[Ffmpeg-devel-irc] ffmpeg-devel.log.20121021

burek burek021 at gmail.com
Mon Oct 22 02:05:02 CEST 2012


[00:13] <ubitux> srt demuxer now outputing subrip packets, check.
[01:11] <ubitux> bot broken?
[01:11] <durandal_1707> i can reproduce mp3 seeking issue (though sample rate does not change for me)
[01:12] <ubitux> (banned by tinyurl? :p)
[01:14] <Compn> we are tired of the tinyurl
[01:14] <Compn> because we dont twitter
[01:28] <ubitux> saste: coverity spotted some stuff in xface
[01:28] <saste> ubitux, where can i find coverity reports?
[01:29] <ubitux> http://ubitux.fr/pub/pics/_xface-coverity.png
[01:29] <ubitux> saste: log in http://scan5.coverity.com:8080/
[01:29] <ubitux> don't you have an account?
[01:29] <saste> ubitux, no
[01:29] <ubitux> ask michaelni :p
[01:30] <ubitux> there are a few spotted issues in your filters ;)
[01:31] <Compn> i dont know why michaelni doesnt just make generic ffmpeg shared login 
[01:31] <Compn> :P
[01:32] <Compn> then anyone who just wants to look at it for 2 mins doesnt have to register
[01:32] <durandal_1707> Compn: in this special case it does not make any sese
[01:32] <durandal_1707> *sense
[01:32] <ubitux> Compn: there is some kind of comment system, with login associated with who fixed the bug etc
[01:33] <saste> goodnight
[01:33] <ubitux> if saste had an account, i could have assigned him some issues ;)
[01:33] <ubitux> 'night saste
[01:33] <Compn> like anyone cares who fixes what bug on a 3rd party bug tracker
[01:33] <Compn> :P
[01:33] <ubitux> well, it's pretty nice :)
[01:40] <durandal_1707> ubitux: jacosub support only some styles?
[01:41] <ubitux> what do you mean?
[01:42] <ubitux> the decoder supports only a limited subset of styles yes
[01:42] <durandal_1707> i mean using mplayer -sub <file>
[01:42] <ubitux> i guess it will use the internal mplayer decoder?
[01:43] <durandal_1707> why why why
[01:43] <ubitux> because mplayer has its own subtitles handling
[01:43] <ubitux> anyway, did you really find jacosub subtitles? @_@
[01:44] <ubitux> i thought i got all the jacosub subtitles in the world
[01:44] <durandal_1707> no, i use one from fate
[01:44] <ubitux> ah, ok
[01:44] <durandal_1707> this sucks that nothing use subtitle code
[01:44] <ubitux> because there are < 70 jacosub files in the world
[01:44] <ubitux> yeah, a bit
[01:44] <ubitux> but it might not be mature enough yet
[01:46] <ubitux> it might be time to look into this new api...
[01:46] <ubitux> but i'd like to get done with the lavfi metadata first
[04:55] <Plorkyeran> @ubitux> because there are < 70 jacosub files in the world <-- there's a lot more than 70 episodes of LoGH
[09:23] <ubitux> Plorkyeran: hehe :)
[09:24] <ubitux> http://samples.ffmpeg.org/sub/jss/ seems i don't have them :P
[09:34] <ubitux> michaelni: btw i was wondering, isn't the split/merge side data problematic for stream copy?
[09:36] <ubitux> (since the demuxer would add the side data to the packet, which is directly copied without the data being split at decoding)
[09:36] <nevcairiel> imho the whole merged side-data thing is problematic. What if you just use avformat and send it so something else entirely afterwards?
[12:39] <j-b> https://trac.videolan.org/vlc/ticket/7625
[12:39] <j-b> Any idea?
[12:43] <michaelni> nevcairiel, in that case, if it doesnt work as is you can set AVFMT_FLAG_KEEP_SIDE_DATA
[12:44] <michaelni> also if we assume the side data did has some purpose then you have to do something with it anyway
[12:44] <michaelni> or that "purpose" will not work anymore
[12:45] <ubitux> well isn't it problematic if for instance you remux a rgb8 avi?
