[Ffmpeg-devel-irc] ffmpeg-devel.log.20170226

burek burek021 at gmail.com
Mon Feb 27 03:05:02 EET 2017


[01:35:19 CET] <mfg> Hi, I submitted a patch to the ML, but I'm having a hard time getting it pulled in -- are people just busy or did I offend someone. Its the strptime changes...
[01:39:22 CET] <JEEB> people can be busy/not notice/etc
[01:39:45 CET] <JEEB> I usually wait a few times and then ping if nobody has noted anything (usually a couple of work days)
[01:40:18 CET] <mfg> ok... its been a week and i sent a reminder mid week. I'll ask again on Monday, but I wanted to make sure I hadn't been black-balled or something -- I accedentally top-posted in one of my emails.
[01:42:54 CET] <memeka> hi, are there any plans on adding v4l2 m2m to ffmpeg?
[01:44:08 CET] <JEEB> heh, just read what I replied to you @ the user channel :P
[01:44:33 CET] <JEEB> mfg: people don't start ignoring you because of that as long as any future correspondence is OK
[01:44:59 CET] <memeka> thanks; i saw there were some patches i see submitted a while ago (2 yrs) but they were refused
[01:45:19 CET] <mfg> alright. I'll assume it just busy. Thanks
[01:46:37 CET] <JEEB> also was going to say that it really depends on which component etc he had posted the patch :P
[01:47:03 CET] <memeka> when i tried them with a more recent ffmpeg version, i had an issue with the h264 decoder: my v4l4 m2m decoder accepts only annexB format
[01:48:01 CET] <memeka> so i added an annexB converter here: https://github.com/mihailescu2m/FFmpeg/commit/d012bc50211fbf8c2cd61624d71fc1dd52f421c3 -- however it didn't change anything, decoder still complains about annexB -- did i do something wrong with the converter?
[01:48:37 CET] <JEEB> check what the error is actually checking for
[01:48:39 CET] <memeka> converting an entire clip to ffmpeg to annexB and then playing it with the m2m decoder works fine
[01:49:17 CET] <memeka> so with or without the patch, converting to annexB then playing the clip works fine
[01:49:24 CET] <memeka> unconverted, it fails
[01:51:41 CET] <memeka> JEEB: -codec:v copy -codec:a none -bsf:v h264_mp4toannexb -f rawvideo
[01:51:49 CET] <memeka> if i add this it works
[01:52:14 CET] <memeka> i thought adding this: https://github.com/mihailescu2m/FFmpeg/commit/d012bc50211fbf8c2cd61624d71fc1dd52f421c3 would do the same thing
[01:54:58 CET] <memeka> JEEB: this is the entire command line that works: ffmpeg -i ./sintel_trailer_1080p.mp4 -codec:v copy -codec:a none -bsf:v h264_mp4toannexb -f rawvideo - | ffplay -an -vcodec h264_v4l2m2m -
[01:55:23 CET] <memeka>  h264_v4l2m2m decoder works if h264 is converted to annexB
[01:58:58 CET] <wm4> libav has autobsf for decoding
[01:59:46 CET] <memeka> libav doesn't do v4l2 m2m :(
[02:01:55 CET] <jamrial> durandal_1707: dbc7f02a727286f353624cf690cc6f430e240e25 broke fate-ffprobe
[02:05:50 CET] <jamrial> wm4: libav doesn't have autobsf. that's something rcombs wrote
[02:06:30 CET] <wm4> jamrial: wut
[02:06:44 CET] <wm4> "autobsf for decoding"
[02:13:16 CET] <memeka> so, anyone can help ... why won't my patch convert to annexB? :(
[02:14:10 CET] <JEEB> memeka: just check what the error you're getting inside the code is checking for
[02:14:18 CET] <JEEB> then check why that is getting triggered
[02:15:13 CET] <memeka> there is no error, i says that it was "successfully converted to annexb" (so av_bitstream_filter_filter is succefful) but then fails to decode
[02:15:36 CET] <JEEB> that's why I'm asking
[02:15:50 CET] <JEEB> "it still complains about annexb"
[02:16:04 CET] <JEEB> what and where complains, and check why
[02:16:25 CET] <JEEB> if you think you are doing it correctly, go check what that thing is checking for
[02:16:40 CET] <memeka> the decoder refuses the frame
[02:16:47 CET] <JEEB> either you will find that you have forgotten to do something, or the check is now incorrect
[02:16:49 CET] <memeka> it doesn't know how to decode it
[02:17:17 CET] <memeka> it's like trying to play a .txt file
[02:17:55 CET] <JEEB> I don't think it's that hard to grasp the concept of "it's complaining of X even though I'm supposedly handling that already - I might need to check what that complaining check is actually checking for"
[02:18:32 CET] <memeka> i was just wondering if i am doing annexB conversion correctly
[02:18:49 CET] <JEEB> and that either leads you to noticing that you have overlooked something, or that the check in that other part of code is incorrect
[02:18:55 CET] <memeka> it's the first time i look at the ffmpeg code, the m2m patches were not done by mw
[02:19:43 CET] <JEEB> yes, I understand, but that's the only way. except for dumping the results of your annex b conversion to a file and trying to read that as -f h264 with ffmpeg.c f.ex.
