[Ffmpeg-devel-irc] ffmpeg-devel.log.20130512

[burek]

[00:11] <saste> ffplay -f lavfi "nullsrc [bg]; color=c=random at 0.95, [bg] overlay"
[00:29] <ubitux> saste: i still don't get the point of nullsrc
[00:30] <ubitux> testing filters with uninitialized buffers?
[00:30] <saste> ubitux, you don't always need to initialize data
[00:31] <saste> if the data will be overwritten
[00:31] <ubitux> should we add/restore a "read" permission?
[00:31] <ubitux> what is the point of nullsrc in practice anyway?
[00:42] <llogan> i guess as a "canvas" that will be overwritten by something else like an overlay or whatever?
[00:43] <ubitux> unless i'm mistaken, you're looking for a transparent frame?
[00:43] <saste> llogan, yes
[00:43] <saste> ubitux, that's not a bug, but the expected behavior
[00:44] <ubitux> okay
[00:45] <llogan> and the advantage of nullsrc over using something like color for a background is that nullsrc is simply unprocessed?
[00:46] <ubitux> color requires almost no computation
[00:46] <ubitux> just once at the beginning
[00:46] <ubitux> i'm not sure displaying uninitialized data with ffplay -f lavfi nullsrc is really a good thing, but well
[00:49] <ubitux> anyway, important question now
[00:49] <ubitux> how do we renamed ow?
[00:49] <ubitux> dwtdnoiz?
[00:49] <ubitux> dwtdenoise?
[00:49] <ubitux> we keep ow?
[00:50] <ubitux> michaelni: btw, i won't use lavc/dwt code for the first version of ow, unless you insist (ETOOMUCHWORK)
[00:51] <ubitux> saste: no comment on dctdnoiz/lut3d/vignette/ow btw?
[00:54] <cone-576> ffmpeg.git 03James Almer 07master:e7814ed839d1: ADPCM DTK decoder
[00:54] <cone-576> ffmpeg.git 03James Almer 07master:b4866f717ced: ADP demuxer
[00:54] <cone-576> ffmpeg.git 03James Almer 07master:22c7784f60c1: adpcm_thp: Allow the use of extradata for the adpcm table
[00:54] <cone-576> ffmpeg.git 03James Almer 07master:d3710c51d9f1: RSD demuxer
[00:54] <cone-576> ffmpeg.git 03James Almer 07master:e5e86db178b7: RedSpark demuxer
[00:54] <cone-576> ffmpeg.git 03James Almer 07master:67fad0d22145: ADPCM IMA Radical decoder
[00:54] <cone-576> ffmpeg.git 03Michael Niedermayer 07master:cbcc5cbbd825: Merge remote-tracking branch 'jamrial/adpcm'
[00:59] <llogan> ubitux: why do the Expressions subsection item sentences start with lowercase? (regarding vignette)
[01:00] <ubitux> copy/paste
[01:00] <ubitux> likely not sentences
[01:00] <ubitux> (so no caps & period)
[01:01] <llogan> but then you could use those reasons for the top level @items
[01:01] <llogan> what is RedSpark anyway?
[01:02] <ubitux> i think they are sentences
[01:02] <ubitux> (in the top level @items)
[01:03] <llogan> anyway your vignette docs make sense to me, so they pass the idiot test.
[01:05] <ubitux> ok :)
[01:05] <ubitux> thx for having a look
[01:05] <ubitux> michaelni: how can i reproduce the issues with trim/atrim and -ss -t?
[01:06] <ubitux> (do you think it's fixable without much effort?)
[01:20] <Compn> hmm
[01:20] <Compn> dumb question
[01:20] <Compn> is -f framecrc / framemd5 multithreaded ?
[01:20] <Compn> that would speed upfate
[01:20] <Compn> assuming you could get the same results every time 
[01:21] <Compn> assuming you run fate on multicore
[01:32] <michaelni> Compn, make already runs fate with multiple threads
[01:33] <Compn> ok then :)
[01:42] <michaelni> ubitux, enable the trim insertion code in ffmpeg and try ./ffmpeg -ss 1 -i matrixbench_mpeg2.mpg -t 1  test.avi
[01:42] <ubitux> ok will do, thx
[01:42] <michaelni> resulting avi is 16frames instead of 25
[01:43] <ubitux> is there a design issue behind this?
