[Ffmpeg-devel-irc] ffmpeg-devel.log.20150828

burek burek021 at gmail.com
Sat Aug 29 02:05:02 CEST 2015


[00:28:53 CEST] <cone-386> ffmpeg 03Zhang Rui 07master:92d378067ee3: ffplay: remove unused include libavutil/colorspace.h
[00:28:54 CEST] <cone-386> ffmpeg 03Michael Niedermayer 07master:e2b19a533dbb: avformat/segment: atomically update list if possible
[01:27:26 CEST] <cone-386> ffmpeg 03rogerdpack 07master:832b4a4a438b: configure: Print out enabled programs
[01:56:32 CEST] <BtbN> philipl, yeah. The comments in the libva headers actualy sound doable. One flag shouldn't matter at all though, as it only affect temporal layering.
[04:16:57 CEST] <cone-386> ffmpeg 03Michael Niedermayer 07master:3322f0d4158f: avcodec/dnxhddata: Fix inconsistent table entry
[04:42:17 CEST] <cone-386> ffmpeg 03Donny Yang 07master:51ca70322296: apng: Support inter-frame compression
[05:58:16 CEST] <rcombs> https://gist.github.com/rcombs/ceb5473baa329ff8153c wat
[05:58:26 CEST] <rcombs> so& the init function locks the roll-your-own-mutex, then loads the .so and initializes that, which... calls the first init function again, which waits forever to take the mutex
[06:04:35 CEST] <cone-386> ffmpeg 03Carl Eugen Hoyos 07master:b009d5b1f73c: configure: Do not let the webm muxer suggest external libraries.
[06:06:34 CEST] <cone-386> ffmpeg 03Thilo Borgmann 07master:2392da164a02: Changelog: Mention the change of the default webm codecs.
[06:37:38 CEST] <rcombs> apparently this happens if you compile mfx_dispatch as a dynamic lib
[07:33:00 CEST] <philipl> rcombs: well, you tried to use mfx, so you're already on the losing side.
[07:34:00 CEST] <rcombs> philipl: in the end, I succeeded!
[07:35:03 CEST] <philipl> well, congratulations. What did you encode? :-)
[07:35:13 CEST] <rcombs> random file I had laying around
[07:35:41 CEST] <rcombs> was mostly just testing to see if I could build it without too much pain
[07:35:48 CEST] <rcombs> and now that I know how, yes, I can
[07:35:54 CEST] <philipl> I think the main challenge there is getting the SDK from intel.
[07:36:31 CEST] <rcombs> that's not too difficult either
[07:36:40 CEST] <rcombs> but again, you need to know where to find it
[07:36:52 CEST] <rcombs> and you need to give them an email address (though it doesn't verify it)
[07:37:20 CEST] <philipl> I haven't succeeded but I also didn't try very hard
[07:37:36 CEST] <rcombs> and then you download like 500MB worth of crap, half of which is open-source stuff that's already on GitHub or elsewhere
[07:38:06 CEST] <rcombs> and open 3 layers deep of .tgz files and navigate an obnoxious dir structure until you finally find the 2 .so's you wanted
[07:38:14 CEST] <rcombs> (really you only need 1 of them)
[07:38:44 CEST] <rcombs> and then you can finally do what you could already have done directly via vaapi if you were a bit more masochistic
[07:39:01 CEST] <philipl> Masochistic enough that no one's written that code and probably never will.
[07:39:12 CEST] <rcombs> I have!
[07:39:23 CEST] <philipl> You wrote vaapi encoding code?
[07:39:30 CEST] <philipl> I'm sorry.
[07:39:43 CEST] <rcombs> I just hacked the sample code into lavc for some testing a while back
[07:40:17 CEST] <philipl> Definitely masochistic.
[07:40:37 CEST] <philipl> Still, I'm glad that someone who doesn't work for Intel has proven this qsv stuff works on linux.
[07:40:48 CEST] <philipl> your sacrifice is appreciated.
[07:41:05 CEST] <rcombs> the libmfx stuff is overall less pain
[07:41:27 CEST] <rcombs> but I think could probably have more things go wrong
[07:41:49 CEST] <rcombs> extra library layer doing weird Intel things, loading proprietary libs&
[07:42:03 CEST] <rcombs> and super awkward from a licensing standpoint
[07:51:24 CEST] <philipl> yeah, makes it all unusable in practice.