[02:19:52 CET] <JEEB> but even that you might get incorrectly
[02:20:01 CET] <JEEB> so better check the darn complaining check
[02:20:13 CET] <JEEB> it will either give you a hint of what you might be doing wrong
[02:20:16 CET] <JEEB> or just missing
[02:20:23 CET] <JEEB> or just plaing incorrect check at this point
[02:20:27 CET] <JEEB> *plain
[02:21:26 CET] <memeka> ok i will try to give more detailed info - i don't have access to the box atm so i couldn't give you the exact error
[02:21:49 CET] <JEEB> *I* don't need the exact error
[02:22:10 CET] <memeka> i noticed tho` that the first packet is not converted, only subsequent packets -- maybe that's my problem
[02:22:24 CET] <JEEB> you know more or less which kind of error you are getting and you have your current FFmpeg code base public it seems, so you have access to that
[02:22:40 CET] <JEEB> yes, that could very well be what the check's checking for :P
[02:23:03 CET] <JEEB> first packet(s) and if they match annex b or AVCc f.ex.
[02:24:29 CET] <memeka> i looked in the libav h264 decoder, and i don't see it converting to annexB either (https://github.com/libav/libav/blob/master/libavcodec/h264dec.c)
[02:24:55 CET] <memeka> so if it's done, it must be done somewhere else
[02:26:53 CET] <memeka> i see that ff_h2645_packet_split has argument is_avc, which i assume is related to annexB?
[06:37:21 CET] <durandal_1707> adeeln_: also add decription to libavcodec/codec_desc.c
[06:40:19 CET] <adeeln_> Done. Does XMP stand for Extreme Memory Profile?
[06:43:56 CET] <durandal_1707> probably no
[06:44:14 CET] <adeeln_> Extensible Metadata Platform?
[06:46:23 CET] <durandal_1707> X PixMap
[06:47:19 CET] <durandal_1707> its in link i sent you
[06:48:22 CET] <adeeln_> Oh, my bad.
[06:49:33 CET] <atomnuker> but its called XMP, it would have to be X MapPix which doesn't make sense
[06:52:18 CET] <adeeln_> Wait. What?
[06:54:06 CET] <adeeln_> It's Extensible Metadata Platform, I think. [https://github.com/fluent-ffmpeg/node-fluent-ffmpeg/issues/484]
[06:57:20 CET] <durandal_1707> lol, ita description wrong, should be XPM on trac
[06:58:06 CET] <adeeln_> Adobe XMP Core 5.5-c021 [https://trac.ffmpeg.org/ticket/4701]
[06:59:48 CET] <durandal_1707> XMP is metadata thing
[07:02:12 CET] <adeeln_> Read here [http://www.adobe.com/devnet/xmp.html], but I guess you already knew that.
[07:02:45 CET] <durandal_1707> it was so long ago that i forgot to type it properly
[07:04:02 CET] <adeeln_> Such silly mistakes...
[07:05:33 CET] <durandal_1707> you found first bug
[07:05:51 CET] <adeeln_> I don't think so.
