[Ffmpeg-devel-irc] ffmpeg-devel.log.20121014

burek burek021 at gmail.com
Mon Oct 15 02:05:02 CEST 2012


[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 07f70a651b3f77: sws-test: raise limigts a bit. * 03http://tinyurl.com/9k2gjvk03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 07670b927aa221: ffv1: make sure gob_count is not 0 * 03http://tinyurl.com/9bgzjeu03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 071037e484f0f1: dnxhddata_ Fix mixup of sizeof() and array elements in ff_dnxhd_find_cid() * 03http://tinyurl.com/9trbf5l03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 076581b6cef47e: dnxhdenc: assert ff_dnxhd_get_cid_table() return value * 03http://tinyurl.com/973zc5v03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 07b3eb4f54c0d0: flashsv: check deflateInit() return value * 03http://tinyurl.com/9nd4nyc03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 078cda27b7537d: avcodec_find_best_pix_fmt_of_2: fix handling or PIX_FMT_NONE * 03http://tinyurl.com/9povcfy03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 07f4d73f0fb55e: mpegvideo: check return value of ff_MPV_common_init() * 03http://tinyurl.com/8w5rkpw03
[00:14] <cone-741> ffmpeg.git 03Michael Niedermayer 0720ec0d2a750a: mpegvideo: fix motion_val checks * 03http://tinyurl.com/8otb5go03
[00:27] <cone-741> ffmpeg.git 03Thilo Borgmann 07feaff427c0df: alsdec: fix clipping of weightings for MCC decoding * 03http://tinyurl.com/97s2kyz03
[00:35] <ubitux> nyuhu: ping
[00:35] <ubitux> nyuhu: i commented on the kerndeint filter for the FATE thing in cast that's what blocking you (it seemed to be last time you asked)
[00:37] <nyuhu> oh thx, I was planning on going back with that very soon (currently workning on eq)
[00:38] <ubitux> hell yeah eq filter!
[00:38] <ubitux> (s/cast/case)
[00:39] <nyuhu> Ill post the histogram normalization filter tomorrow
[00:39] <ubitux> great :)
[00:39] <nyuhu> :)
[00:43] <ubitux> saste: i think Paul's concern was about g_XX[CONSTANT] vs just g_XX[]
[00:43] <ubitux> g_XX[CONSTANT] makes think the array needs zero-padding
[00:44] <ubitux> while it seems every data you use in the code are explicitely in the arrays
[00:44] <saste> ah ok
[00:44] <saste> i used that to document the size of the arrays
[00:44] <saste> well i borrowed that from the original code
[00:44] <saste> (or maybe at some point the data was not static, and thus i needed to declare the size or something)
[00:45] <ubitux> what does 'g' means?
[00:45] <ubitux> global? :)
[01:24] <saste> ubitux, Guess
[01:25] <ubitux> ok :)
[02:21] <cone-741> ffmpeg.git 03Clément BSsch 07dff826b4a8d3: lavfi/select/scene: move out convoluted sad variable init from loop. * 03http://tinyurl.com/8oxyvf603
[02:21] <cone-741> ffmpeg.git 03Clément BSsch 070852648301ad: lavfi/select/scene: fix potential overread. * 03http://tinyurl.com/9jf25nc03
[02:21] <cone-741> ffmpeg.git 03Clément BSsch 07096d96ffda8f: lavfi/select/scene: use pointer increments instead of y*linesize. * 03http://tinyurl.com/8gxzsa503
[02:25] <Daemon404> right
[02:25] <Daemon404> michaelni, 
[02:26] <Daemon404> http://fate.ffmpeg.org/report.cgi?time=20121014002519&slot=x86_64-centos-clang-ioc
[02:27] <Daemon404> thats ansic... gonna move it to c99 for future runs
[02:32] <Daemon404> ohai
[02:32] <michaelni> Daemon404, you need that thing that is under CONFIG_FTRAPV enabled for ioc
[02:33] <Daemon404> can you be a bit more specific as to what and why?
[02:34] <michaelni> there are overflows in the ra144 code, its checked at the end of the loop only
[02:35] <michaelni> iam not 100% sure what is correct
[02:35] <Daemon404> ftrapv is gcc only
[02:35] <Daemon404> iirc
[02:35] <Daemon404> im using clang's ioc patches + some extras from mas
[02:35] <Daemon404> mans*
[02:36] <michaelni> maybe rename CONFIG_FTRAPV to CONFIG_AVOID_ARITHMETIC_ANOMALIES
[02:36] <michaelni> clang supports ftrapv too
[02:36] <Daemon404> why is that  code ifdef'd out?
