[Ffmpeg-devel-irc] ffmpeg-devel.log.20161216

[burek]

[00:01:51 CET] <blue_misfit> hey you responded! more than most folks on #ffmpeg ever did lol!!
[00:02:09 CET] <Compn> thats why they pay me the big bucks!
[02:07:41 CET] <cone-892> ffmpeg 03Martin Vignali 07master:6672a49e5a38: fate/psd: add tests for uncompress and rle samples
[02:07:42 CET] <cone-892> ffmpeg 03Michael Niedermayer 07master:d4c1cc2b876a: avutil/imgutils: Clarify doxy for av_image_check_size2()
[03:52:21 CET] <Zeranoe> Anyone know if FFmpeg works with libjpeg-turbo
[03:52:21 CET] <Zeranoe> , or if openjpeg 2 is supported?
[04:06:31 CET] <jamrial> Zeranoe: the latter is supported
[04:15:35 CET] <Zeranoe> jamrial: But not the former?
[04:15:45 CET] <jamrial> afaics, no
[04:16:06 CET] <Zeranoe> Thanks
[04:56:06 CET] <Zeranoe> FFmpeg doesn't care if x264 was compiled with pthread support right? It uses its own threading methods to run
[05:07:18 CET] <Zeranoe> I think I'm going to switch the Windows builds to native win32 threading instead of pthread
[05:09:47 CET] <Compn> i was unable to run your nightlies on my vista box, Zeranoe
[05:10:15 CET] <Zeranoe> Compn: The latest ones?
[05:10:26 CET] <Compn> no, let me grab an updated build
[05:10:30 CET] <Zeranoe> Compn: And vista.... really....
[05:10:37 CET] Action: blb giggles @ vista
[05:10:47 CET] <rcombs> Zeranoe: x264 just has an internal implementation of the pthread API on top of win32threads
[05:10:51 CET] <rcombs> as does ffmpeg
[05:11:08 CET] <rcombs> neither cares what the other was built with
[05:11:24 CET] <Zeranoe> rcombs: That's what I thought, wanted to make sure
[05:12:46 CET] <Compn> wow some fast speeds from your host
[05:12:55 CET] <Zeranoe> :)
[05:13:50 CET] <Compn> ffmpeg-20161210-edb4f5d-win32-static.zip , reports this on vista the procedure entry point get numa node processor mask ex could not be located in the dll kernel32.dll
[05:14:02 CET] <Compn> then it crashed
[05:14:25 CET] <Zeranoe> Compn: interesting, that's due to MFX
[05:14:25 CET] <Compn> a kernel32 dll function call that isnt available on vista 
[05:14:39 CET] <Compn> mfx ?
[05:15:17 CET] <Zeranoe> qsv
[05:15:25 CET] <Zeranoe> https://ffmpeg.zeranoe.com/forum/viewtopic.php?f=7&t=3728
[05:15:27 CET] <Compn> is there qsv on windows ?
[05:15:48 CET] <rcombs> Compn: dat buttflare
[05:15:53 CET] <Zeranoe> It's in the build
[05:16:08 CET] <Zeranoe> Compn: Are you just a fan of Vista or...
[05:16:32 CET] <Zeranoe> I dropped XP support because it's so old, and I didn't know it effect vista too. I didn't think anyone uses it
[05:16:45 CET] <rcombs> if Compn uses Vista he's the only one
[05:16:53 CET] <rcombs> Chrome even dropped support
[05:16:57 CET] <Compn> rip chrome
[05:17:09 CET] <Zeranoe> Compn: Think it's time for an upgrade?
[05:17:19 CET] <rcombs> did you miss the memo on Win10 having been free
[05:17:24 CET] <Compn> you're talking to the guy who was running win2k into this decade.
[05:17:54 CET] <rcombs> why do you seek to harm yourself and the people around you
[05:18:04 CET] <Zeranoe> Anything is free if you know where to look...
[05:18:17 CET] <Compn> damn you and your latest os upgrading automatically all the time!