[12:45] <ubitux> or any side data thing
[12:45] <ubitux> i don't think that's really a good idea to have that :(
[12:49] <michaelni> do you have some testcase that fails ? if so iam happy to fix it
[12:50] <saste> j-b: ratecontrol trouble
[12:50] <saste> it's weird because i'd expect the rc_eq to be always set, but seems to be NULL
[12:51] <michaelni> j-b, is that using ffmpeg or libav ?
[12:51] <michaelni> i cant help with libav bugs
[12:51] <j-b> michaelni: this is a linux VLC, so your guess is as good as mine
[12:51] <ubitux> michaelni: i don't have any file at hand now, but any media with a rgb8 videostream (and so a palette side data) will fail, right?
[12:51] <j-b> cvlc is only on Linux and so, of course, I don't control the packaging...
[12:52] Action: j-b welcome to my life
[12:52] <saste> j-b: don't you support a --report option or like that
[12:53] <saste> must be an hell to have to guess which are the involved library versions
[12:54] <j-b> saste: yes. As if it was not hard enough with libavcodec going from 52.25 to 54.99, with versions with and without decoders, with fork from Ubuntu-universe and Debian-Multimedia, now you have at least 2 forks...
[12:56] <michaelni> j-b, can this issue be reproduced somehow with command line ffmpeg ?
[13:00] <michaelni> also does vlc do something with rc_eq ? the message would indicate that its NULL which is not the default
[13:09] <saste> michaelni, can you create a coverity account for me?
[13:10] <michaelni> saste, i think i created one days ago, can you check your spam folder ?
[13:15] <saste> michaelni, found none
[13:18] <michaelni> saste, created another one
[13:19] <saste> thanks
[13:22] <saste> michaelni: should I expect a mail or so?
[13:22] <michaelni> saste, yes but i do not know how quick
[13:22] <saste> ok
[13:23] <michaelni> if you got nothing till tomorrow evening then ping me and provide a different email address than the gmail one maybe
[13:38] <saste> in some cases the spotted issues are bogus
[13:38] <saste> what should we do in that case?
[13:39] <saste> for xfaceenc the fix is trivial, but not required
[13:39] <saste> and adds an unnecessary check (so it slows it down, even if in an unnoticeable manner)
[13:40] <saste> [NULL @ 0x2158100]
[13:41] <saste> why do we print the funny hexdigit?
[13:41] <saste> is it somehow useful, or should we replace it with something more useful (e.g. the called function name?)
[13:45] <michaelni> saste, you can mark CIDs as false positive or intentional
[13:45] Action: michaelni misses cone 
[13:47] <michaelni> saste, the idea behind the address for av_log is that if you have 2 abccontext to be able to distinguish them
[14:41] <j-b> michaelni: cone seems to be restarted
[14:47] <cone-618> ffmpeg.git 03Luca Barbato 0771f7b22dba60: ffv1: split decoder and encoder * 03http://tinyurl.com/9qgeb3603
[14:47] <cone-618> ffmpeg.git 03Michael Niedermayer 0769fd0b7adbd2: Merge commit '71f7b22dba60524b2285643ae0b49d8f64977129' * 03http://tinyurl.com/922yrqz03
[14:47] <michaelni> cone :)
[14:48] <ubitux> still no plan to remove that tinyurl thing?
[15:02] <cone-618> ffmpeg.git 03Luca Barbato 074a2a4524a3f5: ffv1: propagate errors * 03http://tinyurl.com/9zlfaln03
[15:02] <cone-618> ffmpeg.git 03Michael Niedermayer 0799ea47fe5a55: Merge commit '4a2a4524a3f50ed302820ba971ddd48e78c7436f' * 03http://tinyurl.com/9clddju03
[16:19] <cone-618> ffmpeg.git 03Luca Barbato 070f13cd318719: ffv1: update to ffv1 version 3 * 03http://tinyurl.com/8rodexv03
[16:19] <cone-618> ffmpeg.git 03Mans Rullgard 07ebe46b8063ae: ARM: reinstate optimised intmath.h * 03http://tinyurl.com/8ber35h03
[16:19] <cone-618> ffmpeg.git 03Derek Buitenhuis 07f2a7236d0c73: doc/platform: Move a caveat down to the notes section * 03http://tinyurl.com/9ef5zg803
[16:19] <cone-618> ffmpeg.git 03Derek Buitenhuis 072d09b36c0379: doc/platform: Add info on shared builds with MSVC * 03http://tinyurl.com/8s5z93e03
[16:19] <cone-618> ffmpeg.git 03Michael Niedermayer 07aa760b173590: Merge commit '2d09b36c0379fcda8f984bc8ad8816c8326fd7bd' * 03http://tinyurl.com/8fa4ort03
[16:36] <ubitux> $(call ALLYES, FFPROBE LAVFI_INDEV SELECT_FILTER AVCODEC MOV_DEMUXER SVQ3_DECODER ZLIB)
[16:37] <ubitux> nice dependencies.
