[Ffmpeg-devel-irc] ffmpeg-devel.log.20140526

burek burek021 at gmail.com
Tue May 27 02:05:02 CEST 2014

[00:00] <Daemon404> not sure how asm is in msvc
[00:01] <kurosu> well, anyway, the problem is more in the C part, but I was wondering how much of mingw I may be using besides running fate
[00:01] <kurosu> (after starting to use msvc)
[00:03] <nevcairiel> you can step through asm and inspect regs, works fine
[00:11] <kurosu> Daemon404, can you do out-of-tree builds with the msvc toolchain ?
[00:11] <Daemon404> yes
[00:11] <kurosu> nice
[00:11] <kurosu> I usually just create a folder and run eg ../configure
[00:12] <kurosu> is that still ok or do I need a special option ?
[00:12] <Daemon404> there i nothing special except --toolchain=msvc
[00:12] <Daemon404> and making sure msys link.exe doesnt screw yu up
[00:12] <Daemon404> (as noted in the docs)
[00:12] <kurosu> yeah saw that part indeed
[00:13] <kurosu> and in addition my .bashrc was resetting LIB & INCLUDE so to avoid the inverse conflict
[00:14] <clever> how are av options like rtmp_conn set from the ffmpeg api level?
[00:18] <cone-679> ffmpeg.git 03Michael Niedermayer 07master:b0d0e29ffbd4: avcodec/sgirledec: fix () in RBG323_TO_BGR8() macro
[00:18] <cone-679> ffmpeg.git 03Michael Niedermayer 07master:c37d17929590: avformat: add format_probesize to allow tuning the maximum amount of bytes to identify the filetype
[00:29] <cone-679> ffmpeg.git 03Michael Niedermayer 07master:d236d4907c93: update APIchanges & version for c37d179
[02:07] <cone-679> ffmpeg.git 03Michael Niedermayer 07master:d2eacbb754ba: doc/APIchanges: fix typo in date
[02:08] <cone-679> ffmpeg.git 03Michael Niedermayer 07master:96470ca22b3b: avformat/mov: allow seeking back to the begin even if nothing is marked as keyframe
[03:49] <jamrial> michaelni: how true is the comment in http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/x86/mpegvideoenc_template.c;h=76a5c5a;hb=HEAD#l112 ?
[03:53] <michaelni> jamrial, the dct is supposed to be inlined there
[03:56] <jamrial> can that line be s->dsp.fdct(block); instead? as in, can these quantize functions use the c version of fdct(), or is the relevant ff_fdct_<opt>() a must?
[03:58] <michaelni> dunno maybe it would work maybe the scaling would differ cant remember 
[04:00] <jamrial> this is regarding the fdct yasm port patch i sent a couple hours ago. these quantize functions are inline asm and currently have a hard dependecy on the fdct functions because of that line
[04:02] <jamrial> so the only options i have are making the quantize functions depend on both yasm and inline asm (ugly and not ok), porting the quantize functions to yasm (a huge task i'm not up to), changing that line so it can fall back to the c version if using --disable-yasm, or dropping the fdct yasm port patch
[14:47] <Daemon404> ubitux, i cannot facepalm harder
[14:57] <ubitux> Daemon404: well i understand the concern
[14:57] <ubitux> but i believe pkg-config is the correct answer
[14:58] <ubitux> and i think it should be used more, as i said in another thread
[14:59] <Daemon404> the concern is bollocks
[14:59] <Daemon404> its a hack to support idiots with broken envs
[15:00] <Daemon404> too stupid to learn the tools
[15:08] <wm4> haha
[15:08] <wm4> what a stupid thread
[15:08] <ubitux> it's still the same story for x264 btw
[15:11] <wm4> why don't you send users to the x264 bug tracker then
[15:11] <ubitux> ?
[15:11] <ubitux> i mean we're not using pkg-config
[15:12] <wm4> well I'd still consider that a bug
[15:12] <ubitux> and ofc now there is (justified) reluctance to change it to pkg-config because it would break with ppl using --extra-{c,ld}flags
[15:12] <wm4> what is this, 1990?
