[Ffmpeg-devel-irc] ffmpeg-devel.log.20181211

burek burek021 at gmail.com
Wed Dec 12 03:05:04 EET 2018


[00:42:58 CET] <cone-680> ffmpeg 03Carl Eugen Hoyos 07master:73abde61bb07: lavf/matroskadec: Do not use strncat() to limit copying a one-char constant.
[00:50:57 CET] <cone-680> ffmpeg 03Carl Eugen Hoyos 07master:73251678c83c: lavd/v4l2: Use "int request" as second parameter for ioctl() on Android.
[00:57:32 CET] <cone-680> ffmpeg 03Carl Hetherington 07master:6190f873eb6d: lavc/tiff: Fix leak of yuv_line in TiffContext.
[01:20:40 CET] <cone-680> ffmpeg 03Lou Logan 07master:0e8eb07980de: doc/indevs: fix hdvbuffer typo in iec61883 example
[06:31:06 CET] <cone-709> ffmpeg 03kjeyapal at akamai.com 07master:1d01ab399fc7: avformat/movenc: Added an option to disable SIDX atom
[06:31:06 CET] <cone-709> ffmpeg 03kjeyapal at akamai.com 07master:876ed08b0d13: avformat/dashenc: Used the movenc option skip_sidx instead of global_sidx
[10:53:19 CET] <Compn> i wonder if you can ssh > psexec localhost hmmmm
[10:53:55 CET] <Compn> ah yes you can
[10:54:02 CET] <Compn> that seems like the workaround then :)
[12:22:49 CET] <cone-456> ffmpeg 03Peter Ross 07master:b41f378fc4dd: os_support: define socket shutdown SHUT_xxx macros if they are not defined
[14:28:01 CET] <Compn> wat
[14:28:06 CET] <Compn> carrrrrrrrl
[16:22:16 CET] <cone-453> ffmpeg 03Gyan Doshi 07master:3b825b2f81e7: fate: add prog suffix in fate-mov-mp4-with-mov-in24-ver
[16:43:36 CET] <cone-453> ffmpeg 03James Almer 07master:9c49933a3e16: configure: require the first stable version of libdav1d
[17:18:04 CET] <jya> with mp4, in a stsd box defining multiple entries, each with a different codec valid? and if so does ffmpeg handle those?
[17:19:18 CET] <JEEB> hmm
[17:19:26 CET] <JEEB> 14496-12 for the first question
[17:19:31 CET] <JEEB> second will be answered after the first
[17:21:22 CET] <JEEB> > quantity: exactly one
[17:21:55 CET] <JEEB> so at least in a single stbl there should be one and only one stsd
[17:22:18 CET] <JEEB> ok, so stbl is also exactly one
[17:22:21 CET] <JEEB> which is in minf
[17:22:34 CET] <JEEB> ok, exactly one minf in mdia
[17:22:48 CET] <JEEB> exactly one mdia in trak
[17:23:13 CET] <JEEB> and of course there can be N traks
[17:23:31 CET] <JEEB> jya: does this answer your first question?
[17:24:19 CET] <jya> JEEB: well, a stbl has a single stsd yes
[17:24:37 CET] <jya> but stsd can have multiple entries, for video a VisualSampleEntry
[17:24:59 CET] <JEEB> right, I missed that part and started going from stsd upwards :P
[17:25:33 CET] <jya> and each VisualSampleEntry can define a codec
[17:25:42 CET] <jya> i've just never seen those files in the wild
[17:26:45 CET] <jya> we started to have to look into this case, because for some encrypted file, we see multiple encryption data stored across multiple entries
[17:28:14 CET] <JEEB> 1) multiple SampleDescriptions are of course valid
[17:28:37 CET] <JEEB> now 2) would be regarding actually different video/audio formats, or just a different identifier for the same stuff?
[17:29:28 CET] <jya> different all together
[17:30:19 CET] <jya> in firefox, we kind of allow different h264 content in the stream, but that only works because we assume it's all h264 and parse the raw bytestream to check if there are inband SPS
[17:30:46 CET] <jya> if you were to say have a group of sample using h264 and nother using vp9 that won't do
