[Ffmpeg-devel-irc] ffmpeg-devel.log.20120917

burek burek021 at gmail.com
Tue Sep 18 02:05:03 CEST 2012


[00:13] <michaelni> Daemon404, http://fate.ffmpeg.org/history.cgi?slot=x86_64-debian-mingw32-gcc-4.6
[00:13] <michaelni> bprint fails
[00:13] <michaelni> i suspect it was one of the msvc commits
[00:13] <Daemon404> interesting
[00:14] <Daemon404> it seems mingw changed their vsnprintf implemenatio nto be conforming at some point
[00:14] <Daemon404> (before the version taht uses)
[00:15] <Daemon404> michaelni, solution is to enable our vsnprintf impl on mingw too, or to resotre the old hack with the ifdef chanegd to mingw
[00:16] <Daemon404> i can test it in a bit
[00:16] <Daemon404> later tonight
[00:16] <Daemon404> and send a patch
[00:16] <michaelni> ok, thx
[00:17] <Daemon404> can you tell me what version of debian and mingw32 that is?
[00:18] <michaelni> which package should i check ?
[00:19] <Daemon404> dpkg --get-selections | grep -i mingw
[00:19] <Daemon404> pastebin it
[00:19] <michaelni> that doesnt contain version numbers (if that was what you want)
[00:20] <Daemon404> nah
[00:20] <Daemon404> i just cant remember which packages exist...
[00:20] <Daemon404> lol
[00:21] <michaelni> http://pastebin.com/tGFTmgGa
[00:21] <Daemon404> mingw-w64 and gcc-mingw-w64
[00:23] <michaelni> http://pastebin.com/kWrVk8mp
[00:23] <Daemon404> thanks
[00:23] <Daemon404> ill check it tonight
[00:33] <Daemon404> michaelni, ok reproduced
[00:33] <Daemon404> but only with debian/ubuntu's toolchain
[00:33] Action: Daemon404 hopes it isnt some dumb debian patch
[00:45] <CIA-54> ffmpeg: 03Marton Balint 07master * rd8f8e911bd 10ffmpeg/ffplay.c: 
[00:45] <CIA-54> ffmpeg: ffplay: dont wait 100ms if data is not yet ready
[00:45] <CIA-54> ffmpeg: Also signal the wait if audio buffer is empty. This fixes jerky audio with
[00:45] <CIA-54> ffmpeg: realtime sources.
[00:45] <CIA-54> ffmpeg: Signed-off-by: Marton Balint <cus at passwd.hu>
[00:45] <CIA-54> ffmpeg: 03Michael Niedermayer 07master * r83438a0db3 10ffmpeg/: 
[00:45] <CIA-54> ffmpeg: Merge remote-tracking branch 'cus/stable'
[00:45] <CIA-54> ffmpeg: * cus/stable:
[00:45] <CIA-54> ffmpeg:  ffplay: dont wait 100ms if data is not yet ready
[00:45] <CIA-54> ffmpeg: Merged-by: Michael Niedermayer <michaelni at gmx.at>
[00:54] <Daemon404> https://github.com/dwbuiten/MinGW-w64/commit/7e3370400278cdc553e42222f3ff69ab40de64b9
[00:54] <Daemon404> possibly related
[00:55] <Daemon404> looks like it is
[03:54] <Ctut_> hello.
[09:49] <mateo`> Tjoppen: can you review my 2 patches on mxfenc (adding support for 23.976 and 24 fps) ?
[10:12] <Tjoppen> sure. got some other things to do first thoguh
[10:30] <ubitux> anyone working on f4v?
[10:30] <ubitux> wbs maybe?
