[Ffmpeg-devel-irc] ffmpeg-devel.log.20121022

burek burek021 at gmail.com
Tue Oct 23 02:05:02 CEST 2012


[00:12] <cone-618> ffmpeg.git 03Michael Niedermayer 07248b1ff26b7c: Fix various uses of av_log_missing_feature() * 03http://tinyurl.com/9ttzpsb03
[00:12] <cone-618> ffmpeg.git 03Michael Niedermayer 0739747d87d090: avfilter_graph_parse: add support for parsing sws_flags * 03http://tinyurl.com/8eevlbz03
[00:12] <cone-618> ffmpeg.git 03Michael Niedermayer 072d11ee4bfc7c: fate: fix fate-filter-metadata-scenedetect * 03http://tinyurl.com/9rorhfu03
[07:35] <ubitux> michaelni: thx for the fix!
[09:08] <ubitux> michaelni: "fix used variable" // "unused", right?
[10:07] <cone-571> ffmpeg.git 03Paul B Mahol 07cb0add3ce961: lavf/flacenc: disallow creation of invalid files with -c copy * 03http://tinyurl.com/9rwjvbk03
[10:42] <ubitux> saste: you forgot to remove gcd in cdea54b4c
[10:42] <ubitux> btw, looking forward tools/ffescape :))
[10:43] <ubitux> oh and i sent a pending patch on the ml for the SAMPLES issue
[10:44] <cjhmdm> hello, can I find libfdk devs here?
[10:44] <ubitux> i don't think so
[10:47] <saste> wbs^^
[10:48] <ubitux> ah, wbs indeed :)
[10:50] <ubitux> btw, just curious, why working on an external encoder?
[10:51] <cone-571> ffmpeg.git 03Stefano Sabatini 07b6e36a424447: lavu/parseutils: remove unused gcd variable in av_parse_ratio() * 03http://tinyurl.com/8cnkneh03
[10:51] <cone-571> ffmpeg.git 03Stefano Sabatini 07ccd6def9b3dc: lavfi/aspect: extend syntax for the setdar and setsar filters * 03http://tinyurl.com/8nserd403
[10:51] <cone-571> ffmpeg.git 03Stefano Sabatini 076752aac6bca7: lavfi/aspect: add max option * 03http://tinyurl.com/962my2b03
[10:52] <ubitux> thx saste 
[10:52] <cjhmdm> ubitux: were you asking me?
[10:52] <ubitux> more wbs i believe :p
[10:53] <cjhmdm> dunno what that is, but ok heh
[10:53] <ubitux> a nickname on this channel :)
[10:57] <saste> ubitux: what is $(FATE_SAMPLES_FFPROBE)?
[10:59] <saste> tested the patch, and seems to work, but i'll leave review to someone more qualified than me
[10:59] <saste> as for ffescape, i can hardly work on it before this night
[11:00] <saste> given a filename, you need to perform this:
[11:00] <ubitux> saste: FATE_SAMPLES_FFPROBE is a different list than FATE_FFPROBE, the same way FATE_SAMPLES_FFMPEG is a different list than FATE_FFMPEG
[11:01] <saste> ubitux, ok bad naming but nothing we can do about
[11:01] <ubitux> and these FATE_SAMPLES_<tool> list are in the FATE_EXTERN list, which is only used when SAMPLES is set
[11:01] <ubitux> do you want me to add this in the comment description?
[11:02] <ubitux> <@saste> given a filename, you need to perform this: // this?
[11:05] <saste> shescape(ffescape(ffescape(filename, ":") + ":$other_options", ",;"))
[11:06] <saste> you need to escape ":" in the filename, then special graph chars in the resulting string, and finally shell escape
[11:06] <saste> we can improve this mess in two ways
[11:07] <saste> 1. by putting the filtergraph description in a file
[11:07] <saste> i proposed that sometimes ago and it was rejected with the argument that sh can do it
[11:07] <saste> (include a file in the command)
[11:07] <ubitux> for the fate failure i believe we will just need to ffescape ':' in $SRC
[11:08] <saste> 2. by creating an alternative syntax (more avisynth like)
[11:08] <ubitux> filtergraph in a description file ’ agree.
