[Ffmpeg-devel-irc] ffmpeg-devel.log.20130403

[burek]

[00:34] <cone-798> ffmpeg.git 03Clément BSsch 07master:e366aec03052: lavfi/edgedetect: add direct path.
[00:34] <cone-798> ffmpeg.git 03Clément BSsch 07master:e7692a9bac23: lavfi/edgedetect: reindent after previous commit.
[04:07] <cone-449> ffmpeg.git 03Michael Niedermayer 07master:aa7d9809c10c: h264pred: fix "warning: initialization discards const qualifier from pointer target type"
[04:07] <cone-449> ffmpeg.git 03Michael Niedermayer 07master:83e22172218a: flvdec: fix seeking in audio only files
[09:15] <durandal_1707> ubitux: what disortion?
[09:16] <ubitux> it doesn't sound like the original
[09:17] <ubitux> i guess that's the purpose of the filter
[09:18] <durandal_1707> ungureanuvladvic: so you want to do gsoc this year?
[09:21] <durandal_1707> ubitux: with s32 now i get only right channel, left is just some gradient in spectrogram (i cant hear anything from left channel)
[09:33] <durandal_1707> i will just push this filter and move on another one
[09:33] <ubitux> durandal_1707: works fine here
[09:33] <ubitux> do you have a sample?
[09:33] <durandal_1707> ubitux: it should not depent on sample...
[09:34] <ubitux> who knows :)
[09:34] <durandal_1707> ubitux: i tried it with several and nothing improved
[09:34] <ubitux> strange that i can't reproduce
[09:46] <durandal_1707> so your new filter just adds fades some pixels?
[09:48] <durandal_1707> if i compile with -O0 ffplay shows nan for pts and output is just silece - this is really getting nowhere
[09:49] <ubitux> my new filter?
[09:49] <ubitux> vignette?
[09:49] <durandal_1707> yes
[09:50] <ubitux> it just adds some shadows in the corners
[09:50] <ubitux> with various eval to make it fun
[09:50] <ubitux> i guess you saw the gone nutty version?
[09:50] <durandal_1707> i saw just video
[09:50] <ubitux> this one http://lucy.pkh.me/gonenutty-vintage-vignette.webm ?
[09:51] <durandal_1707> no, older one
[09:52] <durandal_1707> it's fun, but could other shapes be done?
[09:53] <ubitux> durandal_1707: right now it's just an ellipse (actually 2) 
[09:53] <ubitux> durandal_1707: btw, did you try if it was related to the direct path in your filter?
[09:54] <durandal_1707> nope, what should i do?
[09:55] <ubitux> you could check if you get the same result with read-only and writable frames
[09:57] <ubitux> (make sure there is no auto-inserted filter between aperms and aphaser)
[09:59] <ubitux> erm it seems the default query_formats doesn't accept any config
[10:00] <durandal_1707> i have own query_formats
[10:00] <ubitux> i mean for aperms
[10:01] <durandal_1707> ubitux: you are genius, with ro i get similar problems with other sample fmts
[10:01] <durandal_1707> with s16 for example, but not with flt
[10:01] <ubitux> great
[10:03] <durandal_1707> with s16 and random i have random silence frames
[10:05] <ubitux> mmh, strangely i'm not able to make aperms work with ffplay; but with ffmpeg it's fine
[10:05] <ubitux> for some reason an auto-inserted filter seems to stick between aperms and aphaser
[10:05] <ubitux> but well.
[10:06] <durandal_1707> well i could reproduce original problem with ffmpeg too
[10:07] <ubitux> as long as you found your problem& :)
[10:07] <ubitux> gtg, hf
[10:14] <durandal_1707> i found problem and it's so funny
[10:14] <durandal_1707> dunno why only s32 could show it, and only for me....
