[Ffmpeg-devel-irc] ffmpeg-devel.log.20141220

burek burek021 at gmail.com
Sun Dec 21 02:05:03 CET 2014


[00:00] <aetasx> cone's got your back
[00:06] <michaelni> ubitux, thx
[00:07] <jamrial> why 32 bytes?
[00:07] <jamrial> every other case i saw used 5
[00:07] <Daemon404> because if it isnt printable it does [%c]
[00:07] <Daemon404> er
[00:07] <Daemon404> [%d]
[00:07] <Daemon404> which can be 5 chars long in itself.
[00:08] <Daemon404> mind you that doesnt add to 32.
[00:08] <j-b> ubitux: the split made no sense
[00:08] <ubitux> j-b: any plan to merge them back again? :)
[00:09] <jamrial> ah ok
[00:09] <j-b> ubitux: Flameeyes wants me to do that
[00:09] <ubitux> j-b: i would support that too
[00:09] <j-b> ubitux: it makes sense, mostly because it would reduce the API by 80%
[00:09] <j-b> because dvdread API has so many ifo_* that are only useful for libdvdnav
[00:10] <ubitux> jamrial: i see only 2 other occurences to fix, apng and rsd
[00:10] <Daemon404> it would be fine to use 5 if the tag is known good
[00:10] <Daemon404> ... but then you would know the tag string anyway
[00:10] <ubitux> j-b: yeah, also less confusing
[00:11] <j-b> ubitux: but, sorry, my goal was to clean and fix
[00:11] <jamrial> both are mine, and i remember copying the 5 from some other file, which i guess was fixed
[00:11] <j-b> before doing anything new
[00:11] <j-b> ubitux: stabilization is necessary
[00:11] <ubitux> j-b: sure :)
[00:11] <j-b> ubitux: we are rewriting parts of libdvdread
[00:11] <j-b> mostly the UDF layer
[00:11] <j-b> we did rewrite the md5
[00:11] <ubitux> j-b: btw, did you confirm the problem with the lib working with the ISO file but not the mounted device?
[00:11] <ubitux> (sorry a bit OT)
[00:12] <j-b> never confirmed that
[00:12] <ubitux> strange, ok.
[00:12] <j-b> OT? a bit, sure. but not more than trolling for the sake of trolling
[00:12] <ubitux> i wasn't trolling?
[00:13] <aetasx> looked like the topic at the time was trolling
[00:13] <ubitux> anyway, that's cool if these libs get merged
[00:13] <Daemon404> videolan cgit is 504ing
[00:13] <j-b> ubitux: we are rewriting the udf code, especially for this
[00:13] <j-b> and also because of licensing and libbluray
[00:13] <j-b> Daemon404: ah, crap. Yet another attack
[00:14] <ubitux> j-b: is there a legal concern with dvdcss being merged with the two others?
[00:14] <j-b> Yes.
[00:14] <j-b> It won't be merged.
[00:14] <aetasx> so are most of you also mplayer devs as well?
[00:14] <Daemon404> j-b, skiddies ddosing vlc?
[00:14] <Daemon404> how productive
[00:14] <ubitux> j-b: ah, so you're going to merge only dvdread and dvdnav then?
[00:14] <j-b> ubitux: maybe
[00:14] <ubitux> ok...
[00:14] <ubitux> :P
[00:15] <j-b> Daemon404: no, stupid chinese keep hammering git.videolan.org for ffmpeg tar.gz
[00:15] <j-b> it's really annoying
[00:15] <Daemon404> wtf
[00:15] <wm4> aetasx: mplayer is dead
[00:15] <ubitux> iirc libdvdcss works at "dvdread" level right? like, changing dvdread callbacks?
[00:15] <wm4> aetasx: these days it's more like vlc devs everywhere
[00:15] <j-b> Daemon404: since 3 weeks or so
[00:15] <j-b> Daemon404: then loads get at 200
[00:15] <j-b> and the machine dies
[00:16] <wm4> disable zip downloads?
[00:16] <Daemon404> cant you blacklist IPs after N attempts
[00:16] <Daemon404> at the nginx level
[00:16] <Daemon404> within X minutes
[00:16] <j-b> Daemon404: does not work
[00:16] <Daemon404> that makes me sad
[00:16] <j-b> Daemon404: we're bringing a new machine to do the work
[00:17] <j-b> 40 cores, 128GB
[00:17] <Daemon404> why not just disable tarballs
[00:17] <j-b> because it's not the right philosophy
[00:17] <wm4> can you reenable grep search then?
[00:17] <Daemon404> of 302 redirect to github's tarballs
[00:17] Action: Daemon404 runs
[00:17] <Daemon404> or*
[00:18] <j-b> wm4: after that, yes.
[00:18] <aetasx> isnt that the main role of fail2ban?
[00:18] <j-b> sure
[00:18] <kepstin-laptop> is it just the baidu search engine? I had a friend who noticed that was downloading tarballs from his gitweb, ended up blocking it by robots or user agent or something.