[15:12] <ubitux> wm4: http://ffmpeg.org/pipermail/ffmpeg-devel/2014-May/157646.html
[15:12] <ubitux> see that thread
[15:43] <cone-919> ffmpeg.git 03Christophe Gisquet 07master:86ae0da60c46: x86: hpeldsp: propagate changes across codecs
[16:06] <cone-919> ffmpeg.git 03Neil Armstrong 07master:99d742aac41b: configure: uClibc native pthread requires -ldl
[16:58] <cone-919> ffmpeg.git 03Nicolas George 07master:58a10e0e2ce1: compat/w32pthreads: add return value to pthread_cond_init().
[16:58] <cone-919> ffmpeg.git 03Nicolas George 07master:55cc60cd6d94: lavu: add thread message API.
[16:58] <cone-919> ffmpeg.git 03Nicolas George 07master:fc9c857c2d1b: ffmpeg: use thread message API.
[16:58] <cone-919> ffmpeg.git 03Michael Niedermayer 07master:2db89765f3fc: Merge remote-tracking branch 'cigaes/master'
[17:00] <ubitux> ah, finally!
[17:00] <ubitux> cool.
[17:09] <Daemon404> wow uclibc
[17:38] <BBB> so how do we test this resampling code?
[17:41] <ubitux> make libswresample/swresample-test maybe
[17:41] <ubitux> otherwise... ffmpeg -ar ?
[17:42] <Daemon404> i should not have mention --static
[17:42] <Daemon404> now the neurotic carl eugen is going to spend 3 more days misunderstanding pkh-config
[17:42] <Daemon404> pkg*
[17:46] <JEEBsv> lol
[17:46] <JEEBsv> I hope it's not your having fun with him :P
[17:46] <JEEBsv> *you
[17:47] <nevcairiel> everyone is having fun with carl today
[17:48] <ubitux> well since he's helping out users on ffmpeg-user, it's a good idea if he understands how it works&
[17:49] <ubitux> so go help him Daemon404 ;)
[17:49] <Daemon404> ubitux, heh
[17:49] <Daemon404> i would like to think "how to use pkg-config" is a bit beyond the scope of ffmpeg
[17:49] <Daemon404> but meh
[17:49] <Daemon404> lusers~
[17:50] <nevcairiel> he helps users? doesnt he just go around telling them not to use libav? :p
[17:50] <Daemon404> no nevcairiel 
[17:50] <Daemon404> he also tells them to submit full uncut output
[17:51] <ubitux> this is useful
[17:51] <Daemon404> ubitux, *sometimes*
[17:51] <Daemon404> lots of times knowledgable people submit more than enough info
[17:51] <nevcairiel> report a problem in some code, get asked to provide ffmpeg output
[17:51] <Daemon404> only to be met with :uncutoutput:
[17:51] <Daemon404> bug tracker is absolutely great at alienating power users
[17:51] <Daemon404> or devs
[17:52] <Daemon404> in optimizing for the lowest common denominator
[17:52] <Daemon404> ... much liek a call center.
[17:52] <ubitux> it gives the exact ffmpeg version and configuration
[17:52] <ubitux> so this is important when digging an old report
[17:52] <Daemon404> he tells people ot provide it when teh yeven provide a git hash
[17:52] <ubitux> typically to try reproducing in the exact context
[17:52] <Daemon404> and compiler
[17:52] <Daemon404> etc
[17:52] <Daemon404> its all there.
[17:52] <ubitux> doesn't happen often, and it's simpler for them to just paste the complete output
[17:53] <ubitux> it's not like it's an unbearable request
[17:53] <nevcairiel> it just pisses smart people off
[17:53] <Daemon404> yes
[17:54] <Daemon404> most of them just dont submit bugs anymore
[17:54] Action: compn wonders when Daemon404 or nevcairiel will take over reviewing bug reports on trac
[17:54] <compn> :P
[17:54] <Daemon404> compn, not as long as carl is there to override everythign i do
[17:55] <Daemon404> btw i get teh feeling this is the reason a bunch of people report bugs directly to michael
[17:55] <Daemon404> via email or irc
[17:56] <ubitux> there is quite some activity on the trac you know
[17:56] <Daemon404> i know
[17:56] <Daemon404> but almost none of it is dev
[17:56] <Daemon404> (except that epic aac one)
[17:56] <j-b> pkg-config is bad :)
[17:57] <Daemon404> j-b, vlc is bad
[17:57] <Daemon404> you patch stuff and dont upstream said patches
[17:57] <Daemon404> shame!