[01:43] <ubitux> or it's just a bug in trim filters?
[01:48] <ubitux> michaelni: if you move -ss as output option, it works fine
[01:48] <michaelni> yes i know
[01:49] <michaelni> its probably a issue in ffmpeg*.c not in the trim filters
[01:50] <ubitux> oh well so it's just disabled when using -ss as input option
[01:50] <ubitux> if -ss and -t are as output option, trim is inserted, ok
[11:58] <ubitux> saste: i'll apply the timeline patchset soon
[12:00] <cone-170> ffmpeg.git 03Kostya Shishkov 07release/1.1:31a77177ff32: iff: validate CMAP palette size
[12:00] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:008ae91bcca7: Merge commit '31a77177ff323ef83944c60a8654891213ab6691' into release/1.1
[12:13] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:dc794d7096e3: fate: fetch samples that match the release series
[12:13] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:9d5f16f6fe23: add missed CVE reference in 9.2 release
[12:13] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:2dfe3a7b4dda: update Changelog
[12:13] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:2e0675847965: Prepare for 9.5 Release
[12:13] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:50a3442120e6: Merge commit '2e06758479650f6e2a8820c7105f2d193a701175' into release/1.1
[12:16] <thethruthisoutth> hey
[12:16] <ubitux> saste: ty very much :)
[12:16] <ubitux> any objection on "lavfi: silence a discarded const qualifier warning." btw?
[12:27] <saste> ubitux, ping mails on list, since i'm having an hard time at catching up
[12:27] <saste> don't feel like i should approve all lavfi mails on lists
[12:28] <saste> just I may want to have a look at API changes
[12:46] <cone-170> ffmpeg.git 03Justin Ruggles 07release/1.1:aba56c03b955: flvdec: use the correct audio codec id when parsing metadata
[12:46] <cone-170> ffmpeg.git 03Justin Ruggles 07release/1.1:c6dce259670b: flvdec: read audio sample size and channels metadata
[12:46] <cone-170> ffmpeg.git 03Diego Biurrun 07release/1.1:f074618a9f73: configure: Refactor dxva2api.h dependency declarations
[12:46] <cone-170> ffmpeg.git 03Matt Wolenetz 07release/1.1:fc6825ebb658: vp8: Fix pthread_cond and pthread_mutex leaks
[12:46] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:1ace588f4a94: Merge commit 'fc6825ebb6585138e8ee2bb3484a04542c5d8b6a' into release/1.1
[12:53] <cone-170> ffmpeg.git 03Paul B Mahol 07master:91cae60ecbf3: lavfi/select: unbreak compatibility with older PICT_TYPE syntax
[12:55] <saste> ubitux, are you willing to review zmq filters?
[12:55] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:c0f7df966277: oma: Validate sample rates
[12:56] <cone-170> ffmpeg.git 03Diego Biurrun 07release/1.1:835538380274: configure: Enable hwaccels without external dependencies by default.
[12:56] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:34ecaf6e8888: avfiltergraph: check for sws opts being non-NULL before using them.
[12:56] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:05015d03da1d: matroska: fix a corner case in ebml-lace parsing
[12:56] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:065996b9843b: Merge commit '05015d03da1d745bb92915b5cea92dec16af719f' into release/1.1
[12:56] <ubitux> saste: sure, why not :)
[12:56] <ubitux> do you have other things you need review for?
[12:56] <saste> i'm going to push soon the av_dynarray_alloc_elem() patch
[12:57] <saste> at the end I didn't make use of it
[12:57] <ubitux> haha
[12:57] <saste> but the function may be useful nonetheless
[12:57] <saste> then you could extend your zmqshell script
[12:57] <ubitux> feel free to push the approved patches of the patchset to justify its integration then ;)
[13:09] <cone-170> ffmpeg.git 03Dale Curtis 07release/1.1:b90816d94b0b: matroska: Update the available size after lace parsing
[13:09] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:5353bd028527: Merge commit 'b90816d94b0b5c01f451ff98cfbf1d5ddec9c3c1' into release/1.1
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:59d33eafd19a: lavfi/lut: use FF_CEIL_RSHIFT for chroma w/h rounding.