[11:08] <ubitux> -f lavfi script=myfiltergraph
[11:08] <saste> 2. is much work, so we can ignore it for the moment
[11:09] <saste> ubitux, yes that's a possibility
[11:09] <ubitux> -f lavfi evalscript=myfiltergraph even
[11:09] <saste> ubitux, looks, maybe it is *aready* implemented
[11:09] <ubitux> so we can add a script= later with a scripting thing like you just mentioned
[11:10] <saste> ummhno 
[11:10] <ubitux> haha
[11:10] <saste> but is easy to add
[11:10] <ubitux> no way it is @_@
[11:10] <ubitux> ah no, it's the "str"
[11:10] <ubitux> not a file
[11:10] <saste> check for example how I did it in sendcmd
[11:11] <saste> then I wouldn't be against -{a,v}f_file -filter_complex_file
[11:11] <saste> but it is not required here
[11:12] <saste> gottago, good luck ;-)
[11:12] <ubitux> heh
[11:13] <ubitux> will see later
[11:13] <ubitux> still, generating a file for this in fate looks overkill
[11:13] <ubitux> i'd better just ffescape the ':'
[12:35] <j-b> [mpeg2video @ 0177ab80] a vbv buffer size is needed, for encoding with a maximum bitrate
[12:35] <j-b> any idea?
[12:40] <Compn> -rc_max_vbv_use 1
[12:40] <Compn> j-b : according to this http://www.itbroadcastanddigitalcinema.com/ffmpeg_howto.html anyhow
[12:41] <j-b> weird, this worked before
[12:41] <michaelni> which commit broke it ?
[12:42] <j-b> no clue
[12:42] <j-b> like yesterday, users complaining
[12:45] <michaelni> j-b, how can the problem be reproduced ?
[12:46] <j-b> according to user: "C:\Program Files\VideoLAN\VLC\vlc.exe" -I dummy --dummy-quiet --ffmpeg-hurry-up --no-loop --no-repeat --ffmpeg-skiploopfilter 4 --audio-language=any --start-time 0 --sout "#transcode{venc=ffmpeg,vcodec=mp2v,vb=7000,acodec=a52,ab=384,samplerate=48000,deinterlace}:std{access=file,mux=ts,dst="test.mpg"}" --sout-transcode-width 480 --sout-transcode-height 320 --sout-ffmpeg-keyint=3 "C:\Media\sample.avi" "vlc://quit"
[12:46] <j-b> let it drop
[12:46] <j-b> I will find a fix
[13:21] <cone-571> ffmpeg.git 03Michael Niedermayer 076182e0a6f640: vf_aspect: unbreak avoption system * 03http://tinyurl.com/9escc9y03
[13:21] <saste> damn
[13:24] <Compn> j-b : vlc doesnt have an automated testing system like fate ?
[13:24] <j-b> no
[13:24] <saste> Failed to avformat_open_input ... <= what's this ffspeak?
