitux asks saste 
[23:58] <saste> ubitux: no problem with me
[23:59] <saste> the only problems is that those kind of changes require a version bump and i don't know how we should deal with them in case of releases, maybe you can just skip the version bump
[00:00] --- Thu May 10 2012