[Ffmpeg-devel-irc] ffmpeg-devel.log.20111221

[burek]

[00:00] <michaelni> OanaStratulat, sorry i was in the kitchen eating :)
[00:00] <michaelni> ill look now
[00:00] <OanaStratulat> michaelni: no problem
[00:02] <michaelni> crashes in  mxf_read_header (mxfdec.c:1611) with near null ptr
[00:03] <michaelni> Tjoppen, cbsrobot is this a known bug?
[00:04] <ubitux> michaelni: try [PATCH 1/2] mxfdec: Make sure mxf->nb_index_tables > 0 in mxf_packet_timestamps()
[00:04] <ubitux> could be applied immediately i guess&
[00:05] <ubitux> oh that reminds me i need to open an issue for the mxfenc
[00:05] <OanaStratulat> michaelni: no luck with this one ?:(
[00:05] <michaelni> ubitux, patch makes no difference, it still crashes
[00:05] <cbsrobot> michaelni: you can download 300 mb file: http://ijghv1.dl4free.com/
[00:06] <cbsrobot> and check it out
[00:06] <ubitux> cbsrobot: i think he is refering to another issue
[00:06] <cbsrobot> ah I see
[00:06] <cbsrobot> sorry
[00:08] <michaelni> hmm, noone knows the bug so ill assume its new, let me close the task
[00:09] <OanaStratulat> michaelni: 10x and please accept http://google-melange.appspot.com/gci/task/view/google/gci2011/7217216
[00:19] <OanaStratulat> michaelni: http://google-melange.appspot.com/gci/task/view/google/gci2011/7217216 please review, when run the corruptfile in ffmpeg it says segfault 5 times but in valgrind the mem losses are small hope its good
[00:31] <ubitux> trac is severly slow tonight :)
[00:35] <OanaStratulat> michaelni: http://google-melange.appspot.com/gci/task/view/google/gci2011/7218215 please accept
[00:46] <OanaStratulat> michaelni: http://google-melange.appspot.com/gci/task/view/google/gci2011/7218215 please review
[00:46] <OanaStratulat> michaelni: its a hang, some mem is not freed
[00:47] <ubitux> you should try to quote properly the outputs
[00:47] <ubitux> with {{{ and }}} (you have access to it with a markup special icon)
[00:48] <OanaStratulat> ubitux: on trac?
[00:48] <ubitux> yes
[00:48] <ubitux> just about the description, you have various tools, especially the code block mode
[00:48] <ubitux> also use the preview to check
[00:49] <ubitux> the problem of not doing this is that well, it's not really readable :)
[00:49] <OanaStratulat> ubitux: ok, sorry for this 
[00:49] <ubitux> np, just noticing for next time
[00:54] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r7feb7f16a8 10ffmpeg/libavformat/mxfdec.c: 
[00:54] <CIA-49> ffmpeg: mxfdec: Make sure mxf->nb_index_tables > 0 in mxf_packet_timestamps()
[00:54] <CIA-49> ffmpeg: Only the OPAtom demuxing logic is guaranteed to have index tables, meaning OP1a
[00:54] <CIA-49> ffmpeg: files that lack an index would cause SIGSEGV.
[00:54] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:54] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r134aaa79f7 10ffmpeg/libavcodec/indeo3.c: 
[00:54] <CIA-49> ffmpeg: indeo3: Fix null ptr dereference
[00:54] <CIA-49> ffmpeg: Fixes Ticket804
[00:54] <CIA-49> ffmpeg: Bug found by: Oana Stratulat
[00:54] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:54] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rfc739b79ad 10ffmpeg/libavcodec/golomb.h: 
[00:54] <CIA-49> ffmpeg: golomb: fix end of bitstream check
[00:54] <CIA-49> ffmpeg: Fixes Ticket806
[00:54] <CIA-49> ffmpeg: Bug found by Oana Stratulat
[00:54] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:02] <OanaStratulat> michaelni: are the files submitted good?
[01:15] <OanaStratulat> michaelni: when do you will review my task because its night in Romania and i have school 2morrow, i want to claim the last one and then go to zz
[01:15] <Daemon404> OanaStratulat, michaelni is in germany so he isn't much better off :P
[01:15] <Daemon404> then again, he doesn't sleep.
