[Ffmpeg-devel-irc] ffmpeg-devel.log.20130101

[burek]

[00:07] <ubitux> http://prog21.dadgum.com/161.html
[00:58] <cone-112> ffmpeg.git 03Carl Eugen Hoyos 07master:67abe21c0bdc: mov: fix handling of odd sized yv12
[00:58] <cone-112> ffmpeg.git 03Michael Niedermayer 07master:f105fe5c22c5: rawdec: fix decoding of odd dimensioned raw yv12
[01:05] <nevcairiel> didnt someone tell these people that 4:2:0 with odd dimensions is rather invalid? yet they still create such files :(
[01:06] <wm4> the jpeg decoder creates such output, and these are definitely valid
[01:07] <nevcairiel> how can you subsample chroma by a factor of two if you didnt have an even number of pixels to start with?
[01:08] <nevcairiel> do you create one chroma too much or one too few?
[01:08] <wm4> one too much
[01:09] <wm4> and the last luma row or column is simply missing, I suppose
[01:10] <nevcairiel> isnt jpeg also macroblock based, so its always mod16 coded?
[01:11] <wm4> maybe, but the image dimensions can obviously be arbitrary
[01:11] <wm4> I wonder if ffmpeg has a policy how odd sizes should be handled?
[01:11] <nevcairiel> if you look at those commits up there, they decided to just cut one pixel :)
[01:12] <wm4> the jpeg decoder still outputs odd sizes
[01:26] <iive> nevcairiel: the visible pixels could be subset of the coded pixels
[01:26] <iive> this is why these are separate dimensions.
[02:02] <cone-112> ffmpeg.git 03Nicolas George 07master:2eb2e1798eee: lavfi: add avfilter_get_audio_buffer_ref_from_arrays_channels.
[03:07] <Compn> 4k v210
[03:07] <Compn> sounds like ... high quality
[05:31] <pross-au> michaelni: do you have any thoughts on what to do with bayer pixel formats that use 10 or 12 bits per pixel and are *NOT* padded out to 16. Such formats come from webcams (v4l) and still cameras. Does it make sense to have AV_PIX_FMT_BAYER_xxx for these variants, or better to process them in libavcodec, e.g. get_bits(gb,12) and shifting them to make 16-bits per pixel.
[05:33] <michaelni> pross-au, iam in favor of aligning the MSB and using just a 16bit pix_fmt and storing the true bpp elsewhere
[05:34] <pross-au> ok
[05:48] <cone-750> ffmpeg.git 03Michael Niedermayer 07master:13131685f187: ffmpeg: avoid negative timestamps for DVD
[11:08] <cone-159> ffmpeg.git 03Clément BSsch 07master:c1abfcd7d640: doc/resampler: remove .html from link text.
[12:17] <durandal_1707> ping for my fate patch
[12:40] <durandal_1707> looks like nobody use w64 format
[13:03] <Compn> ehe
[13:12] <ubitux> saste: http://pastie.org/private/flb6qqrsnw8iqvbydaf7q
[13:12] <ubitux> did you see this?
[13:13] <ubitux> saste: also, michael spotted an aliasing problem in ffprobe, which actually sounds fishy for some reason
[13:13] <saste> ubitux, how do you get that warning?
[13:14] <ubitux> "make"
[13:14] <ubitux> :)
[13:14] <saste> which aliasing problem
[13:14] <ubitux> (i'm using gcc 4.7)
[13:15] <ubitux> saste: http://fate.ffmpeg.org/log.cgi?time=20130101081420&log=compile&slot=armv7l-panda-gcc4.4-armv4t
[13:15] <ubitux> look for ffprobe.c
[13:16] <ubitux> and this led me to: 03:31  >>> ubitux wonders where these children_ids are init&
[13:16] <ubitux> but i didn't dig further
[13:26] <durandal_1707> why nobody cares about #1163?
[13:27] <ubitux> i didn't see the patch on the ml
[13:28] <durandal_1707> some kind of patch is in ticket
[13:29] <ubitux> :/
[13:32] <ubitux> erm
[13:32] <ubitux> there is some crap going on here
[13:34] <cone-159> ffmpeg.git 03Clément BSsch 07master:d079d1d368a9: ffprobe: do not escape '"' between single quotes.
[13:38] <ubitux> durandal_1707: that code looks mostly broken and inefficient
[13:38] <durandal_1707> #1539 can be closed
[13:39] <durandal_1707> than fix it :/
[13:39] <ubitux> sounds like another bug for saste ;)
[13:40] <ubitux> i need to fix one of mine now&
[13:55] <durandal_1707> how typos can happen?