[13:24] <Compn> to detect regressions like this
[13:31] <saste> j-b: you could call it "VATE"
[13:31] <saste> http://www.urbandictionary.com/define.php?term=vate
[13:33] <saste> in Italian/Latin it is a word (with a positive connotation, it is the attribution given to a high poet/priest)
[13:38] <ubitux> it's hard to make test suites for video players
[13:41] <av500> for jerkiness and tearing :)
[13:41] <av500> and audio microcuts
[13:44] <durandal_1707> ubitux: give job to several folks which will watch various clips all the time
[13:45] <ubitux> it won't be called "VATE" then, it will be "CHINA" (i let you find an acronym for the joke)
[14:07] <av500> durandal_1707: I watch clips all the time at work :)
[14:07] <durandal_1707> av500: talk with j-b
[14:08] <durandal_1707> av500: lucky you
[14:08] <cone-571> ffmpeg.git 03Michael Niedermayer 07c3778df2d4c0: ffmpeg: fix negative array index * 03http://tinyurl.com/8sa9zdv03
[14:33] <Compn> ubitux : whys that? it has command line interface, and i'm guessing some kind of md5output 
[14:33] <Compn> if you mean , just for sheer amounts of tests, that would be hard to test every option on every build
[14:34] <ubitux> Compn: you would be able to test decoders/demuxers
[14:34] <ubitux> not what actually makes the video player
[14:35] <ubitux> (the user experience, smooth playback, GUI interations etc)
[14:35] <Compn> smooth playback by timing the test
[14:38] <ubitux> i didn't say it was impossible, but it's generally way more complex than just diffing outputs
[14:44] <michaelni> Compn, can you add regression tests to ffplay :) ?
[14:44] <michaelni> i wanted these for a while already ...
[14:45] <ubitux> -vo file!
[14:49] <cone-571> ffmpeg.git 03Nathan Caldwell 07a4aa20fbdbe8: avcodec: prefer decoders without CODEC_CAP_EXPERIMENTAL * 03http://tinyurl.com/8n42naf03
[14:49] <cone-571> ffmpeg.git 03Nathan Caldwell 07a893655bdaa7: avutil: Add AVERROR_EXPERIMENTAL * 03http://tinyurl.com/8se4wgf03
[14:49] <cone-571> ffmpeg.git 03Nathan Caldwell 07c854102da773: avcodec: handle AVERROR_EXPERIMENTAL * 03http://tinyurl.com/8tnjhnu03
[14:50] <cone-571> ffmpeg.git 03Martin Storsjö 07e0d5ac6ae3a8: rtsp: Update a comment to the current filename scheme * 03http://tinyurl.com/96qyhga03
[14:50] <cone-571> ffmpeg.git 03Martin Storsjö 07c3e15f7b39aa: rtpdec: Don't pass a non-AVClass pointer as log context * 03http://tinyurl.com/9mlgx9803
[14:50] <cone-571> ffmpeg.git 03Michael Niedermayer 07e3a91c51f713: Merge commit 'c3e15f7b39aac2012f09ee4ca86d2bc674ffdbd4' * 03http://tinyurl.com/98d379403
[15:01] <cone-571> ffmpeg.git 03Justin Ruggles 0746a86c61940e: alacdec: set bits_per_raw_sample * 03http://tinyurl.com/9zy555a03
[15:01] <cone-571> ffmpeg.git 03Anton Khirnov 07d6f4fe68c8e8: lavc: extend frame_size doxy. * 03http://tinyurl.com/9aolmt203
[15:01] <cone-571> ffmpeg.git 03Anton Khirnov 076173a8fe5d2c: riff: remove a write-only variable * 03http://tinyurl.com/8eovuy903
[15:01] <cone-571> ffmpeg.git 03Michael Niedermayer 07f9cf14c8da2a: Merge remote-tracking branch 'qatar/master' * 03http://tinyurl.com/9mfuo6b03
[15:04] <saste> someone can recommend an image formats which allows to read the values in a human readable way?
[15:04] <saste> sng is a possibility, so i was contemplating the possibility to add native support to ffmpeg
[15:04] <saste> XPM may be another alternative
[15:05] <ubitux> ppm ?
[15:05] <saste> ffmpeg -i tests/lena.pnm lena.ppm
[15:06] <saste> vim lena.ppm... not very much
[15:06] <saste> IIRC PPM has an ASCII option which we don't support
[15:06] <ubitux> https://en.wikipedia.org/wiki/Netpbm_format
[15:06] <ubitux> this is what i call "ppm"
[15:06] <av500> yes
[15:07] <ubitux> https://en.wikipedia.org/wiki/Netpbm_format#PPM_example
[15:07] <saste> we don't support the ASCII format
[15:07] <ubitux> saste: nit++: 's' is before 'v'
[15:07] <ubitux> (ffescape, ffeval -- gitignore)
[15:08] <saste> ubitux, that list is not sorted anyway
[15:08] <saste> locally fixed
[15:08] <ubitux> and yes xpm are indeed another one
[15:09] <saste> so why esr implemented sng?
