[Ffmpeg-devel-irc] ffmpeg-devel.log.20111212

burek burek021 at gmail.com
Tue Dec 13 02:05:02 CET 2011


[00:02] <burek> if i need just to get the latest updates from git (i did git clone ... around 1st december) is it enough to type git update or i need to remove all files and again do git clone?
[00:07] <Compn> a git pull should be enough
[00:07] Action: Compn nto sure what git update does
[00:07] <Compn> svn update ...
[00:07] <burek> yes, git pull sorry :)
[00:07] <burek> thanks :)
[00:12] <ubitux> michaelni: you're not mentionning libass filter in the changelog? :)
[00:12] <michaelni> git: 'update' is not a git command. See 'git --help'.
[00:14] <Compn> i wonder if someone could come up with a little script
[00:14] <Compn> it converts 'svn up' to 'git pull' and 'svn commit' to 'git commit' + 'git push' 
[00:14] <Compn> ;p
[00:32] <_Thomas> Any devs alive currently?   I have the needed docs for the hw codecs of an ARM-based SoC, and want to add support for this in ffmpeg, but I can't seem to find any info on where I should put this in ffmpeg?
[00:33] <_Thomas> the hw codecs does full decode, not just acceleration of some of the steps in h.264 (for instance)
[00:33] <j-b> openmax il?
[00:34] <_Thomas> No, I was hoping to do ffmpeg -> hw, without using any openmax
[00:34] <_Thomas> ;)
[00:35] <Compn> _Thomas : you are looking for the ffmpeg docs in git ?
[00:35] <Compn> or what ?
[00:35] Action: Compn confused
[00:35] <Compn> oh, what file the hw support goes in ?
[00:35] <_Thomas> Compn: I'm just looking for some hints on what parts of ffmpeg I should modify / add libs to to make ffmpeg use hw codecs on the platform 
[00:36] <_Thomas> Compn: yes :)
[00:36] <_Thomas> Compn: My platform does decode of h264, h263, mpeg2, mpeg4, vc1
[00:36] <_Thomas> and encode h264, h264, mpeg4
[00:36] <Compn> i'd look at similar code like crystalhd wrapper or vdpau 
[00:36] <Compn> but you should ask michaelni
[00:36] <Compn> because i'm not sure :)
[00:37] <merbanan> _Thomas: you could look at the stagefright api implementation
[00:37] <_Thomas> stagefright uses ffmpeg?
[00:37] <merbanan> no there is a stagefright_h264 decoder in ffmpeg
[00:38] <merbanan> instead of using the software decoder it uses the hardware decoder through the stagefright api
[00:38] <_Thomas> ah
[00:38] <merbanan> look how that code integrates with the rest of the ffmpeg internal api
[00:38] <_Thomas> The SoC in question is the Exynos 4210, btw
[00:39] <Compn> it probably depends if your hardware does full h264 accel or parts of the codec accel like cabac or stuff like that
[00:39] <merbanan> hopefully you feed it annexb frames
[00:39] <merbanan> with not to much quirks
[00:40] <_Thomas> you feed frames, yes
[00:40] <_Thomas> raw (for the encoder) and annexb (I think) for the decoder
[00:40] <merbanan> anyway, wbs knows a few things about this
[00:40] <_Thomas> seems like crystalhd.c also is a place to look?
[00:40] <merbanan> yes
[00:41] <merbanan> they should have implemented the needed framing
[00:41] <_Thomas> cool
[00:43] <_Thomas> thanks, I think I got what I asked for :)
[00:47] <Compn> _Thomas : btw if you want to submit something that lets other devels know where hwaccel code is/goes, that would be helpful for future people :)
[00:49] <_Thomas> Compn: Good point :)
[00:54] <_Thomas> First step to getting XBMC working is ffmpeg with hw support
[01:02] <merbanan> I think we might need to fix the api so one can choose decoders for a codec
[01:02] <merbanan> now there is only one IIRC
[01:43] <CIA-49> ffmpeg: 03Diego Biurrun 07master * rda9cea77e3 10ffmpeg/ (51 files in 9 dirs): Fix a bunch of common typos.
[01:43] <CIA-49> ffmpeg: 03Martin Storsjö 07master * re2484fb644 10ffmpeg/libavformat/ (movenc.c movenc.h movenchint.c): (log message trimmed)
[01:43] <CIA-49> ffmpeg: movenc: Use the actual converted sample for RTP hinting
[01:43] <CIA-49> ffmpeg: If an annex b bitstream is muxed into mov, the actual written
[01:43] <CIA-49> ffmpeg: sample is reformatted to mp4 syntax before writing.
