[Ffmpeg-devel-irc] ffmpeg-devel.log.20140210

[burek]

[00:46] <saste> why is trac still down?
[00:47] <beastd> saste: Is it?
[00:47] <saste> beastd, here I get a Service Temporarily Unavailable
[00:48] <saste> https://trac.ffmpeg.org/
[00:48] <beastd> saste: sounds cached :(
[00:50] <saste> well time to sleep, will check it tomorrow
[00:50] <saste> god night
[01:28] <cone-814> ffmpeg.git 03Janne Grunau 07master:b37effdc5283: mxfdec: free descriptor extradata in .read_close()
[01:28] <cone-814> ffmpeg.git 03Michael Niedermayer 07master:ec1a1be54fa7: Merge commit 'b37effdc52835e121a0e259e57273a19f9cd61a3'
[02:27] <cone-814> ffmpeg.git 03Ronald S. Bultje 07master:2f6eec65acc9: vp8: fix PPC assembly to work if src_stride != dst_stride
[02:27] <cone-814> ffmpeg.git 03Michael Niedermayer 07master:d84f1a60e07c: Merge commit '2f6eec65acc953faed3951fb8f1eabed830fdd2f'
[03:54] <cone-814> ffmpeg.git 03Kostya Shishkov 07master:e95018b694c0: dnxhd: Support DNx444
[03:54] <cone-814> ffmpeg.git 03Michael Niedermayer 07master:51290068fa12: Merge commit 'e95018b694c0774477abec5bbf86ecc7946a9a28'
[04:11] <cone-814> ffmpeg.git 03Luca Barbato 07master:246d3bf0ec93: hevc: Conceal zero td
[04:11] <cone-814> ffmpeg.git 03Michael Niedermayer 07master:03911c418f1a: Merge remote-tracking branch 'qatar/master'
[10:24] <ubitux> huh, you need to be registered to access relevant ARM documentation?..
[10:24] <ubitux> > NEON Programmers Guide
[10:24] <ubitux> > This document is only available in a PDF version to registered ARM customers.
[10:24] <ubitux> wth.
[11:07] <Mavrik> ;)
[11:56] <cbsrobot>  ubitux: thanks for the ebur128 commit
[11:57] <ubitux> thanks to you :p
[12:26] <cone-890> ffmpeg.git 03Anton Khirnov 07master:d711d839ca67: doc/APIchanges: fill in missing hashes and dates
[12:26] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:2e99867a4e43: Merge commit 'd711d839ca6714a5ef5b53ee51bc7632e62cf203'
[12:29] <cbsrobot> anybody knows a fast way to blend two 48-bit (or 64) per pixel images with alpha channel ?  
[12:31] <cone-890> ffmpeg.git 03Anton Khirnov 07master:cca4742a5e66: doc/APIchanges: mark the place where 9 was branched
[12:31] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:ee2c1e6dd337: Merge commit 'cca4742a5e663cfe4dbd71f01f0c0fb042fddc1d'
[12:39] <cone-890> ffmpeg.git 03Anton Khirnov 07master:38893dc028e4: pcmdec: replace a reachable assert with an error check
[12:39] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:6732f4350d08: Merge commit '38893dc028e458eaf3f906833d4ee515689edb7e'
[12:50] <cone-890> ffmpeg.git 03Anton Khirnov 07master:b5f851ed7e59: FATE: force FLAC in the lavf ogg test
[12:50] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:1e5656bf8a8e: Merge commit 'b5f851ed7e59f88b4130a420033d9fe191ec9e2f'
[12:59] <cone-890> ffmpeg.git 03Guillaume Martres 07master:f548f9f9e7a2: hevc: update conformance streams
[12:59] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:919791f99dc8: Merge commit 'f548f9f9e7a2a08c780e88dac63b9d5d2c55efb0'
[13:03] <cone-890> ffmpeg.git 03Marton Balint 07master:6421111b52de: lavf: allow av_find_best_stream to return AV_DISPOSITION_*_IMPAIRED wanted_streams
