[Ffmpeg-devel-irc] ffmpeg-devel.log.20140302

[burek]

[00:16] <cone-598> ffmpeg.git 03Michael Niedermayer 07master:2adf422ce24b: avformat/aviobuf: factorize buffer_size out
[00:16] <cone-598> ffmpeg.git 03Michael Niedermayer 07master:9600486d10a9: avformat/aviobuf: optimize sequential backward seeking
[00:26] <cone-598> ffmpeg.git 03Michael Niedermayer 07master:171dd67520b6: avformat/mp3dec: optimize mp3_seek() for dir < 0
[01:07] <cone-598> ffmpeg.git 03James Almer 07master:68c3ed936a76: x86/synth_filter: add synth_filter_avx
[01:54] <cone-598> ffmpeg.git 03Keiji Costantini 07master:951444033787: ituh263: reject b-frame with pp_time = 0
[01:54] <cone-598> ffmpeg.git 03Michael Niedermayer 07master:73f930ee71e0: Merge remote-tracking branch 'qatar/master'
[04:46] <cone-598> ffmpeg.git 03Michael Niedermayer 07master:52a213865670: avformat/avidec: Use a buffer with sufficient padding in read_gab2_sub()
[05:14] <cone-598> ffmpeg.git 03Michael Niedermayer 07master:f5d1d1e4667b: avformat/oggparsevorbis: check packet size before reading new_len from it
[12:28] <cone-858> ffmpeg.git 03James Almer 07master:884e085d1ea3: x86/synth_filter: Revert the switch to float ops with SSE2
[12:35] <michaelni> btw, i tend to be sometimes rather quick with applying asm optimization patches, i hope this doesnt prevent anyone from making suggestions on how to improve them further, it definitly isnt intended to. So if someone, Skyler_ ubitux, pengvado, BBB or others still have suggestions, please dont hesitate to post them after the patch got in. Also if preferred and we have a vounteer iam happy to leave asm patch pushing to others,
[12:35] <michaelni>  just dont want patches to get lost/forgotten
[12:36] <ubitux> fine with me, i'm reading cvslog anyway
[13:16] <cone-858> ffmpeg.git 03Michael Niedermayer 07master:0bf882864ecd: Revert "mpegts: do not set pts for missing dts in video streams"
[14:30] <Compn> anyone want to see if these rtmp streams have closed captions inside the flv or if they use a seperate closed caption file ?
[14:31] <Compn> http://media.mtvnservices.com/embed/mgid:uma:videolist:vh1.com:1705091/cp~instance%3Dfullepisode%26autoPlay%3Dfalse%26id%3D1705091%26uri%3Dmgid%3Auma%3Avideolist%3Avh1.com%3A1705091
[15:32] <cone-858> ffmpeg.git 03Michael Niedermayer 07master:6b53c1aa822e: avcodec/g2meet: fix error returns
[15:32] <cone-858> ffmpeg.git 03Michael Niedermayer 07master:8b8ae298afb3: avcodec/g2meet: also reset local got_header when reseting the context got_header
[16:48] <kierank> nevcairiel: is there any way of getting mpc-hc to take in an rtp source?
[16:49] <JEEB> just register LAV Splitter as the source filter for the protocol?
[16:50] <JEEB> I mean, the LAV Splitter source filter
[16:50] <nevcairiel> its internal version should handle RTP
[16:50] <JEEB> yes, but don't you still need to register the source filter thingamajig?