[15:09] <ubitux> (vim even renders directly xpm :p)
[15:17] <saste> replying to myself: SNG supports palette data, so it's ideal for storing/viewing/editing paletted images (like PNGs)
[15:54] <cbsrobot> durandal_1707: what happend to http://ffmpeg.org/pipermail/ffmpeg-devel/2012-June/125668.html ?
[15:55] <durandal_1707> rot, because nobody REing prores alpha part
[15:56] <durandal_1707> it can be added, but nothing will use it, because there is no other codecs that use this format, and anyway it is not really imprtant that much because there is yuva444p16
[15:56] <durandal_1707> and switch can be done at any time once prorse alpha code comes
[16:40] <merbanan> durandal_1707: dnxhd also has an alpha mode that isn't re'd
[16:42] <durandal_1707> isn't that raw?
[16:43] <durandal_1707> but that is 422
[16:43] <merbanan> it is vc3- based
[18:12] <cone-571> ffmpeg.git 03Michael Niedermayer 0722793d7bb36e: ffmpeg/lavc: move experimental warnings to libavcodec. * 03http://tinyurl.com/8uw866r03
[19:44] <cone-571> ffmpeg.git 03Matthieu Bouron 07d8173f264f73: dv: fix a check on dv_extract_timecode return value * 03http://tinyurl.com/8jldf9x03
[19:44] <cone-571> ffmpeg.git 03Matthieu Bouron 07c68a8a1340e7: lavf/aiffdec: don't stop parsing after SSND chunk * 03http://tinyurl.com/8wgqwhz03
[20:46] <durandal_1707> what to use to set jpeg encoder quality/compression?
[20:51] <michaelni> -qscale X
[20:53] <durandal_1707> but from lavc user point of view
[22:14] <ubitux> anyone to have a look to http://ffmpeg.org/pipermail/ffmpeg-devel/2012-October/132854.html ?
[22:15] <ubitux> i added the following description:
[22:15] <ubitux>     FATE_SAMPLES_FFPROBE is a different list than FATE_FFPROBE, the same way
[22:15] <ubitux>     FATE_SAMPLES_FFMPEG is a different list than FATE_FFMPEG; these
[22:15] <ubitux>     FATE_SAMPLES_<tool> lists are in the FATE_EXTERN list, which is only
[22:15] <ubitux>     used when SAMPLES is set.
[22:21] <ubitux> mmh i'm wondering
[22:22] <ubitux> what's wrong actually with the msvc fate instance?
[22:22] <ubitux> the configuration is --samples=/d/Dev/ffmpeg/fate/samples
[22:23] <ubitux> is the ':' also missing on fate display?
[22:23] <ubitux> Daemon404: how is your --samples configured?
[22:24] <Daemon404> my what now?
[22:24] <Daemon404> context?
[22:25] <ubitux> Daemon404: the msvc fate instances are yours, right?
[22:25] <ubitux> i'd like to know you set the samples directory
[22:25] <Daemon404> theyre nev's
[22:25] <ubitux> and if the path is indeed "/d/Dev/ffmpeg/fate/samples"
[22:25] <ubitux> oh, ok, sorry
[22:25] <Daemon404> i have one msvc instance of a shared build, and its for libav
[22:25] <Daemon404> since ffmpeg's doesnt work yet
[22:25] <Daemon404> iirc
[22:26] <ubitux> nevcairiel: ping :)
[22:26] <ubitux> Daemon404: ok :(
[22:26] <nevcairiel> wut
[22:26] <ubitux> nevcairiel: what's your samples path in the msvc instances?
[22:26] <Daemon404> im working on some exotic setups atm...
