[Ffmpeg-devel-irc] ffmpeg-devel.log.20130618

[burek]

[00:04] <cone-390> ffmpeg.git 03Alexander Strasser 07master:c679a1c358c3: lavfi/src_movie: Check pointer is not NULL before deref
[00:04] <cone-390> ffmpeg.git 03Alexander Strasser 07master:ac604e446b5e: lavfi/src_movie: Use movie_common_init instead individual wrappers
[01:51] <cone-390> ffmpeg.git 03Carl Eugen Hoyos 07master:3cbf245f2a2c: vaapi: return early from ff_vaapi_render_picture() without picture
[02:04] <cone-390> ffmpeg.git 03Carl Eugen Hoyos 07master:28bf0d94aaf5: Revert "vaapi: return early from ff_vaapi_render_picture() without picture"
[02:04] <cone-390> ffmpeg.git 03Janne Grunau 07master:2ee380aefbb9: vaapi: return early from ff_vaapi_render_picture() without picture
[02:12] <cone-390> ffmpeg.git 03Janne Grunau 07release/1.2:054779625f76: vaapi: return early from ff_vaapi_render_picture() without picture
[02:31] <cone-390> ffmpeg.git 03Michael Niedermayer 07master:ca2818b88155: swresample/x86/audio_convert: add emms to CONV
[02:54] <cone-390> ffmpeg.git 03Michael Niedermayer 07release/0.11:d5b068d44b4b: swresample/x86/audio_convert: add emms to CONV
[02:54] <cone-390> ffmpeg.git 03Michael Niedermayer 07release/1.0:2ad70f57a88b: swresample/x86/audio_convert: add emms to CONV
[02:54] <cone-390> ffmpeg.git 03Michael Niedermayer 07release/1.1:d9a91dfb542a: swresample/x86/audio_convert: add emms to CONV
[02:54] <cone-390> ffmpeg.git 03Michael Niedermayer 07release/1.2:ae93d3405e5e: swresample/x86/audio_convert: add emms to CONV
[03:16] <lacrymology> excuse me, while trying to build melted, I got this: /usr/bin/ld: /usr/local/lib/libavformat.a(allformats.o): relocation R_X86_64_32 against `ff_a64_muxer' can not be used when making a shared object; recompile with -fPIC  any clues? I'm pretty sure I'm missing some flag to autoconf, but..
[09:43] <cone-631> ffmpeg.git 03Vittorio Giovara 07master:cd0faeee62ff: lavfi: math typo in interlace filter
[09:43] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:e41bf19d2cb0: Merge remote-tracking branch 'qatar/master'
[11:25] <wm4> ubitux: how is libavcodec subtitle charset conversion supposed to work with decoders that take binary data as input, such as movtext?
[11:38] <ubitux> mmh
[11:39] <wm4> the utf-8 bail out makes it impossible to handle such subs if they're not in utf-8 or 7 bit ascii
[11:39] <ubitux> maybe we should add a binary prop flag
[11:40] <ubitux> wm4: can you open a ticket so i don't forget?
[11:40] <wm4> to me it sounds like it'd better to recode the final output... or the input (in libavbformat) if possible
[11:41] <wm4> doing charset autodetection is a pain to do on packets
[11:41] <ubitux> that's what the mode is for
[11:41] <ubitux> in most circomstances it's better to have it at packet level
[11:42] <wm4> though I don't know if movtext will actually contain non-utf-8; but I bet people will break it anyway
[11:42] <wm4> what mode?
[11:42] <ubitux> i *think* mov text has some binary stuff at the beginning
[11:42] <wm4> yes it does
[11:42] <ubitux> sub_charenc_mode
[11:43] <wm4> not sure what you can do with that
[11:47] <wm4> ubitux: so shall I make a ticket? what should I write?
[11:48] <ubitux> i guess it causes trouble when trying to transcode subs from a mov/mp4 to another mov/mp4?
[11:49] <wm4> no, it was just an example
[11:50] <wm4> when using libavcodec to convert subtitles to ass
[11:50] <ubitux> how is ffmpeg doing that?
[11:50] <wm4> maybe there aren't even any other subtitle codecs with this property (binary in, text out)
[11:51] <ubitux> it's strange because we have some mov text tests
[11:51] <ubitux> or maybe it's a lucky shot because the binary data at the beginning is recognized as utf-8
[11:52] <wm4> the utf-8 text is after decoding AFAIK
[11:52] <wm4> but recoding is before
[11:52] <ubitux> ah indeed the utf-8 check is done on the decoded ass
[11:53] <ubitux> ok and so using char enc with mov input will not work
[11:53] <ubitux> wm4: do you have some charset issue with such subs ?
