[Ffmpeg-devel-irc] ffmpeg-devel.log.20130702

[burek]

[00:01] <cone-241> ffmpeg.git 03Paul B Mahol 07master:51f1bf330969: bfi: check return value of av_malloc()
[00:07] <cone-241> ffmpeg.git 03Paul B Mahol 07master:16e0416fa47c: idcin: check return value of av_malloc()
[00:19] <cone-241> ffmpeg.git 03Paul B Mahol 07master:e1b0d3a389ff: westwood_vqa: check return value of av_malloc()
[00:46] <cone-241> ffmpeg.git 03Michael Niedermayer 07master:bf87908cd8da: rmdec: Pass AVIOContext to rm_read_metadata()
[01:48] <wm4> saste: so is the icy patch ok?
[01:49] <saste> wm4: yes
[01:49] <saste> should I push it?
[01:49] <wm4> nice
[01:49] <wm4> sure
[01:49] <saste> i don't know if you have commit rights
[01:49] <wm4> I regret that it's so... specific though
[01:49] <wm4> no I don't
[01:56] <saste> wm4: for "inline" I meant the inline keyword
[01:56] <wm4> saste: I think in this case, the compiler knows better anyway
[01:56] <wm4> it's a static function
[01:58] <saste> wm4, i don't mind (it's a nit and you can ignore it altogether)
[02:20] <cone-241> ffmpeg.git 03wm4 07master:a92fbe16f2dc: lavf/http: add support for reading streamcast metadata
[02:20] <wm4> saste: thanks
[03:28] <cone-241> ffmpeg.git 03Michael Niedermayer 07master:f27b22b4974c: h264: move 444 border xchg under if (deblock_top)
[03:28] <cone-241> ffmpeg.git 03Michael Niedermayer 07master:a35494c56691: h264: fix indention of xchg_mb_border()
[10:23] <saste> michaelni, what's the use case for dump_extradata (bitstream filter)?
[10:25] <j-b> good morning
[10:29] <saste> j-b, mood gorning
[10:31] <j-b> :)
[10:31] <j-b> come stai?
[11:12] <saste> hi j-b, tutto bene grazie :)
[11:32] <JEEB> hmm, I guess I'll make an issue on the tracker for the ogm seeking thing
[11:32] <JEEB> since it seems I've registered onto the tracker at some point o_O
[11:32] Action: JEEB has no recollection altogether of it
[11:35] <JEEB> ahaha
[11:35] <JEEB> that didn't take a while at all
[11:35] <JEEB> to find out what uname/pw combo I was using
[11:36] <wm4> was the password "password"?
[11:38] <JEEB> I have a password that contains 'password' on a certain service because they kept sending me my password back in plain text
[11:41] <cone-468> ffmpeg.git 03Luca Barbato 07master:8bd9039900e6: Revert "indeo5: reject negative motion vectors"
[11:41] <cone-468> ffmpeg.git 03Michael Niedermayer 07master:bc62b24d113e: Merge commit '8bd9039900e67a633cc838d691c449e71c060553'
[11:47] <cone-468> ffmpeg.git 03Loren Merritt 07master:c8b920a9b7fa: lls/x86: use 3-operator vaddpd in ADDPD_MEM
[11:47] <cone-468> ffmpeg.git 03Michael Niedermayer 07master:d1fa6718952c: Merge commit 'c8b920a9b7fa534a6141695ace4e8c2dfcd56cee'
[11:56] <cone-468> ffmpeg.git 03Anton Khirnov 07master:8ad3267ce3a1: oggdec: do not fall back on binary search in the generic code.
[11:56] <cone-468> ffmpeg.git 03Michael Niedermayer 07master:951a36550480: Merge remote-tracking branch 'qatar/master'
[12:13] <JEEB> do I have to provide the text file when uploading to the samples ftp in case I've made a trac issue?
[12:14] <cone-468> ffmpeg.git 03Michael Niedermayer 07master:a478e99a603f: avutil/x86: reenable ff_update_lls_avx()
[13:26] <Compn> JEEB : no, but its nice if you use filename like issue2711.avi or so :P
[13:30] <JEEB> Compn, that's a given :)
[14:04] <durandal_1707> michaelni: why av_parser_init returns NULL when error happens when initializating parser?
[14:04] <durandal_1707> NULL is returned when no parser is found
[14:04] <durandal_1707> no parser is found != error when initializing parser
[14:05] <nevcairiel> what should it return otherwise? :p
[14:05] <durandal_1707> its just stupid design
[14:06] <durandal_1707> now i should need to introduce av_parser_init2 to fix this incosistency
[14:06] <nevcairiel> its not like parser init can really fail
[14:06] <nevcairiel> only in OOM
[14:06] <durandal_1707> and than it fails
[14:07] <durandal_1707> and says that it cant find parser, where there is parser
[14:07] <nevcairiel> feel free to log an extra message in the error case in init so you know that it failed
[14:07] <nevcairiel> adding a init2 just for that is stupid API bloat
[14:11] <JEEB> ok, sample is uploading and https://trac.ffmpeg.org/ticket/2739 is done
[14:16] <wm4> durandal_1707: libavformat doesn't even read mid-stream ogg tags
[14:17] <wm4> it tries to read them as packets instead
[14:17] <wm4> printing funny error messages
[14:26] <durandal_1707> wm4: i don't want to touch that shit
[14:31] <JEEB> and sample is fully uploaded
[15:02] <xlinkz0> has anyone used the ffmpeg4android build scripts?