[13:32] <cone-890> ffmpeg.git 03Anton Khirnov 07master:b36bc81ccaa2: avplay: add support for seeking to chapter marks
[13:32] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:6fb8684a24cd: Merge commit 'b36bc81ccaa2fc85dc4bae7dc546c71e8833573d'
[13:47] <cone-890> ffmpeg.git 03Anton Khirnov 07master:ad7f2b69a546: lavf: rewrite the av_interleaved_write_frame() doxy
[13:47] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:58fdc6db6c72: Merge commit 'ad7f2b69a546ec65ef6b11e1a0fd58e6bb66bc4b'
[13:52] <cone-890> ffmpeg.git 03Anton Khirnov 07master:4c750599509c: lavf: rewrite the av_write_frame() doxy
[13:52] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:6fd7ea44b2e1: Merge commit '4c750599509c3973afbe2850953a1fb963f696e5'
[14:10] <cone-890> ffmpeg.git 03Anton Khirnov 07master:54f7e79d4706: lavf: add some basic documentation of the muxing process
[14:10] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:c84280b09ce6: Merge commit '54f7e79d4706a8343dad1d8da51b7d3d3b2cd3b2'
[14:19] <cone-890> ffmpeg.git 03Anton Khirnov 07master:18c896be3d8e: lavf: extend / improve the AVFormatContext doxy
[14:19] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:ec8e232b53b3: Merge commit '18c896be3d8e926ef806e7de29c4a168d7763389'
[14:31] <cone-890> ffmpeg.git 03Christophe Gisquet 07master:682b2a809b44: build: fix examples compilation when EXESUF is not empty
[14:31] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:1c1bb203ddae: Merge remote-tracking branch 'qatar/master'
[15:22] <Daemon404> is saste away lately?
[15:22] <ubitux> he's not as active as he used to be but he was here yesterday
[15:23] <Daemon404> ah ok
[15:23] Action: Daemon404 will wait
[17:02] <wm4> ubitux: I got a microdvd subtitle file that uses "{1}{1}23.976" to declare the frame rate
[17:16] <wm4> ubitux: it's probably read correctly, but it's also displayed for some reason
[17:16] <ubitux> mmh
[17:17] <ubitux> it should probably be skipped yes
[17:18] <j-b> llogan: btw, the link to the git howto was fixed, as you asked.
[17:18] <ubitux> wm4: can you find someone interested in contributing to subtitles code in ffmpeg?
[17:19] <ubitux> wm4: it seems i'm busy with more fun stuff nowadays :(
[17:20] <wm4> ubitux: I don't know anyone either, but I could write a patch for this issue (if I know how...)
[17:20] <wm4> I'm not entirely sure why this line is displayed at all
[17:20] <wm4> probably my own funny-ass logic for subtitles with no end time
[17:20] <ubitux> so, it's in libavformat/microdvddec.c
[17:21] <ubitux> and then well, i think you would just add a "continue" in the pts_info = ... set thing
[17:21] <ubitux> just like the {DEFAULT}... actually
[17:22] <wm4> sounds like a good idea
[17:22] <ubitux> btw, while at it, i suppose you want to add the "subpts" option just like in aqtitle
[17:22] <ubitux> (libavformat/aqtitledec.c)
[17:23] <ubitux> and eventuallyy add a FRAME_BASED kind of flag
[17:23] <wm4> can't find the string subpts in libavformat/aqtitledec.c
[17:24] <ubitux> subfps
[17:24] <ubitux> sorry
[17:24] <wm4> doesn't look useful
[17:24] <ubitux> why?
[17:24] <ubitux> didn't you open a ticket for this?
[17:25] <wm4> I can use this to override the subtitle fps?