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:60f0e304312d: lavfi/vflip: use FF_CEIL_RSHIFT for chroma w/h rounding.
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:1776177b7f1a: lavfi: replace passthrough_filter_frame with a flag.
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:0652111833e2: lavfi/hqdn3d: use macros instead of hardcoded indexes.
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:0122300c861d: lavfi/hqdn3d: restore timeline feature.
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:9b11c2ef363d: fate/filter: test hqdn3d timeline.
[13:13] <cone-170> ffmpeg.git 03Clément BSsch 07master:09d3449a8f55: fate/filter: test gradfun timeline.
[13:24] <cone-170> ffmpeg.git 03Clément BSsch 07master:cd43a7e94511: doc/filters: split item options which are not aliases.
[13:27] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:09e391abd81c: matroska: pass the lace size to the matroska_parse_rm_audio
[13:27] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:759486829696: xmv: do not leak memory in the error paths in xmv_read_header()
[13:27] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:ba31b72f462c: bmv: check for len being valid in bmv_decode_frame().
[13:27] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:858864d35032: xmv: check audio track parameters validity.
[13:27] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:63235b8d419d: Merge commit '858864d350320dd807e349bda017026e61a47fe0' into release/1.1
[13:41] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:62f9253781fa: dfa: check for invalid access in decode_wdlt().
[13:41] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:094a35aeefea: lavf: make sure stream probe data gets freed.
[13:41] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:5ebdfbe893c4: id3v2: pad the APIC packets as required by lavc.
[13:41] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:8f558c3e1018: af_channelmap: sanity check input channel indices in all cases.
[13:42] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:a367ab657f29: Merge commit '8f558c3e101859aec9adcb4b4b270ae1ef8f88b5' into release/1.1
[13:50] <cone-170> ffmpeg.git 03Paul B Mahol 07master:23f7b14f2851: sws: support endianness conversion for AV_PIX_FMT_RGBA64
[13:50] <cone-170> ffmpeg.git 03Paul B Mahol 07master:87888c043da9: fate: add exr tests
[13:51] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:4c412580fdb6: indeo3: fix data size check
[13:51] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:fa4192e31f4b: rv10: check that extradata is large enough
[13:51] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:b0b33ce14806: indeo3: check motion vectors.
[13:51] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:95db1624ef98: indeo3: switch parsing the header to bytestream2
[13:51] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:395538e0735e: Merge commit '95db1624ef98ccc4ba7ff70d50c4b4d0f8ffed54' into release/1.1
[14:00] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:d0c4d61c8bfd: qdm2: check that the FFT size is a power of 2
[14:00] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:43039f9386d2: svq1dec: check that the reference frame has the same dimensions as the current one
[14:01] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:a3410b5a1fcb: svq1dec: clip motion vectors to the frame size.
[14:01] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:c8462bd17f35: mp3dec: fallback to generic seeking when a TOC is not present
[14:01] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:d2b9da2f3797: Merge commit 'c8462bd17f35f435192281a2ea4ce8008a7398d3' into release/1.1
[14:16] <cone-170> ffmpeg.git 03Paul B Mahol 07master:d93bc4ef5c17: redspark: fix memleak in redspark_read_header
[14:16] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:d70bad04de8a: oggdec: fix faulty cleanup prototype
[14:16] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:c04689019162: riff: check for eof if chunk size and code are 0
[14:16] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:6cad940989d3: update Changelog
[14:16] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:46fd6e4f2ebb: aac: check the maximum number of channels
[14:16] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:9767d7513cbc: Merge commit '46fd6e4f2ebbcd5a00847cdb05fe416466d06d37' into release/1.1
[14:22] <ubitux> durandal_1707: fate doesn't seem happy
[14:23] <ubitux> is it your fault?
[14:30] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:d8745de6ae6c: indeo3: fix off by one in MV validity check
[14:30] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:5aac0811100e: id3v2: check for end of file while unescaping tags
[14:30] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:ddeb6eeeb1c1: afifo: fix request_samples on the last frame in certain cases
[14:30] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:0662967d2bbd: hls, segment: fix splitting for audio-only streams.