[10:31] <wbs> f4v should be pretty normal mp4 iirc? but for the fragmented stuff for HDS, I'm planning to look into it (for producing, not so much for demuxing/decoding)
[10:31] <ubitux> yes it's a kind of fragmented iso base
[10:31] <ubitux> with additional drm and stuff
[10:32] <ubitux> i'm asking because i'll likely work on this in the next days
[10:32] <ubitux> so i wanted to avoid duplicated efforts
[10:32] <ubitux> and for producing as well
[10:33] <wbs> right
[10:33] <wbs> I don't think I'll get to it within that time frame
[10:39] <ubitux> wbs: ok; please tell me if you see someone else or plan to work on this so we don't end up with duplicated work or anything :P
[10:39] <ubitux> it's likely i won't send anything in 2-3 days so... :p
[10:40] <wbs> ok :-)
[10:40] <ubitux> thanks :)
[12:48] <CIA-54> ffmpeg: 03Stefano Sabatini 07master * r85c93d90df 10ffmpeg/libavutil/ (parseutils.c version.h): 
[12:48] <CIA-54> ffmpeg: lavu/parseutils: fix av_small_strptime() whitespace directive parsing
[12:48] <CIA-54> ffmpeg: According to POSIX, strptime() should consume whitespaces in the date
[12:48] <CIA-54> ffmpeg: string everytime a whitespace conversion specification is found in the
[12:48] <CIA-54> ffmpeg: date format specification. Make av_small_strptime() conform with this
[12:48] <CIA-54> ffmpeg: behavior.
[12:49] <CIA-54> ffmpeg: In particular, should fix trac ticket #1739.
[12:49] <CIA-54> ffmpeg: 03Stefano Sabatini 07master * r1ad63ff14a 10ffmpeg/ (libavutil/parseutils.c tests/ref/fate/parseutils): fate: add av_small_strptime() test to fate-parseutils
[13:56] <ubitux> saste: i'd personally avoid setting the same value for the month and the minutes :p
[13:56] <ubitux> "m", "M, ...
[14:02] <saste> ubitux, where?
[14:04] <saste> 20:21:12 should work fine as well
[14:04] <saste> or 21:20:12
[14:04] <saste> or 12:20:21
[14:04] <saste> depending on when do you want the world to end ;-)
[14:06] <Compn> http://www.yebanishedprivateers.com/
[14:06] <Compn> pirate band puts album on piratebay
[14:07] <Compn> oops wrong channel haha
[14:07] Action: Compn hides his pirate shame
[14:07] Action: cbsrobot- wonders what channel compn is joined to
[14:08] <saste> thescene channel?
[14:08] <Compn> i just like pirate music
[14:08] <cbsrobot-> tpb channel ?
[14:08] <Compn> sea shanties and whatnot
[14:08] <cbsrobot-> pirat music sound like a serious thing
[14:08] <Compn> they dress up when they play songs even 
[14:11] <ubitux> saste: "2012-12-21 20:12:21", month=12, minutes=12
[14:11] <ubitux> "%M", "%m" etc
[14:11] <ubitux> but yeah
[14:12] <ubitux> end of the world is a good argument
[14:12] <ubitux> :))
[14:24] <saste> ubitux: I'm rewriting writers, in order to work with a generic concept of "section"
[14:24] <ubitux> saste: shouldn't you add some reg test before? :)
[14:24] <saste> ubitux, maybe I'll do
[14:25] <saste> we should add a file to fate before, with format/frames/streams metadata
[14:26] <saste> and A/V/subtitles
[14:41] <burek> I don't know who is "little mike frrom oakwood", but if anyone finds himself familiar in this email conversation, please respond to the guy, because I don't know what to tell him anymore :) http://pastie.org/private/4qlbtirwnlvmlnbya3wpq
[14:45] <Compn> burek : paste is 404
[14:45] <av500> works here
[14:45] <av500> update your internet
[14:46] <Compn> works now :P
[14:46] <Compn> site threw me an exception
[14:47] <Compn> burek : i can remove him if you have an email addy ?
[14:47] <Compn> or you can tell him to read the instructions on the bottom of each email...
[14:48] <Compn> reading is hard tho, so not a good idea
[14:48] <burek> oh yes his email is not shown in the paste, just a sec
[14:48] <burek> alexjestradais at gmail.com
[14:49] <Compn> oh actually i cant edit that list
[14:49] <Compn> ping michaelni :P
[15:04] <saste> ubitux, uhm it's ending to be almost a complete rewrite
[15:09] <michaelni> Compn, pong, how can i help ?