[01:43] <CIA-49> ffmpeg: Currently, the RTP hints that copy data from the normal video
[01:43] <CIA-49> ffmpeg: track, where the payload data might be offset compared to the
[01:43] <CIA-49> ffmpeg: original sample that the RTP hinting used (when 3 byte
[01:43] <CIA-49> ffmpeg: 03Luca Abeni 07master * re0a3c287b4 10ffmpeg/libavformat/rtpenc_h264.c: 
[01:43] <CIA-49> ffmpeg: rtpenc: Add support for mp4 format h264
[01:43] <CIA-49> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[01:43] <CIA-49> ffmpeg: 03Martin Storsjö 07master * r0ebd4083e1 10ffmpeg/libavformat/sdp.c: 
[01:44] <CIA-49> ffmpeg: sdp: Restore the original mp4 format h264 extradata if converted
[01:44] <CIA-49> ffmpeg: If the sdp is generated before the rtp muxer is initialized
[01:44] <CIA-49> ffmpeg: (e.g. as when called from the rtsp muxer), this has to be done,
[01:44] <CIA-49> ffmpeg: otherwise the rtp muxer doesn't know that the input really is
[01:44] <CIA-49> ffmpeg: in mp4 format.
[01:44] <CIA-49> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[01:44] <CIA-49> ffmpeg: 03Mans Rullgard 07master * r373211d828 10ffmpeg/ (8 files in 4 dirs): 
[01:44] <CIA-49> ffmpeg: Remove extraneous semicolons
[01:44] <CIA-49> ffmpeg: These semicolons cause invalid empty top-level declarations.
[01:44] <CIA-49> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[01:44] <CIA-49> ffmpeg: 03Andrey Utkin 07master * r5b3265a718 10ffmpeg/libavfilter/Makefile: 
[01:44] <CIA-49> ffmpeg: lavfi: install libavfilter/vsrc_buffer.h
[01:44] <CIA-49> ffmpeg: It declares API necessary for filtering apps.
[01:44] <CIA-49> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[01:44] <CIA-49> ffmpeg: 03Mans Rullgard 07master * r3383a53e7d 10ffmpeg/ (23 files in 3 dirs): (log message trimmed)
[01:45] <durandal_1707> anybody have free time to spend it for me?
[01:47] <Daemon404> ?
[01:48] <durandal_1707> i think i finally found cdxl with RGB encoded video
[01:49] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r02a01c2fa1 10ffmpeg/tests/Makefile: 
[01:49] <CIA-49> ffmpeg: fate: replace forgotten avconv by ffmpeg in vsynth / asynth stuff
[01:49] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:05] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rcd2f7ed000 10ffmpeg/libavformat/seek-test.c: 
[02:05] <CIA-49> ffmpeg: seek-test: use standard format specifiers in snprintf()
[02:05] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:06] <michaelni> durandal_1707, if theres anything i can help with just ask ...
[02:07] <durandal_1707> well i'm now attepting to decode rgb data in cdxl
[02:35] <RT|AO> "This channel is publically logged" # who is doing that besides me?
[02:35] <Compn> wheres your logs RT|AO ?
[02:36] <Compn> http://ffmpeg.org/pipermail/ffmpeg-devel-irc/
[02:36] <Compn> seems burek is doing them :)
[02:36] <RT|AO> I see. Mine is password protected. ;-)
[02:36] <RT|AO> http://roy.orz.hm/irclog/%23ffmpeg-devel/latest.log.html
[04:38] <Compn> did ffmpeg ever get midi support ?
[04:38] <Compn> or is that not committed ?
[07:27] <ubitux> Compn: should work with libmodplug wrapper
[07:28] <ubitux> but not when muxed though
[09:35] <Tjoppen> make: *** [fate-utvideo_rgba_left] Error 1
[09:35] <Tjoppen> FYI
[09:36] <ubitux> Tjoppen: even after a fate rsync?
[09:37] <Tjoppen> aha
[09:37] <Tjoppen> yeah, that updated the samples
[09:40] <Tjoppen> success!
[10:01] <Tjoppen> time to post some patche
[10:01] <Tjoppen> s
[14:55] <ubitux> mmh i wanted to extract the frame rate directly in the filtergraph (for timecode burning)
[14:55] <ubitux> the filterchain has access to the time_base 
[14:55] <ubitux> while it seems to work with mxf material (for instance) where i get a 24/1 frame rate
[14:56] <ubitux> it doesn't if it is some classic mpeg2 file, where i have a 1/90000 tb
[14:56] <ubitux> atm the patch use the user specified frame rate, but maybe i could do some specific heuristic?