[15:09] <michaelni> xlinkz0, silly awnser but i guess its author would have used it
[15:09] <xlinkz0> i would guess so too lol
[15:10] <durandal_1707> ubitux: writing AV_OPT_BOOLEAN?
[15:14] <cone-60> ffmpeg.git 03Carl Eugen Hoyos 07master:2492f7774f6c: avformat: Print "Consider increasing the value for analzeduration and probesize" even more often
[15:14] <cone-60> ffmpeg.git 03Michael Niedermayer 07master:ccba9535dc35: av_find_stream_info: initialize ret
[15:14] <cone-60> ffmpeg.git 03Michael Niedermayer 07master:c826c5698b9c: avformat_find_stream_info: try to preserve the error code
[15:14] <cone-60> ffmpeg.git 03Michael Niedermayer 07master:c5ddd753549c: avformat_find_stream_info: Do not consider no streams an error in flush codecs
[15:14] <cone-60> ffmpeg.git 03Michael Niedermayer 07master:d59ec0adbbc4: avformat_find_stream_info: move ret init down from flush codecs.
[16:41] <cone-60> ffmpeg.git 03Jean Delvare 07master:de8d56c6e3f0: MAINTAINERS: clarify the expected format in the document header
[16:41] <cone-60> ffmpeg.git 03Jean Delvare 07master:0472f64c5527: Add a maintainer for the delogo filter
[16:49] <michaelni> saste, do you want to do another review on the delogo SAR or ok if i apply it ?
[17:03] <t4nk188> Hi. I hope this is the right place to ask: Is there a way how I can control FFmpeg to write the same (ideally exactly *one*) number of AAC samples/frames into each "chunk" of a MP4 file?
[17:03] <durandal_1707> t4nk188: i already told you there is no way
[17:03] <durandal_1707> the only question is: is it possible at all...
[17:04] <durandal_1707> anyway you can try implementing it ...
[17:05] <t4nk188> What about -chunk_size and -chunk_duration options?
[17:06] <t4nk188> I see no reason why it should not be possible, in theory. Also, I wonder, how does FFmpeg's MP4 muxer decide it currently?
[17:06] <t4nk188> The chunk sizes are varying, but what is the reasoning?
[17:07] <t4nk188> Implementing it in FFmpeg certainly would be *much* work for us. And currently we want the evaluate different possibilities and FFmpeg is one....
[17:07] <durandal_1707> t4nk188: the chunk_* thing is not used in mp4 for whatever reason
[17:07] <Paranoialmaniac> -chunk_* option is global option for libavformat
[17:08] <durandal_1707> but its not used by mov muxer
[17:08] <Paranoialmaniac> not specific for mov/mp4 muxer
[17:08] <t4nk188> okay, thanks for confirming
[17:08] <t4nk188> there goes that idea...
[17:08] <t4nk188> So, currently FFmpeg's MP4 muxer has no option for the chunk size, right? Only for "fragment" size, right?
[17:09] <Paranoialmaniac> t4nk188: you try just muxing?
[17:09] <Paranoialmaniac> or encoding and muxing simultaneously?
[17:09] <t4nk188> Nope, I do conversion from WMA to AAC in MP4
[17:10] <Paranoialmaniac> so simultaneously
[17:10] <durandal_1707> chunk_size/duration have nothing to do with whay you are talking about...
[17:11] <t4nk188> I do not talk about duration, sorry
[17:11] <durandal_1707> so private mov muxer options have more sense
[17:11] <t4nk188> I talk about number of samples per chunk
[17:11] <t4nk188> Obviously it is NOT constant, but varries
[17:11] <durandal_1707> then write it or ask someone to write it...
[17:11] <t4nk188> But why? And what is it controlled by?
[17:12] <t4nk188> There must be a reason why 3221 samples are in the first chunk and 3343 in the second
[17:12] <durandal_1707> look at code
[17:12] <t4nk188> I don't think it's arbitary...
[17:12] <durandal_1707> usually there are no reason at all
[17:12] <t4nk188> huh?
[17:13] <t4nk188> FFmpeg is not a small project, so "look at the code" is not really helpful
[17:13] <t4nk188> Especially before I know what I am looking for exactly
[17:14] <t4nk188> I do not want to spend a lot of time on this, just to see it doesn't work
[17:14] <t4nk188> Our time is limited and FFmpeg just one option we are evaluating
[17:14] <durandal_1707> i told 10 times already that it can be done with current code
[17:14] <durandal_1707> *can NOT
[17:15] <t4nk188> I think I got that, but still I'd like to know how it is controlled currently
[17:15] <durandal_1707> for some codecs its fixed and for others its not
[17:15] <Paranoialmaniac> ask who knows the control
[17:16] <t4nk188> So the encoder tells the MP4 muxer when the "start" the next chunk?