[22:26] <Daemon404> nerbsd/armv5
[22:26] <Daemon404> netbsd*
[22:26] <nevcairiel> technically the path is D:\Dev\ffmpeg\... however because its running in a msys bash shell, it gets the msys-style path
[22:26] <llogan> nerdsb
[22:26] <nevcairiel> msys should translate that to the real path before ffmpeg ever sees it
[22:27] <nevcairiel> probably d:/dev/...
[22:27] <Daemon404> yes
[22:27] <Daemon404> is this that : problem again?
[22:27] <ubitux> yes likely
[22:28] <ubitux> right now ffmpeg sees '/d/Dev/ffmpeg/fate/samples/svq3/Vertical400kbit.sorenson3.mov'
[22:28] <nevcairiel> i doubt that
[22:29] <ubitux> well: http://fate.ffmpeg.org/report.cgi?time=20121022182926&slot=x86_64-msvc10-windows-native
[22:29] <Daemon404> expecting users to escape : in path names is dumb.
[22:29] <Daemon404> just sayin.
[22:29] <nevcairiel> the script is called with that
[22:29] <nevcairiel> but its translated before ffmpeg ever gets it
[22:29] <Daemon404> yup
[22:29] <nevcairiel> or should be at least
[22:29] <nevcairiel> msys is sometimes funny
[22:30] <cone-571> ffmpeg.git 03Michael Niedermayer 07250fe6eeb4a1: cmdutils: apply option to codec and format contexts if possible.
[22:30] <ubitux> this is a particular case :p
[22:30] <ubitux> i wonder if i shouldn't just define another command in place of "run"
[22:31] <ubitux> ah well that's stupid
[22:31] <ubitux> it wouldn't help either
[22:31] Action: Daemon404 notes a 15 gb image takes forever to dd to an sdhc card
[22:32] <nevcairiel> hm how did you trick msys into not translating the path?
[22:32] <nevcairiel> it did translate it when i looked at the error yesterday
[22:32] <ubitux> i hid it in a string
[22:34] <nevcairiel> i can probably specify a windows-style path in the fate config
[22:35] <Daemon404> [16:32] < nevcairiel> hm how did you trick msys into not translating the path?
[22:35] <Daemon404> this is the wrong "solution"
[22:36] <ubitux> what solution do you see within a filtergraph Daemon404?
[22:36] <Daemon404> ubitux, fix the dang parser
[22:36] <Daemon404> so it handles VALID patsh correctly
[22:36] <ubitux> well, it's just opening the passed path
[22:36] <nevcairiel> in the recent error messages, ffmpeg is actually getting a msys path
[22:36] <ubitux> which is '/d/Dev/ffmpeg/fate/samples/svq3/Vertical400kbit.sorenson3.mov'
[22:37] <nevcairiel> because msys failed to identify the path in the middle of a argument
[22:37] <Daemon404> ah
[22:37] <Daemon404> yes
[22:37] <Daemon404> msys wont tl that
[22:37] <Daemon404> it has specific rules about what it will tl
[22:37] <nevcairiel> so i just change it to windows path with wrong-way-slashes
[22:37] <nevcairiel> work both for win32 and msys apps
[22:38] <ubitux> the problem will then be the ':'
[22:38] <Daemon404> if it cant handle the :
[22:38] <ubitux> (i guess)
[22:38] <Daemon404> thats a bug that should be fixed
[22:38] <Daemon404> period
[22:38] <Daemon404> or rather
[22:38] <Daemon404> colon.
[22:38] <ubitux> how?
[22:38] <Daemon404> ubitux, why cant it handle : in the first place
[22:38] <nevcairiel> because its the key-value pair seperator in filtergraph arguments
[22:38] <ubitux> :)
[22:39] <Daemon404> well you choose that delim poorly.
[22:39] <ubitux> right, but we can't change it now
[22:39] <nevcairiel> i think the whole idea of the movie source is a poor one
[22:39] <Daemon404> me too
[22:39] <nevcairiel> but thats just me
[22:39] <ubitux> again, do you see another solution?