[01:16] <OanaStratulat> Daemon404: he`s 1 hour behind me
[01:16] <michaelni> OanaStratulat, ill look now
[01:16] <Daemon404> probably
[01:20] <michaelni> well the file seems to cause a long or infinite loop
[01:22] <OanaStratulat> michaelni: http://google-melange.appspot.com/gci/task/view/google/gci2011/7218216 please accept the last task of this kind :)
[01:22] <michaelni> done
[01:23] <OanaStratulat> michaelni: 10x
[01:25] <ubitux> what does "10x" mean?
[01:26] <j-b> thanks
[01:29] <OanaStratulat> michaelni: http://google-melange.appspot.com/gci/task/view/google/gci2011/7218216 marked as completed
[01:34] <caytchen> uhm, whats up with AVPacket? can't allocate these on the heap?
[01:34] <michaelni> caytchen, of course you can
[01:35] <caytchen> all I can find allocation related is av_dup_packet and I don't want to new it since thats not using the av_ allocator
[01:51] <OanaStratulat> michaelni: are the files good?
[01:54] <durandal_1707> is it mandatory to check for output buf_size when encoding frame?
[02:05] <OanaStratulat> michaelni: had time to check my task?
[02:06] <michaelni> OanaStratulat, not yet, ill look now
[02:14] <michaelni> OanaStratulat, good, work, segfault reproduced!
[02:14] <OanaStratulat> michaelni: thx
[02:15] <OanaStratulat> michaelni: will more bug finding tasks be posted or the deadline with posting tasks has passed?
[02:17] <michaelni> AFAIK there where only 2 times when tasks became vissible and we are after the second, if you find something from google saying something else then i surely could try to add more tasks
[02:18] <OanaStratulat> michaelni: ah ok, no i did not find anything. so that is all. hope the bugs found by me will help you
[02:19] <michaelni> of course, they helped us alot!
[02:19] <michaelni> and of course you can continue to report issues on trac if you want :)
[03:16] <caytchen> anything else I have to do to an AVFrame other than avcodec_alloc_frame it before passing it on to avcodec_decode_audio4?
[03:17] <caytchen> nevermind, forgot to open the codec for the AVCodecContext
[03:18] <caytchen> avcodec_decode_audio4 should maybe catch that instead of segfaulting, heh.
[03:25] <caytchen> is it optional for codecs to set frame->format? since I now have a decoded frame with no idea what the data in it is.
[03:29] <Daemon404> http://japland.org/img/lol_loud.png <-- so i went to make a test file to fix an audio decoding bug and found this.
[03:29] <Daemon404> sure is mastering.
[03:51] <CIA-49> ffmpeg: 03Alex Converse 07master * rd9b89b23df 10ffmpeg/libavformat/mpegts.c: 
[03:51] <CIA-49> ffmpeg: mpegts: Fix typo in handling sections in the PMT.
[03:51] <CIA-49> ffmpeg: This was an error rebasing 4682a1dc3aa2554ad7077f5db32d0f2d598d018e for
[03:51] <CIA-49> ffmpeg: commit. The "pes" variable guaranteed to be NULL in that block.
[03:51] <CIA-49> ffmpeg: 03Martin Storsjö 07master * rb797f73520 10ffmpeg/libavcodec/vc1dec.c: 
[03:51] <CIA-49> ffmpeg: vc1dec: Use the right pointer type for the tmp pointer
[03:51] <CIA-49> ffmpeg: This fixes warnings about assignment from incompatible
[03:51] <CIA-49> ffmpeg: pointer type.
[03:51] <CIA-49> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[03:51] <CIA-49> ffmpeg: 03Aneesh Dogra 07master * rfd22616c59 10ffmpeg/libavcodec/bytestream.h: 
[03:51] <CIA-49> ffmpeg: bytestream: add a new set of bytestream functions with overread checking
[03:51] <CIA-49> ffmpeg: Signed-off-by: Justin Ruggles <justin.ruggles at gmail.com>
[03:51] <CIA-49> ffmpeg: 03Aneesh Dogra 07master * r1443ea93d9 10ffmpeg/libavcodec/4xm.c: 
[03:51] <CIA-49> ffmpeg: 4xm: Use bytestream2 functions to prevent overreads
[03:51] <CIA-49> ffmpeg: Fixes Bug 110.