[10:15] <durandal_1707> some silly or smart compiler
[10:25] <durandal_1707> oh no, that was fix for another bug
[10:31] <durandal_1707> michaelni: trac is dead
[10:31] <durandal_1707> actually just timeline
[10:32] <durandal_1707> and any way to browse tickets
[10:38] <cone-610> ffmpeg.git 03Paul B Mahol 07master:bf65752848ec: aphaser filter
[10:42] <durandal_1707> Compn: you forget to update APE version in Changelog
[11:14] <Bor0> hi, with libav*, how can I know when the video is finished with playback? what's a good check for this? as I can see ffplay continues to count seconds even when video is completed
[11:15] <av500> using ffplay=
[11:15] <av500> using ffplay?
[11:15] <av500> or your own player?
[11:15] <Bor0> ffplay
[11:16] <av500> you'd need to add that to ffplay then
[11:16] <Bor0> I am building my own player but I don't know which indicator to use
[11:17] <av500> if (av_read_frame( fmt, &packet) < 0) {
[11:17] <av500> is a good indicator that your are done
[11:17] <Bor0> I see
[11:17] <Bor0> cool, thanks.
[11:18] <wm4> isn't there a secret switch that makes ffplay quit on EOF
[11:18] <nevcairiel> there is
[11:18] <nevcairiel> i forgot what it is, though
[11:18] <ubitux> -autoexit
[11:19] <nevcairiel> i see ffplay more of an example application than a player i would use for playback
[11:19] <Bor0> autoexit is a nice way to see what it does in background, thanks
[11:21] <ubitux> nevcairiel: except for muxed subtitles and the default vo, it's mostly complete :p
[11:21] <ubitux> at least for my usage
[11:22] <nevcairiel> its a freaking console player :P
[11:22] <ubitux> do you need to click all over a gui when watching a movie?
[11:23] <wm4> nevcairiel: it's trying to become mplayer
[11:23] <nevcairiel> i never liked mplayer for the same reason, i like some visible gui, and if its only a seekbar or some stream choosing menu
[11:24] Action: ubitux actually likes that
[11:24] <wm4> just wait until they add libavmenu for that ;)
[11:25] <nevcairiel> for my actual media viewing i'm using some 10ft GUI anyway, console windows are so bad to control with a remote control
[12:00] <durandal_1707> nevcairiel: what is so bad with "console windows" with remote control?
[12:02] <nevcairiel> no keyboard in my remote =P
[12:05] <durandal_1707> but there can be menus too
[12:06] <durandal_1707> you don't need gui like gtk/qt to have same or even better ui
[12:07] <wm4> xbmc?
[12:13] <durandal_1707> its so big that i do not have it installed
[12:35] <durandal_1707> what this earwax filter doing on its own? its example where "review" breaks code
[12:39] <durandal_1707> is cehoyous finnaly going to push opencl fix, compilation is currently broken
[12:40] <wm4> this opencl thing sure looked rushed
[12:40] <wm4> even the commit message looked wrong
[12:41] <durandal_1707> blame committer
[12:41] <wm4> "opencl wrapper based on comments on 20130401"
[12:41] <wm4> "deshake opencl based on comments on 20130402 3rd"
[12:42] <durandal_1707> thats how every commit should look like
[12:43] <durandal_1707> eg, you say in commit that only you participate in its review
[12:46] <durandal_1707> michaelni: earwax filter is plain wrong
[12:49] <michaelni> iam not the earwax author but wrong ?
[12:49] <durandal_1707> michaelni: first it's code is different from sox one
[12:51] <durandal_1707> than there is clipping issue, which i gonna fix shortly
[12:54] <cone-610> ffmpeg.git 03Carl Eugen Hoyos 07master:6e90d9e2066a: vf_deshake: Reorder constant arguments in an if() statement.
[12:56] <cone-610> ffmpeg.git 03Paul B Mahol 07master:0ab10f69efd3: lavfi/earwax: clip sample, this is what sox does
[12:59] <durandal_1707> what next filter should i port from sox?