[02:37] <Daemon404> instead of mainlined
[02:37] <michaelni> the ftrapv variant is alot slower
[02:37] <michaelni> like said iam also not sure what is correct
[02:37] <michaelni> vitor said it should be checked per iteration
[02:37] <Daemon404> let me do a run with ftrapv enabled
[02:37] <Daemon404> and c99 enabled
[02:37] <Daemon404> should take ~15 mins
[02:39] <michaelni> to just make the non frapv code pass ioc & ftrapv you just need a few carefully placed unsigend and signed casts but according to vitor IIRC that isnt correct
[02:39] <Daemon404> i see
[02:49] <Daemon404> http://fate.ffmpeg.org/report.cgi?time=20121014004611&slot=x86_64-centos-clang-ioc
[02:49] <Daemon404> wow ok
[02:49] <Daemon404> c99 mode kicked it into gear
[02:49] <Daemon404> i wonder how much of that is valid
[02:50] <ubitux> divVerent: ping
[03:00] <ubitux> is there a way to set decoder avoption from the ffmpeg cmd line?
[03:11] <michaelni> -thisoption thatparameter ?
[03:12] <ubitux> doesn't seem to have any effect :/
[03:12] <ubitux> maybe i was doing it wrong
[03:12] <ubitux> anyway, too much coding for today, i'll dig later
[03:15] <ubitux> michaelni: btw, any comment on the lavfi/select/scene/perm patch, the swf bitslossless, or lavd/lavfi metadata inject are very welcome if you find some time ;)
[03:15] <ubitux> (heh, and even text sub dec now)
[03:18] <ubitux> anyway, 'night ppl
[03:30] <cone-741> ffmpeg.git 03Michael Niedermayer 07d3d715ff1345: mpeg4videodec: Recalculate timebase in case of guessing time increment bits * 03http://tinyurl.com/8cag3em03
[03:30] <cone-741> ffmpeg.git 03Michael Niedermayer 07b12d92efd6c0: avoid "0xFF << 24" as it is considered a integer overflow in C99 * 03http://tinyurl.com/9latfr603
[03:33] <Daemon404> wouldnt that only be an overflow if things are signed?
[03:33] <Daemon404> ah... right.. 
[03:33] Action: Daemon404 just actually looked at the patch
[03:47] <cone-741> ffmpeg.git 03Michael Niedermayer 072fed05f53a88: avoid more "0xFF << 24" as it is considered a integer overflow in C99 * 03http://tinyurl.com/9z8gzjt03
[03:47] <cone-741> ffmpeg.git 03Michael Niedermayer 07014b178f84fd: g723_1: fix overflow in square_root() * 03http://tinyurl.com/8tk2jqd03
[04:08] <cone-741> ffmpeg.git 03Michael Niedermayer 07555f352f9931: swr/rematrix: fix C99 left shift overflow * 03http://tinyurl.com/8enzvh603
[04:08] <cone-741> ffmpeg.git 03Michael Niedermayer 073ee8eefbf262: sws: avoid signed C99 overflows * 03http://tinyurl.com/8n9gwxt03
[04:08] <cone-741> ffmpeg.git 03Michael Niedermayer 07693097c355cb: paf: avoid C99 overflows * 03http://tinyurl.com/8l4pnlf03
[04:09] <michaelni> Daemon404, are there any overflows left ?
[04:10] <Daemon404> michaelni, cron is going to run it in 50 mins from now
[04:10] <Daemon404> well see then
[04:10] <Daemon404> too lazy to connect to work vpn ...
[04:11] <Compn> Daemon404 : yeah but michael is in the mood to fix them now
[04:11] <Compn> gotta work with the master devel
[04:11] <Compn> if you want things worked on quickly ;)
[04:11] <Daemon404> this sounds dirty
[04:11] <michaelni> well i just wanted to fix all before going to bed :)
[04:12] <Compn> michaelni : you work too hard 
[04:12] <Compn> ehe
[04:12] <Compn> so all the dpx samples we have decode correctly ?
[04:12] <Compn> are we still updating the changelog ?
[04:13] <michaelni> all the samples in that zip or what it was should 
[04:17] <Daemon404> tomorrow im going to try and set up an instance with the intel race condition thingy
[04:17] <Daemon404> depending on how slow it is
[04:32] <Daemon404> uh oh
[04:33] <Daemon404> oh its just make segfaulting
[04:49] <michaelni> Daemon404, maybe you could wrap make in something the retries when make segfaults
[04:51] <Daemon404> possible
[04:51] <Daemon404> msys make is really shitty
[04:51] <Daemon404> that's nevcairiel's fate instance, though.