[17:57] <j-b> Do we?
[17:57] <Daemon404> funman does all the time
[17:57] <Daemon404> i troll him about it
[17:57] <Daemon404> he just says hes lazy
[17:57] <nevcairiel> good enough reason
[17:57] <j-b> he is lazy
[17:57] <ubitux> j-b: why?
[17:58] <j-b> Daemon404: but, a contrario from MPlayer, we don't fork hundreds of libraries inside our trees
[17:58] <Daemon404> stuff like "fix mips build of gnutls" and stuff
[17:58] <Daemon404> that definitely does not belong in vlc.
[17:58] <nevcairiel> who cares about mips =P
[17:58] <Daemon404> vlc apparently :P
[17:58] <j-b> Daemon404: have you ever tried to upstream a buildsystem fix?
[17:58] <Daemon404> yes
[17:58] <Daemon404> it used to be my job
[17:58] <nevcairiel> was it fun?
[17:58] <j-b> exactly ?
[17:58] <Daemon404> anyway
[17:58] <j-b> Most people don't understand what a compiler is
[17:58] <Daemon404> oi plan to troll some of contrib/ soon
[17:58] Action: Daemon404 rm -r libmpeg2
[17:59] Action: Daemon404 runs
[17:59] <j-b> what a cross-compiler is
[17:59] <j-b> Daemon404: like I care
[17:59] <Daemon404> j-b, i used to upstream cross-compilation fixes specifically
[17:59] <j-b> Daemon404: 99% of contrib/ patches are about cross-compiling issues
[17:59] <j-b> Daemon404: you were paid to do so.
[17:59] <Daemon404> i used to work on openembedded and its precursor
[17:59] <Daemon404> for a living
[17:59] <Daemon404> ;)
[17:59] <Daemon404> i sitll do
[17:59] <Daemon404> still*
[17:59] <j-b> Daemon404: do you go and say that to all distributions?
[18:00] <j-b> that have 100s of patches?
[18:00] <Daemon404> i call it "debian maintainer syndrome":
[18:00] <Daemon404> it is crap.
[18:00] <j-b> call it the way you want.
[18:00] <j-b> not really my problem.
[18:00] <j-b> but I'm bored explaining to people what pkg-config is abot
[18:00] <j-b> or why a non-autotools setup is stupid.
[18:01] <Daemon404> people who dont know that have no place building stuff really
[18:01] <ubitux> j-b: i wouldn't mind getting lectured on why pkg-config is bad
[18:01] <j-b> ubitux: it's not bad.
[18:01] <j-b> ubitux: that was a joke.
[18:01] <ubitux> ok :)
[18:01] <j-b> ubitux: it used to be very bad in the beginning.
[18:01] <j-b> ubitux: it's now correct.
[18:01] <Daemon404> ubitux, the other thing i want to do btw is remove my great shame from ffmpeg (libutvideo)
[18:01] <Daemon404> i think i will have a hard time convincing people
[18:02] <j-b> ubitux: even if I prefer pkg-config-lite
[18:02] <ubitux> Daemon404: did you see the trac ticket about ut performance?