[07:51:36 CEST] <philipl> although handbrake somehow got into a position to ship with qsv on windows.
[07:55:09 CEST] <rcombs> my biggest concern is whether or not it's actually GPL-compatible
[08:00:25 CEST] <philipl> The Handbrake folks seem quite careful about that (they won't accept nvenc for sure). Maybe they got some special licencing language from Intel.
[09:56:04 CEST] <wm4> j-b: will it be sufficient if I arrive on Saturday 9 AM at the airport in Paris?
[10:10:33 CEST] <nevcairiel> with airports being how they are, you will be slightly late and miss the first event or something, i would s ay
[11:10:55 CEST] <cone-534> ffmpeg 03Carl Eugen Hoyos 07master:24e0b14c4f8b: Changelog: Clarify that the new asf demuxer is optional.
[11:48:48 CEST] <j-b> wm4: might be a bit short
[11:48:52 CEST] <j-b> wm4: but doable.
[11:58:08 CEST] <wm4> I think I'd prefer this over arriving at 10 pm the day before
[13:06:43 CEST] <cone-534> ffmpeg 03Harshit Mittal 07master:53bf32fa4270: doc/examples/filtering_video: better demo ffmpeg filters; demos chaining the filters
[13:15:51 CEST] <BtbN> philipl, so, what about going for 2.8 now, as vaapi hevc is also in?
[13:17:11 CEST] <wm4> 2.8, as in making a new ffmpeg release?
[13:17:15 CEST] <BtbN> yes
[13:17:32 CEST] <wm4> it should be synchronized with Libav if possible
[13:17:36 CEST] <nevcairiel> isnt the bump from libav planned to be soon
[13:17:56 CEST] <nevcairiel> either we still do it shortly before and then one later in sync with libav, or we just wait until the bump now
[13:18:05 CEST] <BtbN> I don't realy follow libav development. If some major bump is planned there, that'd definitely be a good point for the next release.
[15:32:58 CEST] <wm4> what's happening on VDD on sunday?
[15:33:07 CEST] <wm4> in fact, what's happening on VDD at all?
[15:33:18 CEST] <durandal_170> beer
[15:33:36 CEST] <Daemon404> wm4, vdd is a few talks, then an unconference
[15:33:40 CEST] <Daemon404> hacking rooms, etc
[15:33:52 CEST] <Daemon404> schedule generally updated in realtime on the wiki
[15:33:59 CEST] <j-b> Sat morning is going to be talks
[15:34:01 CEST] <Daemon404> e.g. one year, there was an opus-in-mp4 room
[15:34:04 CEST] <Daemon404> where that was hashed out
[15:34:13 CEST] <j-b> It might be also a bit Sat afternoon
[15:34:37 CEST] <j-b> Sat Afternoon then there is the VLC technical meeting with in // the other technical meetings by projects
[15:34:40 CEST] <j-b> then unconference
[15:34:50 CEST] <wm4> unconference = beer?
[15:34:51 CEST] <j-b> 18h30 is the VideoLAN org meeting, with the elections
[15:34:59 CEST] <j-b> and then whole Sunday is unconference
[15:35:22 CEST] <Daemon404> wm4, unconferenc is like rooms and stuff
[15:35:23 CEST] <Compn> wm4 : no, its more like 'hang out in office and chat'
[15:35:27 CEST] <Daemon404> look at previous years wiki
[15:35:30 CEST] <Daemon404> for examples
[15:36:06 CEST] <Compn> wm4 : some companies use this time to recruit developers into jobs :P
[15:36:06 CEST] <j-b> in theory, people list the topics
[15:36:24 CEST] <j-b> well, if someone needs a job, I know plenty
[15:36:26 CEST] <Compn> company spies at vdd no doubt :P
[15:36:37 CEST] <j-b> not to mention that I'd like to hire everyone of you
[15:36:42 CEST] <j-b> well, except Compn 
[15:36:53 CEST] <wm4> use Compn to deal with your enemies?