[17:16] <durandal_1707> nope
[17:17] <t4nk188> Then, can you explain "for some codecs its fixed and for others its not" a bit more, please?
[17:25] <t4nk188> Okay, then, could you at least give me pointer where I have to look? At the AAC encoder (e.g. faac) or at the MP4 muxer (libavformat, I guess)?
[17:25] <durandal_1707> why would you look if you are not interested in changing it?
[17:25] <t4nk188> I did not say I'm not interested in changing it
[17:25] <durandal_1707> obviosly code that covers it is in libavformat/movenc.c
[17:25] <t4nk188> But I cannot decide before I have understood what  the matter is
[17:26] <t4nk188> You see, before we go on with the project, we need to decide the direction
[17:26] <t4nk188> If FFmpeg could can be changed in 1-2 work days, that would be okay
[17:27] <t4nk188> But currently I have no idea how much work it would be (at least I now know FFmpeg can not do it as-is)
[17:27] <t4nk188> I will investigate movenc.c, thanks...
[17:29] <durandal_1707> see mov_write_stsc_tag in movenc.c
[17:30] <Paranoialmaniac> mov_write_stsc_tag does nothing about samples per chunk 
[17:30] <Paranoialmaniac> it just writes stsc atom
[17:31] <t4nk188> But I needs to have the info "sample per chunk" for each chunk
[17:31] <t4nk188> it
[17:31] <t4nk188> so I have to check where it gathers that info...
[17:31] <durandal_1707> but it actually writes samples per chunk
[17:31] <Paranoialmaniac> mov_write_stsc_tag does not decide samples per chunk
[17:32] <durandal_1707> i did not said it does
[18:07] <ubitux> durandal_1707: no i'm not
[18:54] <cone-60> ffmpeg.git 03wm4 07master:c3316ddb83e7: lavf: add support for libgme
[19:02] <saste> who can help with the MP4 format?
[19:02] <saste> i have a custom libavf application, and the resulting MP4 can be played by ffmpeg and vlc, but mp4info complains with:
[19:03] <saste> ReadProperties: atom 'avcC' is too small; overrun at property: configurationVersion (src/mp4atom.cpp,386)
[19:03] <saste> the file is not playable by Quicktime
[19:04] <saste> with MP4Box -info: [iso file] Box "avcC" size 8 invalid (read 909)
[19:05] <durandal_1707> its bad file
[19:05] <JEEBsv> sounds like you are failing to create a correct avcC structure
[19:05] <JEEBsv> time to go read 14496-15
[19:06] <JEEBsv> I mean, you can even put Annex B into "MP4" and ffmpeg and friends will play it :P Doesn't mean it's valid tho
[19:07] <JEEBsv> in other words, testing with ffmpeg-using things only is not a good idea
[19:07] <durandal_1707> because ffmpeg devs are usually lazy
[21:20] Action: durandal_1707 why this crap doesn't work!!!!!
[21:50] <superware> michaelni: hi man. any ETA for http://ffmpeg.org/pipermail/ffmpeg-devel/2013-June/145286.html ? :|
[21:55] <michaelni> superware, if you test it, it will be sooner than if not
[21:56] <michaelni> burek, can you make fflogger use trac.ffmpeg.org instead of ffmpeg.org/trac/ffmpeg
[21:56] <superware> I'm worried I might spend hours before I can compile the working win32 libraries
[21:58] <superware> michaelni, can you please compile the related win32 dll?
[21:59] <superware> if it's not too much trouble
[22:00] <michaelni> superware, you misunderstand something i think
[22:00] <michaelni> we are all volunteers
[22:00] <michaelni> its a issue you want fixed
[22:01] <michaelni> so you do the work or you pay someone to do it for you or you wait
[22:01] <michaelni> the patchset should be tested & reviewed ideally
[22:02] <superware> ok, I want to test it
[22:02] <michaelni> ok, so get ffmpeg git, apply the patchset and build it
[22:02] <michaelni> see our documentation and wiki on how to do all that
[22:03] <michaelni> and please improve the wiki where its not clear enough
[22:03] <superware> ok, but I'm afraid the process might turn out too challenging for me
[22:04] <michaelni> https://trac.ffmpeg.org/wiki/CompilationGuide
[22:05] <superware> ok, thanks
[22:33] <durandal_1707> if bit reader is little endian for specific crap, and others crap in same file use big endian bit reader is placing function in separate file/header best option?
[22:33] <durandal_1707> bswaping buffer is silly
[22:54] <superware> michaelni, building win32 ffmpeg turns out to be too complex for me :(
[22:54] <superware> I'll ask for help in #ffmpeg
[23:24] <durandal_1707> finally bug fixed
[23:32] <durandal_1707> argh, recompiling again....
[23:38] <durandal_1707> wm4: gonna add libmng?
[00:00] --- Wed Jul  3 2013