[14:57] <Tjoppen> I'm fairly sure timecodes can be parsed from mpeg2
[14:57] <ubitux> (so we can avoid the need of a -r)
[14:57] <ubitux> that's not the issue :)
[14:57] <Tjoppen> it shouldn't depend on time_base at all
[14:57] <ubitux> ok
[14:57] <Tjoppen> you should consider timecodes to just be frame metadata
[14:57] <ubitux> so the asking the user to specify the rate looks mandatory to you?
[14:58] <Tjoppen> well, I'm thinking of the case where there actually are timecodes
[14:58] <ubitux> note i'm just talking about timecode burning
[14:58] <Tjoppen> in which case ffmpeg shouldn't go inventing its own
[14:58] <ubitux> i sent a patchset last week for this
[14:58] <ubitux> using drawtext to burn the specified timecode
[14:58] <ubitux> sth like -vf drawtext=font=...:text='TCR ':timecode=00...:r=30000/1001
[14:59] <Tjoppen> I see it now. hm
[14:59] <ubitux> i just wanted to check if the rate couldn't be guessed based on the time base
[14:59] <ubitux> which is available in the filterchain
[14:59] <Tjoppen> avg_frame_rate?
[14:59] <ubitux> is that realiable?
[14:59] <ubitux> afaik it's not in the filterchain anyway
[15:00] <Tjoppen> well, it seems reliable for "normal" files
[15:00] <Tjoppen> codec->time_base is often twice the framerate and stream->time_base is usually whatever
[15:00] <ubitux> how do you guess the frame rate when you have for instance a mpeg with 90000/1 tb ?
[15:01] <Tjoppen> lavf looks at the timestamps. I think it has some magic to properly detect 23.98 fps and such
[15:02] <Tjoppen> also note: timecode != timestamp
[15:03] <Tjoppen> in case you were thinking of computing them like initial_timecode + delta_pts * some_constant
[15:03] <Tjoppen> it can work fine for PAL, but for NTSC timecodes and timestamps are off by 1 ppm
[15:04] <ubitux> no no i know that
[15:04] <Tjoppen> good :)
[15:04] <ubitux> all the code is actually working; i'm just trying to see if i can avoid the need of a rate parameter :p
[15:04] <ubitux> reliably
[15:04] <ubitux> but i guess i'll just go with the current version
[15:04] <ubitux> requesting the user to specify the input rate
[15:05] <Tjoppen> you could perhaps have certain values for avg_frame_rate be OK
[15:05] <Tjoppen> if you can access it. or the ffmpeg CLI can look at the value and set the option accordingly?
[15:05] <ubitux> avg_frame_rate is not available in the lavfi scope
[15:06] <ubitux> i don't understand what you're saying about the cli
[15:06] <Tjoppen> it could set the optional automagically if avg_frame_rate is any of the outer product {24,25,30}*{1/1,1000/1001}
[15:06] <Tjoppen> *options
[15:07] <Tjoppen> so tricky stuff like VFR still requires the option
[15:08] <ubitux> ok, i'll eventually add optionnal specific heuristics later
[15:10] <Tjoppen> AVFrame doesn't have metadata yet?
[15:10] <ubitux> michaelni: could you check if this looks better for ff_ to avpriv_ bump? http://blackhole.ubitux.fr/tc/0004-timecode-rename-internal-ff_-symbols-to-avpriv_.patch
[15:10] <ubitux> Tjoppen: AVFrame is not in the filterchain
[15:11] <Tjoppen> ah
[15:31] <ubitux> haha Nicolas G. :D
[15:31] <ubitux> Maybe lose the "0.": 0.8.7 becomes 8.7, 0.9 becomes 9.0, and the next major release would be 10.0.
[15:31] <ubitux> and then 100.0, left shift each time
[15:31] <ubitux> 1000.0, 10000.0...
[15:35] <michaelni> it shouldnt be hard to add a avg_frame_rate field to some struct in lavfi if that helps
[15:36] <ubitux> i don't think it's worth the effort, at least at first
[15:36] <Tjoppen> ubitux: do all your use cases use sane containers and/or codecs?
[15:36] <michaelni> ff/avpriv patch LGTM 
[15:36] <ubitux> Tjoppen: it should& but i hope i won't be the only user ;)
[15:36] <Tjoppen> as in, do all of them have proper timecodes and could be exposed somehow?