[08:35:22 CET] <adeeln_> durandal_1707: What about the changes made in other files? like the changes made in parseutils.c on [https://ffmpeg.org/pipermail/ffmpeg-devel/2012-June/126647.html] are not available on GitHub[https://github.com/FFmpeg/FFmpeg/blob/master/libavutil/parseutils.c]. You only sent the difference file.
[09:12:49 CET] <adeeln_> So, I have created extracted the xmpdec.c file, shared here[https://gist.github.com/adl1995/fda8021a256396a62c9606cfc65e6314]. Pastebin blocked in my country. I have made changes in the origianal Makefile in /examples directory and linked this with it. 
[09:13:07 CET] <adeeln_> original*
[09:24:16 CET] <adeeln_> What should I do next?
[09:54:50 CET] <durandal_1707> get it build
[10:18:47 CET] <adeeln_> It's finally XPM :D
[16:48:43 CET] <michaelni> durandal_1707, dbc7f02a727286f353624cf690cc6f430e240e25 seems to break fate-ffprobe_*
[16:49:12 CET] <michaelni> http://fate.ffmpeg.org/report.cgi?time=20170226003803&slot=x86_64-netbsd-gcc45
[17:07:25 CET] <cone-272> ffmpeg 03Paul B Mahol 07master:3a003cc3817f: fate: update ffprobe stuff after dbc7f02a727286f3
[17:16:53 CET] <cone-272> ffmpeg 03Micah Galizia 07master:489c09ad4b1a: add locale month names to av_small_strptime
[17:16:54 CET] <cone-272> ffmpeg 03Michael Niedermayer 07master:0716bcce5bdc: avcodec/ituh263dec: Check for the bitstream end in ff_h263_decode_mb()
[17:16:55 CET] <cone-272> ffmpeg 03Michael Niedermayer 07master:aff8cf18cb0b: avcodec/mpeg12dec: Fix runtime error: left shift of negative value -2
[17:16:56 CET] <cone-272> ffmpeg 03Michael Niedermayer 07master:87eb3749708c: avcodec/eac3dec: Fix runtime error: left shift of negative value -3
[17:16:57 CET] <cone-272> ffmpeg 03Michael Niedermayer 07master:7e9ba78f6bd1: avcodec/flacdsp: Fix: runtime error: signed integer overflow: -1027555328 + -1226681270 cannot be represented in type 'int'
[21:12:13 CET] <BBB> michaelni: please specify what SUINT means, its very unclear to me
[21:14:03 CET] <BBB> michaelni: e.g. grep -r SUINT $srcdir is empty for me?
[21:14:22 CET] <kierank> https://patchwork.ffmpeg.org/patch/1792/
[21:16:11 CET] <BBB> Im going to vomit for a second, hold on - brb
[22:11:21 CET] <cone-272> ffmpeg 03Paul B Mahol 07master:6d856b257997: avcodec/scpr: add support for older version
[22:12:25 CET] <durandal_1707> RiCON: ^
[22:32:51 CET] <Compn> durandal_1707 : any chance you want to tackle vivo codecs? h263 based...
[22:33:16 CET] <JEEB> I should rebase some of my patches...
[22:33:26 CET] <JEEB> (unrelated to that specific format family)
[22:33:55 CET] <durandal_1707> Compn: have encoder and codec binary?
[22:34:43 CET] <durandal_1707> JEEB: what patches?
[22:34:52 CET] <JEEB> earthsoft dv demuxer
[22:35:00 CET] <JEEB> I got as far as being able to play back audio
[22:35:38 CET] <durandal_1707> good ;) being useful to community
[22:36:11 CET] <JEEB> aye :)
[22:36:18 CET] <JEEB> not something like the ISML shit :DDDDDDDD
[22:36:26 CET] Action: JEEB hates that shit with passion by now
[22:36:55 CET] <kierank> JEEB: i feel so bad
[22:36:56 CET] <kierank> for you
[22:37:46 CET] <JEEB> kierank: meanwhile I have a stash of booze for the time when someone tells me that I have to implement MMTP
[22:38:00 CET] <kierank> lol
[22:38:08 CET] <kierank> that thing
[22:38:13 CET] <kierank> in.sane
[22:38:35 CET] <JEEB> is it actually required by ATSC 3?