[00:19] <j-b> but when they take a 50MB tarball from every revision or so
[00:19] <Daemon404> aetasx, doesnt work if the chinese have a bazillipn IPs rotating
[00:19] <Daemon404> or a botnet
[00:19] <j-b> kepstin-laptop: already blacklisted
[00:19] <aetasx> true
[00:19] <ubitux> if china is ever dropping its great firewall, we'll have to rebuild it ourselves
[00:20] <aetasx> just add the entire IP range for the country ;)
[00:21] <BBB> ubitux: is there a reason you do so many stores in the vp9 lpf asm?
[00:21] <BBB> I wonder if I should rewrite that a bit to be more reg-only-friendly
[00:22] <BBB> (particularly in the h/v transitions)
[00:22] <ubitux> which store in particular are you thinking about?
[00:22] <j-b> Daemon404: should be fixed
[00:22] <Daemon404> yeah i noticed
[00:23] <BBB> line 670-706
[00:23] <BBB> 745-752
[00:23] <j-b> Daemon404: sorry for the issue
[00:24] <BBB> 376-389
[00:24] <Daemon404> dunno why youd apologize, it's a free servicd you provide to us.
[00:24] <BBB> also maybe DEFINE_REAL_P7_TO_Q7 takes one more reg than necessary (if you have a range of -4 to +4, 2 dst regs should be enough to load a range of -7 to +7)
[00:25] <ubitux> BBB: give me a moment to dig again into this
[00:26] <BBB> hehe ok
[00:26] <BBB> (btw I fully recall that lpf is very complex and mind-torning to write in asm ...)
[00:27] <cone-112> ffmpeg.git 03Clément BSsch 07master:d60fb4f79462: avformat/apngdec: make tag_buf string larger
[00:27] <BBB> also 753-789, have you noticed movhps?
[00:27] <cone-112> ffmpeg.git 03Clément BSsch 07master:edbbb11488e1: avformat/rsd: make tag_buf string larger
[00:27] <cone-112> ffmpeg.git 03Martin Storsjö 07master:4895aa65c6dd: libavformat: Allow calling av_write_trailer with a NULL AVIOContext
[00:27] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:fdc297bcf374: Merge commit '4895aa65c6ddec2e33ec3c023f221b1bafcbaf9f'
[00:27] <BBB> movhps is basically the same as punpckhqdq; movh
[00:27] <BBB> but saves an instruction
[00:27] <ubitux> you're concerned about the loads or the stores?
[00:28] <ubitux> i suppose i'm loading them all because i can
[00:28] <ubitux> (already using up to 16 from previous code probably)
[00:28] <jamrial> ubitux: lol, i just tried to push the same thing and of course didn't let me
[00:28] <jamrial> thanks for fixing
[00:28] <ubitux> jamrial: sorry :)
[00:29] <BBB> both
[00:31] <BBB> Ill look into it if youre tired, its been a while
[00:31] <BBB> partially trying to get 8px wide loopfitler functions assemblified, partially trying to get 32bit support
[00:31] <BBB> both are essentially the same thing
[00:32] <cone-112> ffmpeg.git 03Martin Storsjö 07master:4895aa65c6dd: libavformat: Allow calling av_write_trailer with a NULL AVIOContext
[00:33] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:3a15168a6e97: Merge commit '4895aa65c6ddec2e33ec3c023f221b1bafcbaf9f'
[00:36] <ubitux> BBB: about DEFINE_REAL_P7_TO_Q7, jungling between 2 pointers was really annoying
[00:36] <ubitux> and it required updating them
[00:36] <ubitux> here i can have immediate accesses
[00:36] <ubitux> and it made the code faster
[00:37] <ubitux> anyway, not sure what you mean; try to change them you'll see :D
[00:37] <BBB> updating?
[00:37] <ubitux> you're making it 32-bit compliant?
[00:37] <BBB> yeah
[00:37] <BBB> ok
[00:37] <ubitux> good luck... :(
[00:37] <BBB> Ill try it :)
[00:37] <ubitux> yeah
[00:37] <ubitux> it was such a pain to do even with 16 regs, i'm not exactly willing to do that again haha
[00:38] <wm4> recently I switched my system back to 32 bits
[00:38] <ubitux> wat
[00:38] <wm4> because 2GB of RAM is just not enough for 64 bits
[00:39] <ubitux> that's the minimum to open 2 tabs in browsers nowadays 
[00:39] <ubitux> (careful, it allocates more memory when you close the tab, so better keep them open)
[00:40] <BBB> lol
[00:40] <BBB> you mean you dont have 3TB memory?