[15:37] <ubitux> you can have timecode at different levels
[15:37] <ubitux> codecs & formats
[15:37] <ubitux> user might want to choose between them
[15:37] <Tjoppen> yeah. like mpeg2 can have timecodes per frame but in mxf it's just StartTimecode + delta from there
[15:39] <Tjoppen> for op1a at least.. though there needs to be at least one contiguous timecode track
[15:40] <ubitux> michaelni: i'll push the whole patchset soon then, if you don't have a better idea for the user input frame rate in drawtext filter
[15:40] <ubitux> michaelni: and thx for having a look
[15:42] <michaelni> you can calculate the average framerate from the timestamps of the past frames but then it would not be available at first
[15:42] <ubitux> Tjoppen: also, avg_frame_rate might not be enough; i'm not sure you can guess accuratly all the frame rate in rationnal form from this
[15:43] <ubitux> michaelni: i was thinking of using time_base from the filter chain at first
[15:43] <Tjoppen> why not just burn pts and time_base as-is?
[15:43] <Tjoppen> and maybe a running mean FPS
[15:43] <ubitux> but while it seems to work sometimes (got a 24/1 with a mxf file for instance) it doesn't for a mpeg2 one
[15:43] <ubitux> (90000/1)
[15:44] <Tjoppen> considering timecode is just a glorified frame counter
[15:45] <ubitux> i don't understand what you propose :p
[15:47] <michaelni> when you have a frame with timestamp 0 and the next with timestamp 3750 and tb=1/90000 then the avg fps is 24
[15:47] <ubitux> yes i got that
[15:48] <ubitux> but unfortunately i need the first frame to be processed too
[15:49] <michaelni> that either requires user input or passing the avg fps through the chain or some added init pass for filters where they get the first frames twice so they can learn these parameters before actual filtering
[15:51] <ubitux> then user input will be just fine imho
[15:51] <CIA-49> ffmpeg: 03Martin Storsjö 07master * r8349dbfe46 10ffmpeg/libavcodec/x86/cabac.h: 
[15:51] <CIA-49> ffmpeg: x86: Require 7 registers for the cabac asm
[15:51] <CIA-49> ffmpeg: The change in 599b4c6ef didn't turn out to work properly on
[15:51] <CIA-49> ffmpeg: i386 on OS X, where it broke building with PIC enabled.
[15:51] <CIA-49> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[15:51] <CIA-49> ffmpeg: (cherry picked from commit f1dba9e4988e78738ad9065e4639b82b8355774a)
[15:51] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:51] <ubitux> i'll push the patchset in a moment
[17:22] <ubitux> someone is asking for a gradual fade on the mailing list afaiu
[17:22] <ubitux> i wonder if the alpha parameter of the fate filter couldn't be an eval
[17:22] <ubitux> so it can be computed according to the pts
[17:27] <Compn> thats all anyone wants
[17:28] <Compn> gapless playback, fading between videos
[17:28] <Compn> uhh
[17:34] <ubitux> shouldn't be too hard to do
[17:37] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * ree731c1ab0 10ffmpeg/ffmpeg.c: 
[17:37] <CIA-49> ffmpeg: ffmpeg: fix codecids with ffserver + ffmpeg + ffm
[17:37] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:53] <CIA-49> ffmpeg: 03Clément BSsch 07master * r66160bdb01 10ffmpeg/libavcodec/ (libx264.c proresdec2.c): cleanup: remove two extraneous semicolons.
[17:53] <CIA-49> ffmpeg: 03Clément BSsch 07master * r98aca4bbef 10ffmpeg/libavcodec/timecode.c: timecode: better input checks in init function.
[17:53] <CIA-49> ffmpeg: 03Clément BSsch 07master * rad5ade6615 10ffmpeg/libavcodec/ (timecode.c timecode.h): timecode: add avpriv_timecode_to_string().
[17:53] <CIA-49> ffmpeg: 03Clément BSsch 07master * rfa4e30af7e 10ffmpeg/ (doc/filters.texi libavfilter/vf_drawtext.c): drawtext: allow burning the timecode.
[17:53] <CIA-49> ffmpeg: 03Clément BSsch 07master * r80914cde6f 10ffmpeg/ (7 files in 3 dirs): 
[17:53] <CIA-49> ffmpeg: timecode: rename internal ff_* symbols to avpriv_*.
[17:53] <CIA-49> ffmpeg: Those functions are shared between libs.
[17:53] <CIA-49> ffmpeg: Also fix a typo in function names: smtpe ’ smpte.