[13:58] <cone-159> ffmpeg.git 03Martin Storsjö 07master:c76daa89ab91: rtmp: Return a proper error code in handle_invoke_error
[13:58] <cone-159> ffmpeg.git 03Martin Storsjö 07master:33f28a3be309: rtmp: Add a function for writing AMF strings based on two substrings
[13:58] <cone-159> ffmpeg.git 03Michael Niedermayer 07master:c047a41bc5ff: Merge commit '33f28a3be3092f642778253d9529dd66fe2a014a'
[13:59] <saste> durandal_1707, the invalid UTF-8 codes are stored in the metadata, fixing them at the application level is not a good idea
[13:59] <saste> we should have some validation/fixing function in the demuxer maybe
[13:59] <durandal_1707> +1
[14:00] <cone-159> ffmpeg.git 03Paul B Mahol 07master:8e4e532bb28e: lavfi/avcodec: fix typo
[14:09] <cone-159> ffmpeg.git 03Martin Storsjö 07master:08225d01262b: rtmp: Add support for adobe authentication
[14:09] <cone-159> ffmpeg.git 03Martin Storsjö 07master:c1ea44c54d83: rtmp: Add support for limelight authentication
[14:09] <cone-159> ffmpeg.git 03Michael Niedermayer 07master:9336a980bb26: Merge remote-tracking branch 'qatar/master'
[14:50] <ubitux> oh.
[14:50] <ubitux> nevcairiel: i think i found a better way :)
[14:51] <ubitux> and btw, your patch doesn't really fix the problem
[14:59] <ubitux> i think i'm going to simplify that whole timecode stuff
[15:01] <ubitux> mmh wait your patch might be ok :)
[15:02] <ubitux> almost
[15:05] <cone-159> ffmpeg.git 03Paul B Mahol 07master:8c33cb5c77b9: fate: add nistsphere demuxer test
[15:18] <nevcairiel> ubitux: well it fixed my problem, of chap nolonger being parsed :)
[15:18] <cone-159> ffmpeg.git 03Hendrik Leppkes 07master:765158dd82a2: mov: fix parsing of the chap atom.
[15:18] <ubitux> nevcairiel: yes it is good
[15:18] <ubitux> thanks :)
[15:19] <nevcairiel> the context variables are named a bit oddly
[15:19] <ubitux> yes
[15:19] <nevcairiel> should rather be tmcd_count or something like that
[15:19] <ubitux> i was planning to change this
[15:22] <ubitux> i don't think i can reference more than one timecode track per stream anyway so&
[15:27] <cone-159> ffmpeg.git 03Nicolas George 07master:82541d833016: lavfi: merge all filtering code into ff_filter_frame.
[15:36] <ubitux> is there a function to get a stream from its stream->id or we have to write the loop all the time?
[15:53] <cone-159> ffmpeg.git 03Michael Niedermayer 07master:9b53306178ca: cmdutils: we have 2013
[15:54] <durandal_1707> ubitux: what is id used for?
[15:58] <ubitux> it helps me identify the references
[16:06] <saste> ubitux: you confused nicolas with the AV_BASE64_SIZE thing!
[16:06] <ubitux> haha
[16:06] <ubitux> my stupidity is so evil!
[16:07] <saste> maybe a longer name AV_BASE64_ENCODE_SIZE was better
[16:10] <ubitux> saste: any objection to ffprobe stream->id?
[16:11] <saste> ubitux, no but let me read the patch
[16:17] <cone-159> ffmpeg.git 03Nicolas George 07master:69a1667b064b: ffmpeg: mark attachment streams finished.
[16:22] <cone-159> ffmpeg.git 03Clément BSsch 07master:8d398f40b4a1: lavf/mov: simplify timecode track ref.
[16:22] <cone-159> ffmpeg.git 03Clément BSsch 07master:c14f8a52b0aa: lavf/mov: fix timecode track reference usage.
[16:43] <cone-159> ffmpeg.git 03Clément BSsch 07master:1dd4a21cb787: fate: re-enable fate-sub-mpsub.
[16:43] <cone-159> ffmpeg.git 03Clément BSsch 07master:7d66bc792024: lavc/movtextdec: keep the min size instead of max to fix overread.
[17:11] <saste> ubitux: ffprobe already prints the stream id
[17:12] <ubitux> oh&
[17:12] <ubitux> i'm stupid
[17:12] <ubitux> thanks :P
[17:17] <saste> ubitux, no, you're working too much ;-)
[17:18] <ubitux> if only&
[17:24] <cone-159> ffmpeg.git 03Clément BSsch 07release/1.0:68e48ed72e05: lavc/movtextdec: keep the min size instead of max to fix overread.
[17:40] <cone-159> ffmpeg.git 03Hendrik Leppkes 07release/1.0:5f95b300c121: mov: fix parsing of the chap atom.