[00:40] <BBB> whats wrong with you guys
[00:41] <cone-112> ffmpeg.git 03Martin Storsjö 07master:e2ce16392205: mpegts: Support running the write_trailer function without an AVIOContext
[00:41] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:6b0f54ddc746: Merge commit 'e2ce16392205d8efe9143329ed3fb5fcb15498fa'
[00:42] <wm4> x32 is the future
[00:42] <ubitux> :)
[00:43] <BBB> x32 is hype
[00:48] <cone-112> ffmpeg.git 03Martin Storsjö 07master:fc308b30bb24: rtpenc_mpegts: Call write_trailer for the mpegts muxer even if no output buffer exists
[00:48] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:a701a9cff4d5: Merge commit 'fc308b30bb24e623fed042ec78b10803b2362a18'
[00:58] <wm4> hm why is there nnedi discussion in the vf_uspp thrad
[00:58] <wm4> *fspp thread
[00:59] <wm4> didn't follow most of it
[00:59] <wm4> nnedi should be available as vapoursynth filter
[00:59] <wm4> which would solve the issues about file sizes
[00:59] <wm4> because it can just be external
[01:10] <wm4> just accept and welcome the greatness of vapoursynth
[01:11] <ubitux>   543 mpv/video/filter/vf_lavfi.c
[01:11] <ubitux>  1025 mpv/video/filter/vf_vapoursynth.c
[01:11] <ubitux> i thought our api was complicated :X
[01:11] <jamrial> ubitux: i'm going to backport the two tag_buf changes
[01:12] <ubitux> sure..
[01:12] <wm4> ubitux: actually, vf_vapoursynth.c contains a complete scripting backend for Lua
[01:12] <wm4> ubitux: additionally to calling the VS API for using Python
[01:13] <jamrial> just letting you know since both commits were yours :p
[01:13] <wm4> the VS API isn't complicated, but admittedly translating the threading models is a bit of work
[01:13] <ubitux> jamrial: also, one thing i was thinking
[01:13] <ubitux> jamrial: if the buf_size < 21, we can just use another method
[01:13] <cone-112> ffmpeg.git 03Martin Storsjö 07master:8ebf02f8f530: libavformat: Only use MoveFileExA when targeting the desktop API subset
[01:13] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:3d6545d51e6c: Merge commit '8ebf02f8f530240edf7e45f35f7647ef9dd44a58'
[01:13] <ubitux> such as printing '.' instead of unprintable chars
[01:13] <ubitux> so for api users sending 5 it would "work"
[01:15] <jamrial> sounds good i guess
[01:17] <jamrial> do i backport the fixes already in the tree, or do you want me to wait and backport that change instead?
[01:18] <ubitux> i'm very happy when i don't need to do boring stuff like this
[01:18] <ubitux> so feel free to do it
[01:19] <ubitux> http://pastie.org/9790763 completely untested
[01:19] <ubitux> anyway, gonna sleep and forget about it
[01:19] <ubitux> 'night
[01:19] <jamrial> haha, ok
[01:19] <jamrial> night
[01:23] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.0:8a35f24ca943: avformat/rsd: make tag_buf string larger
[01:23] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.1:ea1d0bc02919: avformat/rsd: make tag_buf string larger
[01:23] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.2:b807e987f1c3: avformat/rsd: make tag_buf string larger
[01:24] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.3:c2517fb363f5: avformat/rsd: make tag_buf string larger
[01:24] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.4:f30a89e15f9f: avformat/rsd: make tag_buf string larger
[01:24] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.5:f295f9488a30: avformat/apngdec: make tag_buf string larger
[01:24] <cone-112> ffmpeg.git 03Clément BSsch 07release/2.5:7e130ca5b42c: avformat/rsd: make tag_buf string larger
[02:10] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:4b45aa517c30: avformat/asfenc: replace != 0 error check by <0
[02:10] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:172865931951: avformat/westwood_aud: replace != 0 error check by <0
[02:24] <cone-112> ffmpeg.git 03Luca Barbato 07release/2.4:3ea49fc5081d: vc1: Do not assume seek happens after decoding
[02:25] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:03fbc282ff58: Merge commit '3ea49fc5081d63277ecbc12ed440af4b02ddfdf9' into release/2.4
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:c4e18917d4f7: avformat/aviobuf: Fix infinite loop in ff_get_line()
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:117dcc209796: avformat/rmdec: Check codec_data_size
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:f36b3c5df97d: swscale/x86/rgb2rgb_template: fix crash with tiny size and nv12 output
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:82db2f2ac8c3: ffmpeg: drop usage of coded_frame
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:fd72ff6f31f9: doc/examples/transcoding: check encoder before using it
[02:47] <cone-112> ffmpeg.git 03wm4 07release/2.4:95c298b12502: avformat/matroskadec: fix handling of recursive SeekHead elements
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:787e4d12daf5: avformat/flvdec: fix potential use of uninitialized variables
[02:47] <cone-112> ffmpeg.git 03Rob Sykes 07release/2.4:3b332ef33cec: swresample/soxr_resample: fix error handling
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:0d277be45ae9: avformat/aviobuf: Check that avio_seek() target is non negative
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:0fb2b616142e: avformat/utils: Do not update programs streams from program-less streams in update_wrap_reference()
[02:47] <cone-112> ffmpeg.git 03wm4 07release/2.4:14d6ea0c459c: lavu/frame: fix malloc error path in av_frame_copy_props()
[02:47] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:820f41e1a1b4: Add FFMPEG_VERSION into the binary libs
[03:10] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:5d1d143a4eb4: Update for FFmpeg 2.4.5
[03:51] <Daemon404> https://github.com/mpv-player/mpv/blob/master/video/filter/vf_vapoursynth.c#L790-804
[03:51] <Daemon404> very nice
[03:59] <Compn> gasp :P
[03:59] <Compn> anyone who can read mandarin want to check an email that was sent to projects at mphq ?