[03:51] <CIA-49> ffmpeg: Signed-off-by: Justin Ruggles <justin.ruggles at gmail.com>
[03:51] <CIA-49> ffmpeg: 03Diego Biurrun 07master * r606c9297b9 10ffmpeg/configure: 
[03:52] <CIA-49> ffmpeg: build: fix standalone compilation of mpc7/mpc8 decoders
[03:52] <CIA-49> ffmpeg: These decoders depend on the mpegaudio DSP code.
[03:52] <CIA-49> ffmpeg: 03Luca Barbato 07master * r432f0e5b7d 10ffmpeg/libavformat/nut.c: 
[06:28] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r077b086dea 10ffmpeg/libavformat/avio.h: 
[06:28] <CIA-49> ffmpeg: avio: remove ENOTSUP mention in doxy as its not available on all supported platforms
[06:28] <CIA-49> ffmpeg: and thus cannot be used
[06:28] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[09:21] <lionaneesh> Hello!
[09:21] <lionaneesh> Just wanted to ask >> Does a HANG count as a bug?
[09:21] <lionaneesh> michaelni: ^^
[09:30] <Tjoppen> cbsrobot: uhm, libav master doesn't even probe 2011_DCPTEST_24FPS.V.mxf as being jpeg2000
[09:31] <cbsrobot> Tjoppen: weired
[09:32] <cbsrobot> Tjoppen: if you checkout one commit before ddcf3e0535191a776728d0ecabc71c6dddf05dd1 you see it worked
[09:37] <cbsrobot> Tjoppen: this is the last one that worked: 5e67e3eac2fed771ef7363face3ff4691dd4ff2d
[09:37] <cbsrobot> if i'm not mistaken
[09:40] <cbsrobot> btw - whats the best way to find the commits before and after a specific commit in git ?
[09:42] <TimN> cbsrobot:  I use gitk
[09:48] <cbsrobot> TimN: is there a shell only solution ? When connected to a remote host ?
[09:50] <kcm1700> git log --graph | less ?
[09:50] Action: TimN thinks about this
[09:52] <lionaneesh> Can someone review my task http://www.google-melange.com/gci/task/view/google/gci2011/7217215 , michaelni 
[10:00] <TimN> yes, and once in less /<sha-1> to take you to the one you want...
[10:06] <Tjoppen> git log --all --oneline --graph hash   is gitk-ish
[10:16] <lionaneesh> Bye Guyz!
[10:23] <TimN> "git log --all --oneline --graph hash " doesn't take me to the hash in question though...
[10:55] <Tjoppen> TimN: /hash
[11:16] <cbreak-work> I use alias.graph=log --graph --decorate --oneline --all
[11:17] <cbreak-work> if you want to get commits before a certain commit, replace --all in that command with that commit
[11:17] <cbreak-work> commits after are harder, there's no automatic easy solution
[11:18] <Tjoppen> isn't the something similar to FOO~1234, but forward?
[11:19] <cbreak-work> no
[11:19] <cbreak-work> it's impossible to do that in git in general
[11:19] <cbreak-work> there are specific solutions, like if you know on which branch the commits you are interested in are
[11:26] <Tjoppen> ~ is also impossible in general, no?
[12:02] <cbreak-work> Tjoppen: no, it is possible
[12:02] <cbreak-work> because git history is a DAG
[12:02] <cbreak-work> a directed acyclic graph.
[12:03] <cbreak-work> each commit knows all its parents
[12:03] <cbreak-work> but it knows nothing about children
[12:57] <Tjoppen> cbreak-work: it's computable. as long as it only has one child commit of course
[12:58] <Tjoppen> else it could complain
[13:23] <Tjoppen> 58000 zzuf ffprobe mxfdec runs. still going strong
[13:53] <Tjoppen> I already have a fix for corruptfile2 btw
[15:00] <ubitux> michaelni: do you think an audio stream merging helper has its place in libswr?
