[Ffmpeg-devel-irc] ffmpeg-devel.log.20180203

burek burek021 at gmail.com
Sun Feb 4 03:05:03 EET 2018


[00:17:45 CET] <atomnuker> huh, a patch from someone with a @google.com address
[00:18:19 CET] <BtbN> there are a lot of them
[00:19:02 CET] <atomnuker> last one I remember was the lavf mp4 editing thing
[00:26:19 CET] <Compn> atomnuker : google has several places they use ffmpeg in. at least google video, google docs, google drive... also whatever security fuzz oss testing they are doing in secret
[00:26:24 CET] <Compn> plus youtube team
[00:28:43 CET] <jamrial_> And Chromium
[00:29:54 CET] <Compn> oh yeah them :D
[00:30:09 CET] <Compn> and android
[00:30:37 CET] <JEEB> chromium seems like the bigges thing, which also seems to have prompted the fuzzing
[00:55:37 CET] <cone-117> ffmpeg 03Jérôme Martinez 07master:e3d946b3f48f: avcodec/ffv1enc: mark RGB48 support as non-experimental
[01:41:29 CET] <Chloe> why is this guy trying to optimise iteration of components
[01:41:42 CET] <Chloe> is O(k) lookup really that bad
[02:04:16 CET] <SortaCore> I guess it's not O(k) with him
[02:04:26 CET] <DHE> *groan*
[12:56:36 CET] <wm4> michaelni: I've caught swresample corruption memory/crashing with very large compensation distance in mpv (not sure if my fault or libswresample)
[13:02:53 CET] <michaelni> will look later, but how can i reproduce it ?
[13:27:10 CET] <wm4> michaelni: mpv 196950e905, mpv something.mkv --no-config --audio-pitch-correction=no, hit [ very quickly during playback (or just keep it pressed)
[13:30:58 CET] <wm4> also resampling with an extreme ration (100x smaller output rate than input) and then feeding a small frame leaves you with no returned sample data, even if the intention is to drain the resampler instead of feeding more data into it
[14:10:21 CET] <JEEB> was there a patch posted for libaom for FFmpeg yet?
[14:10:28 CET] <JEEB> I can find one for the IVF muxer but that's it
[14:10:55 CET] <RiCON> JEEB: not for ffmpeg, only for libav, iirc
[14:11:47 CET] <JEEB> right
[14:16:43 CET] <jamrial> JEEB: BBB ported/updated it here https://github.com/rbultje/ffmpeg/commits/libaom_av1_decoder
[14:23:06 CET] <JEEB> cool
[14:40:46 CET] <JEEB> jamrial: nice, I just forward-ported it to current HEAD, let's see if this works with the latest libaom
[14:47:57 CET] <JEEB> seems to work, nice
[14:54:46 CET] <JEEB> oh huh
[14:55:00 CET] <JEEB> it just silently took my enable-libaom
[14:55:08 CET] <JEEB> but then failed in the check :/
[14:55:55 CET] <RiCON> configure seems to do that a lot
[14:56:11 CET] <wm4> configure turns out to be a fragile mess!
[14:56:36 CET] <RiCON> or silently disable explicitely enabled libs missing some deps
[15:02:30 CET] <JEEB> ok, now I know why it fails
[15:02:40 CET] <JEEB> it doesn't add the include flags for some reason from pkg-config
[15:03:15 CET] <JEEB> right, the pkg-config based check is disabled in the patch
[15:06:55 CET] <jamrial> it's using check_lib() instead of require() or require_pkg_config()
[15:07:01 CET] <jamrial> which doesn't abort on failure
[15:07:17 CET] <JEEB> yup
[15:10:15 CET] <wm4> rcombs: FOSDEM just had a stream with a talk about HEIF, and there was someone who asked the presenter "are you happy with the format?" and mentioned your patch
[15:11:05 CET] <kierank> wm4: you should come to FOSDEM
[15:11:14 CET] <wm4> too late
[15:11:43 CET] <nevcairiel> FOSDEM somehow always happens to be on a vacation weekend for me, I just back back an hour ago from denmark, last year I was in australia =p