[22:39] <Daemon404> if it was sane
[22:40] <ubitux> having sources in the filtergraph is nice IMO
[22:40] <Daemon404> hwy?
[22:40] <Daemon404> i dont see ANY benefit
[22:40] <Daemon404> from doing -i
[22:40] <ubitux> you can describe completely your filtergraph with a string, which simplifies the API a lot
[22:40] <ubitux> not all tools want to reproduce the filter complex thing from ffmpeg
[22:41] <ubitux> and that's indeed not done in ffprobe, and ffplay
[22:41] <ubitux> as the name suggest, it's "complex"
[22:41] <Daemon404> yes, one horribly ugly string
[22:41] <Daemon404> :|
[22:41] <Daemon404> i dont see a benefit to having the src specified in it
[22:41] <ubitux> i just told you... :/
[22:41] <Daemon404> [16:40] <@ubitux> not all tools want to reproduce the filter complex thing from ffmpeg
[22:42] <Daemon404> thats called design failure
[22:42] <Daemon404> but not making that an api
[22:42] <ubitux> well it really is simpler
[22:42] <ubitux> you only need to send a string to your app
[22:42] <Daemon404> there is nothing simpel about lavfi's api.
[22:42] <ubitux> to describe a filtergraph
[22:42] <Daemon404> or even its input strings
[22:43] <Daemon404> its akin to reading lisp
[22:43] <Daemon404> for me
[22:43] <Daemon404> ;)
[22:43] <ubitux> i wonder if we shouldn't just add a CONFIG_SANE_ENV and put SANE_ENV in the deps of these tests
[22:45] <nevcairiel> the inability to be able to deal with windows path is something that should be fixed
[22:45] <ubitux> anyway, if the path is transformed to a "d:\foo\bar" syntax, it should be possible to handle it without much trouble when ffescape is in
[22:45] <ubitux> nevcairiel: it is able to deal with windows path
[22:45] <Daemon404> if it cant handle C:/depr.mov
[22:45] <Daemon404> no
[22:45] <Daemon404> no it cannot.
[22:45] <ubitux> it can, you just have to escape the : afaict
[22:45] <Daemon404> thats not handling it
[22:45] <Daemon404> at all
[22:45] <Daemon404> >_>
[22:45] <ubitux> please
[22:46] <Daemon404> im serious
[22:46] <Daemon404> that is insanely dumb
[22:46] <ubitux> we can't just change the whole syntax of filtergraph just because you don't want to add a '\'
[22:46] <Daemon404> or add some basic checks?
[22:46] <ubitux> or maybe that's actually a solution mmh
[22:46] <Daemon404> for e.g. /,.:/
[22:46] <Daemon404> or something
[22:46] <ubitux> like setting a global separator
[22:46] <Daemon404> yes
[22:46] <Daemon404> like sed :P
[22:47] <Daemon404> s#a#b#
[22:47] <Daemon404> or s/a/b/
[22:50] Action: ubitux wonders about what other character to use
[22:50] <ubitux> mmh actually
[22:51] <ubitux> would "movie='d:\foo\bar':xx=..." not working?
[22:51] <ubitux> it's using av_get_token so it might work
[22:52] <michaelni> hmm
[22:52] <michaelni> why is this failing anyway ?
[22:52] <nevcairiel> damnit, there is some bash magic in the fate test that checks if the samples path starts with a slash
[22:52] <nevcairiel> unportable shit ftl
[22:52] <michaelni> there is surely no "d" option in the filter
[22:52] <michaelni> so d: could easily be handled
[22:53] <ubitux> michaelni: right now because it's not passing "d:\foo\bar" but the other path
[22:53] <ubitux> and then it will fail because of ':' escaping which IMO can be avoiding by just adding '' around $(SRC)
[22:54] <michaelni> that should work indeed
[22:55] <michaelni> something quite different, i think filter graph descriptions in genera need more whitespace and newlines, that would make them more readable
[22:56] <ubitux> https://github.com/ubitux/FFmpeg/compare/master...fate
[22:56] <Daemon404> inb4 add a \
[22:56] <Daemon404> and hit return
[22:56] <ubitux> spaces should be skipped yes
[22:56] <cone-571> ffmpeg.git 03Michael Niedermayer 070de41ead6f90: qt-faststart: check fseeko() return codes
[22:57] <ubitux> < nevcairiel> damnit, there is some bash magic in the fate test that checks if the samples path starts with a slash // huh?