[16:37] <ubitux> :)
[16:47] <ubitux> michaelni: i'll commit soon the lavfi metadata injection with fate tests for scene & silence detection, unless you want me to wait more
[17:04] <cone-618> ffmpeg.git 03Diego Biurrun 076cfca5b6ae5a: ffv1: Add missing #includes to header file * 03http://tinyurl.com/8kn9fhp03
[17:04] <cone-618> ffmpeg.git 03Diego Biurrun 07886087829157: fate: Introduce ENCMUX macro for tests that require encoders and a muxer * 03http://tinyurl.com/8sduptg03
[17:04] <cone-618> ffmpeg.git 03Diego Biurrun 07a7d2861d3675: svq3: K&R formatting cosmetics * 03http://tinyurl.com/8arttqk03
[17:04] <cone-618> ffmpeg.git 03Michael Niedermayer 070e097616865e: Merge commit 'a7d2861d36756b913e85681b86ed3385274e8ced' * 03http://tinyurl.com/92ne85o03
[17:05] <ubitux> https://github.com/ubitux/FFmpeg/compare/master...lavfi-metadata
[17:15] <cone-618> ffmpeg.git 03Diego Biurrun 07af1ede069cb6: svq3: cosmetics: Drop useless parentheses * 03http://tinyurl.com/9njxlrz03
[17:15] <cone-618> ffmpeg.git 03Diego Biurrun 072e0c410485ef: fate: adpcm: Add dependencies * 03http://tinyurl.com/8dsklj803
[17:15] <cone-618> ffmpeg.git 03Diego Biurrun 07620345f930d1: fate: adpcm: cosmetics: Sort test entries * 03http://tinyurl.com/995aqz803
[17:15] <cone-618> ffmpeg.git 03Michael Niedermayer 07d7b8a9a589d9: Merge remote-tracking branch 'qatar/master' * 03http://tinyurl.com/9ptywg503
[17:15] <ubitux> meh i need to rebase again
[17:16] <michaelni> ubitux, make sure you dont access the new fields directly from outside avcodec
[17:16] <michaelni> i think i saw at least one acces from lavfi
[17:17] <ubitux> huh, really?
[17:17] <michaelni> AVFrame.metadata i think
[17:18] <ubitux> oh!
[17:18] <ubitux> ok, will fix
[17:18] <ubitux> thank you
[17:21] <ubitux> fixed in avfilter_copy_frame_props()
[17:49] <ubitux> wtf i just -c copy and the h264 input got transformed into rawvideo..
[17:50] <ubitux> oh my bad.