[18:35] <CIA-49> ffmpeg: 03Tomas Härdin 07master * rc0ee522df6 10ffmpeg/libavformat/mxfdec.c: 
[18:35] <CIA-49> ffmpeg: mxfdec: Bring back accumulated_offset that was removed in 5e67e3e
[18:35] <CIA-49> ffmpeg: I thought it had to do with file offsets, but's actually the offset inside
[18:35] <CIA-49> ffmpeg: the essence container.
[18:35] <CIA-49> ffmpeg: In other words, unbreak multiple EditUnitByteCounts.
[18:35] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:35] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r1e14bfc117 10ffmpeg/libavformat/mxfdec.c: 
[18:35] <CIA-49> ffmpeg: mxfdec: Add support for Avid indexes
[18:35] <CIA-49> ffmpeg: These indexes duplicate every entry and have the total size of the essence
[18:35] <CIA-49> ffmpeg: container as the last entry.
[18:35] <CIA-49> ffmpeg: This patch also computes the size of the packets when unknown.
[18:35] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:35] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r0bcada9f61 10ffmpeg/libavformat/mxfdec.c: 
[18:35] <CIA-49> ffmpeg: mxfdec: Add Avid's essence element key
[18:35] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:35] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r8a2cb2abbf 10ffmpeg/libavformat/mxfdec.c: 
[18:35] <CIA-49> ffmpeg: mxfdec: Only sort IndexTableSegments once
[18:36] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:36] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r7d531e83b6 10ffmpeg/ (libavformat/mxfdec.c tests/ref/seek/lavf_mxf): 
[18:36] <CIA-49> ffmpeg: mxfdec: Set DTS and PTS on video packets
[18:36] <CIA-49> ffmpeg: This involves computing DTS/PTS from the index tables.
[18:36] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:36] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r89cc8c17df 10ffmpeg/libavformat/mxfdec.c: 
[18:36] <CIA-49> ffmpeg: mxfdec: Check slice_count before freeing slice_offset_entries
[18:36] <CIA-49> ffmpeg: This fixes SIGSEGV on sliceless (single stream) VBR files
[18:36] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:36] <luke-jr> Why does ffmpeg 0.8.7 ignore CFLAGS, and leave out -mfpu=neon, for the NEON configure test? <.<
[19:38] <Compn> luke-jr : because it hates you :P
[19:38] <luke-jr> Compn: sure, even if I use --extra-cflags to add it, it fails to assemble ;)
[19:39] <luke-jr> in fact, I can't seem to get ffmpeg to build on ARM at all
[19:39] <Compn> why .8.7 and not master ?
[19:39] <Compn> does libav build ? mru does the arm/neon stuff 
[19:39] <luke-jr> I don't know how stable master is.
[19:40] <Compn> master is stable , not api stable, but then again, nothing is api stable anymore
[19:43] <luke-jr> Compn: well, if I'm just interested in mplayer/mplayer2, will master work? >_<
[19:45] <Compn> mplayer svn automatically pulls ffmpeg git master if you compile mplayer manually...
[19:46] <Compn> so yes, it will work
[19:46] <Compn> well, on x86
[19:46] <Compn> i dont think mplayer has anyone testing on arm lately
[19:46] <Compn> mplayer is lacking devels something bad.
[19:46] <Compn> will older ffmpeg work on newer mplayer? dunno
[20:01] <michaelni> luke-jr, check fate.ffmpeg.org there are 2 ARM clients 
[20:02] <michaelni> http://fate.ffmpeg.org/report.cgi?time=20111212162021&slot=qemu-armv7-ubuntu-gcc4.5
[20:02] <michaelni> above lists the used flags
[20:02] <Compn> michaelni : luke-jr wants you to fix 8.7 on arm tho
[20:03] <michaelni> 8.7 is worse than master ?
[20:03] <michaelni> 0.8.7 :)
[20:03] <Compn> he said it fails to assemble 
[20:03] <Compn> 0.8.7
[20:05] <luke-jr> in many different ways :/
[20:06] <luke-jr> 0.7.7/8 isn't any better either
[20:06] <Compn> luke-jr : does master fail ?
[20:06] <luke-jr> Compn: still waiting for it to clone
[20:06] <Compn> shouldnt take that long
[20:06] <michaelni> master has to work because well its working for me :)
[20:06] <Compn> :D
[20:06] <michaelni> id test 0.8.7 but arm-qemu is SLOOOOOOOW
[20:07] <ubitux> same here, so it definitely works
[20:07] <luke-jr> Compn: well, 5 MB/s disk bandwidth and 600 MHz CPU&
[20:07] <Compn> ah
[20:07] <Compn> you guys need to cross compile :P
[20:07] <michaelni> a single compile + fate under qemu is 6h IIRC so i wont start testing before i know what to test
[20:07] <ubitux> hey btw i tried to make fate with valgrind, and got a lot of issues :)
[20:08] <ubitux> in various places
[20:08] <luke-jr> fate?