[22:57] <nevcairiel> well there is
[22:58] <nevcairiel> it uses that to try to check if its a relative path
[22:58] <ubitux> ah, ok
[22:58] <nevcairiel> and something like d:/... looks relative to it :P
[22:58] <ubitux> haha
[22:59] <nevcairiel> arent there bash-builtins to do this stuff
[22:59] <nevcairiel> why the manual hackery
[22:59] <nevcairiel> oh well
[22:59] <nevcairiel> no fixing the path then
[22:59] <ubitux> ok so, how are we supposed to handle that '/d/...' thing?
[23:00] <nevcairiel> you're not, fix fate :p
[23:00] <ubitux> so much trolling just to realize that there is nothing wrong with lavfi :(
[23:01] <nevcairiel> it still failed on the : path before you added those sws flags
[23:01] <ubitux> what ':'?
[23:01] <nevcairiel> http://fate.ffmpeg.org/report.cgi?time=20121021161916&slot=x86_64-msvc10-windows-native
[23:01] <ubitux> if we add the '' quoting it will work just fine
[23:02] <ubitux> i can push my branch right now to fix that
[23:02] <ubitux> i'd just need a ok for 01f2386 and maybe the others
[23:02] <nevcairiel> i should figure out a more portable way to figure out if a path is abolute in bash
[23:02] <nevcairiel> i'm sure there is a real function for that
[23:03] <nevcairiel> all the quick solutions on google check for damn slashes
[23:04] <ubitux> it's simple
[23:04] <ubitux> we should just check for damn ':'
[23:06] <nevcairiel> i could just leave out the drive from the path, its all on the same one anyway, so /Dev/... would work
[23:06] <nevcairiel> but thats not a universal solution
[23:06] <nevcairiel> and it probably wouldnt work with any msys tools
[23:09] <ubitux> wait.
[23:09] <ubitux> Error initializing filter 'movie' with args 'd:/Dev/ffmpeg/fate/samples/svq3/Vertical400kbit.sorenson3.mov'
[23:09] <ubitux> so this is what you actually pass to the filter
[23:09] <nevcairiel> if msys manages to translate the path
[23:09] <nevcairiel> or the native path is passed
[23:10] <ubitux> ah and the sws_flags thing broke that detection?
[23:10] <nevcairiel> yeah, it works if its the first argument
[23:10] <nevcairiel> but the sws flags appeared before it
[23:11] <ubitux> how does this detection work?
[23:11] <nevcairiel> a lot of magic, i guess
[23:11] <ubitux> i'm not even sure it will detect it if we enclose it within ''
[23:11] <nevcairiel> they might expect quoting
[23:12] <ubitux> well we can see it with the silencedetect one
[23:12] <ubitux> since it hasn't sws flag thing
[23:13] <ubitux> michaelni: do you mind if i push my branch?
[23:13] <michaelni> ubitux, if it fixes things sure push it
[23:13] <ubitux> ok
[23:14] <cone-571> ffmpeg.git 03Clément BSsch 077be9c0c10fc5: fate: fix SAMPLES dependency for ffprobe.
[23:14] <cone-571> ffmpeg.git 03Clément BSsch 07ae69c683e58c: fate: add scale filter to the scene detect dependencies.
[23:14] <cone-571> ffmpeg.git 03Clément BSsch 072649b78384a7: fate: improve metadata filter deps readability.