[05:01] <Daemon404> ok...
[05:02] <Daemon404> so clearly i have no idea what im doign re: cron
[05:02] <Daemon404> it didnt run
[05:02] <Daemon404> i guess it helps if crond is running.
[05:14] <Daemon404> michaelni, well it looks like most are in swscale.c now
[05:14] <Daemon404> if not all
[05:14] <Daemon404> re: failures
[05:14] <Daemon404> (if youre not asleep)
[05:15] <michaelni> Daemon404, i dont sleep, its waste of time
[05:16] <Daemon404> lul
[05:18] <sandsmark> sleep is a luxury most people can't afford
[05:21] <cone-741> ffmpeg.git 03Michael Niedermayer 077a32ab5ed0b0: sws: get rid of C99 signed overflows try #2 * 03http://tinyurl.com/9b9ghl203
[05:34] <Daemon404> only 2 left... by why arent all 1255 test being run?
[05:35] <Daemon404> only 1149
[05:36] <Daemon404> ah. no seek tests due to lavf failures
[05:53] <cone-741> ffmpeg.git 03Michael Niedermayer 072217a2249dd7: dpxenc: fix signed c99 overflows * 03http://tinyurl.com/9boo5ok03
[05:53] <cone-741> ffmpeg.git 03Michael Niedermayer 07d31098113fab: mp3enc: fix signed C99 overflow * 03http://tinyurl.com/995s29n03
[06:16] <Daemon404> and ioc is all green now...
[09:52] <durandal_1707> michaelni: dnxd can assert in init, because table have holes and probe does not check for this
[09:52] <durandal_1707> *dnxhd
[12:45] <cone-20> ffmpeg.git 03Paul B Mahol 07e2bf1082cc0d: svq1dec: return more meaningful error codes * 03http://tinyurl.com/9z7enxe03
[12:45] <cone-20> ffmpeg.git 03Paul B Mahol 078f4020d8a4b2: svq1dec: check return value of get_bits1() in more common way * 03http://tinyurl.com/9ej6fca03
[12:45] <cone-20> ffmpeg.git 03Paul B Mahol 077b9fc769e40a: svq1dec: use more common way to check if result is not zero * 03http://tinyurl.com/9nj9avg03
[12:45] <cone-20> ffmpeg.git 03Paul B Mahol 07dd5aff001e1e: svq1dec: use log level that have more common sense * 03http://tinyurl.com/8ptx4qp03
[12:45] <cone-20> ffmpeg.git 03Paul B Mahol 07039341eb4394: wv: do not report invalid stream duration * 03http://tinyurl.com/9fjq7qf03
[12:45] <durandal_1707> aww
[12:46] <durandal_1707> sorry, svq1dec commits should not be there....
[12:51] <saste> durandal_1707, I'm used to to: git push --dry-run
[12:52] <saste> then I check with git show HASH
[12:52] <durandal_1707> i put all stuff in same branch
[12:52] <saste> and finally i commit, that's boring but sometimes it prevented me to screw up things
[12:53] <durandal_1707> anyway svq1dec changes are cosmetics
[12:53] <saste> i have several branch for each "topic", then i usually cherry pick to master before pushing
[12:53] <saste> yes, they don't seem to be harmful at all
[12:54] <saste> before using git I was using quilt, and I had a single stack where I put all the patches
[12:55] <saste> and rebasing was a huge PITA every time
[14:09] <cone-20> ffmpeg.git 03Michael Niedermayer 0703760297b1d3: doc/nut: Clarify what/where is the official nut specification * 03http://tinyurl.com/8ngbgvc03
[14:12] <durandal11707> michaelni did you merged libav nut changes?
[14:17] <michaelni> durandal11707, not yet, why do you ask ?
[14:18] <durandal11707> i thought it was pushed and you skipped it
[14:20] <michaelni> anyway the changes dont look like a terribly good idea, what are they good for anyway ?
[14:21] <durandal_1707> they added tags for planar and float raw codecs
[14:23] <michaelni> sure but whats the point of the other changes ?
[14:23] <durandal_1707> like?
[14:24] <michaelni> changing order of preferrance of tags
[14:24] <michaelni> also i have my doubts their code works as is
[14:36] <cone-20> ffmpeg.git 03Luca Barbato 071bd442c276e6: nut: prioritize native tags * 03http://tinyurl.com/933rmlw03
[14:36] <cone-20> ffmpeg.git 03Michael Niedermayer 07db51c65961bd: Merge commit '1bd442c276e6688b43777a198cad0d7e3a92123f' * 03http://tinyurl.com/8zhqqj303
[14:39] <durandal_1707> michaelni: how do you merge code?