[12:59] <wm4> none
[12:59] <wm4> write a sox wrapper
[13:00] <durandal_1707> wm4: wrong. sox and because of that wrapper to is slower
[13:01] <wm4> it's slower? awesome argument
[13:02] <durandal_1707> yes, write code right - its not same as NIH, which is just rewrite that could be just slower
[13:02] <wm4> so why don't you fix it upstream
[13:02] <durandal_1707> in sox?
[13:02] <wm4> yes
[13:02] <durandal_1707> more effort
[13:03] <wm4> hurr
[13:03] <durandal_1707> i would basically copy paste sws assembly code
[13:11] <cone-610> ffmpeg.git 03Paul B Mahol 07master:6fcb4ae1d81d: lavfi/afade * biquads: call av_frame_copy_props()
[13:42] <Compn> durandal_1707 : i was watching an old video captured from an old vhs tape, the audio was very scratchy. it would be nice if there was a filter to fix that :)
[13:42] <Compn> probably i deleted the file already, doh
[13:44] <durandal_1707> sox have some noise* effects
[13:44] <durandal_1707> they are 2 pass. first effect profile noise and second does actual filtering
[13:46] <durandal_1707> i dunno if that is possible with current lavfi
[14:07] <durandal_1707> hmm ffmpeg is mirrored on mplayerhq?
[14:08] <Compn> what do you mean mirrored ?
[14:08] <Compn> they are hosted on the same box
[14:08] <Compn> i meant, which part is mirrored?
[14:09] <durandal_1707> Compn: ahh, me feels stupid
[14:24] <durandal_1707> hmm, who is working on smacker regression(s)?
[14:29] <durandal_1707> that regression is 2 years old
[14:47] <Compn> kostya is smacker dev iirc :P
[14:54] <durandal_1707> the fix was not a fix
[14:54] <durandal_1707> like each bug fix introduces two new bugs
[14:56] <cone-610> ffmpeg.git 03Anton Khirnov 07master:f9f6402e9c9c: configure: prettify atomics handling.
[14:56] <cone-610> ffmpeg.git 03Michael Niedermayer 07master:6ce1d873552a: Merge commit 'f9f6402e9c9ce3642df981b48507df3a2d956f65'
[15:02] <cone-610> ffmpeg.git 03Paul B Mahol 07master:e3cc92a623a6: smacker: fix off by one error
[15:13] <cone-610> ffmpeg.git 03Martin Storsjö 07master:817dff578f13: configure: Check for the atomic.h functions used in the suncc atomics header
[15:13] <cone-610> ffmpeg.git 03Michael Niedermayer 07master:4d636947b4e2: Merge commit '817dff578f13ce97576c609ab141508b9dc782e9'
[15:34] <cone-610> ffmpeg.git 03Anton Khirnov 07master:6e3c13a559e9: avfiltergraph: check for sws opts being non-NULL before using them.
[15:34] <cone-610> ffmpeg.git 03Michael Niedermayer 07master:d3fc2db47e85: Merge remote-tracking branch 'qatar/master'
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/0.10:1aa50348d8bc: smacker: fix off by one error
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/0.11:708101c6f5c5: smacker: fix off by one error
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/0.7:8c5140df2b53: smacker: fix off by one error
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/0.8:537c17385350: smacker: fix off by one error
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/1.0:83d5871cf100: smacker: fix off by one error
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/1.1:5ee539f69d9a: smacker: fix off by one error
[15:34] <cone-610> ffmpeg.git 03Paul B Mahol 07release/1.2:d553a522b969: smacker: fix off by one error
[15:36] <durandal_1707> aww my clones
[15:40] <cone-610> ffmpeg.git 03Carl Eugen Hoyos 07master:c484b3af15f1: Fix make checkheaders without opencl headers.