[17:56] <cone-618> ffmpeg.git 03Clément BSsch 076fb2fd895e85: lavc: add lavfi metadata support. * 03http://tinyurl.com/8bosxvf03
[17:56] <cone-618> ffmpeg.git 03Clément BSsch 07de23953de2d8: lavfi/select: store scene score in buf ref metadata. * 03http://tinyurl.com/9b3v8l203
[17:56] <cone-618> ffmpeg.git 03Clément BSsch 07fbedce6b8512: lavfi/silencedetect: export silence info to metadata. * 03http://tinyurl.com/9qwcoex03
[18:08] <cone-618> ffmpeg.git 03Michael Niedermayer 0738797dc31b65: compat/getopt: add {} to complex ifs * 03http://tinyurl.com/94ugnq903
[18:08] <cone-618> ffmpeg.git 03Michael Niedermayer 074ce9312d7617: libavcodec/cook: add {} to complex ifs * 03http://tinyurl.com/9ygau6d03
[18:08] <cone-618> ffmpeg.git 03Michael Niedermayer 07f72b735d41f9: libavcodec/vorbisenc: add {} to complex ifs * 03http://tinyurl.com/95pqdv803
[18:09] <cone-618> ffmpeg.git 03Michael Niedermayer 07217193facd3e: bavformat/mov: add {} to complex ifs * 03http://tinyurl.com/8lkucus03
[18:09] <cone-618> ffmpeg.git 03Michael Niedermayer 07fefe9bd78266: libavformat/movenc: add {} to complex ifs * 03http://tinyurl.com/9j2bq9903
[18:09] <cone-618> ffmpeg.git 03Michael Niedermayer 07b8a64d69a927: libavformat/mpegvideodec: add {} to complex ifs * 03http://tinyurl.com/8wbjlqw03
[18:09] <cone-618> ffmpeg.git 03Michael Niedermayer 07c2c066784f19: libavformat/mxfenc: add {} to complex ifs * 03http://tinyurl.com/94gbsnq03
[18:09] <cone-618> ffmpeg.git 03Michael Niedermayer 0740ceb6d49f19: libavformat/oggenc: add {} to complex ifs * 03http://tinyurl.com/8rsyoxt03
[18:12] <ubitux> michaelni: is there some warnings with some compiler?
[18:26] <cone-618> ffmpeg.git 03Marton Balint 07d6e95669496a: h264: add support for AFD detection * 03http://tinyurl.com/8scfs9h03
[18:31] <michaelni> ubitux, i dont know but that kind of code is just fragile, the develoer sees the association based on indention but the compiler doesnt care about the indention at all
[18:33] <michaelni> its a recipe for bugs
[18:33] <ubitux> of course
[18:33] <ubitux> i was just wondering how you spotted them all
[18:37] <michaelni> i dont know if i spotted all, it was just an ugly grep command that i used
[18:42] <Compn> michaelni : you should put it in patcheck :P
[18:42] <Compn> if possible
[18:42] <Compn> and all other sneaky code things
[18:42] <Compn> does libav use patcheck ?
[18:43] <michaelni> patcheck works on diffs, i greped *.c so its not going to work 
[18:44] <Compn> for new code i mean
[18:44] <Compn> but ok
[18:45] <saste> michaelni, ping on  "propagate or return meaningful error codes in avcodec_open2()"
[18:54] <saste> really, we should properly document all the quantization stuff (a dedicated libavcodec manual would fit)
[19:02] <cone-618> ffmpeg.git 03Marton Balint 07da569556217f: ffplay: use framedrop by default when sync is not done to video * 03http://tinyurl.com/9ukuag203
[19:02] <cone-618> ffmpeg.git 03Marton Balint 07eaa91ed863a7: ffplay: fix nosync threshold check in synchronize_audio * 03http://tinyurl.com/8dyzujs03
[19:02] <cone-618> ffmpeg.git 03Marton Balint 0766bb5b1bc9b1: ffplay: initialize audio and video pts drift * 03http://tinyurl.com/8ebxfra03
[19:02] <cone-618> ffmpeg.git 03Marton Balint 0777bd595ad2ff: ffplay: fix external time sync mode * 03http://tinyurl.com/8wb66wm03
[19:02] <cone-618> ffmpeg.git 03Marton Balint 072a4c7e6540c4: ffplay: add serial field to PacketQueue entry and populate it * 03http://tinyurl.com/96rp3a503
[19:02] <cone-618> ffmpeg.git 03Marton Balint 07b2a8850969b8: ffplay: only check external clock if current frame serial matches the displayed frame serial * 03http://tinyurl.com/8smuwbz03
[19:02] <cone-618> ffmpeg.git 03Marton Balint 07fca16a15712b: ffplay: add get_master_sync_type function * 03Error03
[19:02] <cone-618> ffmpeg.git 03Marton Balint 07d30c69251f04: ffplay: use get_master_sync_type where necessary * 03Error03
[19:02] <cone-618> ffmpeg.git 03Marton Balint 073166a6fc3797: ffplay: if there is no audio stream, use external clock by default * 03Error03
[19:02] <cone-618> ffmpeg.git 03Michael Niedermayer 07747a00b688d7: Merge remote-tracking branch 'cus/stable' * 03Error03
[19:20] <saste> ubitux: http://fate.ffmpeg.org/report.cgi?time=20121021160939&slot=x86_32-msvc10-windows-native
[19:20] <saste> damned windows path
[19:21] <saste> but it raises an unconfortable issue
[19:24] <ubitux> oh :(
[19:24] <ubitux> saste: any idea how i should fix it?