[20:08] <av500> michaelni: cross compile
[20:08] <ubitux> ffmpeg automated testing env (iirc)
[20:08] <michaelni> ubitux, what issues ?
[20:08] <luke-jr> compile doesn't take *that* long on my system <.<
[20:08] <ubitux> michaelni: invalid read and similar
[20:08] <Compn> fate is ffmpeg's complete testing system to look for regressions or errors in compilation
[20:08] <ubitux> michaelni: i'll make a fate instance "soon"
[20:08] <michaelni> :)
[20:10] <ubitux> TARGET_EXEC=valgrind --error-exitcode=1 --malloc-fill=0x2a --undef-value-errors=no
[20:10] <ubitux> you can try with this
[20:10] <luke-jr> fwiw, I'm using (for the rest of my system, ffmpeg ignores it&) CFLAGS='-O2 -pipe -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard'
[20:10] <michaelni> av500, if i cross compile then theres no single person on earth who tests native compile ;)
[20:10] <luke-jr> michaelni: does qemu really count as native? ;)
[20:11] <av500> michaelni: you can do both
[20:11] <michaelni> if i find the time i might setup both ...
[20:12] <michaelni> my fate client uses -extra-cflags='-mfpu=neon -mfloat-abi=softfp' --disable-debug --cpu=cortex-a8 
[20:12] <michaelni> and --cc='ccache gcc-4.5'  of course
[20:12] <burek> was there any effort about creating an ffmpeg/ffplay player as a browser plugin?
[20:13] <luke-jr> ./configure --prefix=/usr --libdir=/usr/lib --shlibdir=/usr/lib --mandir=/usr/share/man --enable-shared --cc=armv7a-hardfloat-linux-gnueabi-gcc --disable-optimizations --disable-static --enable-gpl --enable-version3 --enable-postproc --enable-avfilter --disable-stripping --disable-debug --disable-doc --disable-vaapi --disable-vdpau --enable-libmp3lame --enable-libvo-aacenc --enable-libtheora --enable-libvorbis --enable-libxvid --
[20:13] <luke-jr> disable-indev=v4l --disable-indev=jack --enable-x11grab --enable-libfreetype --enable-pthreads --enable-libgsm --enable-libspeex --enable-libopenjpeg --disable-amd3dnow --disable-amd3dnowext --disable-altivec --disable-avx --disable-mmx --disable-mmx2 --disable-ssse3 --cpu=cortex-a8 --enable-hardcoded-tables
[20:13] <luke-jr> ^ full configure :p
[20:13] <luke-jr> (that's for 0.8.7)
[20:13] Action: luke-jr ponders why he's disabling v4l
[20:14] <av500> burek: there is mplayer and vlc for the browser, no?
[20:14] <luke-jr> NEON enabled              no <-- master fails to detect NEON at least
[20:14] <burek> yes, but they are kinda unstable :/
[20:15] <michaelni> for me with master (and my configrue command) NEON enabled              yes
[20:16] <Compn> michaelni : what did you ask about the ffmpeg website again? link to something ?
[20:16] Action: Compn has time to work now
[20:16] <michaelni> also naively --disable-optimizations somehow looks like a bad idea
[20:16] <michaelni> Compn, there are some new texi files that get changed to html automatically but i think we have no links to them
[20:16] <Compn> luke-jr : can you paste the errors during make to pastebin ?
[20:16] <Compn> oh
[20:17] <michaelni> like http://ffmpeg.org/platform.html
[20:17] <Compn> in /doc/ then :)
[20:18] <michaelni> none of this is new "info" its just splited out of existing texi files or converted from .txt
[20:18] <Compn> right , i see ubitux committed something already, i think
[20:18] <michaelni> http://ffmpeg.org/documentation.html has some links to the generated html but not complete
[20:19] <luke-jr> michaelni: why does ffmpeg ignore CFLAGS, and require --extra-cflags?
[20:19] <Compn> er no, derek buitenhuis made a patch that is...