[18:02] <Daemon404> ubitux, its idiots talkign to idiots
[18:02] <ubitux> j-b: -lite? what? :)
[18:02] <Daemon404> nobody has any clue wtf theyre talking about
[18:02] <nevcairiel> yeah that guy complained to me
[18:02] <Daemon404> and i cant be arsed to fix them
[18:02] <nevcairiel> i sent him to carl
[18:02] <nevcairiel> :P
[18:03] <j-b> ubitux: a version that does not need glib
[18:03] <ubitux> sounds good
[18:03] <Daemon404> j-b, good news: freebsd wroye pkgconf
[18:03] <Daemon404> which needs nothing
[18:03] Action: Daemon404 runs
[18:03] <ubitux> iirc pkg-config embed a glib copy in its repository or something
[18:03] <ubitux> but yeah ok :)
[18:03] <Daemon404> i dont really understand why pkg-config needs glib in the first place
[18:03] <nevcairiel> someone liked glib?
[18:04] <j-b> Daemon404: and that depends on a libc that is only in FreeBSD, like 99% of BSD stuff?
[18:04] <Daemon404> j-b, not sur.w
[18:04] <Daemon404> sure*
[18:04] <j-b> Daemon404: because of the RedHat/Gnome Glib fetishism
[18:04] <Daemon404> j-b, will people get pissy if i upstream random vlc patches
[18:04] <Daemon404> (with proper author attribution)
[18:04] <j-b> Daemon404: of course not.
[18:04] <j-b> Daemon404: and noone cares about attributions for those things
[18:05] <Daemon404> i learned never to say "noone"
[18:05] <j-b> most patches are from funman or me
[18:05] <Daemon404> some are remk
[18:05] <Daemon404> remi*
[18:05] <j-b> just don't try to upstream my fontconfig patch :)
[18:06] <Daemon404> well also
[18:06] <Daemon404> i cant upstream anythign to GNU
[18:06] <Daemon404> because i cant sign a CLA for someone else
[18:06] <Daemon404> obviously.
[18:06] <compn> wait
[18:06] <compn> vlc doesnt fork libs ?
[18:06] <j-b> GNU still needs CLA?
[18:06] Action: compn looks at libdvdread
[18:06] <j-b> compn: of course not.
[18:06] <j-b> compn: dvdread is not in our tree.
[18:06] <Daemon404> j-b, yes
[18:06] <Daemon404> they rewrote a whole bunch of my patch submissions
[18:06] <Daemon404> because they couldnt be arsed to get me to sign a CLA
[18:06] <j-b> compn: and I forked dvdread, because MPlayer maitainers are lazy.
[18:06] <compn> :D
[18:07] <Daemon404> kind of annoying, since i spent hours trackign down and fixing these bugsd
[18:07] <Daemon404> but 0 credit
[18:07] <Daemon404> =p
[18:07] <j-b> and dvdread is the biggest crashers ever.
[18:07] <compn> probably yes
[18:07] <j-b> because it's coded by incompetent
[18:07] <compn> its old old old code
[18:07] <compn> and dvds are corrupted now
[18:07] <j-b> not only
[18:07] <compn> anti-pirate dvd protections
[18:07] <j-b> not only
[18:08] <j-b> "Oh, I can't parse this IFO file. I couldn't reject and stop, but wait, It's better to assert(0) and abort()!"
[18:08] Action: wm4 would like a libdvd* with non-retarded API
[18:08] <wm4> PS: libdvdnav has a retarded API
[18:08] Action: compn would like dvd:// support in ffmpeg
[18:08] <j-b> I agree with wm4, this should be rewritten from scratch
[18:09] <wm4> compn: well, that would be nice, but it wouldn't really fit into libavformat
[18:09] <compn> libavdevice surely
[18:10] <compn> j-b : how is quality of bluray code ?
[18:10] <j-b> compn: quite good, to be honest.
[18:11] <wm4> libbluray has the same API problems as libdvdnav IMO
[18:11] <j-b> compn: parsing of files is done with flex/bison, license is correct, cross-OS is correctly done
[18:11] <ubitux> maybe that's the reason we have a bluray reader and not a dvd one in ffmpeg :p
[18:11] <j-b> wm4: libbluray APIs are way better.
[18:11] <compn> mplayer dvd stuff works cross-os
[18:11] <compn> :P
[18:11] <compn> probably lots of ifdefs 
[18:12] Action: compn probably just trolling now
[18:12] <wm4> j-b: my main problem with both of them is that it's hard to do correct seeking
[18:13] <j-b> bd_seek_time ?