[17:30:49 CET] <JEEB> yea, I think we should have support for multiple initialization data
[17:30:55 CET] <JEEB> for the same codec
[17:31:11 CET] <JEEB> but not the "lol these samples have completely different content"
[17:31:15 CET] <jya> yes, same codec can do. different opens a bug can or worms
[17:31:19 CET] <jya> of worms
[17:31:32 CET] <JEEB> yes
[17:31:41 CET] <JEEB> I'm not sure if that's technically allowed in 14496-12
[17:31:49 CET] <JEEB> will have to take a look a bit at where different IDs are
[17:32:08 CET] <JEEB> (of course in the worst case it'll be "lol ISOBMFF" and someone goes get his whiskey)
[17:33:41 CET] <jya> JEEB: i can't see anything in there related to the use of different values for the codec type
[17:34:01 CET] <JEEB> yea
[17:34:02 CET] <jya> now I only have the 2015 version of it, that may have changed
[17:34:08 CET] <JEEB> because it's in stsd
[17:34:17 CET] <JEEB> yes, technically it ain't saying anything
[17:34:19 CET] <JEEB> (jesus christ)
[17:34:53 CET] <jya> looking at their stsd summaries: sample descriptions (codec types, initialization etc.)
[17:34:58 CET] <jya> its all plural :)
[17:35:08 CET] <JEEB> yes, there's no question about that
[17:35:32 CET] <JEEB> I mostly had to verify where the format identifiers were again by looking at boxdumper's output in less :P
[17:35:34 CET] <jya> not in 8.5.2.1 however
[17:35:46 CET] <jya> "The sample description table gives detailed information about the coding type used, and any initialization information needed for that coding."
[17:35:50 CET] <jya> that's singular
[17:36:23 CET] <jya> we've written our own moov parser in rust
[17:36:27 CET] <JEEB> yes, but the actual ID is in the sample descriptor
[17:36:48 CET] <jya> due to the multiple security issues we got when we unfortunately decided to use stagefright
[17:36:48 CET] <JEEB> > Multiple descriptions may be used within a track.
[17:37:08 CET] <JEEB> which is in 8.5.2.1 on the next page
[17:37:10 CET] <jya> (reason for stagefright was that ffmpeg had no support for fragmented mp4 back then)
[17:38:09 CET] <jya> luckily there can only be 65536 entries :)
[17:39:45 CET] <JEEB> and yea, stsc lets you switch around
[17:39:48 CET] <JEEB> christ this is perverse
[17:40:03 CET] <JEEB> at least if you take all the freedoms the spec (seems to) give you
[17:48:58 CET] <jya> we're going to parse all entries for the various configuration available, but error if the codec ever change.
[17:49:01 CET] <jya> thanks for your help
[17:49:52 CET] <JEEB> yes, that sounds like the correct way to do it unless you want a world of pain
[17:56:27 CET] <durandal_1707> who bumps versions ALLLL THE TIME, STOOOOOOOOOOOOP!!!!!!!!!
[17:57:52 CET] <cone-453> ffmpeg 03Paul B Mahol 07master:dd7d6034f1fe: avfilter/vf_paletteuse: fix flags for alpha_threshold option
[19:14:18 CET] <cone-453> ffmpeg 03Carl Eugen Hoyos 07master:464aa60eb5e5: tests/api-flac-test: Rename NUMBER_OF_FRAMES as NUMBER_OF_AUDIO_FRAMES.
[20:12:45 CET] <kierank> durandal_1707: nice project for you
[20:12:58 CET] <kierank> dealing with masterpiece that is mpegvideo.c, h263.c and ituh263.c
[20:15:04 CET] <atomnuker> I wonder if you can make an mpeg4 version of https://pics.onsizzle.com/the-the-father-is-not-ll-on-is-god-0-34080567.png
[22:03:39 CET] <durandal_1707> kierank: i do not want crappy toys for xmas
[00:00:00 CET] --- Wed Dec 12 2018


More information about the Ffmpeg-devel-irc mailing list