[05:18:19 CET] <rcombs> it's okay, there are people who care about your well-being
[05:18:40 CET] <rcombs> just, nobody wants to maintain backwards-compatibility with your ancient OS
[05:18:45 CET] <Zeranoe> I have to agree with that. Win10 likes to f me all the time
[05:19:15 CET] Action: rcombs rings doorbell
[05:19:19 CET] <Zeranoe> "Oh you're clearly working but stepped away for two milliseconds? Time for a reboot."
[05:19:21 CET] <rcombs> HELLO!
[05:20:00 CET] <rcombs> My name is elder rcombs! And I would like to share with you this most amazing code!
[05:20:20 CET] <rcombs> It's some code to replace Windows and sucks slightly less
[05:20:51 CET] <Zeranoe> Compn: https://ffmpeg.zeranoe.com/forum/viewtopic.php?f=13&t=3572
[05:21:01 CET] <rcombs> &turns out I'm far too lazy to turn the opening number from The Book of Mormon into a song about Linux
[05:21:34 CET] <Zeranoe> Compn: Turns out there's a community of you update refusers
[05:21:49 CET] <rcombs> but yeah if you hate current Windows so much you use old Windows& have you considered not-Windows
[05:23:15 CET] <Zeranoe> Compn: I didn't design my site for IE8, so lemme know how many issues you find.
[05:26:23 CET] <rcombs> issues -> 🗑
[05:26:34 CET] <rcombs> but you can't read that character because your font doesn't support it
[05:26:58 CET] <Compn> boxes :D
[05:27:07 CET] <Zeranoe> lol
[05:27:10 CET] <Compn> also im using a web browser that isnt even made anymore
[05:27:15 CET] <Compn> old opera
[05:27:33 CET] <Zeranoe> I'm out, I can't deal with this on my Win10, bleeding edge workstation.
[05:55:13 CET] <Zeranoe> I'm sure this will upset no one, but schrodinger support was dropped from the builds.
[12:32:54 CET] <durandal_1707> BBB: do you have wmavoice specification?
[13:23:40 CET] <cone-199> ffmpeg 03Mark Harris 07master:a5cf600ccb7f: avformat/wavdec: Eliminate goto for clang -O0 DCE
[13:37:01 CET] <BBB> durandal_1707: https://wiki.multimedia.cx/index.php/Windows_Media_Audio_Voice
[13:38:17 CET] <BBB> durandal_1707: I never wrote down the postfilter, but its fairly straightforward if you read the code, its just a combination of gain control, kalman smoothening filter, wiener denoise filter and DC filter, I dont remember details
[13:40:34 CET] <durandal_1707> BBB: but it doesn't mention decoding of wmapro superframes in bitstream
[13:41:04 CET] <durandal_1707> how that one looks?
[13:41:11 CET] <BBB> its literally one wmapro superframe
[13:41:16 CET] <BBB> its not hard to wrap, I had example code once
[13:41:24 CET] <BBB> but I only ever saw it used for silence so I didnt care
[13:41:30 CET] <BBB> (sorry)
[13:41:48 CET] <BBB> it doesnt implement some other features also
[13:41:52 CET] <BBB> but it does some things correctly :)
[13:41:59 CET] <BBB> got a sample?
[13:43:16 CET] <durandal_1707> BBB: see trac timeline
[13:43:40 CET] <durandal_1707> #5426
[13:46:21 CET] <BBB> too many things in that bug report...
[13:46:26 CET] <BBB> is there a summary?
[13:48:59 CET] <BBB> ffplay(56101,0x7fff740d3000) malloc: *** error for object 0xbdbbe8443e2b635c: pointer being freed was not allocated
[13:49:00 CET] <BBB> ...
[13:52:53 CET] <jkqxz> nevcairiel:  I think it wouldn't be unreasonable to add a weird tiled format as a pixfmt if there is more than one thing which uses it.  They can be passed around, but most components will just treat them like opaque hwaccel pixfmts (so maybe a new flag or otherwise make clear in the component information that the layout is not sane).