[14:30] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:204c8798a896: Merge commit '0662967d2bbdbe90540eaa8c847f521fa4b75aab' into release/1.1
[14:32] <durandal_1707> why are those packed yuyvs failing
[14:36] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:a6f7fc8f3b1a: Prepare for 9.6 Release
[14:36] <cone-170> ffmpeg.git 03Martin Storsjö 07release/1.1:5772cbb3435e: swscale: Use alpha from the right row in yuva2rgba_c
[14:36] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:52ab9e898485: wav: Always seek to an even offset
[14:36] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:1ab4578c88dc: lavc: Fix assignments in if() when calling ff_af_queue_add
[14:36] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:93f80cf2cc50: Merge commit '1ab4578c88dc3e1407da15471bd323ba40c3ebbb' into release/1.1
[14:45] <durandal11707> rgba64 appeared in vflip and hflip but that is completly bogus as hashes are same
[14:55] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:7f8b55b560bb: indeo3: use unaligned reads on reference blocks.
[14:55] <cone-170> ffmpeg.git 03Luca Barbato 07release/1.1:77a2f4cbcfc3: oma: properly forward errors in oma_read_packet
[14:55] <cone-170> ffmpeg.git 03Anton Khirnov 07release/1.1:600bc1debaad: af_asyncts: fix offset calculation
[14:55] <cone-170> ffmpeg.git 03Reinhard Tartler 07release/1.1:d2d38531d608: update Changelog
[14:56] <cone-170> ffmpeg.git 03Michael Niedermayer 07release/1.1:898ce4d6e2ef: Merge remote-tracking branch 'qatar/release/9' into release/1.1
[15:13] <cone-170> ffmpeg.git 03Michael Niedermayer 07master:63725d158d34: Revert "sws: support endianness conversion for AV_PIX_FMT_RGBA64"
[15:15] <michaelni> durandal_1707, i reverted the commit that broke fate, feel free to recommit it once the issue is fixed, but i dont want to leave fate broken for everyone while this is being investigates
[15:16] <michaelni> investigateD
[15:16] <durandal_1707> now exr is broken
[15:16] <durandal_1707> on big endian
[15:17] <michaelni> fate did not pass on big endian before the revert
[15:20] <michaelni> durandal_1707, can you fix the rgba64 / exr issue or should we revert it too ?
[15:20] <michaelni> i mean the exr tests that need the rgba64 endianness code
[15:21] <durandal_1707> michaelni: i'm working on it
[15:22] <durandal_1707> and big endian is less relevant....
[15:22] <michaelni> thanks and yes
[15:28] <durandal_1707> michaelni: for some reason i get error log and black screen
[15:29] <durandal_1707> and similar may hapend with xyz
[15:47] <durandal_1707> michaelni: i tried with rgba64 png and hflip works as expected, vflip gives black only, this still does not explain same hash, so swscale is mess
[15:58] <durandal_1707> scale filter set i/o for endianess conversion which is flawed
[15:59] <durandal_1707> scale filter can do more than unscaled stuff 
[16:00] <durandal_1707> michaelni: ^
[16:06] <durandal_1707> michaelni: are you gonna to say something?
[16:27] <ubitux> TheRyuu: yes we have more tests, about 200 more
[16:30] <ubitux> TheRyuu: current icc instance is 13.1.1.163, i can update if necessary
[16:35] <durandal_1707> huh vflip for rgb48be->rgb48le gives crap here
[16:45] Action: ubitux wonders if he is the only one to consider #2559 *not* a normal behaviour
[16:46] <durandal_1707> ubitux: does rgb48be->rgb48le with vflip works for you?
[16:47] <ubitux> how to test?
[16:51] <michaelni> durandal_1707, sorry i was afk
[16:51] <michaelni> how can i help ?
[16:52] <durandal_1707> i tried hflip with real 64bit rgb png file and it works fine, vflip gives black
[16:52] <durandal_1707> but that does not explains same hash in [v/h]flip for pixfmts
[16:53] <durandal_1707> doing rgb48be->rgb48le with vflip (png->tiff) gives broken output
[16:53] <durandal_1707> hflip works
[16:55] <michaelni> maybe something fails with negative stride
[16:58] <michaelni> i cannot confirm a problem with rgb48 and vflip
[17:01] <durandal_1707> try with http://trac.ffmpeg.org/attachment/ticket/639/64bit.png converted to 48bit
[17:01] <michaelni> hflip == vflip because theres no output in either case
[17:02] <durandal_1707> yes, something related to full chroma H ...