[20:20] <Compn> luke-jr : at least it helps show up in the ffmpeg info how it was compiled :P
[20:20] <Compn> but i dont know if that was the reason
[20:20] <luke-jr> usually CFLAGS does too ;)
[20:20] <luke-jr> err, well in config.log at least
[20:20] <luke-jr> no reason ffmpeg couldn't include it in its own stuff too
[20:22] <michaelni> luke-jr, i dont know at all, maybe its so that host and target cflags dont get mixed, but iam just guessing
[20:22] <luke-jr> doesn't explain it for native builds ;)
[20:23] <michaelni> maybe its  a bug for native builds
[20:26] <michaelni> CFLAGS=FUN ./configure fails for me so they are not ignored
[20:27] <michaelni> CFLAGS='-O0' ./configure seems working as expected, so where is the problem ?
[20:27] <Compn> luke-jr : what make, bash, gcc and as version are you running ?
[20:29] <luke-jr> GNU Make 3.82
[20:29] <luke-jr> GNU bash, version 4.1.9(1)-release (armv7a-hardfloat-linux-gnueabi)
[20:29] <luke-jr> gcc (Gentoo 4.5.3-r1 p1.0, pie-0.4.5) 4.5.3
[20:29] <luke-jr> GNU assembler (GNU Binutils) 2.21.1
[20:31] <luke-jr> master is still building fwiw
[20:32] Action: Compn wonders why libav moved platform specific out of general docs anyhow
[20:33] <Compn> michaelni : want to just revert 1de6e14efe1f3966023ad1121a6ce3fcbbb1b573 
[20:33] <Compn> theres really no reason to move it out to a new file. just means another click a user has to make to find documentation :P
[20:34] <Compn> and they put build rules under 'developer documentation' , but building isnt devel-only instructions... but i guess this is just me talking to myself
[20:34] <michaelni> Compn, if people agree that the platform split was a bad idea i can revert it
[20:34] <Compn> i'll ask on list then
[20:34] <michaelni> ok
[20:37] <Compn> i mean if you want to seperate it so it makes changes from them easier, i can understand that
[20:37] <Compn> i'm not really paying attention to the merges recently, so i dont know the plan.
[20:41] <michaelni> I doubt there will be enough changes to platform.texi to make much of a difference to the merge work
[20:43] <Compn> gotta replace ramiro's url on the platform page too
[20:51] <CIA-49> ffmpeg: 03Reimar Döffinger 07master * r6c723f3f9d 10ffmpeg/tests/ (10 files in 2 dirs): 
[20:51] <CIA-49> ffmpeg: lavf-regression: minimal metadata test.
[20:51] <CIA-49> ffmpeg: This tests writing the global "title" metadata.
[20:51] <CIA-49> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[21:13] <luke-jr> http://paste.pocoo.org/show/519905/
[21:17] <michaelni> luke-jr, is this with disabled optimizations ?
[21:20] <luke-jr> michaelni: no idea :D
[21:20] <luke-jr> # ./configure --prefix=/usr --libdir=/usr/lib --shlibdir=/usr/lib --mandir=/usr/share/man --enable-shared --cc=armv7a-hardfloat-linux-gnueabi-gcc --disable-optimizations --disable-static --enable-gpl --enable-version3 --enable-postproc --enable-avfilter --disable-stripping --disable-debug --disable-doc --disable-vaapi --disable-vdpau --enable-gnutls --enable-libmp3lame --enable-libvo-aacenc --enable-libtheora --enable-libvorbis --
[21:20] <luke-jr> enable-libxvid --disable-indev=v4l --disable-indev=jack --enable-x11grab --enable-libv4l2 --enable-libfreetype --enable-libass --enable-pthreads --enable-libgsm --enable-libmodplug --enable-libspeex --enable-libopenjpeg --disable-amd3dnow --disable-amd3dnowext --disable-altivec --disable-avx --disable-mmx --disable-mmx2 --disable-ssse3 --disable-vis --disable-iwmmxt --cpu=cortex-a8 --enable-hardcoded-tables
[21:20] <michaelni> --disable-optimizations <-- try without this
[21:21] <michaelni> disabling optimizations will break asm on x86 so i would not be surprised if itz does the same on arm
[21:32] <Compn> should have a warning when some smart user tries to --disable-optimizations
[21:33] <kierank> or some silly distro
[21:46] <MT`> hi all
[21:47] <luke-jr> it seems Gentoo always tries to --disable-optimization <.<
[21:47] <MT`> im trying to track down why my ffmpeg build is segfaulting when trying to transcode sd mpeg-2 to x264, but having difficulty getting gdb to cooperate
[21:47] <MT`> im currently using these options to configure: 
[21:48] <MT`> --extra-cflags="-g -O0" --enable-debug=3 --disable-stripping
[21:49] <MT`> which is at least getting me a decent stack trace, but in each frame, gdb doesnt recognise any symbols
[21:49] <MT`> so i have the entry arguments, all globals, but no locals
[21:49] <MT`> any ideas?