[04:00] <Compn> not sure where our chinese friends are
[04:00] <Compn> >wonder if wm4 is chinese
[04:00] <Compn> >so mysterious
[04:00] <Daemon404> wm4 is not real
[04:00] <Compn> ghost in the budokai ?
[04:01] <Daemon404> troll in the shell
[04:01] <Compn> lol
[04:01] <Compn> thats a great title
[04:04] <pross> Compn: link please
[04:10] <Compn> its an email. not sure if this works or not . ??!???ffmpeg???????url??avformat_open_input??,ffmpeg?????,??url???pageUrl ,swfUrl?????,??????????????????????,????wireshark??????,?????????????????,
[04:10] <Compn> looks like rtmp question
[04:11] <Compn> http://pastebin.com/FavgBX2T
[04:11] <Compn> displays correctly ^
[04:12] <Compn> google translates ni hao to 'hello there' :)
[04:12] <Compn> google tranlsation : Hello there! Then I assign directly to the url when using ffmpeg do live avformat_open_input, ffmpeg do internal processing, because there url pageUrl, swfUrl and some other parameters, logically before it is sent to the server will these parameters extracted, but I caught wireshark package looked and found these parameters are not parsed out to the server,
[04:12] <wm4> why the fuck does this guy think it's ok to send a chinese mail to a community that speaks english only
[04:13] <Compn> why do you think sending non-english is an insult ?
[04:13] <Compn> many developers here are bilingual
[04:13] <wm4> not necessarily, but obviously nobody will be able to read it
[04:15] <pross> thx
[04:16] <Compn> pross : want me to forward email to you ?
[04:16] <Compn> if you want to reply to person...
[04:16] <pross> ok
[04:16] <Compn> pross at xvid.org ?
[04:16] <pross> dui4
[04:16] <Compn> ?
[04:16] <pross> yes
[04:17] <Compn> sent
[04:18] <Compn> is he saying that ffmpeg sends pageurl when not requested ?
[04:18] <Compn> while wireshark shows flash player does not 
[04:18] <Compn> hard to tell from translation. oh well
[04:20] <pross> no no, hes not seeing the parameters with wireshark (ffmpeg->server)
[04:21] <Compn> ah
[04:22] <Compn> wm4 : so what do you do when some non english pops up in your project? :)
[04:22] <Compn> thanks for looking at it pross :)
[04:22] Action: Compn wonders why english was chosen for these projects
[04:23] <pross> excellent question
[04:23] <Compn> fabrice is french , arpi is hungarian
[04:24] <wm4> so obviously french and hungarian
[04:24] <wm4> would make a nice mixture
[04:24] <wm4> maybe french function names and hungarian comments
[04:24] <wm4> it's natural
[04:24] <Compn> vlc was made in france
[04:24] <Compn> most of its developers ... french
[04:24] <jamrial> it would seem like half of api users are chinese or indian, so public functions should be in one of those
[04:25] <Compn> india has a lot of english , i guess from being british colony for a long time
[04:27] <Compn> i do not know world history much :\
[04:29] Action: pross blames popularity of the qwerty keyboard
[04:30] <pross> programmers are inherently lazy. switching ime (input method engines) when using vi or any other editor is painful. and, some of dont even both with capitalisation.
[04:31] <wm4> w shld skp ll vwls
[04:33] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:4ae87554f3c8: Makefile: add dependencies which require ffversion.h
[04:34] <Compn> so its all the keyboards fault :D
[04:34] <Compn> ehe
[04:38] <pross> or blame 7-bit ASCII
[04:39] <Compn> blame windows? ehe
[04:41] <Compn> if only bill gates made better language support, we wouldnt have been able to communicate with each other and open source would have never worked! 
[04:42] <pross> no no, ascii was created <1970
[04:42] <pross> blame AMERICANS
[05:04] <cone-112> ffmpeg.git 03Michael Niedermayer 07release/2.4:4afe2684d8f5: Makefile: add dependencies which require ffversion.h
[05:15] <cone-112> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n2.4.5': unknown revision or path not in the working tree.
[05:15] <cone-112> Use '--' to separate paths from revisions
[05:15] <cone-112> refs/tags/n2.4.5:HEAD: Makefile: add dependencies which require ffversion.h
[05:54] <t_g> pross: you can read Chinese?
[06:13] <t_g> Just finished running fate on Haiku nightly x86_64. Everything passes with a patch to Haiku (due to a bug in the definition of INT64_MIN)
[06:32] <pross> t_g: ù
[07:22] <t_g> pross: `ÀHö
[07:23] <t_g> pross: try to put "wm4¯šùìˆØt" into google translate: "wm4 certainly we are very happy" LOLOL
[07:42] <pross> t_g my terminal fucked that up!