[15:02] <cbreak-work> Tjoppen: the problem is that you might not even have the child commit
[15:02] <cbreak-work> it could be on some other computer, in an other repository
[15:02] <cbreak-work> there might be thousands of them
[15:03] <cbreak-work> so in general it is impossible to find child commits
[15:03] <cbreak-work> of course, you can find child commits on known branches
[15:03] <cbreak-work> but only by checking all of them
[15:04] <cbreak-work> git log for example can be used
[15:17] <Tjoppen> .. yes, of course you can only look at local commits
[15:18] <Tjoppen> I mean, it could traverse from all heads down until it reaches the specified commit, then back up
[15:19] <Tjoppen> being able to look at the garbage would also be nice
[15:20] <cbreak-work> garbage?
[15:20] <cbreak-work> you can look at reflog
[15:20] <cbreak-work> or you can find dangling commits with fsck
[15:20] <cbreak-work> but I wouldn't use the dangling ones, they might no longer be valid
[15:21] <Tjoppen> yeah, not in this case. hence why I say "heads"
[15:21] <Tjoppen> but reflog in gitk could be useful
[15:36] <michaelni> ubitux, i think it could make sense, yes
[15:36] <ubitux> ok
[16:03] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r085249a379 10ffmpeg/libavformat/mxfdec.c: 
[16:03] <CIA-49> ffmpeg: mxfdec: Ignore the last entry in Avid's index table segments
[16:03] <CIA-49> ffmpeg: The last entry is the total size of the essence container.
[16:03] <CIA-49> ffmpeg: Previously a TemporalOffset error would be logged, even though segments like
[16:03] <CIA-49> ffmpeg: these are expected.
[16:03] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r5b718009b2 10ffmpeg/libavformat/mxfdec.c: 
[16:03] <CIA-49> ffmpeg: mxfdec: Sanity-check SampleRate
[16:03] <CIA-49> ffmpeg: This avoids a SIGFPE if SampleRate is missing or set to naughty values.
[16:03] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r54cbbdd25a 10ffmpeg/libavformat/mxfdec.c: mxfdec: Remove unused variables
[16:03] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rc3846e3eba 10ffmpeg/libavcodec/utils.c: 
[16:03] <CIA-49> ffmpeg: lavc: Do not fail audio decoding if user has set a custom get_buffer.
[16:03] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:03] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r1888bd1b6d 10ffmpeg/libavformat/mxfdec.c: 
[16:03] <CIA-49> ffmpeg: mxfdec: Make sure mxf->nb_index_tables > 0 in mxf_packet_timestamps()
[16:03] <CIA-49> ffmpeg: Only the OPAtom demuxing logic is guaranteed to have index tables, meaning OP1a
[16:03] <CIA-49> ffmpeg: files that lack an index would cause SIGSEGV.
[16:03] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r1fbb859cdf 10ffmpeg/: (log message trimmed)
[16:03] <CIA-49> ffmpeg: Merge remote-tracking branch 'tjoppen/mxf_fixes_20111220'
[16:03] <CIA-49> ffmpeg: * tjoppen/mxf_fixes_20111220:
[16:03] <CIA-49> ffmpeg:  mxfdec: Sanity-check SampleRate
[16:03] <CIA-49> ffmpeg:  mxfdec: Make sure mxf->nb_index_tables > 0 in mxf_packet_timestamps()
[16:03] <CIA-49> ffmpeg:  mxfdec: Remove unused variables
[16:03] <CIA-49> ffmpeg:  mxfdec: Make sure x < index_table->nb_ptses
[16:03] <CIA-49> ffmpeg: 03Tomas Härdin 07master * r6e9081f4d8 10ffmpeg/libavformat/mxfdec.c: 
[16:03] <CIA-49> ffmpeg: mxfdec: Make sure x < index_table->nb_ptses
[16:03] <CIA-49> ffmpeg: Without this the demuxer will SIGSEGV on files with IndexEntryCount < IndexDuration
[16:56] <Compn> http://www.adweek.com/video/advertising-branding/freakiest-ads-2011-mcac-rhian-touches-herself-136806
[16:58] <kierank> lol
[17:34] <ubitux> TimN: are you Tim Nicholson?