[19:24] <saste> automatic escaping
[19:24] <saste> that's why i implemented tools/ffescape
[19:25] <saste> but then we need to wrap the test in a script, which performs automatic escaping
[19:25] <saste> that's why the graph syntax sucks hard, compared to avisynth
[19:25] <saste> since requires up to two levels of escaping, three considering shell escaping
[19:25] <saste> most users get it wrong
[19:26] <saste> ubitux: also BSD is failing
[19:27] <saste> http://fate.ffmpeg.org/report.cgi?time=20121021163915&slot=x86_64-freebsd8.2-clang2.8
[19:27] <saste> though I have no idea why
[19:27] <ubitux> -video|1|14100|23.500000|14100|23.500000|1|0.001667|2006431|320|240|rgb24|1:1|I|0|0|0|0|0|0|0.850000 +video|1|14100|23.500000|14100|23.500000|1|0.001667|2006431|320|240|rgb24|1:1|I|0|0|0|0|0|0|0.840000
[19:27] <ubitux> looks like yet another floating thing
[19:28] <saste> what is the last value?
[19:28] <ubitux> the scene score
[19:28] <saste> ah ok
[19:29] <ubitux> saste: so how/where would you implement the automatic escaping?
[19:30] <saste> ubitux, I have a patch pending
[19:30] <ubitux> ah, great :)
[19:30] <saste> i want to fix a few things before committing i
[19:30] <saste> *it
[19:31] <saste> but then deploying it in the test system may be complicated
[19:31] <cone-618> ffmpeg.git 03Carl Eugen Hoyos 07fa190b3cd736: Improve MPEG-PS-in-MOV detection. * 03http://tinyurl.com/8hqxwdl03
[19:32] <ubitux> i wonder how i'm supposed to fix the freebsd one :(
[19:32] <saste> avoiding float computations?
[19:33] <ubitux> i want the value in the [0;1] range
[19:34] <saste> ubitux: you may add an option where you specify the value in integer units
[19:35] <saste> and return 0-100 integer values
[19:35] <saste> maybe someone has a better idea (this may complicate the internal code)
[19:47] <michaelni> ubitux, distclean + fate-filter-metadata-scenedetect fails with no ffprobe having being build
[19:47] <michaelni> the 0.84/0.85 can be reproduced with --disable-asm --disable-yasm on linux
[19:47] <michaelni> make sure swscale gets its bitexact flags!
[19:48] <ubitux> i added -bitexact
[19:48] <ubitux> mmh CONFIG_FFPROBE isn't enough, i will add the ffprobe dep
[19:48] <ubitux> thanks
[19:48] <ubitux> i'll fix all of this asap
[19:48] <michaelni> try -sws_flags +accurate_rnd+bitexact if its not there yet
[19:50] <ubitux> mmh ok
[20:06] <cone-618> ffmpeg.git 03Clément BSsch 07c0d56bf8a4c6: fate: fix filter metadata dependency to ffprobe. * 03http://tinyurl.com/8f39mdf03
[20:13] <ubitux> the sws flags don't seem to help :/
[20:17] <cone-618> ffmpeg.git 03Clément BSsch 07e168165489fe: doc/muxers: document mov faststart option. * 03http://tinyurl.com/9okq7vw03
[20:51] <michaelni> ubitux, is the input to the filter the same and the output score differs or the input isnt the same in the first place ?
[20:52] <ubitux> i'm going to check in a moment
[20:53] <ubitux> i was just fooling around with something right now :p
[20:53] <ubitux> saste: just added the sliding mode to showspectrum :P
[20:53] <saste> nice
[20:56] <ubitux> just sent
[20:56] <ubitux> now serious matter, fate.