[14:41] <michaelni> magic
[14:42] <durandal_1707> not magic at all, in many cases you undo you own change, instead you should merge by hand conflicts
[14:46] <michaelni> now sure what you mean ?
[14:46] <michaelni> noT
[14:48] <durandal_1707> see dc945b1fa8ae65a <- that is not best solution neither is commit after that that use b2 but you got the point
[14:49] <durandal_1707> b2 commit removed your code (which was not wrong in such case that is should be removed....)
[14:49] <cone-20> ffmpeg.git 03Luca Barbato 0792281850a2d8: nut: support pcm codecs not mapped in avi * 03http://tinyurl.com/92hvu6m03
[14:49] <cone-20> ffmpeg.git 03Justin Ruggles 0710e645e9cb06: lavr: handle clipping in the float to s32 conversion * 03http://tinyurl.com/9rrl3l903
[14:49] <cone-20> ffmpeg.git 03Mans Rullgard 0774c39bc68271: eval-test: make table static const * 03http://tinyurl.com/9y8hbw703
[14:49] <cone-20> ffmpeg.git 03Mans Rullgard 07ac17ccf73ad0: configure: work around bug in ash shell * 03http://tinyurl.com/9txhvbk03
[14:49] <cone-20> ffmpeg.git 03Mans Rullgard 07741a8b724e47: configure: recognise Minix as OS * 03http://tinyurl.com/8lnnw9y03
[14:49] <cone-20> ffmpeg.git 03Mans Rullgard 07b5198a2637b7: configure: tms470: add mapping for -mfpu=vfpv3-d16 flag * 03http://tinyurl.com/9fnwjxj03
[14:49] <cone-20> ffmpeg.git 03Michael Niedermayer 0782c0055c5e71: Merge commit 'b5198a2637b7b45b0049a1d4b386a06f016f2520' * 03http://tinyurl.com/8apw9bm03
[14:53] <michaelni> what do you mean by b2 commit ?
[14:54] <durandal_1707> 1255eed533b4
[14:59] <cone-20> ffmpeg.git 03Paul B Mahol 07de3b1116dab1: eatgq: fix small overread * 03http://tinyurl.com/9stwpqx03
[15:06] <saste> michaelni, anyway nut changes should be discussed on nut-devel
[15:06] <michaelni> saste, yes
[15:07] <saste> i don't think libav has the right to properly change a format which was not designed by them
[15:07] <saste> did someone tried to contact them about the issue?
[15:08] <michaelni> libav doesnt talk with me when my name appears in From: so no i didnt try
[15:09] <michaelni> doesnt talk with me NORMALLY (there are exceptions) ...
[15:09] <saste> anyone is welcome to contribute... meh
[15:09] <saste> enough for today...
[15:14] <durandal_1707> michaelni: there is code in mov/avi that calls dvi_produce_packet but it does not give any audio
[15:14] <durandal_1707> perhaps it worked at some time and then stopped
[15:14] <cone-20> ffmpeg.git 03Mans Rullgard 075ab432fa3500: configure: improve tms470 compiler usage with glibc * 03http://tinyurl.com/9uuxqbb03
[15:14] <cone-20> ffmpeg.git 03Mans Rullgard 07b6f8d635f27b: build: tms470: work around glibc math.h problems * 03http://tinyurl.com/9vpagga03
[15:14] <cone-20> ffmpeg.git 03Diego Biurrun 0752d113ee06ba: avutil: Rename ff_set_systematic_pal2() ---> avpriv_set_systematic_pal2() * 03http://tinyurl.com/8n6yw6903
[15:14] <cone-20> ffmpeg.git 03Diego Biurrun 077638f0b2fef0: avutil: Do not make ff_ symbols globally visible. * 03http://tinyurl.com/9zfp45903
[15:15] <cone-20> ffmpeg.git 03Michael Niedermayer 07d6f6a7557c8d: Merge remote-tracking branch 'qatar/master' * 03http://tinyurl.com/8mddqmd03
[15:15] <michaelni> durandal_1707, possible
[15:37] <durandal_1707> michaelni: actually avi sample have wav tag
[15:43] <durandal_1707> vlc does not support dts in mov...
[15:44] <durandal_1707> vlc is one big mess
[15:45] <durandal_1707> michaelni: the only way i see to support such old avi files is to let avi demuxer read wav tags
[15:46] <durandal_1707> but is there any wav file with ulead audio.....
[15:46] <saste> durandal_1707, which software is not a big mess?