[07:43] <pross> google translate knows wm4 better than we all do
[07:51] <t_g> pross: lol
[11:57] <nevcairiel> look, carl found another pkg-config case to lobby with :d
[12:09] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.4:577f1feb3fd1: mov: avoid a memleak when multiple stss boxes are present
[12:09] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.4:aa7a19b41774: mjpegdec: check for pixel format changes
[12:09] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:a058cbcfdf49: avformat/mov: Fix memleaks for duplicate STCO/CO64/STSC atoms
[12:09] <cone-49> ffmpeg.git 03Clément BSsch 07release/2.4:47e4a1ac6d1c: avformat/mov: strengthen some table allocations (cherry picked from commit 5ab882d7283f57560c889919c35f2688253b1d9c)
[12:09] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:36e6842aa63f: Merge commit '577f1feb3fd1e51fd14af7ce6d79d468faa3b929' into release/2.4
[12:09] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:81754d8f8b77: Merge commit 'aa7a19b41774ce5f8a4e43f3692a4f9d90aa5c92' into release/2.4
[12:20] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.4:55788572ea7b: jvdec: check frame dimensions
[12:20] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.4:69a930b988ff: mmvideo: check frame dimensions
[12:20] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:eac49477aa95: gifdec: refactor interleave end handling
[12:20] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:a7c18b628ffe: Merge commit '55788572ea7b89cdd77bab1cf4bf06d14ead34f5' into release/2.4
[12:20] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:54825047ab7f: Merge commit '69a930b988ff4f88ae27e4fc24ff6ed116840b5e' into release/2.4
[12:20] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:9b9411c1cf97: Merge commit 'eac49477aa95cf727d87d2741ee8e60be59d394b' into release/2.4
[12:31] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:58dc526ebf72: smc: fix the bounds check
[12:31] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:48952116352a: on2avc: check number of channels
[12:31] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:b86c2640f654: Merge commit '58dc526ebf722d33bf09275c1241674e0e6b9ef1' into release/2.4
[12:31] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.4:849f55368c36: Merge commit '48952116352ab03565cc14805e0f1d63cf0318fe' into release/2.4
[12:58] <cone-49> ffmpeg.git 03Julien Ramseier 07release/2.2:e7fdd6aa0d11: avconv: Use the mpeg12 private option scan_offset
[12:58] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.2:931f5b235112: mov: avoid a memleak when multiple stss boxes are present
[12:58] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:2d855c94b62b: avformat/mov: Fix memleaks for duplicate STCO/CO64/STSC atoms
[12:58] <cone-49> ffmpeg.git 03Clément BSsch 07release/2.2:c494be64119f: avformat/mov: strengthen some table allocations (cherry picked from commit 5ab882d7283f57560c889919c35f2688253b1d9c)
[12:58] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:3cd2c33d36cb: Merge commit 'e7fdd6aa0d11c3e144988d7090f972224799e07f' into release/2.2
[12:58] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:ede5946b7095: Merge commit '931f5b235112f1c2a09dead36f0a228061d23942' into release/2.2
[13:11] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.2:da4f5d9d7788: mjpegdec: check for pixel format changes
[13:11] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.2:8f238dd9bdd9: jvdec: check frame dimensions
[13:11] <cone-49> ffmpeg.git 03Anton Khirnov 07release/2.2:3f10a779b465: mmvideo: check frame dimensions
[13:11] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:bf00132a14b9: Merge commit 'da4f5d9d77882bee568266d764b95b51f81b7871' into release/2.2
[13:11] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:7fd20799d785: Merge commit '8f238dd9bdd9eba569fcaa564a07fbdd89412a14' into release/2.2
[13:11] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:475d68e0f501: Merge commit '3f10a779b465fd22d3aec1b744ca8544bc2da970' into release/2.2
[13:17] <BBB> ubitux: https://github.com/rbultje/ffmpeg/commits/vp9-32bit top 3 commits ok?
[13:20] <BBB> Im even wondering if (for the 48/84) it makes sense to do partial 4x8+8x8 transposes instead of a full 16x8
[13:21] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:92888e9ed4ea: gifdec: refactor interleave end handling
[13:22] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:f249e9889155: smc: fix the bounds check
[13:22] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:0d31a01d3a5c: Merge commit '92888e9ed4ea4e761ae953bbe28c85cc658abc8f' into release/2.2
[13:22] <cone-49> ffmpeg.git 03Michael Niedermayer 07release/2.2:c52c9f534e6a: Merge commit 'f249e9889155599ee3ad0172832d38f68b0c625d' into release/2.2
[14:27] <BBB> ubitux: poke
[14:45] <FreeSpirit-Girl> i dont think anyone is in ffmpeg channel .. so can someone help me reinstall the latest version of ffmpeg. i am using Ubuntu?