[23:14] <cone-571> ffmpeg.git 03Clément BSsch 077c2d5eec29e0: fate: quote file paths in movie/amovie filtergraphs.
[23:14] <ubitux> let's see now.
[23:15] <cbsrobot> so now cone wont give an url anymore ?
[23:15] <ubitux> haha :)
[23:15] <ubitux> better than a tiny url!
[23:16] Action: cbsrobot stares at gnafu 
[23:17] <ubitux> nevcairiel: i see you're doing some experiments with one msvc instance :D
[23:18] <nevcairiel> i tried to change the path, but i found that stupid absolute-path check :p
[23:19] <ubitux> i just pushed the '' btw, if you want to give it a try
[23:23] <ubitux> nevcairiel: i'd like to have a look to the code doing the path translation; what part of what project is responsible for this?
[23:23] <nevcairiel> in msys? i have no clue
[23:23] <ubitux> msys from mingw?
[23:24] <nevcairiel> yeah its somewhere in one of the msys runtime libs that get linked into the executables
[23:24] <nevcairiel> no idea where
[23:24] <ubitux> ok, thanks
[23:25] <ubitux> git clone git://mingw.git.sourceforge.net/gitroot/mingw/mingw
[23:25] <ubitux> Cloning into 'mingw'...
[23:25] <ubitux> warning: You appear to have cloned an empty repository.
[23:25] <ubitux> :(
[23:25] <Daemon404> mingw uses svn
[23:25] <Daemon404> why are you cloning mingw
[23:25] <Daemon404> >____>
[23:25] <Daemon404> nothing good can come of this
[23:26] <ubitux> http://sourceforge.net/projects/mingw/develop
[23:26] <ubitux> i see cvs & git :p
[23:26] <nevcairiel> you probably want http://mingw.git.sourceforge.net/git/gitweb.cgi?p=mingw/msys-runtime;a=summary
[23:26] <ubitux> thx
[23:26] <Daemon404> again
[23:26] <Daemon404> why
[23:26] <Daemon404> <_<
[23:26] <nevcairiel> he wants to find the path translation code :p
[23:26] <Daemon404> ..
[23:26] <Daemon404> theres an entire damn page
[23:26] <Daemon404> dedicated to it
[23:26] <Daemon404> on their site
[23:27] <Daemon404> http://www.mingw.org/wiki/Posix_path_conversion
[23:27] <ubitux> oh great
[23:27] <ubitux> thx
[23:27] <ubitux> http://mingw.cvs.sourceforge.net/viewvc/mingw/msys/rt/src/winsup/cygwin/path.cc?view=markup *_*
[23:27] <Daemon404> viewvc is horrible
[23:27] <Daemon404> end of story
[23:28] <Daemon404> :3
[23:29] <nevcairiel> my fate finished building, nothing changed
[23:31] <nevcairiel> anyway, time for sleep
[23:31] <ubitux> mmh this looks weird
[23:31] <ubitux> night :)
[23:33] <ubitux> Missing key or no key/value separator found after key '/Dev/ffmpeg/fate/samples/amrwb/seed-12k65.awb'
[23:33] <ubitux> this is pretty weird
[23:37] <ubitux> as if the msys escaping was just breaking everything
[23:43] <ubitux> mmh something will need to be changed in src movie maybe
[23:45] <ubitux> ok get it
[23:59] <ubitux> "movie=f='c:\foo\bar'" doesn't actually receive "f='c:\foo\bar'" as arg, but "f=c:\foo\bar"
[23:59] <ubitux> that's what causing the issue
[23:59] <durandal_1707> heh sws cant handle 4x4 images
[23:59] <ubitux> hey saste, perfect time! :)
[23:59] <saste> ubitux, i want to tell that the ' should work
[23:59] <ubitux> i commited
[23:59] <ubitux> but it doesn't
[23:59] <ubitux> because of what i just said
[00:00] --- Tue Oct 23 2012


More information about the Ffmpeg-devel-irc mailing list