[17:25] <ubitux> so you could force the subfps to the video frame rate
[17:25] <wm4> I don't even know the video frame rate when opening the demuxer
[17:25] <ubitux> well i guess you would do that to override the pts_info thing
[17:26] <wm4> I have to know two things: 1. is the format timestamp based or frame based? 2. if it's frame based, how do I get the frame number from the (made up) timestamps
[17:26] <wm4> 2. could be solved by always using a fixed timebase (like 23.976)
[17:26] <wm4> for 1., a flag has to be added, as you said
[17:27] <wm4> which is an api change
[17:27] <ubitux> the pts are generated for the given FPS
[17:28] <ubitux> so it's just like if the subtitles format was containing timestamps for you
[17:28] <ubitux> it shouldn't be different
[17:28] <ubitux> so i don't understand the problem
[17:28] <wm4> a bit roundabout, but could work
[17:31] <ubitux> wm4: the time base is set to 1/fps, and pts are actually frame number
[17:31] <ubitux> but if you rescale those pts to that timebase
[17:31] <ubitux> you actually get the timestamps
[17:31] <wm4> I know
[17:31] <BtbN> should i set the pts or the dts to AV_NOPTS_VALUE? Currently i set dts to AV_NOPTS_VALUE, but i get "Encoder did not produce proper pts, making some up.", but the pts looks ok.
[17:31] <ubitux> just like any other subtitles format
[17:31] <ubitux> so what's the problem?
[17:32] <BtbN> should i just set the pts to AV_NOPTS_VALUE, and set the dts to the current pts instead?
[17:32] <ubitux> i mean what this difference is breaking?
[17:32] <wm4> ubitux: well, I need a way that works and that is uniform across subtitle formats
[17:32] <ubitux> wm4: it's uniform?
[17:32] <wm4> aqtitle apparently uses 25 fps by default, microdvd uses 23.976
[17:32] <ubitux> it doesn't matter for you
[17:32] <wm4> unless you guarantee that the timestamp is always the frame number
[17:32] <ubitux> it's 25 by default because the format stat so iirc
[17:33] <ubitux> wm4: ??
[17:33] <ubitux> wm4: with any subtitle format you're supposed to rescale the pts based on that timebase
[17:33] <ubitux> frame based or not that will give you the same correct ts
[17:33] <ubitux> most subtitles format use different tb
[17:34] <ubitux> srt will use 1/1000, ass 1/100, aqt 1/25, ..
[17:34] <ubitux> or maybe i'm missing something?
[17:34] <wm4> I need to know the case when a subtitle (format) depends on the video FPS
[17:35] <ubitux> wwhy?
[17:35] <wm4> which is the case with microdvd _if_ it doesn't set the fps
[17:35] <wm4> because that's how shit works
[17:35] <wm4> download a microdvd file for PAL, and it's implictly 25 fps
[17:35] <wm4> download one for NTSC, it's implicitly 23.976
[17:35] <wm4> unless the file contains a FPS header
[17:35] <ubitux> ah you're trying to actually support the frame-based thing without the timestamp abstraction from ffmpeg?
[17:36] <wm4> what
[17:36] <ubitux> if you mean you want to change that hardcoded 23.976 value, you would just add a "subfps" option just like aqtitle
[17:37] <ubitux> and try to set it all the time if you know that value
[17:37] <wm4> but I don't know the video fps when opening the subtitle file
[17:39] <ubitux> i'm sorry, i'm probably under neurosis because i just don't understand why you want to know that
[17:40] <ubitux> and how knowing that the subtitles are actually frame based will allow you to do
[17:40] <wm4> I can process the subtitle later
[17:40] <ubitux> unless you actually want to be very frame exact and actually blend the subtitles per frame
[17:41] <ubitux> wm4: btw, are you able to change the time base of the demuxer?
[17:41] <ubitux> (i wonder if the subfps option is actually necessary if you can do that)
[17:41] <wm4> ubitux: I don't think just setting the timebase from outside is allowed
[17:41] <ubitux> ok
[17:42] <ubitux> sounds legit
[17:42] <ubitux> then we may want a subfps in microdvddec to force a different fps
[17:42] <ubitux> and also, we can add a FRAME_BASED flag in the codec desc. for those codec
[17:42] <ubitux> telling the frame->pts will actually be the frame number
[17:42] <wm4> currently I have this cute hack to detect frame based: http://privatepaste.com/d9a464bcfc
[17:43] <ubitux> what?