[14:49] <FreeSpirit-Girl> i think i accidentally uninstalled 
[14:52] <rcombs> that sounds more like a problem with your package manager than with ffmpeg
[14:53] <cone-49> ffmpeg.git 03James Almer 07master:7696e429c760: x86/vp3dsp: port put_vp_no_rnd_pixels8_l2_mmx to yasm
[15:05] <FreeSpirit-Girl> I thought i deleted it.. but I'm getting an error message stating impossible to link plugin pad
[15:09] <BBB> FreeSpirit-Girl: #ffmpeg-users please
[15:09] <BBB> FreeSpirit-Girl: this channel is for ffmpeg development discussions, not end-user assistance
[15:10] <FreeSpirit-Girl> thought i was helping the developers by telling you guys i got an error, sorry
[15:15] <michaelni> BBB, the user channel is #ffmpeg, i dont think theres a  #ffmpeg-users
[15:15] <BBB> oh, oops
[15:15] <BBB> oh right it says so in the channel topic
[15:21] <FreeSpirit-Girl> you dont have to get nasty. BBB maybe you told someone a person was asking in the help channel.. who knows.. I thought this statement applied to me "Questions about using FFmpeg or developing with libav*"
[15:24] <michaelni> FreeSpirit-Girl, you said you unsínstalled it, so things worked before ? if so do you remember what you did exactly to uninstall ?
[15:25] <FreeSpirit-Girl> i still got the error message
[15:28] <FreeSpirit-Girl> then i uninstalled it using synaptic pack .. hoping i could install the newest version, then
[15:30] <FreeSpirit-Girl> but please tell BBB to please dont get hostile over ffmpeg's crappy code and misdirect his anger at me.. it wasnt me who coded it.. Ill leave this channel,  Have a wonderful day 
[15:31] <michaelni> FreeSpirit-Girl, your errors are from brasero
[15:31] <michaelni> i dont know what the issue is
[15:32] <FreeSpirit-Girl> oh.. then i will look at that program.. thank you very much michaelni 
[15:47] <arwa> which filter should I proceed with ?
[15:48] <arwa> And are there any further changes required in fspp?
[15:49] <BBB> & omg
[15:59] <michaelni> arwa, let me check, but i think the width argumet also needed ptrdiff_t
[16:01] <michaelni> ohh the last already has width ptrdiff_t, sorry i missed that
[16:08] <michaelni> arwa, "./ffplay_g   -i matrixbench_mpeg2.mpg -vf fspp=1:1" crashes
[16:08] <michaelni> also crashes with cpuflags 0
[16:10] <arwa> ohh...!!
[16:11] <arwa> I will check it
[16:12] <arwa> But it is crashing with mp=fspp also!
[16:22] <michaelni> arwa,  fspp=4:40 and mp=fspp=4:40 work but look quite different
[16:26] <arwa> yes, they look quite different! I will look for the mistake.
[16:28] <michaelni> arwa, i see 2 calls to mul_thrmat_s() in the mp filter but only one in the new code, this may be related
[16:33] <kierank> yeah all BBB's fault
[16:33] <kierank> Crappy FFmpeg code =p
[16:42] <BBB> totally
[16:49] <michaelni> arwa, about the fspp=1:1 crash, the mp-fspp filter limits log2_count to 4 or 5 so 1 was silently changed to 4, while the new code left it at 1 i suspect 
[16:50] <arwa> yeah, i was just looking into the same thing
[16:51] <arwa> so should i limit log2_count between 4 and 5?
[16:52] <arwa> * to be
[16:57] <ubitux> BBB: sorry, i'm back
[16:58] <arwa> And regarding different ouputs, it is because, they are taking the range of qp to be from -15 to 32.
[16:59] <arwa> and in the current code, i am taking it to be from 0 to 63.
[16:59] <ubitux> BBB: about the first commit, you need to do 10 ’ 9 for the num of reg i suppose
[17:02] <ubitux> i wonder if that saved register can be re-used
[17:04] <cone-49> ffmpeg.git 03Michael Niedermayer 07master:081913aec53b: avformat/id3v2: Use av_freep() avoid leaving stale pointers in memory
[17:04] <cone-49> ffmpeg.git 03Michael Niedermayer 07master:bcd4447173c8: avformat/http: Use av_freep() avoid leaving stale pointers in memory
[17:04] <cone-49> ffmpeg.git 03Michael Niedermayer 07master:8db0c2ffe60d: avformat/mmst: Use av_freep() avoid leaving stale pointers in memory
[17:06] <michaelni> arwa, yes limiting it the same way as mp-fspp does should make comparision easier
[17:06] <arwa> I tried it that way also, but the output is still different!
[17:08] <ubitux> BBB: in the second commit, i think you're putting the 256B allocation for both functions, i don't think we want that
[17:10] <BBB> oh right
[17:10] <BBB> Ill fix that
[17:11] <BBB> fixed
[17:12] <ubitux> BBB: in the commit about ABSSUB_CMP (well spotted) i think you want to fix the "prototype" of the macro
[17:13] <BBB> fixed
[17:13] <BBB> (I swear I did that earlier and it disappeared!)