[15:36:58 CEST] <Compn> i do not work well with others. :P
[15:37:01 CEST] <Daemon404> Compn should be the ideal hitman
[15:37:04 CEST] <Daemon404> he lives in detroit
[15:37:08 CEST] <Daemon404> gun expert
[15:37:09 CEST] <Compn> raycist.
[15:37:14 CEST] Action: Daemon404 runs
[15:37:15 CEST] <durandal_170> when will be next irc meeting?
[15:37:33 CEST] <Compn> i do not like guns, too loud 
[15:38:03 CEST] <Daemon404> the loudness bothers you, but not the kill-y stuff?
[15:38:12 CEST] <Compn> durandal_170 : irc meeting for what
[15:38:23 CEST] <wm4> where's that VDD wiki?
[15:39:29 CEST] <Compn> wm4 : https://wiki.videolan.org/VDD14/Sessions/
[15:39:39 CEST] <Daemon404> s/14/XX/ for otehr years
[15:39:57 CEST] <wm4> Daemon404: yeah, but I didn't find it anyway
[15:41:01 CEST] <kierank> Compn: nobody tried to recruit anyone
[15:41:46 CEST] <Compn> i saw youtube talking to and offering jobs to people 
[15:42:58 CEST] <Daemon404> that's because youtube has The Google PRoblem
[15:43:06 CEST] <Daemon404> hiring PHDs who cant code for real
[15:43:08 CEST] <Daemon404> or lack common sense
[15:43:18 CEST] <Compn> mencoder jobs :P
[15:43:22 CEST] <Compn> bwahaha
[15:43:32 CEST] <Daemon404> "they dont know the difference between pts and dts" -- actual quote from thierry
[15:43:40 CEST] <wm4> but those PHDs are good in tricky interview puzzles?
[15:43:43 CEST] <kierank> youtube was not hiring
[15:43:51 CEST] <Daemon404> kierank, not now, no
[15:43:51 CEST] <kierank> they were just talking to reimar about the binary loader
[15:43:58 CEST] <kierank> whatever that may or may not do at youtube
[15:44:01 CEST] <Daemon404> however 50 otehr departments of google are
[15:44:10 CEST] <Daemon404> i get regular spam from some tv startup thei acquired
[15:44:13 CEST] <Daemon404> they*
[15:44:15 CEST] <kierank> which one
[15:44:37 CEST] <nevcairiel> Didn't get a Google offer this year yet
[15:44:44 CEST] <Daemon404> their was an artice lately...
[15:44:45 CEST] <Daemon404> cant remember
[15:44:49 CEST] <Daemon404> they open source it recently i think
[15:45:10 CEST] <nevcairiel> SageTV?
[15:45:21 CEST] <Daemon404> yeah that.
[15:46:11 CEST] <nevcairiel> The old leader of that project contacted me once about recruitment, but I dont know for what exactly at google
[15:46:23 CEST] <nevcairiel> Didn't really bother to find out
[16:16:41 CEST] <nevcairiel> So libav pushed their bump, shall I just merge it as is and enjoy flame posts after? =p
[16:20:19 CEST] <wm4> of course
[16:20:45 CEST] <nevcairiel> the only really controversial changes seem to be the pix_fmts, which are truely butt-old
[16:20:50 CEST] <nevcairiel> and the avcodec frame API
[16:21:29 CEST] <nevcairiel> both are trivial replacements, at the very least
[16:21:56 CEST] <aballier> branch 2.8 before if there's anything you to be want widely used since 2.7
[16:22:16 CEST] <aballier> want to be*
[16:23:06 CEST] <nevcairiel> The major FFmpeg users will adapt anyway, all the flaming is about all those things noone even knows about it exists
[16:23:58 CEST] <aballier> distributions dont update until all those things noone knows are good or killed :)
[16:24:01 CEST] <aballier> s/dont/cant/
[16:24:17 CEST] <aballier> so this usually gives a few months delay
[16:24:25 CEST] <aballier> at least
[16:25:06 CEST] <aballier> (and i excpect major ffmpeg users to have already adapted to be fair)
[16:25:10 CEST] <BtbN> If the merge is easy and doesn't break a lot of API, i'd just go for it.