[17:43] <wm4> indeed
[17:43] <ubitux> why didn't you hardcode the codec name instead? :p
[17:43] <wm4> yeah, maybe it should _also_ check for the codec name
[17:43] <ubitux> anyway, i still don't get it
[17:44] <ubitux> also? meh
[17:44] <wm4> I'll try to explain it again
[17:44] <wm4> microdvd files normally contain frame numbers, right?
[17:44] <ubitux> yes
[17:44] <wm4> these are video frame numbers (during which the subs should be displayed)
[17:44] <wm4> so, at which time a subtitle appears obviously depends on the video frame rate
[17:45] <wm4> crappy design, but this is how this format works
[17:45] <ubitux> so far, i understand
[17:45] <wm4> so if you use the "wrong" video file with the sub file, the subs will be mistimed
[17:45] <wm4> (but that's by design)
[17:45] <ubitux> right
[17:46] <wm4> to do this correctly, the subtitle code in the applciation has to know that this format is frame based
[17:46] <wm4> because libavformat always returns timestamps, not frame numbers
[17:46] <wm4> you could special case this by checking the codec name
[17:47] <wm4> say, if the codec name is microdvd, then assume it's frame based
[17:47] <ubitux> right right
[17:47] <ubitux> so that was my initial suggestion
[17:47] <wm4> now here starts the problem: some microdvd files actually declare a FPS
[17:47] <ubitux> you actually want to handle the pts as frame id yourself
[17:47] <wm4> this is the FPS of the video the subs were timed against
[17:47] <ubitux> and blend the subtitle based on frame
[17:48] <wm4> and in this case, you don't want to treat this subtitle as frame based, even though it's microdvd
[17:48] <ubitux> mmh
[17:48] <wm4> and yes, setting the subfps option (like aqtitle has it) to the video frame rate would handle this correctly
[17:48] <wm4> because microdvd.c just overrides the FPS with the value in the file header
[17:49] <wm4> but my application opens all files at start, so I don't know the video FPS yet
[17:49] <wm4> now that I think about it, I could set the subfps option to something absurd
[17:49] <wm4> and detect the frame based case by this
[17:49] <ubitux> haha
[17:50] <ubitux> you can just add a flag to the codec desc
[17:50] <ubitux> that's fine with me
[17:50] <wm4> like I just explained, whether it's frame based or not depends on the _file_
[17:50] <ubitux> but that won't give you the default fps, which you seem to need
[17:50] <wm4> at least for microdvd
[17:52] <ubitux> ah, hell yeah i understand now
[17:53] <ubitux> mmh
[17:54] <ubitux> wm4: do you see a pratical solution for this except the subfps hack to detect explictely set fps?
[17:56] <wm4> hm maybe AV_DISPOSITION_ could be abused to set a frame-based flag?
[17:56] <wm4> there's already random crap in it (like AV_DISPOSITION_ATTACHED_PIC), so why not
[17:56] <ubitux> and that flag would be set by microdvd (and unset if FPS is set in the file)?
[17:57] <wm4> yes
[17:57] <wm4> as for the default frame rate; I'm not sure
[17:57] <ubitux> i'm not sure about AV_DISPOSITION but that could be discussed
[17:57] <wm4> subfps might also work, but it'd have tricky semantics...
