[Ffmpeg-devel-irc] ffmpeg-devel.log.20190127

burek burek021 at gmail.com
Mon Jan 28 03:05:03 EET 2019


[04:34:08 CET] <agrecascino> hi i'm trying to fix a mess in mov_read_sidx and i'm not sure really what it's entirely _doing_
[04:35:40 CET] <agrecascino> like my brain is like, ok offset should be bounded within the limits of the start of the littlethingywhateveryoucallit and the that plus the size
[04:35:47 CET] <agrecascino> but no it's just kinda going everywhere
[04:36:24 CET] <agrecascino> and then mov_read_sidx declares it's done its day's work before even reading the last sidx thingywhatever and suddenly we're in an infinite loop
[04:37:10 CET] <agrecascino> like on line 5010 we have offset == avio_size(pb) and at this point i'm just really not sure what makes up offset
[04:37:59 CET] <agrecascino> like it's read from sidx right?
[04:44:21 CET] <agrecascino> like offset + first_offset + itemsizes = the file size is the exit condition but
[04:44:41 CET] <agrecascino> that apparently not correct?
[04:44:47 CET] <agrecascino> or at least some part of this isn't
[04:45:05 CET] <agrecascino> even if this is a malformed mp4 i feel like that shouldn't infloop ffmpeg
[04:45:21 CET] <agrecascino> which is even stranger since it was made by ffmpeg
[04:47:00 CET] <agrecascino> what even makes it stranger is that if you just comment out setting the frag index as complete
[04:47:01 CET] <agrecascino> it works
[04:47:12 CET] <agrecascino> so something is busted i really just don't know what
[04:51:54 CET] <agrecascino> ah no wait i got it
[04:52:05 CET] <agrecascino> i think you guys are adding the size of the block twice
[04:52:49 CET] <agrecascino> if this version of ffmpeg still works then
[04:52:53 CET] <agrecascino> i'm sorry this is so stupid
[04:58:41 CET] <agrecascino> nvm figured it out
[04:58:46 CET] <agrecascino> this is beyond stupid
[04:59:06 CET] <agrecascino> actually what am i saying nevermind for
[04:59:09 CET] <agrecascino> that's the dang issue
[13:55:32 CET] <cone-235> ffmpeg 03Paul B Mahol 07master:795af110f70b: avcodec: add ARBC decoder
[14:17:33 CET] <cone-235> ffmpeg 03Carl Eugen Hoyos 07master:8a788ff5d635: lavf/cafdec: Do not fail for unknown atoms with negative size.
[14:18:50 CET] <cone-235> ffmpeg 03Gyan Doshi 07master:b5b6f6ad59ef: avfilter/buffersrc: print relevant info when skipping filter reinit
[14:26:40 CET] <cone-235> ffmpeg 03Carl Eugen Hoyos 07master:d3a69438049b: lavf/subviewerdec: Skip leading BOM.
[17:01:22 CET] <agrecascino> how do i like
[17:01:27 CET] <agrecascino> submit a patch correctly
[17:01:49 CET] <agrecascino> i tried to send one and it hasn't showed up on the mailing list and i'm worried i did something wrong
[17:02:21 CET] <kierank> agrecascino: if you're not subscribed it needs manual approval
[17:02:30 CET] <agrecascino> ah
[17:02:47 CET] <agrecascino> also i messed up the title format, should i resend it with a fixed title?
[19:23:23 CET] <cand> how does one apply for commit access to ffmpeg? I searched, but found no docs
[19:24:05 CET] <durandal_1707> you need to be maintainer and have already commits in tree
[19:24:08 CET] <cand> bg: I've been sending power patches for a few months, but the interest is not high. Only cehoyos and michaelni have ever reviewed, and applying has all fallen on michael
[19:24:16 CET] <cand> http://ffmpeg.org/pipermail/ffmpeg-devel/2019-January/239357.html
[19:24:31 CET] <cand> I asked whether I should apply, and cehoyos told here I should
[19:25:01 CET] <JEEB> there's a maintainer file that contains a list of people
[19:25:14 CET] <JEEB> post a patch that adds yourself to the PPC side of things I guess?
[19:25:32 CET] <cand> atm I have 7 commits in the tree
[19:25:43 CET] <cand> oops, 5, 7 in my local tree
[19:25:51 CET] <JEEB> then poke michaelni if he trusts you and your key might get added somewhere
[19:26:16 CET] <JEEB> but yes, ppc in general is a very limited field :P
[19:26:28 CET] <JEEB> which is why f.ex. I haven't reviewed that stuff
[19:26:40 CET] <JEEB> I don't have things to test patches on, and I lack thematic knowledge
[19:26:59 CET] <cand> for my first such patch, michael said altivec is not his area either, and he'd be ok if someone else checked it out
[19:27:04 CET] <cand> naturally, no such someone else appeared
[19:27:21 CET] <JEEB> yup, unfortunately very few people have PPC (esp. power8+)
[19:28:40 CET] <cand> are there responsibilities for a maintainer? It's a heavy word, in that I can't guarantee I have time to review ppc things always, etc
[19:29:14 CET] <iive> cand, you might be asked to review ppc patches other people are sending.
[19:30:15 CET] <cand> that's totally fine, I was wondering about any guidelines or expectations
[19:31:28 CET] <cand> some old irc logs talked about commit frequency requirements, but they were many years ago
[19:31:31 CET] <JEEB> this is IMHO a general thing (not specific to maintainership) that if you break something, you should either fix it or revert it. and you are going to be the one probably CC'd onto relevant patches. the amount of activity required depends.
[19:32:46 CET] <cand> yep, so ffmpeg maintainership sounds a lot lighter than kernel, etc
[19:32:59 CET] <durandal_1707> you need to donate 5k bitcoins
[19:33:12 CET] <cand> aww crap, I only have 4k
[19:34:34 CET] <kierank> i have a power9 at work but it's being used for other things
[19:34:57 CET] <JEEB> and yes, being on IRC lets you get access to people who might not otherwise respond as easily on the ML for one reason or another
[19:35:20 CET] <JEEB> (busy and thus feel a bigger commitment when an e-mail has to be written f.ex.)
[19:36:04 CET] <JEEB> the patches get posted on the ML of course, that's (currently) the official way
[19:53:21 CET] <JEEB> ubitux: OK, I think I fixed the subtitle time bases - time to see if I can get anything out of this :)
[19:53:51 CET] <JEEB> AVSubtitle's pts to AV_TIME_BASE_Q, duration is 1, 1000
[19:54:20 CET] <JEEB> (talking of results of av_rescale_q of course
[20:16:14 CET] <JEEB> well, ffprobe looks all correct http://up-cat.net/p/588125e5
[20:23:00 CET] <agrecascino> hey
[20:23:06 CET] <agrecascino> i use power9
[20:23:11 CET] <agrecascino> power9 solidarity!
[20:23:51 CET] <JEEB> and webvtt with -copyts looks OK
[20:24:35 CET] <JEEB> (ASS just lacks timestamps over 10 hours)
[21:26:42 CET] <JEEB> ubitux: ok, now I know why mpv wasn't working - the ASS subtitle specific code wasn't considering duration == 0 as unknown duration from the AVPacket
[21:26:57 CET] <JEEB> jesus christ I wasted time on this
[23:49:07 CET] <cone-653> ffmpeg 03Mark Thompson 07master:44bcccb7f0c4: vaapi_encode_h265: Ensure that ref pics are always in the RPS
[00:00:00 CET] --- Mon Jan 28 2019


More information about the Ffmpeg-devel-irc mailing list