[17:37] <kierank> that's him
[17:37] <ubitux> ok :)
[17:37] <ubitux> TimN: you can ask git question here if you're not confident
[18:03] <caytchen> what happened to avcodec_encode_audio2? I can see references to it on the web, but its not in master
[18:10] <ubitux> there is no avcodec_encode_audio2 and never was afaik
[18:11] <ubitux> maybe you're mixing up with avcodec_decode_audio*?
[18:11] <caytchen> it seems like the stuff I found was pertaining to the libav fork
[18:11] <ubitux> there might be a work in progress for avcodec_encode_audio2() yes
[18:12] <ubitux> but atm there is only avcodec_encode_audio()
[18:14] <caytchen> mhm libav doesn't have it either. which is odd, given that there are quite a few patches from years back on google adding it. maybe they never got into upstream
[18:15] <nevcairiel> there is a new encode_audio2 on libav mailing list just from a week back or so
[18:16] <caytchen> just seems odd that the one central encoding function isn't yet using the AVFrame paradigma, instead forcing you to hand in s16 samples
[18:17] <nevcairiel> decode audio was spitting out the same just until a few weeks ago
[18:17] <nevcairiel> it only very recently got AVFrame'ed with decode_audio4
[18:20] <caytchen> that might be the time it got in, but the idea and implementation even seem to be much older than that
[18:21] <TimN> ubitux: Thanks, I have gained help from #git in the past. But now I'm kind of juggling git usage at the same time as ffmpeg-devel netiquette so, as the English say; "Please excuse my size nines"
[18:22] <caytchen> e.g. theres a patch at https://lists.libav.org/pipermail/ffmpeg-devel/2010-July/092924.html for both encode_audio2 and decode_audio4 (same function prototype as the one in master today) from july 2010
[18:22] <TimN> ...added to which having to using a web mail client to avoid corporate rubbish...
[18:24] <ubitux> :D
[18:24] <nevcairiel> caytchen: it probably was talked to death like it usually happens, and took about a year until someone bothered to pick it up =p
[18:25] <caytchen> if so they only picked up half, heh. sorry I keep going on and on about this, I was just suprised at the inconsistency and don't really have an intimate knowledge into the ffmpeg development process.
[18:26] <nevcairiel> well decode4 was added now, and encode2 is still being worked on i guess
[18:28] <nevcairiel> https://lists.libav.org/pipermail/libav-devel/2011-December/017170.html
[18:28] <nevcairiel> just 3 days ago.
[18:28] <ubitux> caytchen: the reasons for a encode_audio2() a while ago might not be the same as for now btw
[18:28] <nevcairiel> not only does it take a AVFrame, it also outputs an AVPacket. :d
[18:39] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * ra40f6a5c69 10ffmpeg/libavcodec/mpeg12.c: 
[18:39] <CIA-49> ffmpeg: mpeg12dec: reset first_field with picture_structure changes.
[18:39] <CIA-49> ffmpeg: Fixes Ticket809
[18:39] <CIA-49> ffmpeg: Bug found by: Oana Stratulat
[18:39] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:44] <Daemon404> michaelni, that 32-bit wavpack problem seems to be more of a pain than i thought. apparently wavpack.c doesn't even read post_shift correctly in the first place...
[20:00] <CIA-49> ffmpeg: 03Nicolas George 07master * r20af3af575 10ffmpeg/libavformat/mov.c: mov: replace AVERROR(EOF) with AVERROR_EOF.
[20:00] <CIA-49> ffmpeg: 03Nicolas George 07master * r1f797a5826 10ffmpeg/doc/errno.txt: doc: add a list of errno codes on some systems.
[21:07] <durandal_1707> somehow when playing file with ffplay i get red errors from demuxer but when using ffmpeg no errors pop up
[21:46] <Compn> durandal_1707 : different versions ?
[21:47] <durandal_1707> nope
[21:47] <durandal_1707> i must be related to seeking
[21:48] <durandal_1707> but i do not seek when using ffplay
[21:48] <durandal_1707> so ffplay is doing something extra which ffmpeg does not
[00:00] --- Thu Dec 22 2011