[17:13] <ubitux> :)
[17:13] <ubitux> these commits are really interesting
[17:13] <BBB> Im currently trying to see if I can remove the pb_80 part of CMP_LTE
[17:13] <BBB> and possibly make it 2 instead of 3 instructions
[17:14] <ubitux> :)
[17:14] <BBB> (by using pxor 0xffff.. combined with pcmpgt instead of pcmpgt, pcmpeq, por)
[17:16] <ubitux> i should have a look to this stuff again when you're done
[17:16] <ubitux> saving these regs is really interesting
[17:16] <BBB> I really want to work on 32bit, but its using so many regs that its kind of hard
[17:16] <BBB> (like, filter14, ...)
[17:19] <ubitux> BBB: yeah, sorry about that :D
[17:21] <ubitux> BBB: the tricky part is that you can not re-read lines you've written
[17:22] <ubitux> (i was fucked by this the first time :p)
[17:27] <arwa> Can I access the value of input arguments to the filter from config_input?
[17:27] <ubitux> yes
[17:28] <ubitux> AVFilterContext *ctx = inlink->dst;
[17:28] <ubitux> then get the fspp context from ctx->priv
[17:28] <ubitux> or directly inlink->dst->priv
[17:32] <BBB> ubitux: you mean you need to split in and out?
[17:32] <BBB> I can see how that sucks yes
[17:37] <ubitux> BBB: yeah :p
[17:37] <ubitux> is freenode on lsd again?
[17:37] <ubitux> ah, seems so.
[18:44] <allengates> can ffmpeg can vcxproj file for visual studio 2013?
[18:44] <allengates> can ffmpeg generate vcxproj file for visual studio 2013?
[18:45] <JEEB> no, you use the cross-platform build system with the MSVC compiler
[18:45] <allengates> j-b
[18:46] <allengates> yes, I can build on command line.
[18:46] <allengates> but it is easy to study, read, debug in IDE
[18:47] <JEEB> you can debug in MSVS just fine
[18:47] <JEEB> no need for a project file for that
[18:47] <JEEB> regarding reading you don't need a project file, either.
[18:48] <JEEB> the debug symbols created will note where the source files once were, and MSVC will thus show you the source code when debugging
[18:50] <allengates> a project is more convenient in my opinion.
[18:51] <nevcairiel> you are free to just create one for yourself, but maintenance will fall to you
[18:51] <wm4> obviously ffmpeg should switch to cmake
[18:51] <JEEB> I really don't see the reason behind a project file. A project file also means that you are hardcoding a configuration there, too :P
[18:52] <cone-118> ffmpeg.git 03Michael Niedermayer 07master:bb852f006688: avcodec/fdctdsp: use the 10bit code for 9 bit too
[18:52] <cone-118> ffmpeg.git 03Michael Niedermayer 07master:88c1869afe70: avcodec/idctdsp: use the 10bit code for 9 bit too
[18:52] <cone-118> ffmpeg.git 03Michael Niedermayer 07master:988ca9bd53c0: avfilter/vf_spp: add support for 9bit YUV and GBR as well as GBRP10
[18:52] <j-b> a project file is an extremely bad idea
[18:52] <JEEB> the only plus is that you can double-click something and hit ctrl+alt+f7 instead of opening a terminal and hitting make distclean && make -jX
[18:52] <j-b> it's unMaintainable
[18:53] <wm4> michaelni: is there any particular point to adding all pixel formats to vf_spp?
[18:54] <allengates> switch a file, make to breakpoint at the keypoint,  and run, and break, and you can see the stacktrace, and you can switch between each frames, and see the vars
[18:54] <allengates> s values.
[18:54] <wm4> practically all video is 8 bit yuv 420
[18:54] <j-b> GBRP10..
[18:58] <michaelni> wm4, no particular point just so its possible to run spp on videos which are in other colorspaces and also to just filter for other purposes than deblocking
[19:04] <j-b> how good is spp compared to postproc?
[19:14] <michaelni> spp should be alot better quality wise but its also alot slower at high quality settings
[19:24] <cone-118> ffmpeg.git 03Kieran Kunhya 07master:9cfa68c560bd: mpegts: add support for Opus
[19:24] <cone-118> ffmpeg.git 03Michael Niedermayer 07master:5318cf521f66: Merge commit '9cfa68c560bdec82d2d5ec079f9c5b0f9ca37af0'
[19:52] <jamrial> wm4: most anime encodes are 10bits lately, to the dismay of mobile users for that matter
[19:52] <BBB> ubitux: one more; I was able to get rid of the pxor in ABSSUM_CMP for the <=1 and H comparisons, the I/E ones are harder because they exceed [0-0x7f] range, I believe
[20:50] <cone-118> ffmpeg.git 03Michael Niedermayer 07master:1ac5abb1d062: avfilter/vf_spp: fix overflows with depth > 8
[22:16] <ubitux> BBB: sounds cool; just send the patchset to the mailing list when you're done, i'll have a deeper look
[22:29] <t_g> just because I'm a troll...
