[Ffmpeg-devel-irc] ffmpeg-devel.log.20160717

burek burek021 at gmail.com
Mon Jul 18 02:05:02 CEST 2016


[04:48:43 CEST] <omerjerk> Hi,
[04:49:03 CEST] <omerjerk> is there a way of reading bits in the reverse order using something in get_bits.h ?
[13:02:55 CEST] <firewated> Hi, I've been trying to get HLS subtitles reading working in ffmpeg and although it's working, when the stream is probed for format detection it detects as WEBVTT with a score of 1 because I strip the WEBVTT header that appears at the beginning of every segment. The WEBVTT headers need to be removed because otherwise they will appear throughout the stream
[13:02:55 CEST] <firewated> and be treated as subtitles. Any ideas on how to make it more robust?
[13:51:23 CEST] <Compn> the probe or removing of webvtt headers ?
[13:51:33 CEST] <Compn> either case i dunno, firewated. stick around for answers :)
[13:54:58 CEST] <rcombs> u wot m8
[13:55:06 CEST] <rcombs> HLS subtitles are meant to have the WEBVTT header
[14:11:33 CEST] <firewated> Yeah but if I leave them in the stream they'll be treated as subs and appear in the output, once for every segment in the HLS playlist, they need to be removed
[14:12:51 CEST] <nevcairiel> sounds like thats the bug you should fix instead
[14:13:36 CEST] <firewated> Maybe I'm not explaining it well but I don't think that's a bug
[14:13:49 CEST] <firewated> I'm going to try and leave the first segment alone and strip the headers from the rest of the segments and see how that works
[14:14:04 CEST] <nevcairiel> isnt hls also supposed to support random access
[14:14:19 CEST] <nevcairiel> ie. repeat headers in every segment for that purpose
[14:14:23 CEST] <JEEB> yes
[14:14:35 CEST] <JEEB> same for PPS/SPS etc in video
[14:15:16 CEST] <JEEB> so basically the parser should handle it that each subtitle segment has the headers and that thos aren't subtitles
[14:15:22 CEST] <rcombs> yeah, if your player is displaying the "WEBVTT" text as subtitles then you have a bug
[14:15:25 CEST] <rcombs> and you should fix that
[14:15:46 CEST] <rcombs> each segment should have that header
[14:18:37 CEST] <firewated> To be clear, ffmpeg doesn't currently read HLS subtitles, it's something I'm trying to add, I'm testing it by using ffmpeg to read a HLS playlist and output a mp4 file with subtitles
[14:19:00 CEST] <JEEB> yes, which is why we noted that it's most probably your parser handling the stuff incorrectly :)
[14:24:38 CEST] <firewated> Yeah, I'm going to try stripping the header from all the segments except the first, just wanted to get a feel for the IRC and what not, first time here :)
[14:25:50 CEST] <rcombs> that is not the correct solution
[14:26:05 CEST] <firewated> What do you suggest?
[14:33:33 CEST] <rcombs> fix whatever's treating "WEBVTT" as part of the subtitle text to not do that
[14:34:04 CEST] <rcombs> sounds like something's just concatenating segments as text files instead of demuxing them as independent files and concatenating their packets
[14:34:11 CEST] <rcombs> that's incorrect behavior for HLS subs
[15:05:04 CEST] <firewated> That sounds better, I'll look into it, thanks
[15:08:02 CEST] <omerjerk> hey how do i calculate crc of a test file to put it inside tests/ref/fate/
[15:08:33 CEST] <JEEB> the test will be outputting the values it got
[15:08:48 CEST] <JEEB> and telling you there's nothing there in the file you're comparing against
[15:08:57 CEST] <JEEB> that kind of does give you the values, no? ;)
[15:09:07 CEST] <nevcairiel> you can use make fate-my-test GEN=1 to generate the files
[15:09:16 CEST] <rcombs> oh huh
[15:09:17 CEST] <rcombs> TIL
[15:09:30 CEST] <omerjerk> I didn't get it. :|
[15:09:43 CEST] <omerjerk> I've to add a new file
[15:11:06 CEST] <omerjerk> I did this - https://github.com/omerjerk/FFmpeg/commit/e172e333807b4b3b2558a1ffa735ade79a3f3e36
[15:11:44 CEST] <omerjerk> so I also need to add the crc of this new file. 
[15:11:48 CEST] <omerjerk> I have it locally as well
[15:12:10 CEST] <JEEB> shouldn't we use something better than CRC for stuff like that nowadays?
[15:12:25 CEST] <JEEB> I think there was at least md5 (frame/full stream) available
[15:12:37 CEST] <omerjerk> I have no idea.
[15:12:48 CEST] <omerjerk> trac.ffmpeg doesn't mention anything
[15:14:38 CEST] <JEEB> https://trac.ffmpeg.org/wiki/framemd5%20Intro%20and%20HowTo
[15:15:10 CEST] <JEEB> you can git grep the fate tests for usage examples
[15:18:59 CEST] <omerjerk> so I got the framemd5 output. Now it just needs to go inside tests/ref/fate/mpeg4-als-conformance-07 ?