[16:50] <JEEB> or will it override
[16:50] <nevcairiel> there is an internal mapping for rtp, i think
[16:50] <JEEB> if it will override then you don't have to do that, but just have it enabled in MPC-HC's "internal filters" part
[16:51] <JEEB> k
[17:02] <kierank> but how do I specify rtp input
[17:02] <kierank> I have packets coming in on port 1234
[17:02] <kierank> in vlc you use rtp://@:1234
[17:03] <nevcairiel> however you would tell libavformat to do that, if you even can :)
[17:03] <kierank> you can iirc
[17:06] <nevcairiel> any rtp:// url is just given to avformat to handle, so if you figure out its syntax, then it s hould work
[17:06] <nevcairiel> i've only ever connected to a remote, not listened for incoming traffic
[17:08] <nevcairiel> not that streaming in general works really good through mpc-hc, but it should spit out some sort of result
[17:09] <kierank> basically I want to use gpu deinterlacing
[17:09] <JEEB> too see the stuff that gets output into stderr, just stick a debugger to the process :)
[17:09] <nevcairiel> mpc-hc can read from stdin if that helps you :P
[17:09] <JEEB> could help possible debugging of why stuff doesn't work a bit
[18:20] <kierank> i am not even sure how stderr works in windows
[18:21] <JEEB> kierank, in GUI apps it ends up in debug output
[18:21] <kierank> ah
[18:21] <JEEB> so sticking a debugger kind of thing or a debug log reader to the process should output that
[18:22] <nevcairiel> you can cheat and get it if you use something like more
[18:22] <nevcairiel> which probably does something similar to the debugger, just attach something to the stdout/stderr, which otherwise go to NUL
[18:27] <Waraqa> Hi, I found a bug related to an audio filter. Do I need to upload the file to be able to report the bug?
[20:01] <llogan> Waraqa: if the sample is required to reproduce the bug then you need to upload it
[20:01] <Waraqa> llogan: It's 4 Gb
[20:01] <llogan> that's big.
[20:02] <llogan> can you still reproduce the issue if you make a smaller sample with dd?
[20:02] <Waraqa> I will check
[20:02] <llogan> also you should test with ffmpeg from git master if you can. if you can't compile then try a recent release via links on the FFmpeg Download page.
[20:03] <llogan> ...if you haven't already
[20:04] <Waraqa> It's still crashing with latest ffmpeg build
[20:17] <llogan> Waraqa: did it also crash in the same way with the small sample from dd?
[20:18] <Waraqa> llogan: It's on the other laptop, I have not checked yet
[20:25] <Waraqa> llogan: It's not crashing but there is an error
[20:25] <Waraqa> Assertion zeros != nsamples failed at libavfilter/af_atempo.c:540
[20:27] <Waraqa> I get also the same error if I take a sample using -vcodec copy
[20:27] <llogan> if you think it's a bug then go ahead and report it. might want to perform a search first to see if an existing, related report exists. then include the full ffmpeg command and the complete ffmpeg console otuput.
[20:27] <Waraqa> No need to upload any input?
[20:27] <llogan> ...and include the sample if it is required to duplicate the issue
[20:28] <llogan> unless the issue can also be encountered with ffmpeg generated audio or a file from the FATE test suite.
[20:28] <wm4> ubitux: you here? I'm still not sure what exactly I should do about the microdvd framerate issue
[20:28] <wm4> ubitux: I think the API my last patch added was disliked
[20:28] <llogan> Waraqa: "ffmpeg -f lavfi -i sine=d=5 <options> output" for example
[20:29] <ubitux> wm4: weren't you going to add that private option and read it to check if it changed?
[20:29] <wm4> ubitux: a "subfps" option was discussed
[20:29] <wm4> but aqtitle also has this option
[20:29] <ubitux> what's the problem?
[20:29] <Waraqa> llogan: Is this example enough to generate audio?
[20:30] <llogan> Waraqa: probably. i have to go now, but go ahead and report the bug if you can. at worst it will be closed as a duplicate/invalid or you may be asked for more information.
[20:30] <wm4> so what exactly should it do when the user sets it?
[20:30] <ubitux> override, juste like in aqtitle
[20:30] <wm4> if the subtitle has a framerate, and the subfps option is set, should the option be ignored and overwritten?
[20:30] <ubitux> the difference being that if it's set in the file it will set the option as well
[20:30] <ubitux> unless specified by the user
[20:31] <ubitux> well, whatever behaviour you would prefer
[20:31] <Waraqa> llogan: thanks
[20:32] <wm4> ubitux: does aqtitle usually use 25 fps, or does it use video framerate?