[15:11] <ubitux> saste: have fun
[15:12] <ubitux> saste: i'm wondering how you plan to keep compatibility with the old stuff
[15:12] <ubitux> especially the tags one
[15:12] <saste> ubitux: i'm not gonna break it
[15:12] <saste> that's the hard part
[15:12] <ubitux> :)
[15:13] <ubitux> yeah, that was the hard part already
[15:13] <ubitux> :D
[15:13] <saste> I have to convey the structure to the print_section_header/footer
[15:13] <saste> for example for format tags now I have tags/format/tags
[15:14] <saste> and I have to tell the default writer that it has to print the section header for format, but not for tags
[15:15] <saste> at the end of the pain the code would be much cleaner/more generic
[15:15] <saste> allowing some fun stuff like subsections selection
[15:17] <ubitux> ok
[15:17] <ubitux> less code? :)
[15:18] <saste> uhm not yet sure about that
[15:18] <ubitux> :D
[15:18] <ubitux> ok :)
[15:18] <saste> writers should be simpler, I'm cutting down show_header/footer, show_chapter_footer/header
[15:18] <saste> they are particular cases of show_section_header/footer
[15:21] <CIA-54> ffmpeg: 03Luca Barbato 07master * r4ed5ac50d3 10ffmpeg/libavformat/file.c: file: return proper error on seek failures
[15:21] <CIA-54> ffmpeg: 03Michael Niedermayer 07master * r291b220377 10ffmpeg/: (log message trimmed)
[15:21] <CIA-54> ffmpeg: Merge remote-tracking branch 'qatar/master'
[15:21] <CIA-54> ffmpeg: * qatar/master:
[15:21] <CIA-54> ffmpeg:  avconv: flush filtered frames before reconfiguring filters
[15:21] <CIA-54> ffmpeg:  mov: stsd entries must be at least 16 byte
[15:21] <CIA-54> ffmpeg:  mov: detect EOF in mov_read_dref()
[15:21] <CIA-54> ffmpeg:  file: return proper error on seek failures
[15:22] <CIA-54> ffmpeg: 03Michael Niedermayer 07master * ra5ea623b36 10ffmpeg/libavformat/mov.c: 
[15:22] <CIA-54> ffmpeg: mov: stsd entries must be at least 16 byte
[15:22] <CIA-54> ffmpeg: Fix near infinite loop in stsd parsing.
[15:22] <CIA-54> ffmpeg: Bug found by: Diana Elena Muscalu
[15:22] <CIA-54> ffmpeg: The size is unsigned according the specification.
[15:22] <CIA-54> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:22] <CIA-54> ffmpeg: Signed-off-by: Luca Barbato <lu_zero at gentoo.org>
[15:22] <CIA-54> ffmpeg: 03Janne Grunau 07master * r2f34021d57 10ffmpeg/avconv.c: 
[15:22] <CIA-54> ffmpeg: avconv: flush filtered frames before reconfiguring filters
[15:22] <CIA-54> ffmpeg: This prevents lost frames after a resolution change.
[15:22] <CIA-54> ffmpeg: 03Michael Niedermayer 07master * r9db67bedf0 10ffmpeg/libavformat/mov.c: 
[15:22] <CIA-54> ffmpeg: Signed-off-by: Luca Barbato <lu_zero at gentoo.org>
[15:46] <ubitux> what's the difference between VP6[012AF]?