[16:25:37 CEST] <nevcairiel> it intentionally removes a bunch of old deprecated shit
[16:26:20 CEST] <wm4> also this was discussed to death
[16:29:19 CEST] <aballier> im just pointing facts: what this implies, not discussing whether its good or not (which i'm more in favor of good anyway)
[16:29:54 CEST] <Daemon404> nevcairiel, the pix fmts are also the easiest to fix
[16:29:58 CEST] <Daemon404> literally one line of sed
[16:30:03 CEST] <nevcairiel> <nevcairiel> both are trivial replacements, at the very least
[16:30:09 CEST] <Daemon404> yea.
[16:34:20 CEST] <wm4> Daemon404: still too hard for debian
[16:34:23 CEST] Action: wm4 runs
[16:34:31 CEST] <wm4> actually there's nothing to run from
[16:50:58 CEST] <philipl> BtbN: waiting for the bump seems reasonable. Is this (equivalent to) the 57 bump?
[16:51:13 CEST] <philipl> Have we finished marking everything we wanted for deprecation?
[16:51:24 CEST] <Daemon404> philipl, this stuff has been deprecated for years...
[16:51:44 CEST] <philipl> Of course.
[16:51:59 CEST] <wm4> also that means shit I have missed the window to attempt to deprecate and provide replacements for some pet peeves of mine, like AVPicture
[16:52:19 CEST] <philipl> Ok, I mean two things. 1) Anything like old-vdpau which wasn't properly marked. 2) Is anyone still fighting any of the deprecations?
[16:53:06 CEST] <philipl> wm4: There's always next time :-P
[16:53:26 CEST] <Daemon404> what acpicture stuff even needs replacements besides avpicture_layout?
[16:54:00 CEST] <wm4> AVSubtitle
[16:54:12 CEST] <Daemon404> nah.
[16:54:44 CEST] <nevcairiel> the window for marking things is still open until the first release, imho
[16:55:00 CEST] <Daemon404> indeed
[16:55:25 CEST] <philipl> nevcairiel: So just to make sure I'm in sync, the libav bump is the bump that we were busy preparing for the last few weeks? Not the one after that one right? :-)
[16:55:37 CEST] <nevcairiel> yes
[16:55:45 CEST] <philipl> K. All good then.
[17:00:52 CEST] <cone-534> ffmpeg 03Paul B Mahol 07master:9f2fa95bd8e8: avfilter/vf_histogram: 9 and 10 bit depth support
[17:08:51 CEST] <wm4> a concise bug report
[17:09:12 CEST] <wm4> durandal_170: you should write what it's needed for
[17:10:19 CEST] <durandal_170> for exr and to rule them all
[17:16:55 CEST] <cone-534> ffmpeg 03Philip Langdale 07master:1e50f953fac7: Changelog: Add VDPAU HEVC to the list
[17:36:38 CEST] <wm4> dammit arranging flights is so annoying
[17:37:05 CEST] <Daemon404> trains are an option too
[17:37:17 CEST] <wm4> 2x-3x slower
[17:37:27 CEST] Action: Daemon404 will have flown 65000 miles, and 53 flights at the end of 2015
[17:37:28 CEST] <Daemon404> so.
[17:38:06 CEST] <Daemon404> i assume you'd grab a lufthansa or germanwings flight from $germancity
[17:39:55 CEST] <wm4> yep, and which is 1 hour away from where I live
[17:40:54 CEST] <Daemon404> lufthansa does regional train connections doesnt it
[17:41:27 CEST] <wm4> probably, but doesn't make it faster
[17:41:37 CEST] <wm4> also I'm starting out from Florence for reasons
[17:42:13 CEST] <j-b> take the night train from Florence on Friday night.
[17:42:17 CEST] <j-b> clearly the best option.
[17:43:15 CEST] <Daemon404> im going to get looks from feepk or j-b for my flights
[17:43:22 CEST] <Daemon404> i took a weird routing that was the same price as direct
[17:43:31 CEST] <Daemon404> because miles.
[17:43:35 CEST] <JEEB> lol
[17:44:11 CEST] <Daemon404> adding a intra-europe connection is usually free
[17:44:18 CEST] <Daemon404> or very cheao
[17:44:19 CEST] <Daemon404> cheap
[17:44:53 CEST] <j-b> Daemon404: aren't you coming from UK?