[18:14] <wm4> j-b: that's not very precise, and also associating precise timestamps with a raw ts byte stream is kind of hard
[18:16] <j-b> wm4: well, dvdnav doesn't even provide that.
[18:16] <wm4> j-b: maybe I'm missing something, but to me it seems both provides equivalent APIs in this aspect
[18:16] <wm4> you can get a timestamp that is valid for a specific byte position
[18:17] <j-b> nope, you cannot
[18:17] <wm4> and you can seek "somewhere" using a timestamp
[18:17] <j-b> see the jump_to_sector_by_time commit from lately
[18:18] <wm4> and libbluray does that better?
[20:02] <kurosu> how do you people debug using msvc? create an empty projet and just manually add the binary? or do you have a solution file at the ready ?
[20:03] <Daemon404> mostly people who debug are API users
[20:03] <Daemon404> and hence have their own rojects
[20:03] <Daemon404> projects*
[20:05] <kurosu> for some reason the express version does show me win64/x64/whatever targets but accepts to debug just fine whatever is thrown at it (ie even a win64 binary in a win32 solution)
[20:05] <kurosu> *doesn't show
[20:06] <nevcairiel> express includes 64-bit compilers though
[20:06] <nevcairiel> maybe the IDE just doesnt show it
[20:06] <nevcairiel> my fate runs on express
[20:06] <nevcairiel> both 32 and 64 bit
[20:08] <kurosu> hum, actually, I don't know what it built, as I just supplied --toolchain=msvc
[20:10] <nevcairiel> that depends what your env variables contain really
[20:10] <kurosu> duh, so it is win64
[20:11] <kurosu> I just wrote a batch file to invoke win64 + msys bat directly
[20:33] <cone-919> ffmpeg.git 03Carl Eugen Hoyos 07master:2d56f0d05475: Write the actual mp3 frame size when muxing into caf if available.
[20:34] <funman> Daemon404: did i not upstream some patches?
[20:44] <ubitux> ok so, valgrind drd went from 1199 to 1215 passing tests (on a total of 1937) after the thread changes
[20:44] <ubitux> better, but still not complete :P
[21:05] <cone-919> ffmpeg.git 03Michael Niedermayer 07master:d35000c2dcb7: avcodec/g723_1: fix writing into input frame data and warning
[21:13] <ubitux> alpa_astero: that's just bits here and there to diversify, don't worry ;)
[21:15] <alpa_astero> ubitux: what do you mean, precisely?
[21:16] <ubitux> i'm just being sarcastic, you can ignore me :p
[21:28] <cone-919> ffmpeg.git 03Clément BSsch 07master:53bac1b4bd2a: doc: add a tutorial for writing filters.
[21:32] <ubitux> wm4: still didn't find motivation to get that utf-16 done? :)
[21:32] <wm4> no
[21:32] <wm4> I hate subtitles
[21:33] <ubitux> :D
[21:33] <wm4> but I think I should get it done before the next release
[21:33] <TheFluff> fuck subtitles; subitles suck; subtitles are dying; subtitles are dead to me; subtitles hit WTC
[21:33] <wm4> yes
[21:58] <cone-919> ffmpeg.git 03Lou Logan 07master:2dbf5e7038f4: configure: s/solving/solve
[22:12] <cone-919> ffmpeg.git 03Lou Logan 07master:5b1a56b34b3f: doc/ffmpeg: note that -t is also an input option
[22:20] <cone-919> ffmpeg.git 03Lou Logan 07master:60ef61242c33: MAINTAINERS: add Trac maintainers
[22:37] <cone-919> ffmpeg.git 03Michael Niedermayer 07master:52887aa409bf: avformat: move default for max_analyze_duration into utils.c
[22:37] <cone-919> ffmpeg.git 03Michael Niedermayer 07master:e6a084641aad: avformat/utils: set max analyze duration depending on the circumstances
[00:00] --- Tue May 27 2014

More information about the Ffmpeg-devel-irc mailing list