[15:48] <durandal_1707> saste: ffmpeg
[15:49] <saste> durandal_1707, less than other software, but still a mess ;-)
[15:50] <durandal_1707> saste: where is it?
[15:50] <saste> many places, do you want specific examples?
[15:50] <durandal_1707> yes
[15:51] <saste> not that i want to blame anyone for that, in my philosophy being messy is a natural state for software (in rapid development, and with a long story such as ffmpeg)
[15:52] <saste> my favourite: libavcodec/imgconvert.c and how things are split between libavcodec and libavutil with regards to pixfmts
[15:52] <saste> color space handling in libswscale
[15:52] <saste> ah the new entry, pseudo_pal in libavutil/pixdesc (that hack doesn't belong to lavu)
[15:53] <saste> but overall i agree, ffmpeg is pretty clean in comparison with most software projects
[16:04] <cone-20> ffmpeg.git 03Peter Ross 070ea55365b91c: ansi: by design nb_args may exceed the size of the args array, so guard accordingly * 03http://tinyurl.com/97yxcg803
[16:04] <cone-20> ffmpeg.git 03Peter Ross 076bf43441f0a7: eatgq: raise error on unsupported macroblock mode * 03http://tinyurl.com/8detkr603
[16:45] <durandal_1707> michaelni: lavfi/buffer.c: line 225 gets number of channels from channel layout
[16:46] <durandal_1707> this crash if avctx->channels  is smaller number
[16:51] <cone-20> ffmpeg.git 03Hendrik Leppkes 07572781b25f98: h264: fix parsing of old lossless profile (profile_idc == 144) * 03http://tinyurl.com/9gpmwmn03
[19:30] <cone-20> ffmpeg.git 03Michael Niedermayer 0745bd0d15f453: lavf: fix fps detection with PAFF H.264 * 03http://tinyurl.com/9g7rt6n03
[21:07] <cone-20> ffmpeg.git 03Thilo Borgmann 0741bf943f709e: bgmc: fix sizeof arguments (should fix CIDs: 700724 and 608084) * 03http://tinyurl.com/94okot703
[21:07] <cone-20> ffmpeg.git 03Michael Chinen 07c73b1a1d8010: flac_parser.c: fix issue with looping output for small files. * 03http://tinyurl.com/8fk9j8603
[21:34] <j-b> Compn: freaking about?
[21:46] <Compn> j-b : dvaudio decoder :P
[21:47] <j-b> how so?
[22:01] <Compn> someone said it was committed to vlc
[22:01] <Compn> not important
[22:01] <j-b> I think it was, but quite some time ago
[22:20] <cone-20> ffmpeg.git 03Michael Niedermayer 0793ef29b6f47e: noise_bsf: fix division by 0 * 03http://tinyurl.com/928qbgd03
[22:20] <cone-20> ffmpeg.git 03Michael Niedermayer 07a6cac64f69bf: compute_lpc_coefs: assert that normalize and fail have a supported combination * 03http://tinyurl.com/8tmt9h303
[22:20] <cone-20> ffmpeg.git 03Michael Niedermayer 070de0ae5a20cf: tag_tree_decode: check node being non null * 03http://tinyurl.com/8uz5wm203
[22:20] <cone-20> ffmpeg.git 03Michael Niedermayer 071e6cbb01e911: j2kdec: make curtilenum integer. * 03http://tinyurl.com/9kytlkh03
[22:20] <cone-20> ffmpeg.git 03Michael Niedermayer 07bc077ca6b5f8: j2kdec: remove unneeded operation * 03http://tinyurl.com/997mm7903
[22:37] <ubitux> it would be nice to have a teletext decoder..
[22:38] <j-b> rofl
[22:38] <ubitux> ? :(
[22:39] <j-b> complete teletext decoder is hard
[22:39] <j-b> see libzvbi
[22:39] <ubitux> is that what you use @vlc?
[22:40] <j-b> ubitux: yes, because a complete decoder was too much of a mess
[22:40] <ubitux> ok
[22:40] <ubitux> only for teletext?
[22:41] <j-b> decoding just teletext subitles is not that hard
[22:41] <j-b> complete teletext pages is ...
[22:43] <nevcairiel> the subtitles is the important part for most people
[22:48] Action: ubitux adds a new TODO list entry
[22:48] <ubitux> i'll never reach peace :(
[22:50] <nevcairiel> we would all be bored if we had empty lists
[22:51] <ubitux> sure, 2-3 entries is fine
[22:51] <ubitux> 90-100 isn't :(
[00:00] --- Mon Oct 15 2012


More information about the Ffmpeg-devel-irc mailing list