[13:53:51 CET] <nevcairiel> if its fully opaque,  but it isnt, its trying to be shoehorned into existing functions like swscale and imgutils, which leads to a lot of ifs all around
[13:53:53 CET] <jkqxz> It needs to be much clearer what hardware this applies to and what uses it, though.  (Certainly can't be called NV12_TILED, because there are many such formats.)
[13:54:50 CET] <BtbN> could just act like the CUDA format. With an upload and download filter, so it doesn't mess up swscale
[13:56:41 CET] <nevcairiel> and even worse if there is various different variants from different hardware
[13:57:15 CET] <jkqxz> Some swscale support doesn't seem objectionable, though it needs to be cleaner than the patch currently.
[13:57:32 CET] <BtbN> nvenc also supports tiled NV12 as input, but I never saw a point in bothering with it
[13:58:11 CET] <jkqxz> I assume this specific version is the format used by some IP core, and the SoC they integrated it on doesn't have a linear remapper for it.
[14:00:48 CET] <jkqxz> (Most hardware which actually uses tiled formats also has a linear remapper in the MMU so that you can access it sensibly.)
[14:11:30 CET] <durandal_1707> BBB: i use mpv for playback, have ffplay disabled
[14:13:09 CET] <iive> tile format could be handled as pixel formats, if they do contain pixels
[14:13:18 CET] <iive> e.g. hm12
[14:20:12 CET] <BBB> its quite irritating if people who have no idea start patching decoders but dont do it correctly...
[14:23:41 CET] <BBB> durandal_1707: let me see if I can get wmapro-in-wmavoice working at least in a hacky way
[14:23:48 CET] <BBB> durandal_1707: maybe others can finish it cleanly then
[14:33:30 CET] <wm4> jkqxz: pixdesc can't represent tiled formats, so it'll have to be opaque
[14:56:59 CET] <jkqxz> Yeah, it has to be opaque in lavu and common code.  swscale could deal with it, though.
[14:58:19 CET] <Compn> just woke up, whats DCE
[14:58:35 CET] <JEEB> Dead Code Elimination
[15:04:29 CET] <Compn> does not compute, back to sleep
[15:07:56 CET] <JEEB> oh fucking wow
[15:08:08 CET] <JEEB> someone ported the lavc PGS decoder to java
[15:08:17 CET] <JEEB> and is trying to push it to google's player library for android :D
[15:08:33 CET] <JEEB> https://github.com/google/ExoPlayer/pull/2143/commits/7dc4e73cae2f327aca9b8de378bbe57789f35837#diff-f63bea117f099b5045ee7e9da9bf69d4
[15:08:43 CET] <JEEB> https://github.com/google/ExoPlayer/pull/2143/commits/7dc4e73cae2f327aca9b8de378bbe57789f35837#diff-f63bea117f099b5045ee7e9da9bf69d4R50
[15:08:47 CET] <JEEB> this is just fabulous
[15:09:07 CET] <JEEB> AVSubtitleRect just pushed into java
[15:19:28 CET] <wm4> JEEB: I think this is a LGPL violation
[15:19:36 CET] <wm4> the license only talks about Apache 2
[15:22:20 CET] <JEEB> yes, the exoplayer license is Apachev2
[15:23:46 CET] <wm4> it's just a PR, commented
[16:45:47 CET] <BBB> speaking of copied code & https://github.com/webmproject/libvpx/blob/master/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
[16:49:06 CET] <wm4> BBB: " Part; of the functions are originally derived from the ffmpeg project."
[16:49:12 CET] <wm4> but license is not LGPL
[16:49:16 CET] <BBB> indeed
[16:49:28 CET] <wm4> copyright violation by google? lol
[16:55:39 CET] <kierank> LOL
[17:06:00 CET] <durandal_1707> sue them!
[17:09:54 CET] <durandal_1707> to anyone interested: on my github branch pixlet is some code I'm looking if it can be simplified
[17:14:10 CET] <BBB> direct link?