[11:53] <wm4> not yet
[11:54] <wm4> I just noticed it as weakness in the architecture
[11:54] <ubitux> that stuff is not finalized, so well yeah
[11:54] <ubitux> i wanted nicolas to continue but he doesn't seem much active lately :(
[11:55] <ubitux> so well mmh i don't know
[11:55] <ubitux> i need to think about this, and currently i don't want to :p
[11:56] <wm4> easiest is to recode the ass output
[11:56] <wm4> though that assumes that with multibyte encoding, the state is reset to ascii mode where it's needed
[11:57] <ubitux> pre recoding is necessary
[11:57] <ubitux> post could be done as well too yes
[11:57] <ubitux> that's what the mode thing was for initially
[11:58] <wm4> not sure why you would have both
[12:02] <ubitux> wm4: currently the internal representation of decoded subs is ass; it means it could be recoded without much issue; but what if it becomes a binary format of some sort?
[12:02] <wm4> then you have to change your architecture completely
[12:04] <ubitux> no, with pre-recoding it works fine
[12:04] <ubitux> (as it is currently)
[12:04] <ubitux> except for the issue you just pointed out of course
[12:05] <wm4> so how is binary data output?
[12:05] <wm4> will there be binary rectangles?
[12:06] <ubitux> no idea
[13:41] Action: durandal_1707 got multichannel working
[16:58] <cone-753> ffmpeg.git 03Michael Niedermayer 07master:9626d0e906f6: avutil/frame: Try to align width to achive linesize[0] alignment
[17:16] <michaelni> ubitux, whats the status of https://ffmpeg.org/trac/ffmpeg/ticket/1898 ? (you said there "Going to look into this") 
[17:16] <ubitux> mmh indeed
[17:17] <ubitux> i need to honor the review i got
[18:05] <michaelni> any volunteers who want to review the mips patches ?
[18:07] <gnafu> I could /look/ at them, but it wouldn't do any good :-D.
[18:08] <michaelni> we need look AND review :)
[18:09] <michaelni> there are also some non mips FFT things in there
[18:10] <gnafu> michaelni: I lack skills in all necessary areas, but I didn't want to leave you in silence since I was here :-).
[18:11] <michaelni> nevcairiel, ubitux wm4, any of you interrested in reviewing them ?
[18:12] <ubitux> mips, not really
[18:12] <michaelni> or maybe BBB, as the fixed point decoder patches may help on ARM .... android ARM
[18:15] <michaelni> or maybe some rockbox devels ight be interrested in the fixed point patches, do we have any in here ?
[18:20] <gnafu> Looks like funman isn't here, and he's the only one I know by name :-P.
[18:24] <cone-753> ffmpeg.git 03Matthew Heaney 07master:59b13750d2cd: Add maintainer for WebVTT demuxer and muxer
[18:24] <cone-753> ffmpeg.git 03Timothy Gu 07master:e3580449221c: doc/decoders: Document libcelt
[18:50] <ubitux> "In addition to specifying JPEG data, DefineBitsJPEG2 can also contain PNG image data and non-animated GIF89a image data"
[18:50] <ubitux> ._.
[18:51] <Daemon404> lulz
[18:52] <ubitux> there is also DefineBitsJPEG3 with the same "features"
[18:53] <ubitux> the only one being jpeg only is... "DefineBits"
[18:53] <ubitux> (let's not take risk mentioning jpeg)
[18:53] <Daemon404> DoAllTheThings()
[19:20] <burek> someone mentioned before something about having our own Q&A system, similar to stackoverflow
[19:20] <burek> if we are interested in that stuff, here is an open source thing that can help us bring it up and running very quickly http://www.osqa.net/
[19:21] <gnafu> Looks like Fedora uses Askbot for something like that.
[19:21] <gnafu> Seems nice.
[19:38] <kierank> i still haven't found a good OSS community thing
[19:38] <kierank> apart from irc
[19:40] <beastd> ml and irc. some also use bb and irc. i do not see those combos getting replaced soon.
[19:41] <kierank> bb?
[19:41] <JEEB> bulletin board
[19:41] <JEEB> probably
[19:41] <kierank> oh 
[19:42] <kierank> wiki's get spammed to pieces
[19:42] <kierank> like mine :(
[19:42] <burek> install fail2ban
[19:43] <kierank> fail2ban is for ssh
[19:43] <burek> not specifically
[19:43] <nevcairiel> how would fail2ban detect spam?