[22:38:49 CET] <Compn> durandal_1707 : vivoactive producer for encoder
[22:39:18 CET] <durandal_1707> you have it?
[22:40:16 CET] <durandal_1707> is it vfw kind?
[22:40:27 CET] <Compn> uhhhh
[22:42:32 CET] <Compn> hmmmmm
[22:42:35 CET] <Compn> not vfw i dont thin
[22:43:09 CET] <durandal_1707> just encode full white or black frame
[22:51:03 CET] <kierank> i wonder if I can troll durandal_1707 into finishing mpeg-4 sstp
[22:51:53 CET] <durandal_1707> mpeg4 is one big mess in lavc
[22:52:23 CET] <durandal_1707> kierank: whats missing?
[22:52:58 CET] <kierank> dpcm mode and painful integration into mpegvideo.c
[22:53:09 CET] <kierank> and samples
[22:53:48 CET] <durandal_1707> thats why i prefer working on stuff i can freely encode
[22:56:01 CET] <atomnuker> durandal_1707: write some encoders then, lol
[22:56:15 CET] <JEEB> lulzifications
[22:56:34 CET] <JEEB> since you'd still not know if your output would be interoperable
[22:56:47 CET] <durandal_1707> i wanted to add some rle encoders
[22:56:56 CET] <kierank> durandal_1707: https://wiki.videolan.org/Bounties/#Libavcodec_bugs
[22:57:30 CET] <kierank> or add missing cineform features
[22:57:58 CET] <durandal_1707> i hate sony
[22:58:29 CET] <kierank> cineform is not sony
[22:58:35 CET] <kierank> it is gopro
[22:58:40 CET] <JEEB> ^
[22:58:44 CET] <RiCON> durandal_1707: nice, thanks
[22:59:01 CET] <durandal_1707> cineform interlaced , looks easy 
[22:59:13 CET] <kierank> interlaced is easy yes
[22:59:21 CET] <kierank> just modify code to do the transform on each field
[22:59:32 CET] <kierank> but there are some samples which have some kind of p-frame
[22:59:34 CET] <kierank> that I never did
[22:59:49 CET] <durandal_1707> huh?
[23:00:37 CET] <durandal_1707> i still need to motivate myself to work on atrac9
[23:02:39 CET] <atomnuker> durandal_1707: I'll help you myself, just give me some quantized coefficients
[23:03:43 CET] <durandal_1707> i still need to find how to decompress them :)
[23:10:47 CET] <cone-272> ffmpeg 03Marton Balint 07master:e0eb0bdab948: avdevice/decklink_enc: convert to codecpar
[23:10:48 CET] <cone-272> ffmpeg 03Marton Balint 07master:48f8ad329010: avdevice/decklink_enc: add support to specify field order
[23:10:49 CET] <cone-272> ffmpeg 03Marton Balint 07master:418ce915071e: avdevice/decklink_enc: convert AVFMT_RAWPICTURE to AV_CODEC_ID_WRAPPED_AVFRAME
[23:10:50 CET] <cone-272> ffmpeg 03Marton Balint 07master:80b644c6ee3e: avdevice/decklink_dec: remove AVFMT_RAWPICTURE
[23:16:25 CET] <RiCON> durandal_1707: without using illegal at9tool you mean?
[23:17:21 CET] <Compn> probably he means something easy software to encode without installing 500mb program suite
[23:17:41 CET] <durandal_1707> psst
[23:18:43 CET] <Compn> durandal_1707 : do you encode on linux or windows ?
[23:18:48 CET] <RiCON> Compn: it's easily found, but also comes with headers and code
[23:18:52 CET] <RiCON> from sony
[23:19:11 CET] <durandal_1707> i downloaded >2gb just to find out wine crashes
[23:25:08 CET] <durandal_1707> could add dnxhr encoder as gsoc task :)
[23:25:55 CET] <atomnuker> >yet another 8x8 dct codec
[23:26:19 CET] <durandal_1707> it have alpha too
[23:26:56 CET] <durandal_1707> can be stored losslesly
[23:27:31 CET] <durandal_1707> or even raw - no mention in spec - how odd
[00:00:00 CET] --- Mon Feb 27 2017


More information about the Ffmpeg-devel-irc mailing list