[17:14:17 CET] <BBB> is it dsp code or is it bitreading code?
[17:14:33 CET] <BBB> I fixed the fate issue with wmavoice btw, but its a little ugly
[17:14:36 CET] <BBB> need to make it nicer
[17:14:49 CET] <BBB> Im thinking maybe I should split the bitreading from the synthesis
[17:14:56 CET] <BBB> it looks like itd make everything better
[17:25:53 CET] <wm4> lol this andreas guy has no sense for nice code
[17:33:19 CET] <durandal_1707> BBB: bitreading, dsp is multiplication with floats
[17:37:35 CET] <durandal_1707> BBB: https://github.com/richardpl/FFmpeg/tree/pixlet
[17:38:18 CET] <BBB> another_param :D
[17:39:05 CET] <durandal_1707> hey, I could pick our nicks
[17:39:24 CET] <BBB> I didnt say it was bad
[17:39:43 CET] <BBB> I did what youre using elsewhere, just v%d with reg/mem location in comments
[17:39:54 CET] <BBB> although I probably used another character, probably d
[17:39:56 CET] <BBB> I dont know why
[17:40:21 CET] <durandal_1707> look at other function which have nice names
[17:40:38 CET] <durandal_1707> ignore one with gotos
[17:40:51 CET] <durandal_1707> it's in rewritting stage
[18:08:40 CET] <Compn> another_param :D
[18:18:37 CET] <durandal_1707> Compn: renamed to compn
[18:32:58 CET] <Compn> YES
[18:33:17 CET] <Compn> theres a lot of compN functions out there tho
[19:35:08 CET] <Timothy_Gu> Compn: I'm < 18 so nah
[19:36:00 CET] <Timothy_Gu> (re mailing hard drives)
[19:36:41 CET] <Timothy_Gu> BtbN: did you see the weird PARSE_ERRORS in Coverity?
[19:37:30 CET] <BtbN> No PARSE_ERRORS yes, but some of the stuff it reported today seems odd, didn't take an in depth look yet though
[19:37:34 CET] <BtbN> *yet
[19:38:23 CET] <BtbN> uh, now I see them.
[19:38:33 CET] <BtbN> Only looked at high impact ones
[19:38:59 CET] <BtbN> "Please see the build output log for more information." just, where is it?
[19:44:10 CET] <nevcairiel> it looks like their parser dies from usage of c11 atomics =p
[19:44:47 CET] <BtbN> https://travis-ci.org/FFmpeg/FFmpeg-Coverity#L1244 it's at least not this build log.
[20:15:08 CET] <Timothy_Gu> BtbN: I think there's a separate log file in cov-int 
[20:15:27 CET] <Timothy_Gu> cov-int/build-log.txt
[20:15:35 CET] <BtbN> hm
[20:16:15 CET] <Timothy_Gu> it's really big though, so I wouldn't try to `cat` it in the build script
[20:16:25 CET] <Timothy_Gu> maybe try building locally?
[20:16:41 CET] <BtbN> can just run the docker container locally
[20:17:11 CET] <Timothy_Gu> yep
[22:15:50 CET] <durandal_1707> how can compile ffmpeg with integer sanitizer?
[22:16:17 CET] <Compn> hmm sounds like special gcc option
[22:17:14 CET] <Compn> durandal_1707 : https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html
[22:37:49 CET] <durandal_1707> Compn: i use clang
[22:48:34 CET] <durandal_1707> I get linking errors
[22:49:12 CET] <durandal_1707> perhaps using undefined will do same
[22:52:26 CET] <BBB> you need cflags and ldflags to use -fsanitize=integer
[22:52:42 CET] <BBB> so extra-cflags=.. and extra-ldflags (or extra-libs, I Cant remember)=..
[22:58:50 CET] <Compn> tgg
[22:58:58 CET] <Compn> wrong window.
[00:00:00 CET] --- Sat Dec 17 2016