[15:47] <av500> yes
[15:48] <ubitux> thx av500 
[15:50] <av500> :)
[16:02] <merbanan> VP6[012] uses different features
[16:04] <JEEB> yeah, he got a nice explanation from kostya :)
[16:08] <ubitux> merbanan: yeah thx
[16:08] <ubitux> here is the log sample for those interested: http://pastie.org/private/lobme5ingewpa6ekoasqrg
[16:09] <merbanan> sorry got distracted
[16:09] <merbanan> A adds an alpha channel
[16:09] <merbanan> F is used in flash and codes the picture upside down, compared to the 012 ways
[16:13] <merbanan> VP6F is only found in flv
[16:13] <merbanan> or originates from there
[16:31] <Compn> michaelni : remove  alexjestradais at gmail.com from ffmpeg-devel-irc mailing list
[16:32] <Compn> ubitux : i cant believe someone is working on vp6a finally. even tho i only found one sample ever ;d
[16:32] <Compn> some people found some more samples anyhow
[16:53] <michaelni> Compn, done
[17:05] <ubitux> Compn: you should be able to find some in .f4v files
[17:23] <ubitux> is it me or the Makefile mov-like muxer are broken?
[17:23] <ubitux> like let's say, there is no ipod config for instance
[17:24] <ubitux> mov-like muxer deps*
[17:24] <Daemon404> is there any reason that ffplay works fine, but transcoding only encodes 1 frame
[17:24] <Daemon404> always wmv files
[17:24] <Daemon404> always with:
[17:24] <Daemon404> comment         : Movie/video screen capture
[17:24] <Daemon404> (wmv3)
[17:25] <Daemon404> avconv merely fails with
[17:25] <Daemon404> Application provided invalid, non monotonically increasing dts to muxer in stream 1: 4 >= 3
[17:25] <Daemon404> my old favorite error...
[17:26] <JEEBsv> :D
[17:26] <ubitux> oh i love this one as well
[17:26] <ubitux> second favorite after operation not permitted
[17:27] <Daemon404> ubitux, apparently all these files used to work with an ffmpeg from jan, before we updated
[17:27] <Daemon404> supposedly.
[17:28] <ubitux> regression from 0.11?
[17:28] <Daemon404> we dont use point releases
[17:28] <Daemon404> let me grab the hash for our old build
[17:29] <Daemon404> N-36919-g4259d1e
[17:29] <Daemon404> how do i match this with a git hash?
[17:29] <ubitux> 4259d1e ?
[17:29] <JEEBsv> yup, that
[17:29] <Daemon404> oh.
[17:29] <Daemon404> g.
[17:29] <Daemon404> right
[17:32] <Daemon404> yup
[17:32] <Daemon404> confirmed it works fine
[17:32] <Daemon404> regression.
[17:32] <JEEBsv> also is stream 1 audio or video?
[17:33] <Daemon404>     Stream #0:0: Audio: wmav2 (a[1][0][0] / 0x0161), 48000 Hz, 2 channels, s16, 192 kb/s
[17:33] <Daemon404>     Stream #0:1: Video: wmv3 (Main) (WMV3 / 0x33564D57), yuv420p, 640x480, 5500 kb/s, 29.97 tbr, 1k tbn, 1k t
[17:33] <JEEBsv> so audio
[17:33] <JEEBsv> try with -async -1 ?
[17:33] <Daemon404> >_>
[17:33] <Daemon404> k
[17:33] <Daemon404> but it happens when i do -an
[17:33] <Daemon404> so i doubt thats why
[17:34] <Daemon404> yeah no boat.
[17:34] <Daemon404> regression is also in libav
[18:18] <Daemon404> bisecting ffmpeg is fun
[18:18] <Daemon404> cause of the merges, i never know if ill have to run avconv or ffmpeg
[19:49] <saste> ubitux: 1 files changed, 200 insertions(+), 261 deletions(-)
[20:25] <ubitux> saste: \o/
[20:25] <ubitux> now make it works!