[15:19:57 CEST] <nevcairiel> we told you, just run "make fate-whateveryourtestiscalled GEN=1" and it generates the file
[15:20:24 CEST] <omerjerk> okayy
[15:21:35 CEST] <omerjerk> just to be sure, what should "whateveryourtestiscalled" be in my case?
[15:21:50 CEST] <nevcairiel> how would we know, its your test
[15:22:48 CEST] <JEEB> omerjerk: what fate outputs when it runs your test ;)
[15:22:53 CEST] <JEEB> also how you defined it :)
[15:24:47 CEST] <omerjerk> how do I run my test? that's the problem. 
[15:24:56 CEST] <omerjerk> :/
[15:25:14 CEST] <omerjerk> tried with make fate-mpeg4-als-conformance-07, doesn't work
[15:26:05 CEST] <JEEB> there's not too many als tests so I wonder if something a la make fate-als would work?
[15:27:29 CEST] <omerjerk> okay. it's doing something. with make-als
[15:29:16 CEST] <omerjerk> it did not run for  mpeg4-als-conformance-07 
[15:29:34 CEST] <omerjerk> let me see what's the problem 
[15:29:55 CEST] <omerjerk> where do we define which files to download?
[15:30:11 CEST] <omerjerk> in make fate-rsync
[15:30:13 CEST] <JEEB> omerjerk: probably because you defined it after the foreach?
[15:30:31 CEST] <JEEB> looking at https://github.com/omerjerk/FFmpeg/commit/e172e333807b4b3b2558a1ffa735ade79a3f3e36
[15:30:38 CEST] <omerjerk> I don't think that should be the problem. let me just see again.
[15:30:42 CEST] <JEEB> oh wait, no
[15:30:47 CEST] <JEEB> yeah that probably isn't an issue
[15:30:55 CEST] <JEEB> it just loops through the first ALS samples
[15:35:46 CEST] <omerjerk> so the problem is that there is no als_07_2ch192k32bF.mp4 inside fate-suit/lossless-audio
[15:36:29 CEST] <JEEB> when your test is fully done and you're sure it's actually correct
[15:36:39 CEST] <JEEB> you will ask it to be uploaded to the FATE servers
[15:36:54 CEST] <JEEB> then after 24h or so the thing can be merged code-wise
[15:38:26 CEST] <omerjerk> oh okay.
[15:40:20 CEST] <omerjerk> btw I'm at where I started now. 
[15:40:36 CEST] <omerjerk> how do I generate the crc/md5 for the patch?
[15:40:38 CEST] <omerjerk> :|
[15:41:34 CEST] <JEEB> you put the file there?
[15:41:40 CEST] <JEEB> locally
[15:41:43 CEST] <JEEB> and run the test
[15:45:33 CEST] <omerjerk> tried that. still doesn't run for 07
[15:46:48 CEST] <omerjerk> ah. found the problem. had an error in the als.mak file
[15:46:54 CEST] <omerjerk> thanks for all the help. :)
[15:48:47 CEST] <omerjerk> all fine now. got the crc. 
[15:49:48 CEST] <Chloe> In the HEVC decoder the function name "transform_add" is fairly unclear, it'd make more sense to have it as "add_residual" or similar; what do people think of this?
[17:50:16 CEST] <cone-829> ffmpeg 03Jörn Heusipp 07master:f1eb6ddcb34f: libavformat/libopenmpt: Fix memory leak in error path in read_header_openmpt().
[18:07:26 CEST] <bencoh> .53
[18:07:29 CEST] <bencoh> woops
[18:12:36 CEST] <cone-829> ffmpeg 03James Almer 07master:d688f39dc4bc: cosmetics: fix some misspelled words
[19:14:16 CEST] <FishPencil> So OpenH264 is compatible with the LGPL 2.1?
[19:15:11 CEST] <nevcairiel> its BSD licensed, so yes
[19:15:18 CEST] <FishPencil> nice
[19:15:30 CEST] <nevcairiel> but it supports only a very small subset of h264 streams
[19:30:11 CEST] <FishPencil> What a pain to compile
[20:15:39 CEST] <cone-829> ffmpeg 03Michael Niedermayer 07master:af13967b6451: MAINTAINERS: Add Aman Gupta for ccaption_dec
[20:46:57 CEST] <ubitux> michaelni: re:2b14204 http://sprunge.us/DCQg 
[20:47:02 CEST] <ubitux> i guess "consistency" but well
[20:55:57 CEST] <jamrial> ubitux: any decent compiler will make that optimization. the thing is that there used to be a few that apparently didn't
[20:56:21 CEST] <jamrial> at this point it's mostly a cosmetic thing for consistency with the rest of the codebase, i'd say
[21:27:19 CEST] <cone-829> ffmpeg 03Aman Gupta 07master:309322ef65ab: avcodec/ccaption_dec: change write_char() to void as return value is unused
[21:29:56 CEST] <michaelni> ubitux, i didnt expect it to make a difference with gcc -O3 but i tend to always replace % by & if i can because the auto replace only works for unsigned or proofen to be positive values
[21:30:41 CEST] <michaelni> and keeping a mix of & and % depening on gcc being able to auto replace % is more work and easier to get wrong
[21:30:45 CEST] <michaelni> than replacing all
[23:39:08 CEST] <cone-829> ffmpeg 03James Almer 07master:64d16fd7f5f3: tools/patcheck: add a couple more common typos
[00:00:00 CEST] --- Mon Jul 18 2016


More information about the Ffmpeg-devel-irc mailing list