[15:12:01 CET] <nevcairiel> got back*
[15:13:00 CET] <JEEB> grr, need to remember the parameter order for require_pkg_config
[15:13:31 CET] <nevcairiel> just do what everyone does and copy some other library
[15:13:33 CET] <JEEB> ok, the internal name, the pkg-config name
[15:13:45 CET] <JEEB> and header and func
[15:13:47 CET] <JEEB> ooh-kay
[15:16:07 CET] <wm4> man HLS live streams always either fall behind or underrun (due to being too far ahead) for me
[15:16:49 CET] <JEEB> yayifications, that worked
[15:16:55 CET] <JEEB> let's see if I can make it now
[15:17:08 CET] <JEEB> muh battery
[15:20:35 CET] <JEEB> ok, now I finally got to libaom API having changed :)
[15:20:41 CET] <JEEB> and my battery is dying
[15:20:47 CET] <JEEB> so I guess this is most of it for today
[15:29:25 CET] <jamrial> so the wrapper is already outdated? :p
[15:29:43 CET] <JEEB> yea
[15:29:51 CET] <JEEB> probably just arg count thing
[15:30:08 CET] <jamrial> no point bothering then until they tag a stable release
[15:31:13 CET] <JEEB> yea
[15:31:56 CET] <BBB> JEEB: I disabled the pkg-config because Im using uninstalled libs
[15:31:59 CET] <BBB> sorry for not documenting that
[15:32:08 CET] <BBB> my use case is kind of weird that way
[15:32:29 CET] <JEEB> BBB: it was quick so np
[15:35:15 CET] <atomnuker> Tempted to see how many frames per day I get with using libaom on my first gen rpi but don't want to contribute industrial amounts to global warming
[15:35:49 CET] <RiCON> wouldn't the rpi be more "frame per month"?
[15:39:40 CET] <TD-Linux> realtime with rav1e :^)
[16:05:40 CET] <JEEB> ok, this seems to work with current libaom HEAD https://github.com/jeeb/ffmpeg/commits/libaom_av1_decoder
[16:22:41 CET] <funman> atomnuker: you did not upload your slides on fosdem?
[17:29:38 CET] <durandal_1707> ffmpeg description should also contain word filter, not just stream and convert
[17:37:17 CET] <BBB> JEEB: ty
[17:37:32 CET] <BBB> TD-Linux: x264 preset=ultrafast is also realtime, doesnt mean its good :-p
[17:39:01 CET] <RiCON> even medium is more or less realtime, iirc
[17:41:20 CET] <DHE> that's highly CPU dependent though. my core i7, sure for not-too-busy action
[18:03:59 CET] <atomnuker> funman: I did, check the talk's page
[18:42:00 CET] <rcombs> wm4: lol link?
[18:43:23 CET] <wm4> rcombs: I think recordings come out later
[18:43:28 CET] <rcombs> ah
[18:43:56 CET] <rcombs> I'm internet famous \o/
[18:44:07 CET] <wm4> he didn't say your name
[18:44:16 CET] <wm4> just that your patch didn't work with most of their samples
[18:44:29 CET] <rcombs> well yeah, 'cause tiling
[18:44:39 CET] <rcombs> always knew bitching about shitty format design would get me places
[18:45:02 CET] <durandal_1707> nobody mentioned me? :(
[18:45:23 CET] <rcombs> have you bitched about shitty format design in an HEIF patch's commit message
[18:55:22 CET] <durandal_1707> atomnuker: dont get drunk!
[19:14:24 CET] <KGB> [13FFV1] 15dericed opened pull request #102: support method to output lines only to version0,1,3 or version4 (06master...06split-ffv1-version-outputs) 02https://git.io/vNxce
[22:53:52 CET] <cone-481> ffmpeg 03Nikolas Bowe 07master:c86490c49a3d: avfilter/af_join: Fix crash in join filter
[22:53:52 CET] <cone-481> ffmpeg 03Michael Niedermayer 07master:8e50bd61e4ff: avcodec/wavpack: Fix integer overflow in FFABS
[00:00:00 CET] --- Sun Feb  4 2018


More information about the Ffmpeg-devel-irc mailing list