[17:57] <wm4> I mean, it's not obvious that a demuxer must ignore subfps if the data is not frame based
[18:00] <ubitux> so anyway, the semantic of microdvd is not 23.976  when unset
[18:01] <ubitux> microdvd can actually be fps based only when FPS are explicit.. that's pretty news to me :(
[18:02] <ubitux> wm4: and btw, the AVOptions don't allow you to read values from the context? :-°
[18:02] <wm4> sure
[18:03] <wm4> that could be used
[18:03] <ubitux> because in that case we can use those avoptions
[18:03] <ubitux> to communicate the explicit set fps
[18:03] <wm4> you just have to make sure the used option names don't collide with anything else
[18:05] <ubitux> gtg bbl
[18:15] <cone-890> ffmpeg.git 03Zhang Rui 07master:4958628ba69e: avformat/concatdec: pass the interrupt callback on
[18:56] <llogan> j-b: thanks!
[19:05] <Skyler_> BBB: by the way you can steal any of my ideas to stuff into your patches if you'd like, I don't care about credit (e.g. the cache thing)
[19:06] <j-b> llogan: tell me if it does not work
[19:13] <llogan> j-b: looks good to me
[19:53] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:e93674c4a322: doc/APIchanges: fill in FFmpeg hashes & versions
[20:32] <cone-890> ffmpeg.git 03Anton Khirnov 07master:394fb56c29ee: lavf: always unref the packet passed to av_interleaved_write_frame() on error
[20:32] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:8560fdc45195: Merge commit '394fb56c29eee7f4f8f0334d8b5d30d3c54ac703'
[20:32] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:3868e97c6b68: ffmpeg: fix vstats
[20:41] <cone-890> ffmpeg.git 03Anton Khirnov 07master:e1218ce9149a: rpza: switch to bytestream2
[20:41] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:8b0befe45195: Merge commit 'e1218ce9149aad0ae9770a32a78d77ad2210bd64'
[20:46] <cone-890> ffmpeg.git 03Anton Khirnov 07master:c7769df178fd: rpza: use fixed-width int types
[20:46] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:0a4311dee55b: Merge commit 'c7769df178fd670156f1c11f6c8410ddd59bbfee'
[21:02] <cone-890> ffmpeg.git 03Anton Khirnov 07master:a46dc49744bd: rpza: move some variables to the blocks where they are used
[21:02] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:996fdca8fbc1: Merge commit 'a46dc49744bdc4f2e31725b63ac8e41f701e4fa1'
[21:38] <cone-890> ffmpeg.git 03Anton Khirnov 07master:77bb0004bbe1: rpza: limit the number of blocks to the total remaining blocks in the frame
[21:38] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:dfc99ca04d76: Merge remote-tracking branch 'qatar/master'
[22:04] <cone-890> ffmpeg.git 03Christophe Gisquet 07master:c3390fd56cf5: ra144: use scalarproduct_int16
[22:04] <cone-890> ffmpeg.git 03Michael Niedermayer 07master:8f92edf6f86c: avcodec/ra144: remove redundant AV_ZERO128
[23:01] <michaelni> llogan, theres no link to twitter anywhere on our homepage or am i missing it ?
[23:08] <llogan> michaelni: i don't think there was ever a link.
[23:08] <michaelni> i think you should add one
[23:08] <michaelni> maybe where the g+ and rss icons are there also could be a twitter icon
[23:09] <llogan> fine with me. i'll add one in a few days unless you get to it first.
[23:10] <michaelni> llogan, this is the correct url ? https://twitter.com/FFmpeg (asking because its one field in the gsoc form)
[23:10] <llogan> yes
[23:10] <michaelni> thx
[23:11] <michaelni> do we have a link to a smaller tan 65x65 logo ?
[23:11] <llogan> i'm not sure
[23:12] <llogan> what size and logo variant do you need exactly?
[23:12] <llogan> square? with or without "FFmpeg"? "3D"? non-3D?
[23:16] <michaelni> llogan, "URL to the logo of the organization. Please ensure that the provided image is smaller than 65px65px."
[23:16] <llogan> so they can upscale it to 650x650 and print it
[23:16] <llogan> i'll send you something
[23:46] <michaelni> llogan, i think smaller than 65x65 implies not 65x65
[23:47] <michaelni> also the logo isnt mandatory
[23:53] <michaelni> we have no facebook page ?
[00:00] --- Tue Feb 11 2014