[17:44:56 CEST] <Daemon404> j-b, i am
[17:45:01 CEST] <j-b> Eurostar?
[17:45:08 CEST] <Daemon404> plane was cheaper
[17:45:11 CEST] <Daemon404> euostar is so hit or miss
[17:45:14 CEST] <Daemon404> IME
[17:45:40 CEST] <Daemon404> planes from london to paris (even with connectiosn) are cheao
[17:45:44 CEST] <Daemon404> because they compete with eurostar
[17:45:54 CEST] <Daemon404> same to brussels
[17:46:21 CEST] <kierank> meh
[17:46:22 CEST] <j-b> OK, I won't ask.
[17:46:30 CEST] <j-b> London goes to Disneyland directly :)
[17:46:33 CEST] <Daemon404> lol
[17:46:40 CEST] <kierank> yes but the timings were not good for the disneyland train
[17:46:46 CEST] <kierank> you arrived mid-afternoon
[17:46:47 CEST] <kierank> iirc
[17:46:48 CEST] <Daemon404> even coursmich said my plane routing was insane
[17:46:52 CEST] <Daemon404> that should tell you everything
[17:48:33 CEST] <Daemon404> j-b, well also
[17:48:41 CEST] <Daemon404> last time i was on teh eurostar, i very nearly got in a fistfight
[17:48:43 CEST] <Daemon404> so.
[17:48:45 CEST] <Daemon404> <_<
[17:48:51 CEST] <kierank> why
[17:49:13 CEST] <Daemon404> i told off a guy who was harassing the bartender
[17:49:22 CEST] <Daemon404> one of those pish rich spoiled kids from Chelsea
[17:49:25 CEST] <Daemon404> posh*
[17:49:31 CEST] <Daemon404> got a "do you know who my dad is?"
[17:49:33 CEST] <Daemon404> go from there.
[17:49:40 CEST] <j-b> No, and I don't care.
[17:49:54 CEST] <Daemon404> ofc
[17:49:55 CEST] <Shiz> "unless he's on the train right now, why should i care"
[17:50:37 CEST] <Daemon404> anyway point: im not fleecing videolan for money
[17:50:41 CEST] <Daemon404> im just insane.
[17:51:39 CEST] <kierank> "No, but your mum on the other hand..."
[17:52:15 CEST] <Daemon404> that guy got remove from the car in the end, anyway
[17:52:28 CEST] <Daemon404> bartender called security on him
[17:53:07 CEST] <kierank> should have kicked him off at ashford
[17:53:14 CEST] <kierank> or calais
[17:53:25 CEST] <Daemon404> i dont know what they did after
[17:53:29 CEST] <Daemon404> they very well could have
[18:33:46 CEST] <kierank> i bet that ticket could just be solved by putting the encode in a different thread
[18:33:47 CEST] <kierank> job done
[18:36:18 CEST] <wm4> many of these issues are probably because the API was designed for synchronous operation
[18:36:29 CEST] <wm4> while these OS-level things are designed for asynchronous
[20:20:54 CEST] <peloverde> Is there a decimation filter that just supports a dumb frame pattern and doesn't try to do matching?
[20:32:01 CEST] <durandal_1707> peloverde: select?
[20:33:18 CEST] <peloverde> that's the one, thanks
[20:58:42 CEST] <cone-534> ffmpeg 03Ronald S. Bultje 07master:2fb593dcb90c: Put remaining pieces of CODEC_FLAG_EMU_EDGE under FF_API_EMU_EDGE.
[21:00:45 CEST] <cone-534> ffmpeg 03Ronald S. Bultje 07master:7e12a5425114: ffserver: use -b instead of -ab for setting audio bitrate.
[21:05:45 CEST] <cone-534> ffmpeg 03Paul B Mahol 07master:777df1ff746e: avfilter/vf_dejudder: use the name 's' for the pointer to the private context
[23:20:16 CEST] <cone-534> ffmpeg 03Michael Niedermayer 07master:628a73f8f376: ffmpeg: force 128k default audio bitrate if nothing is specified and there is no specific default
[00:00:00 CEST] --- Sat Aug 29 2015


More information about the Ffmpeg-devel-irc mailing list