[15:40] <cone-610> ffmpeg.git 03Michael Niedermayer 07master:c208576cef0a: Merge remote-tracking branch 'cehoyos/master'
[15:42] <durandal_1707> this is second time cahoyos "stole" almer patch
[17:42] <durandal_1707> michaelni: i think that listing pass/fail ratio of previous gsoc is relevant for being accepted
[17:45] <durandal_1707> and what was past ratio?
[17:50] <michaelni> durandal_1707, i dont know
[17:50] <michaelni> there where many past gsocs also
[17:52] <durandal_1707> np i know who to blame ;)
[17:58] <michaelni> durandal_1707, the proposal was posted on the ML before saste/logan submitted it
[17:58] <michaelni> i dont remember if it contained the pass/fail ratios
[17:59] <durandal_1707> i take a look and it did not
[17:59] <durandal_1707> they usually just put N/A
[18:01] <michaelni> its a bit late that you notice this 
[18:21] <cone-610> ffmpeg.git 03Carl Eugen Hoyos 07master:db2d3a908250: Skip padding in an id3 tag in aiff files.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:79d8cfacf078: lavfi: loop on request_frame if necessary.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:69d67fb622b6: lavfi: reindent after last commit.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:77fa554b6e08: lavfi/vf_tile: simplify request_frame loop.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:9566ba91abf4: lavfi/f_ebur128: remove request_frame hack.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:b570f24d7600: lavfi/af_asetnsamples: simplify request_frame loop.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:d38c173dfb4b: lavfi/af_atempo: simplify request_frame loop.
[18:36] <cone-610> ffmpeg.git 03Nicolas George 07master:464f94b206b0: lavc: do not init frame with guessed layout.
[18:36] <cone-610> ffmpeg.git 03Michael Niedermayer 07master:5c6f7ed667d6: Merge remote-tracking branch 'cigaes/master'
[18:41] <durandal_1707> micro bump for libavfilter is missing
[18:41] <mateo`_> Tjoppen: hello, any comments on the mxfdec: handle identification metadata patch ?
[19:07] <cone-610> ffmpeg.git 03Paul B Mahol 07master:9cb25440b38c: doc/filters: fix error in blend description
[19:11] Action: durandal_1707 finds out that people actually use blend filter
[19:12] <Compn> people actually use the filters ? yay!
[19:12] <ubitux> durandal_1707: bullshit, they are all useless filters, lavfi is shit, and no one uses it, it should be rewritten from scratch
[19:13] <Compn> lol
[19:13] <Compn> i thought we were supposed to rm the whole thing and use lav filters or vapoursynth or something 
[21:10] <ubitux> writing/porting the fieldmatch/decimate documentation is so much pain :(
[21:10] <cone-610> ffmpeg.git 03Stefano Sabatini 07master:36babfdc399f: doc/fate.texi: fix typo
[21:10] <ubitux> Myrsloik: any reason the vdecimate doesn't support cycleR?
[21:10] <Myrsloik> ubitux: I just didn't feel like adding it
[21:10] <Myrsloik> and it's not needed that often
[21:11] <ubitux> except for 1/5 frame drop, what other common settings are used?
[21:13] <Myrsloik> none, ever
[21:13] <ubitux> why is cycle even configurable then? :)
[21:13] <Myrsloik> because it's always been like that
[21:13] <ubitux> i guess having both configurable could make sense if you want to, let's say, 2/10?
[21:14] <Myrsloik> yes, but finding a clip where that makes sense will be hard
[21:14] <ubitux> i thought bad edits were common
[21:14] <ubitux> :p
[21:14] <ubitux> well anyway, ok
[21:16] <Myrsloik> usually stuff doesn't have 2 cuts/10 frames
[21:28] <ubitux> saste: man ffmpeg-all sounds better than the -monolithic suffix which i'm unable to write in one row without a typo
[21:28] <saste> yes indeed
[21:29] <ubitux> saste: so we're going to have ffmpeg-all ffplay-all etc?