[20:32] <ubitux> wm4: actually, following the logic that the user will not set the fps unless it's wrongly guessed (default value, or specified in the file), if the option is set, the demuxer should not update it from the value in the file
[20:32] <ubitux> wm4: i don't remember honestly
[20:33] <ubitux> wm4: and didn't have much real sample with a video
[20:33] <ubitux> currently aqtitle assume 25 fps, and you can override that value
[20:33] <wm4> well we could just say "fuck aqtitle"
[20:33] <ubitux> sure
[20:33] <ubitux> as long as you don't break it
[20:34] <ubitux> the difference with aqtitle option is that you will set a 0/0 frame rate (or 0/1) by default
[20:34] <ubitux> it will be overwritten by either the user, or the file content
[20:35] <wm4> so the file contents should just override subfps?
[20:35] <ubitux> yes, so you can read it back
[20:35] <ubitux> note that the default value (29something) should not be the default value of the option - and should never set -, so you can make a distinction between the two
[20:36] <ubitux> and i'd say a user setting the option should override the value in the file
[20:36] <wm4> so that means if I want to handle other, timestamp based subtitle formats, I first have to check whether subfps exists
[20:36] <ubitux> (so users can adjust/fix themselves if it's not ok)
[20:36] <ubitux> wm4: i'm still waiting for another example
[20:36] <ubitux> so far only aqtitle looks like a candidate
[20:36] <wm4> generally: if subfps exists, but is not set (set to 0/0?), then assume the subtitle uses video framerate
[20:36] <ubitux> and we pick a same option
[20:37] <ubitux> probably yes
[20:37] <wm4> currently, only aqtitle has a subfps option
[20:37] <ubitux> that's the only sub frame based afaict
[20:37] <wm4> anyway, I bet if I write a patch, that will again trigger lots of discussion
[20:37] Action: wm4 grumbles
[20:37] <ubitux> oh well, now that i think about it
[20:37] <ubitux> jacosub might have a frame-based info
[20:38] <ubitux> wm4: if you're only playing with the private option of microdvd and aqtitle, that should be pretty fine :)
[21:34] <wm4> ubitux: do I have to document the option somewhere outside of micodvddev.c?
[21:34] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:dfefc9097e9b: avcodec/hevcpred_template: also initialize top[-1] for constrained intra prediction
[21:34] <ubitux> you can update doc/demuxers.texi if you feel like it but i don't think that's worth the effort
[21:35] <wm4> aqtitle isn't documented either, so I'll skip that one
[21:57] <cone-897> ffmpeg.git 03Patrice Clement 07master:5c9c305d1473: ass: K&R formatting cosmetics
[21:57] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:411f8f4c9a05: Merge commit '5c9c305d14730faf2af206e5d519df4948de0ed2'
[22:41] <ubitux> wm4: so the input frame rate takes over the one the user requested?
[22:41] <wm4> ubitux: *shrug*
[22:41] <wm4> it does, should it not? I have no clue and actually don't care
[22:42] <wm4> I tried to make it somewhat symmetric with aqtitle, but maybe that failed
[22:43] <ubitux> you override microdvd->frame_rate before checking if it was set by the user
[22:43] <wm4> maybe the avoption should just export a single flag (the information I really need)
[22:43] <ubitux> i was assuming the user would want to customize it whatever the input says (since it's probably wrong)
[22:43] <wm4> ubitux: the idea was that writing the option sets the fallback
[22:43] <wm4> not that it always forces the framerate
[22:44] <wm4> always forcing it would be useless too, because you can just play with the timebase yourself
[22:44] <ubitux> mmh, ok..
[22:45] <ubitux> in that case, the description should says "set fallback movie frame rate"
[22:45] <ubitux> i guess
[22:45] <ubitux> otherwise i guess that patchset is fine
[22:45] <ubitux> if that achieves what you need
[22:46] <wm4> setting the subfps option to the video framerate in fact achieves the logic you need to get proper timestamps
[22:46] <wm4> but still allows an api user to do this manually
[23:38] <Snowrider_> Can anyone give me any insight on how to access rgb values for a frame of a picture? 
[23:39] <Snowrider_> an AvFrame*
[23:46] <BtbN> Snowrider_, most frames from normal video files don't have any rgb data.
[00:00] --- Mon Mar  3 2014