[21:06] <cone-618> ffmpeg.git 03Michael Niedermayer 07571309181859: mpegaudiodec: Fix buffer handling on random access * 03http://tinyurl.com/8gwvuxw03
[21:12] <cone-618> ffmpeg.git 03Stefano Sabatini 078c2dbc380500: lavc/utils: provide more feedback in case of experimental codec * 03http://tinyurl.com/8qrq7vo03
[21:13] <ubitux> michaelni: the input isn't the same in the first place indeed, maybe the flag are not properly transmited to swscale
[21:13] <ubitux> most likely a stupid mistake from me
[21:14] Action: Compn waits for autocrop filter
[21:14] <Compn> :P
[21:15] Action: Compn afk
[21:18] <ubitux> michaelni: any simple check to see if the flags reach sws?
[21:23] <michaelni> ubitux, print the flags from sws_scale()
[21:34] <ubitux> ok they don't seem to be transmited
[21:38] <ubitux> scale_args=[0:0:(null)]
[21:38] <ubitux> mmh.
[21:39] <ubitux> wasn't this fixed?
[21:39] <saste> ubitux, is it a problem?
[21:39] <ubitux> ok it's done in ffmpeg & ffplay.
[21:39] <ubitux> but not ffprobe
[21:39] <ubitux> saste: i see at least two problems :p
[21:40] <saste> i mean with your fate test
[21:40] <ubitux> ah yes it's a problem
[21:40] <ubitux> the bitexact flags for sws are not honored
[21:40] <ubitux> and so the input differs
[21:40] <ubitux> (since select is converting to rgb when using scene detection)
[21:41] <saste> ok i'll commit them soon
[21:41] <saste> i just want to test them another time
[21:41] <ubitux> ah you have a fix for that? i wasn't inventing thing?
[21:50] <ubitux> well i guess i'll wait for your patches then
[21:55] <saste> ubitux: 6fb2fd895e858ab93f46e656a322778ee181c307
[21:55] <saste> wow, longest commit message in FFmpeg history
[21:55] <ubitux> :)
[21:56] <saste> btw fate is failing on your last test
[21:56] <ubitux> you mean the scene detect?
[21:56] <saste> yes
[21:56] <ubitux> yes that's what we were talking about
[21:57] <saste> amovie=/amrwb/seed-12k65.awb,silencedetect=d=.1: No such file or directory
[21:57] <ubitux> huh?
[21:58] <saste> i believe it is because I'm not using SAMPLES
[21:58] <ubitux> ah i use --samples at configure time
[21:58] <ubitux> mmh i guess i'll need to add them to some kind of samples dep
[22:00] <ubitux> anyway, for the scene detect, there are multiple things to do
[22:00] <ubitux> first make sure ffprobe is transmitting the sws flags, and then add the bitexact flags to the FILTER_METADATA_COMMAND
[22:00] <ubitux> but you told me you had some pending patches for this
[22:01] <ubitux> about the SAMPLES issue, i don't know how i'm supposed to fix that properly..
[22:04] <ubitux> saste: the sws flags thing looks a bit messy and inconsistent
[22:05] <ubitux> it's only set in ffplay and ffmpeg, but not in ffprobe; and maybe it should be set in the lavfi device?
[22:12] <saste> ubitux, yes
[22:12] <michaelni> btw if you cant fix this, the not working test should be disabled
[22:13] <Daemon404> didable broken things to make stuff appear to pass isnt a very good idea...
[22:13] <Daemon404> disabling*
[22:14] <ubitux> michaelni: i'll fix as soon as saste push what he was talking about
[22:14] <ubitux> it can be fixed without much trouble i believe
[22:15] <ubitux> we just need to pass the bitexact flags properly to sws :p
[22:20] <saste> ubitux: can you fix the samples thing in the meanwhile
[22:20] <ubitux> yes, trying to figure out how to :p
[22:23] <michaelni> Daemon404, i think we speak about different things
[22:26] <michaelni> if code is commited that breaks something it should be fixed or reverted, ubitux is working on that. But if noone had time, the code that broke something should be reverted
[22:26] <nevcairiel> isnt it new code that just doesnt work everywhere?
[22:26] <nevcairiel> its not a regression, just open TODOs, basically =)
[22:27] <ubitux> michaelni: what's my deadline for fixing fate?