[21:29] <ubitux> (in addition to the other current manpages)
[21:41] <saste> ubitux, yes
[21:42] <ubitux> ok with me then
[21:43] <ubitux> i hope we can compete with:
[21:43] <ubitux> - man zshall|wc -l
[21:43] <ubitux> 22382
[21:49] <llogan> michaelni: is there any bounce information for this user? http://ffmpeg.org/pipermail/ffmpeg-user/2013-April/014334.html
[21:49] <llogan> in the mail logs i mean
[21:50] <llogan> although i have a feeling it's his fault, but he's generally too much of an asshole to believe otherwise.
[21:52] <gnafu> Ugh, /that/ guy.
[21:53] <llogan> lol. a spam in mod queue has a 392 word count for subject
[22:01] <ubitux> 0001-Added-ability-to-do-video-fade-based-on-time.patch.tar.gz arg :(
[22:01] <durandal_1707> why?
[22:02] <ubitux> .tar.gz
[22:02] <saste> how is the patch?
[22:02] <durandal_1707> you dont have gunzip and tar installed
[22:02] <durandal_1707> saste: he removes .shortand .....
[22:02] <saste> fun
[22:03] <ubitux> he breaks compat afact
[22:03] <ubitux> on the form, he also adds a lot of whitespace but that's secondary
[22:03] <ubitux> anyone going to review or i should do it?
[22:04] <llogan> i was just about to tell him to attach no archived version
[22:05] <ubitux> i can do that in the review
[22:05] <ubitux> to avoid one pointless iteration
[22:05] <llogan> i was too slow anyway
[22:05] <ubitux> :)
[22:05] <ubitux> ah well, i guess i'll wait for the next iteration
[22:06] <durandal_1707> hmm is he subscribed, it looks like message is too old
[22:07] <durandal_1707> michaelni: can you add bot to report is someone does something funny with github mirror (like reporting bugs or send pull request)
[22:08] <michaelni> durandal_1707, thats more something for burek to setup, he is the master of fflogger
[22:10] <ubitux> https://github.com/FFmpeg/FFmpeg/pulls ow.
[22:10] <durandal_1707> michaelni: there is github bot already
[22:10] <durandal_1707> no need to setup new one...
[22:10] <durandal_1707> one just need to enable it/turn it on
[22:11] <michaelni> durandal_1707, there are 500 github bots, which one is it ?
[22:12] <llogan> durandal_1707: the vf_fade guy isn't subscribed. i had to accept the message in mod queue.
[22:12] <llogan> is why it arrived later
[22:14] <ubitux> durandal_1707: ah haha thx for correcting me
[22:14] <durandal_1707> perhaps we should migrate to github?
[22:14] Action: ubitux imagine the guy removing all spaces
[22:14] <durandal_1707> :)))
[22:14] <ubitux> "faster!
[22:15] <ubitux> hey cehoyos 
[22:15] <ubitux> nice to see you here more often
[22:15] <cehoyos> Hi! Do we know already what c99wrap does with av_rescale and elseif ?
[22:15] <cehoyos> Merci!
[22:17] Action: ubitux scratches nevcairiel 
[22:19] <durandal_1707> why libav did not pushed that s302m encoder? perhaps sender did not bumped thread enough times?
[22:23] <durandal_1707> why i cant fetch github pull request as patch?
[22:25] <cehoyos> durandal_1707: Wasn't this a patch on ffmpeg-devel ?
[22:25] <durandal_1707> cehoyos: wav one?
[22:25] <cehoyos> It was on both mailing lists, found both patches
[22:27] <cehoyos> It was actually send to ffmpeg-devel, it was posted by one of the thieves on avconv-devel
[22:27] <durandal_1707> there is no avconv-devel
[22:27] <cptspiff> durandal_1707: you mean in general, or some particular weird issue?
[22:28] <durandal_1707> are you talking about wav or encoder one?