[20:26] <saste> i'm toiling over evil fate
[20:26] <saste> it's more convoluted every time i look at it
[20:26] <saste> also creating the test file looks more pain than expected
[20:27] <saste> keep having weird codec-specific/format specific errors
[20:27] <saste> multimedia is a mess
[20:27] <ubitux> :)
[21:01] <CIA-52> ffmpeg: 03Derek Buitenhuis 07master * r5c95a93411 10ffmpeg/libavcodec/dirac_arith.h: 
[21:01] <CIA-52> ffmpeg: dirac_arith: Fix unprotected inline asm
[21:01] <CIA-52> ffmpeg: Signed-off-by: Derek Buitenhuis <derek.buitenhuis at gmail.com>
[21:02] <Daemon404> ^ ffmpeg, now with 100% more x86_64 msvc support
[21:03] <nevcairiel> i enabled the fate for it last night without even checking if  it works :D
[21:03] <Daemon404> lawl
[21:04] <Daemon404> http://fate.ffmpeg.org/history.cgi?slot=x86_64-msvc10-windows-native
[21:04] <Daemon404> lol.
[21:04] <Daemon404> next run will work
[21:04] <Daemon404> whenever that is
[21:04] <Daemon404> your times seem to be sporatic
[21:05] <nevcairiel> fate starts every 2 hours
[21:05] <nevcairiel> but it has some more fate jobs
[21:05] <Daemon404> oic
[21:05] <nevcairiel> so depending on if libav also needs a rebuild, it may take a while until it gets to this one
[21:06] <Daemon404> i thought your fate was fast?
[21:06] <nevcairiel> takes like 8 minutes per instance still
[21:06] <Daemon404> ;p
[21:06] <Daemon404> 3 libav instaces, 2 ffmpeg?
[21:06] <nevcairiel> yes
[21:06] <nevcairiel> didnt bother doing a vs2012 here
[21:06] <Daemon404> i am using vs2012 to test ffmpeg
[21:06] <Daemon404> so it works
[21:07] <nevcairiel> i wish we could somehow figure out the borked debug info
[21:07] <Daemon404> oh
[21:07] <Daemon404> poke wbs
[21:07] <Daemon404> im betting its due to line differences
[21:07] <Daemon404> for structs and stuff
[21:08] <nevcairiel> but doesnt the normal C preprocessor also cause that stuff
[21:08] <Daemon404> no no
[21:08] <Daemon404> i mean the tmp___derp structs
[21:08] <Daemon404> the converter adss
[21:08] <nevcairiel> oh
[21:09] <nevcairiel> i should look at a converted file some day
[21:09] <Daemon404> ;p
[21:10] <Daemon404> i thought it was possible to keep all teh src in pdb
[21:10] <nevcairiel> i havent found the option to do that
[21:11] <nevcairiel> maybe the msdn doesnt mention it and using /Zi over Z7 does that already
[21:11] <Daemon404> idn
[21:18] <CIA-52> ffmpeg: 03Reimar Döffinger 07master * r07d2d06395 10ffmpeg/libavformat/utils.c: 
[21:18] <CIA-52> ffmpeg: Remove code that duplicates existing code a few lines down.
[21:18] <CIA-52> ffmpeg: This also fixes a potential crash since s->pb can be NULL.
[21:18] <CIA-52> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[21:35] <Daemon404> hey nevcairiel 
[21:35] <Daemon404> why do you build with opencore-amr?
[21:35] <Daemon404> is it better than ffmpeg's internal amr dec?
[21:35] <nevcairiel> internal amr sucks
[21:35] <Daemon404> o ok
[21:35] <nevcairiel> it doesnt support a mode called "dpx" or what it was
[21:35] <nevcairiel> users complained
[21:35] <nevcairiel> so meh
[21:35] <Daemon404> ll
[21:35] <Daemon404> lol
[21:36] Action: Daemon404 wonders wtf liborc is
[21:36] <ohsix> it's a code generator library
[21:36] <Daemon404> i wonder why ffmpeg has a check for it
[21:36] <ohsix> makes it easy to manage specializations too
[21:37] <ohsix> libvpx? there are a few things that actually use it that ffmpeg might link to
[21:37] <gnafu> libschro uses it./
[21:37] <ohsix> it's actually pretty neat
[21:37] <Daemon404> gnafu, ah
[21:38] <Daemon404> well im mostly only using decoders that we need libs for, or where teh internal impl is bad
[21:38] <Daemon404> like amr
[21:38] <Daemon404> and opus
[21:38] <gnafu> I don't think libvpx uses it.  I'm not sure what else does.