[21:50] <teratorn> MT`: pastebin?
[21:50] <MT`> of the traceback? sure
[21:52] <MT`> http://pastebin.com/5my7M3g6
[21:53] <MT`> i'd like to be able to do fr 0; p po
[21:53] <MT`> but gdb wails :(
[21:55] <CIA-49> ffmpeg: 03Reimar Döffinger 07master * r31a68ad554 10ffmpeg/tests/ref/seek/ (7 files): 
[21:55] <CIA-49> ffmpeg: Update seek reference for added metadata to lavf tests.
[21:55] <CIA-49> ffmpeg: As far as I could see the only change is increased pos values,
[21:55] <CIA-49> ffmpeg: which is as expected with additional metadata in the files.
[21:55] <CIA-49> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[21:55] <pasteeater> kierank: tell me that's not for real.
[21:56] <Compn> MT` : well, try to get a smaller command line, could help in reproducing :P
[21:57] <MT`> heh, true :)
[21:57] <Compn> looks like it crashes in copying ac3 audio to mp4 
[21:57] <MT`> I think i can track down that problem, I really just want to get a debuggable build
[21:57] <Compn> so just ffmpeg -vn -acodec copy -i blah out.mp4
[21:57] <Compn> btw that trace looks complete to me, but i dont know what i'm talking about sometimes
[21:58] <MT`> ah, but if i attempt to look at any locals inside each frame, they've been optimised away
[21:58] <Compn> each frame of what ?
[21:58] <MT`> in gdb
[21:58] <Compn> oh i havent gone that far in gdb so i dunno ;)
[21:58] <Compn> did you verify things compiled with -O0 ?
[21:59] <MT`> http://pastebin.com/TLEWrPrW
[21:59] <MT`> that sort of thing
[21:59] <MT`> yar, I passed this to configure:
[21:59] <MT`> --extra-cflags="-g -O0" --enable-debug=3 --disable-stripping
[21:59] <Compn> wouldnt you want p flags ?
[21:59] <Compn> but i mean did you verify that it ran gcc -O0 on each file...
[22:00] <MT`> no, that was going to be my second question :) how can i turn off the quiet compilation mode :)
[22:00] <MT`> i do a gmake clean before each build
[22:00] <MT`> btw, p flags?
[22:01] <Compn> instead of p po
[22:01] <Compn> p flags
[22:01] Action: Compn has no clue about programming or debugging
[22:02] <MT`> aha I see! I thought you meant compile time flags, -p/-pg is for profiling :D
[22:02] <Compn> no
[22:02] <Compn> i meant it says if po->flags
[22:03] <Compn> so type p flags into gdb
[22:04] <Compn> gmake? what os are you on ?
[22:04] <MT`> freebsd
[22:04] <Compn> what make / gmake version
[22:04] <Compn> ah
[22:04] <Compn> its nice to see someone using master on bsd :)
[22:05] <MT`> I think what im trying to do is illegal, and its dieing before it gets to tell me
[22:05] <michaelni> MT`, make sure its not using -fomit-frame-pointer
[22:05] <MT`> its not as daunting as people make out, I keep meaning to make a blog post
[22:05] <MT`> michaelni: thx, ill kick off a build with -fno-omit-frame-pointer
[22:06] <Compn> we have some bsd instructions, send a patch to update those if you want to... http://ffmpeg.org/platform.html#BSD
[22:06] <Compn> our bsd instructions just say to use gmake, but more info would be nice :)
[22:09] <MT`> yes, thats pretty much it, install latest binutils and gcc46 from ports, configure with --with-cc= --with-as=, use gmake
[22:09] <MT`> might get interesting when we get a bsd clang/llvm toolchain :)
[22:31] <durandal_1707> does ffmpeg have some api to handle different ways of storing graphics info in memory?
[22:46] <MT`> thx for the help guys (and for ffmpeg obv), found the problem - PEBKAC, I had some old profile files installed, and was still trying to use them, which new ffmpeg did not agree with
[22:46] <MT`> do you want a bug report, it probably shouldnt have seg faulted on it
[22:59] <michaelni> MT`, yes please report it
[23:03] <luke-jr> michaelni: without --disable-optimizations, my system crashes during build <.<
[00:00] --- Tue Dec 13 2011


More information about the Ffmpeg-devel-irc mailing list