[22:28] <durandal_1707> cptspiff: context? i'm little lost now
[22:28] <cptspiff> github pull as a patch
[22:29] <durandal_1707> i want it to download with my belowed web browser
[22:29] <cptspiff> right. append .patch to the url
[22:29] <durandal_1707> huh
[22:29] <durandal_1707> dark magic
[22:29] <cptspiff> https://github.com/xbmc/xbmc/pull/2444 -> https://github.com/xbmc/xbmc/pull/2444.patch
[22:30] <cptspiff> .diff for a flat one.
[22:31] <durandal_1707> about that wav patch, it could be written better one by making next_tag return FFALIGN(avio_rl32(pb), 2)
[22:40] <durandal_1707> cptspiff: thanks for hint (i will try to not forget it)
[22:41] <cptspiff> np
[22:52] <durandal_1707> sent comment to that wav patch
[22:54] <nevcairiel> cehoyos / ubitux: i build a reproducable test case for the c99-c89 converter bug, lets see if someone wants to fix it
[22:55] <durandal_1707> what about this mov change, it looks straightforward except it change/remove some tags
[22:55] <durandal_1707> and it sits there for 6 months with nobody replying
[22:56] <cehoyos> nevcairiel: Is there no work-around?
[22:58] <nevcairiel> i don't care, i want the issue in the converter fixed, thats all
[22:58] <michaelni> durandal_1707, about pull reqs i looked over the bots one can enable on github but i dont see anything for pull requests
[22:59] <durandal_1707> wm4 have one on mpv-player channel
[23:00] <durandal_1707> wm4: what you did?
[23:03] <nevcairiel> cehoyos: easy workaround would be to assign AV_TIME_BASE_Q to a variable first, instead of using it inline, because thats whats choking it
[23:07] <cehoyos> nevcairiel: " i want the issue in the converter fixed" <- of course, but for the time being, the variable would allow to continue testing msvc compilation.
[23:08] <durandal_1707> cehoyos: wasn't #2430 fixed? (aiff padding)
[23:09] <ubitux> cehoyos: if you commit a workaround, please comment the code with XXX/TODO/whatever so we don't forget to revert it later
[23:09] <cehoyos> Thank you for reminding me, I have to backport it.
[23:09] <cehoyos> ubitux: I don't like untested commits...
[23:29] <cone-610> ffmpeg.git 03Carl Eugen Hoyos 07release/1.0:13cf445358a1: Skip padding in an id3 tag in aiff files.
[23:29] <cone-610> ffmpeg.git 03Carl Eugen Hoyos 07release/1.1:bb46240cbb5d: Skip padding in an id3 tag in aiff files.
[23:29] <cone-610> ffmpeg.git 03Carl Eugen Hoyos 07release/1.2:c351d8781dc5: Skip padding in an id3 tag in aiff files.
[23:33] <cehoyos> nevcairiel: I sent you a patch, could you test?
[23:34] <nevcairiel> thats the wrong spot, the problem is the two av_rescale_q calls above that
[23:34] <nevcairiel> specifically, the first of those calls
[23:35] <nevcairiel> but i need sleep now, maybe i can find some time tomorrow to check
[23:38] <cehoyos> Sorry, new patch incoming...
[23:40] <ubitux> __gb__: btw, it would be nice to have a v4l2 "muxer" :)
[23:42] <nevcairiel> cehoyos: patch works
[23:43] <nevcairiel> and now off to bed
[23:44] <nevcairiel> oh and i reported the issue with the converter here in case anyone wants to track it: https://github.com/libav/c99-to-c89/issues/7
[23:44] Action: nevcairiel leaves
[23:53] Action: ubitux dreams of a day when mozilla wouldn't try to write broken a/v code and just uses ffmpeg
[23:57] <cone-610> ffmpeg.git 03Michael Niedermayer 07master:f51ce34e5e89: matroskadec: fix display_width/height = 0 handling
[23:59] <Compn> ubitux : you mean v4l loopback device ?
[23:59] <cehoyos> michaelni: Work-around for msvc to gitbug committed, please merge.
[00:00] --- Thu Apr  4 2013