[22:27] <michaelni> its causing a regression for "make fate"
[22:27] <nevcairiel> a regression is existing tests breaking, not new tests not passing, imho
[22:28] <nevcairiel> anyhow fixes seem to be on their way
[22:28] <michaelni> ubitux, no idea, it doesnt matter, just if you give up / work on something else please disable it first
[22:28] <ubitux> ok
[22:28] <ubitux> i can go to sleep tonight without fixing it? :)
[22:29] <michaelni> if you continue dreaming about how to solve it :)
[22:29] <michaelni> nevcairiel, its a regression in "make fate"
[22:29] <michaelni> it worked before
[22:30] <ubitux> of course ;)
[22:31] <michaelni> "make fate" is used to test if code is ok, for example before push, if it fails developers work is hindered
[22:33] <michaelni> thats IMHO not so much about "hiding a problem" more "not having all alarm bells ringing for days while its being worked on"
[22:35] <michaelni> and these alarm bells could hide other new problems or regressions
[22:37] <Daemon404> michaelni, i thought you were talkign about a fate test
[22:37] <Daemon404> by bad
[22:37] <nevcairiel> considering its only broken on one os/toolchain combination, i think we'll survive a day until its fixed
[22:39] <nevcairiel> i think its not even the change itself, just the way the test is setup with the movie source, yes?
[22:40] <michaelni> i think its broken on all non x86 platforms
[22:41] <nevcairiel> fate looks rather normal
[22:42] <nevcairiel> except the fail on native windows builds which is new
[22:45] <michaelni> many didnt test the problematic revission yet i think
[22:45] <cone-618> ffmpeg.git 03Stefano Sabatini 072969abd908f1: lavfi/graphparser: fix parsing error in case of NULL sws_opts addition * 03http://tinyurl.com/8tk9gkf03
[22:45] <cone-618> ffmpeg.git 03Stefano Sabatini 078f37a1e8dc93: lavfi/avfiltergraph: avoid to print "(null)" in the scale args * 03http://tinyurl.com/8otm4q403
[22:46] <nevcairiel> didnt it just uncover a parsing problem when a string parameter includes a :, which happens in windows path strings?
[22:46] <ubitux> saste: thx!
[22:46] <nevcairiel> are we talking about the same problem? :P
[22:46] <saste> nevcairiel, you need to escape it
[22:46] <saste> the problem is doing it in an automated fashion
[22:47] <nevcairiel> thats rather unintuitive though :p
[22:47] <nevcairiel> how do you escape in that place even? normal backslash?
[22:56] <ubitux> ok almost fixed it
[23:00] <ubitux> ok, fixed
[23:02] <ubitux> i wonder how safe this is though
[23:03] <ubitux> saste: http://b.pkh.me/0001-lavd-lavfi-honor-sws-flags.patch  i guess that needs to be made safer, right?
[23:04] <ubitux> (with some #if CONFIG_...)
[23:10] <michaelni> ubitux, you are accessing global variables that reside in ff* user tools from the libs
[23:11] <michaelni> that is definitly not safe
[23:12] <ubitux> mmh it's in the cmdutils, missed that
[23:14] <ubitux> i guess i'll have to do that in ffprobe
[23:15] <ubitux> but i don't have access to the filtergraph..
[23:16] <michaelni> quick solution is to add a AVOption to lavfi that allows setting the sws flags
[23:16] <michaelni> libavdevice/lavfi.c that is
[23:16] <ubitux> what about getting a new sws context in the lavfi device?
[23:17] <ubitux> mmh wait i'll need the link with the options
[23:19] <michaelni> another way is to make avfilter_graph_parse() somehow set the "global" flags from the graph string
[23:22] <ubitux> my brain is dead, i'll see that tomorrow
[23:22] <michaelni> ok no hurry, maybe ill look at it while you sleep
[23:24] <ubitux> you're a real guardian ;)
[23:24] <ubitux> 'night :)
[23:25] <ubitux> this is a test case btw: http://b.pkh.me/0002-fate-add-bitexact-flags-to-filter-metadata-test-comm.patch
[23:25] <ubitux> and i have another patch to add the scale filter dep to scene detect as well
[23:26] <ubitux> anyway, cya
[23:26] <michaelni> good night ubitux
[23:31] <Compn> i thought boxx codec was dead, whats it doing with 10bit ?
[23:41] <saste> since when "being experimental" is an error?
[23:41] <saste> what about AVERROR_PINK?
[00:00] --- Mon Oct 22 2012


More information about the Ffmpeg-devel-irc mailing list