[17:02] <michaelni> d41d8cd98f00b204e9800998ecf8427e is the md5 of a 0 byte string
[17:03] <durandal_1707> yes, it fails output because conversion is unscaled
[17:03] <durandal_1707> and scale filter enable such formats 
[17:04] <durandal_1707> and afaik pixfmts tests use scale filter
[17:06] <durandal_1707> so such hack in filter should be removed
[17:06] <durandal_1707> also nonsense flags for swscale
[17:07] <durandal_1707> anyway such design in swscale sucks
[17:07] <durandal_1707> because it would be much better if table for NxN flags is used
[17:08] <durandal_1707> eg. rows are inputs and columns are outputs....
[17:08] <michaelni> partial support is bad formats should be fully supported
[17:09] <durandal_1707> partial support is better than no support at all
[17:09] <durandal_1707> its just swscale was never designed for such thing in mind
[17:10] <durandal_1707> i already had partial rgba64 output working...
[17:11] <durandal_1707> will turn to that, as curren state with endianess only conversion is dead end
[17:30] <cone-170> ffmpeg.git 03Michael Niedermayer 07master:91f4a44ff4fa: sws/packed_16bpc_bswap: fix handling of negative stride
[17:48] <durandal_1707> so anyone can reproduce rgb48 issue with link to file i mentioned?
[19:14] <cone-170> ffmpeg.git 03Claudio Freire 07master:e41cd3cdeb49: aacenc: Fix ticket #1784: erasure of surround channels
[19:59] <durandal_1707> michaelni: i get strange results with yuv4subsampled -> rgba64 (yuv*16 bits works fine)
[20:08] <burek> :) a guy reported my post on the forum because of "The message contains links to illegal or pirated software." :)))
[20:08] <burek> http://ffmpeg.gusari.org/viewtopic.php?f=16&t=919
[20:08] <burek> he considered this link to be illegal/pirated software: http://bit.ly/10IhBsZ :D
[20:09] <Daemon404> many forums do not allow ffmpeg binaries
[20:09] <Daemon404> due to patent/whatever concerns
[20:10] <burek> did you see the link? :D
[20:10] <burek> (btw, im an admin on that forum) :D
[20:12] <durandal_1707> than ban yourself
[20:13] <beastd> burek:  maybe like me he does not like following random URLs --- also your post had some scent of rickrolling to it ;D
[20:13] <burek> :)
[20:15] <durandal_1707> maybe he had software installed that reported that link as fraud
[20:17] <burek> anyway it was funny :)
[20:19] <durandal_1707> michaelni: sent big patch (i really, really, really started to reallly hate sws)
[20:21] <burek> hate leads to anger :)
[20:22] <durandal_1707> not always.....
[20:30] <durandal_1707> michaelni: so you reproduced and fixed that vflip issue?
[20:42] <michaelni> durandal_1707, i think ive fixed the vflip issue yes
[21:27] <xlinkz0> can anyone direct me to some resources that teach about the mp4 headers in binary?
[21:28] <kierank> the spec
[21:29] <xlinkz0> what's it called?
[21:34] <JEEB> 14496-1, 12
[21:34] <JEEB> aka MPEG-4 Part 1 and Part 12
[21:35] <JEEB> and 14 it seems
[21:43] <xlinkz0> thanks
[22:10] <cone-170> ffmpeg.git 03Paul B Mahol 07master:e1746d057b22: swscale: RGBA64 output
[22:10] <cone-170> ffmpeg.git 03Michael Niedermayer 07master:145e644aeed4: sws/output: remove offset/coef for alpha in rgba64
[22:14] <durandal_1707> michaelni: thx for fixing that packed_16bpc_bswap (i failed to see obvious)
[22:22] <michaelni> np
[22:44] <durandal_1707> is cehoyos subscribed to trac or something?
[00:00] --- Mon May 13 2013