[17:45] <saste> int len = strlen(p) * 6 / 8;
[17:45] <saste> ^ seriously??
[17:46] <wm4> let me guess, this code calculates how large a buffer allocation is?
[17:47] <saste> 6/8 baffles me
[17:47] <durandal_1707> where?
[17:47] <ubitux> simplifying fraction can be hard
[17:47] <ubitux> durandal_1707: libavformat/rtp_asf.c
[17:47] <ubitux> base64 stuff
[17:47] <saste> yes
[17:48] <saste> i'm not sure that size is correct
[17:48] <ubitux> 6/8 looks bigger than 3/4 so it should be safe
[17:49] <wm4> I like how it doesn't check the return value of the malloc either
[17:49] <durandal_1707> like?
[17:49] <saste> ubitux, lol
[17:54] <saste> but yes the use of the function is correct
[17:55] <saste> on the other hand the API is inconsistent
[17:55] <saste> decode writes a not nul-terminated buffer
[17:55] <saste> encode writes a nul-terminated buffer
[17:56] <saste> AV_BASE64_SIZE() macro name is confusing
[17:56] <saste> if you just want to encode a buffer (no null termination) you're out of lack
[17:58] <saste> ^^ that's what happens when you "steal" VLC code :-)
[18:06] <cone-159> ffmpeg.git 03Piotr Bandurski 07master:fde13052fb82: aiffdec: set block duration for QDM2
[18:34] <cone-159> ffmpeg.git 03Clément BSsch 07release/0.10:c3d7c805bc9c: lavc/ass_split: check for NULL pointer in ff_ass_split_override_codes().
[18:34] <cone-159> ffmpeg.git 03Clément BSsch 07release/0.11:8bc0127b3d59: lavc/ass_split: check for NULL pointer in ff_ass_split_override_codes().
[18:34] <cone-159> ffmpeg.git 03Clément BSsch 07release/0.7:20940780c726: lavc/ass_split: check for NULL pointer in ff_ass_split_override_codes().
[18:34] <cone-159> ffmpeg.git 03Clément BSsch 07release/0.8:80b8dc30dc96: lavc/ass_split: check for NULL pointer in ff_ass_split_override_codes().
[18:34] <cone-159> ffmpeg.git 03Clément BSsch 07release/0.9:42abae1125d7: lavc/ass_split: check for NULL pointer in ff_ass_split_override_codes().
[18:38] <cone-159> ffmpeg.git 03Clément BSsch 07release/1.0:20c121c00747: lavc/ass_split: check for NULL pointer in ff_ass_split_override_codes().
[18:54] <durandal_1707> omg, why is bitrate int
[18:57] <cone-159> ffmpeg.git 03Paul B Mahol 07master:ba1cbf40cebf: lavfi/lut: use ff_fill_rgba_map()
[19:38] <cone-159> ffmpeg.git 03Nicolas George 07master:ecda1d3cbe1c: lavf: add data: URI scheme.
[19:39] <cone-159> ffmpeg.git 03rogerdpack 07master:10a927756235: better filter error message when accidentally reusing outputs
[20:00] <cone-159> ffmpeg.git 03Nicolas George 07master:951a03b3a8f2: ffmpeg: move stream_copy flags handling after attachments.
[20:00] <cone-159> ffmpeg.git 03Nicolas George 07master:5bab99baea4f: ffmpeg: apply flags to attachments too.
[20:00] <cone-159> ffmpeg.git 03Nicolas George 07master:8dbbaf568ed9: lavf/matroskaenc: respect bitexact for attachments.
[20:17] <cone-159> ffmpeg.git 03Nicolas George 07master:5ed5e90f2ae2: lavc/dvdsubenc: improve color distance function.
[20:20] <durandal_1707> anyone use PA-AF format ?
[20:21] <kierank> what is that
[20:22] <kierank> oh some mpeg thing
[20:22] <kierank> never heard of it
[20:22] <kierank> durandal_1707: i think pa-af is just a metadata store for other containers, no?
[20:23] <durandal_1707> http://www.mb-1.jpn.org/paaf/
[20:25] <durandal_1707> it some crap that is misdetected as mov
[20:26] <durandal_1707> it looks to be withdrawn
[20:27] <durandal_1707> i found it on page of als 
[20:56] <kierank> withdrawn?
[20:57] <kierank> looks active to me
[20:57] <durandal_1707> but i cant get spec
[20:57] <kierank> iso specs cost $$$$
[21:00] <durandal_1707> right, i ignored newer version here: http://www.iso.org/iso/home/store/catalogue_ics/catalogue_detail_ics.htm?csnumber=61312
[21:46] <Compn> ack durandal left
[21:46] <Compn> wanted to ask him how far he got with evrc 
[00:00] --- Wed Jan  2 2013