[22:29] <ubitux> this new nickname is disturbing for french people
[22:29] <ubitux> :)
[22:31] <t_g> http://www.urbandictionary.com/define.php?term=ta%20gueule ?
[22:31] <t_g> ah
[22:31] <ubitux> yeah
[22:31] <ubitux> :)
[22:31] <t_g> whoops
[22:31] <ubitux> :)
[22:32] <tim_g> better?
[22:32] <ubitux> sure
[22:32] <ubitux> i mean "t_g" was fine, it was just funny :)
[22:32] <tim_g> a German person pointed out libcaca has an offensive name a few days ago on #haiku
[22:33] <tim_g> when I said "somehow libcaca is such an interesting library..."
[22:37] <ubitux> you mean in german it means something else? or it's about the french meaning?
[22:40] <tim_g> ubitux: no "just" the French meaning. I didn't know that before even though I wrote the libcaca wikipedia article o_O
[22:41] <ubitux> haha
[22:41] <ubitux> well that's the naming signature of sam :P
[22:42] <tim_g> lol
[22:42] <tim_g> also libcucul, toilet
[22:48] <ubitux> ah btw, michaelni, since you wanted some new filters in ffmpeg
[22:48] <ubitux> http://people.csail.mit.edu/mrub/vidmag/  this one might be fun to do
[22:49] <ubitux> (the demo are impressive, but the annoying part is that it needs a lot of user inputs to make it useful)
[22:49] <ubitux> that's nevertheless a cool filter
[22:51] <iive> ENHANCE!
[22:53] <ubitux> iive: https://www.youtube.com/watch?v=Vxq9yj2pVWk 
[22:55] <iive> that's what I meant ;)
[22:55] <tim_g> ubitux: really? "This video is unavailable with Safety Mode enabled. To view this video, you will need to disable Safety Mode"
[22:56] <ubitux> i doubt it
[22:56] <ubitux> i'm never signed in youtube
[22:56] <ubitux> and i can watch it
[22:57] <tim_g> ubitux: I can't watch it *because* I signed in as a minor
[22:58] <tim_g> It's just that one scene at 1:11
[23:02] <iive> i think you can watch any video with youtube.com/v/<hashcode> or just use youtube-dl
[23:02] <ubitux> iive: not all the time unfortunately
[23:03] <ubitux> i sometime use the /embed/ trick
[23:03] <ubitux> but sometimes it's also blocked for $randomreason
[23:03] <iive> if it is taken down, nothing works... afaik
[23:05] <ubitux> i mean for video where you need age approval
[23:05] <ubitux> and you need to login to see it
[23:05] <iive> well, youtube-dl always gets it, so far.
[23:05] <ubitux> ah? interesting
[23:05] <ubitux> will try next time it happens
[23:10] <ubitux> http://www.ipol.im/pub/art/2014/104/ another interesting filter
[23:20] <michaelni> ubitux, great ideas, do they have GPL compatible source too ? because porting could be hard without that
[23:20] <ubitux> the video magnification has only a code in matlab iirc
[23:20] <ubitux> they're talking about c++ but never released the code (or at least i haven't found it)
[23:21] <ubitux> about the ipol one, the demo code is gpl
[23:22] <ubitux> but i'm not exactly fan of "porting" code like this
[23:22] <ubitux> i'd better have someone write it while understanding it
[23:23] <ubitux> i mean there is a paper for this, better provide an alternate implementation
[23:24] <michaelni> i prefer porting 5 filters than reimplementing 1
[23:27] <ubitux> i prefer 1 we can maintain that 5 no ones understand
[23:29] <michaelni> If one doesnt understand some code then reimplementing it generally isnt going to improve it
[23:31] <michaelni> anyway if implementation from scratch is the goal and assuming arwa is ok with that then the obvious next filter would be the spatial FFT one, we already have FFT code so it would be reasonable easy and surely would be usefull that is IMHO at least
[23:45] <iive> one usually understands the code by reimplementing it, that's why the software world is full of NIH things.
[23:49] <ubitux> i think it's kind of important to have different implementation of the same algorithms
[23:51] <iive> if the code implements a standard, then it is supposed that different implementations would have different bugs...
[23:51] <iive> what's the benefit for filters?
[23:52] <ubitux> iive: different bugs is a good thing; it means one security issue might not affect the others
[23:52] <ubitux> it has similar drawbacks though
[23:53] <ubitux> iive: and for filters, it means someone might think of a better implementation, simply by coming up with his own ideas, and not following what everyone does
[23:54] <iive> that was quick... (koda) 
[23:55] <iive> ubitux: researchers usually don't think about optimizations. :)
[23:57] <ubitux> that's why the one implementing the filter from ipol paper should probably not copy paste the code from the researchers.
[23:57] <ubitux> and actually reimplement it
[00:00] --- Sun Dec 21 2014


More information about the Ffmpeg-devel-irc mailing list