[21:38] <ohsix> opus :>
[21:38] <Daemon404> no need for libvpx
[21:38] <Daemon404> it has no use to me
[21:40] <CIA-52> ffmpeg: 03Lyubomir Marinov 07master * rfd9cd8cd5a 10ffmpeg/libavcodec/libx264.c: 
[21:40] <CIA-52> ffmpeg: libx264: connect AVCodecContext.rtp_payload_size to x264_params_t.i_slice_max_size
[21:40] <CIA-52> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:40] <CIA-52> ffmpeg: 03Lyubomir Marinov 07master * r63951aec2b 10ffmpeg/libavcodec/libx264.c: 
[21:40] <CIA-52> ffmpeg: libx264: feed x264_param_apply_profile from AVCodecContext.profile in absence of other information
[21:40] <CIA-52> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:41] <burek> have any of you heard of any project dealing with programming tools that use 3d goggles and gloves for virtual reality thingy?
[21:41] <ohsix> you might reference it in a sentence, that's some use
[21:48] <iive> burek: have you watched john cartmark presentation at quakecon 2012?
[21:48] <burek> iive no I didn't
[21:48] <Daemon404> you should
[21:48] <iive> it is inspiring. watch the one from the next day too. 
[21:52] <burek> any links?
[21:53] <iive> youtube :)
[21:54] <burek> oh, ok :)
[21:54] <burek> I'll take a look now, thanks :)
[21:54] <iive> the first one is 3 hours and half.
[21:57] <gnafu> "libvpx is dead to me.  Do you hear me?  Dead!"
[22:41] <ubitux> michaelni: did you have a chance to look at the oggdec patch for fixing the regression?
[22:45] <michaelni> ubitux, moment, ill look
[22:46] <Daemon404> orite i should also update that mingw hac--- i mean patch
[22:48] <ubitux> oh and i forgot to reset the state in ogg_reset
[22:52] <michaelni> ubitux, how does the patch avoid the save+add stream+restore problem ?
[22:52] <ubitux> just re-sent a patch
[22:52] <ubitux> which should fix the case you're talking about
[22:52] <ubitux> afaict
[22:54] <michaelni> how does it avoid it ?
[22:55] <ubitux> the state of the stream got_data is reset after in ogg_reset, so.. ?
[22:55] <ubitux> or maybe again i'm confused with something else?
[22:58] <CIA-52> ffmpeg: 03Andrey Utkin 07master * rb0fabed20f 10ffmpeg/libavcodec/avpacket.c: 
[22:58] <CIA-52> ffmpeg: When copying pkt side_data, don't forget size, type
[22:58] <CIA-52> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:59] <michaelni> i meant ogg_save() + new stream added + ogg_restore() -> nstreams != nb_streams
[22:59] <michaelni> how does reset stop this ?
[23:00] <ubitux> i'm sorry i don't get how this is different from the current situation
[23:03] <ubitux> does the patch makes possible such code path which wasn't possible before?
[23:04] <michaelni> ogg->headers is 1 after the loop in ogg_read_header() ogg_new_stream() isnt called for headers!=0 and ogg_save() isnt called in or before the loop
[23:06] <michaelni> from a quick look id say that a check on if a state was saved and refusing to add a new stream if thats the case might solve this
[23:08] <ubitux> mmh unfortunately, doing something like if (data_packets_seen(ogg)) replace() else if (!ogg->headers) new() else abort() won't solve the flac+theora.ogg sample
[23:08] <ubitux> if that's was what you meant
[23:09] <ubitux> (or even adding a if(ogg->headers)return 1 in data_packets_seen)
[23:09] <michaelni> i was more thinking of if(ogg->state)
[23:14] <ubitux> that code looks so fragile and complex... 
[00:00] --- Tue Sep 18 2012


More information about the Ffmpeg-devel-irc mailing list