[19:43] <burek> it can analyze any log file you give it
[19:43] <beastd> yes, i meant bulletin board.  they are not really working for me though :)
[19:43] <burek> for example access.log
[19:43] <kierank> spammers try and make accounts
[19:43] <kierank> and then fail
[19:44] <kierank> because they can't do the captcha i guess
[19:44] <nevcairiel> well then they also dont manage to spam you
[19:44] <nevcairiel> so thats good? :)
[19:44] <kierank> some get through
[19:44] <kierank> what i don't understand is i also have anoynmous edits enabled
[19:44] <kierank> but they don't do that
[19:45] <burek> usually they make too fast sequential requests, which can trigger your iptables
[19:45] <JEEB> <kierank> wiki's get spammed to pieces <- yeah, I ended up disabling registrations and anonymous editing at a mediawiki setup I admin in '11
[19:45] <nevcairiel> i would probably rather use apache mod_security instead of trying to catch that with fail2ban
[19:45] <JEEB> because it suddenly got rather bad
[19:45] <nevcairiel> mod_security is quite neat for such things
[19:45] <burek> use whatever helps you :)
[19:46] <burek> but, you can't quite compare fail2ban with apache module, because it's more generic
[19:46] <kierank> what does mod_security do
[19:46] <nevcairiel> analyse requests, apply rules, block users if so configured
[19:47] <nevcairiel> can do simple things like number of requests per time, or even content based
[19:47] <nevcairiel> like someone trying to inject html in certain places might always be a bad guy, or stuff like that
[19:48] <burek> this might go offtopic, but the possibility to inject a html is actually a bad written server side script
[19:48] <nevcairiel> its really flexible when you want to protect a website, but of course that also requires careful config or you block the innocent
[19:48] <nevcairiel> burek: well sure, but someone trying anyway can help you to flag spammers :D
[19:49] <burek> i agree :)
[22:27] <wm4> hm wasn't ffmpeg supposed to be able to mux webvtt into matroska, or is that still in the pipeline?
[22:27] <Daemon404> into webm
[22:28] <Daemon404> "matroska"
[22:44] <ubitux> wm4: yes i need to review some stuff
[22:44] <ubitux> the thing is, webvtt is used for chaptering, metadata and stuff
[22:45] <ubitux> so there is some slight api changes along with it
[22:45] <ubitux> which i want to be cautious with
[22:45] <wm4> sounds terrible
[22:46] <ubitux> so far the trick is to mark the stream with a disposition
[22:47] <ubitux> and use that disposition in the webm muxer
[22:47] <ubitux> that's simple and relatively straightfoward, but i'm not sure that's the best way
[22:47] <ubitux> i don't like to have it separated from the chaptering system for instance
[22:53] <MadTBone> What's the state of VC-1 Progressive Segmented Frame decoding support?  I have sample files if they're needed.
[22:59] <cone-753> ffmpeg.git 03Michael Niedermayer 07master:9d73e3025371: ff_gen_search: make step 64it to prevent hypothetical integer overflow
[22:59] <cone-753> ffmpeg.git 03Michael Niedermayer 07master:b3dd50cd0dbb: ff_gen_search: fix limit
[22:59] <cone-753> ffmpeg.git 03Michael Niedermayer 07master:f8ca8138f01d: ff_gen_search: fix infinite loop
[23:00] <michaelni> MadTBone, if its not supported yet then a patch is welcome
[23:00] <michaelni> and samples are welcome too if there are none yet
[23:00] <michaelni> on samples.ffmpeg.org
[23:03] <Compn> ubitux : what about native speex decoder ?
[23:03] <ubitux> why do you ask me?
[23:03] <Compn> oh i think paul was asking me 
[23:03] <Compn> nevermind
[23:30] <cehoyos> michaelni: Hi, is the problem that the sample from ticket 2639 cannot be seeked the same as ticket 1170 ?
[23:51] <michaelni> cehoyos, i think they are unrelated but i didnt really investigate
[23:56] <cone-753> ffmpeg.git 03Michael Niedermayer 07release/1.2:04d69466003f: ff_gen_search: make step 64 bit to prevent hypothetical integer overflow
[23:56] <cone-753> ffmpeg.git 03Michael Niedermayer 07release/1.2:c77b3737b9d1: ff_gen_search: fix limit
[23:56] <cone-753> ffmpeg.git 03Michael Niedermayer 07release/1.2:c0445df5b359: ff_gen_search: fix infinite loop
[00:00] --- Wed Jun 19 2013