[Ffmpeg-devel-irc] ffmpeg-devel.log.20160624

[burek]

[00:03:47 CEST] <mateo`> michaelni: I'll fix that "tomorrow" morning (in 10h), is it only reproducible with this particular input ?
[00:05:44 CEST] <iive> i might remember incorrectly, but I think that the matrix sample had some b-frames at the start with missing references.
[00:06:59 CEST] <michaelni> ~/videos/angels.mpg fails the same way, ~/videos/aletrek-gray.mkv doesnt, didt try more
[00:10:11 CEST] <michaelni> "./ffmpeg -i ~/videos/matrixbench_mpeg2.mpg -movflags dash   fail.mp4" <-- fails too
[00:10:28 CEST] <michaelni> same way "Assertion next_dts >= 0 failed at libavformat/movenc.c:874"
[00:12:08 CEST] <michaelni> also same failure with "./ffmpeg -i ~/videos/matrixbench_mpeg2.mpg -frag_duration 200k  fail.mp4"
[03:42:57 CEST] <cone-026> ffmpeg 03Michael Niedermayer 07master:6935aaa74841: avcodec/flac_parser: Increase FLAC_MAX_SEQUENTIAL_HEADERS by 1
[09:55:56 CEST] Action: mateo` working on fixing the regression b1e1da52fe2e7320363fdaa0cdd276847e25bf53 has introduced
[10:31:13 CEST] <ubitux> michaelni: doesn't assert here
[10:35:44 CEST] <mateo`> michaelni: can't reproduce either
[10:38:50 CEST] <cone-763> ffmpeg 03Muhammad Faiz 07master:1b7eaead0cd4: avfilter/showcqt: set some frame properties
[10:43:39 CEST] <cone-763> ffmpeg 03Martin Storsjö 07master:d34826c33d40: mov: Add a comment referring to the standard that defines the loci box
[10:43:40 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:6dd80591b883: Merge commit 'd34826c33d401929b2ff8aee161fd39ad0a73613'
[10:46:58 CEST] <cone-763> ffmpeg 03Michael Niedermayer 07master:704d2bd18be8: mov: Print reason of loci parsing failure
[10:46:59 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:9c4af99583a1: Merge commit '704d2bd18be8f134cb9ab19463a3c4eb63233dbd'
[10:49:01 CEST] <cone-763> ffmpeg 03Michael Niedermayer 07master:f126365cdad6: mov: Fix parsing short loci
[10:49:02 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:c8c402201ad9: Merge commit 'f126365cdad6327e6def8ceb918eaf538fbb97b3'
[10:51:54 CEST] <cone-763> ffmpeg 03Michael Niedermayer 07master:94d8141c8cd9: mov: Append place name instead of overwriting for loci
[10:51:55 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:b1b5ba034f38: Merge commit '94d8141c8cd96c176f7d4bdb74a95331ec7db19b'
[10:52:56 CEST] <cone-763> ffmpeg 03Michael Niedermayer 07master:bc43131997a4: mov: Also export loci altitude
[10:52:57 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:7414ae4d8ae5: Merge commit 'bc43131997a456ecdff160eab288395f5adb8d72'
[10:54:27 CEST] <cone-763> ffmpeg 03Michael Niedermayer 07master:daec1651f1c0: movenc: Write ?xyz geotag metadata for mov files
[10:54:28 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:8e934e4c3305: Merge commit 'daec1651f1c0f8f8e704db375a2ecb364a722fa1'
[11:01:13 CEST] <cone-763> ffmpeg 03Martin Storsjö 07master:f12a705ee570: movenc: Factorize a function for finding a metadata entry and the associated language
[11:01:14 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:7a1785014b3b: Merge commit 'f12a705ee570e16ca692c66b62821a2dbdf82566'
[11:05:17 CEST] <andrey_turkin> I wonder if reverse merge is possible at all
[11:05:29 CEST] <andrey_turkin> or feasible. or makes sense at all
[11:08:28 CEST] <cone-763> ffmpeg 03Martin Storsjö 07master:b84f3a40034d: movenc: Write 'loci' geotag metadata for 3gp and mp4
[11:08:29 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:2c45254a773f: Merge commit 'b84f3a40034d28828c4ca639d012479a7eaace28'
[11:13:01 CEST] <cone-763> ffmpeg 03Diego Biurrun 07master:a86ef80494d7: blowfish-test: Return different values for different errors
[11:13:02 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:1ca383c8d4e8: Merge commit 'a86ef80494d759807a5ef8dd12838f0690092e37'
[11:14:57 CEST] <cone-763> ffmpeg 03Diego Biurrun 07master:8131bd4c22ad: blowfish-test: Use struct allocation functions from the API
[11:14:57 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:b6310ab6b5bf: Merge commit '8131bd4c22ad59b040173f67a5b2edb75863e801'
[11:16:22 CEST] <cone-763> ffmpeg 03Diego Biurrun 07master:ded4a6543dfb: xtea-test: Use struct allocation functions from the API
[11:16:23 CEST] <cone-763> ffmpeg 03Clément BSsch 07master:b77f62e27fd7: Merge commit 'ded4a6543dfb5a2f1d9f7fb8d3ab27a1872c22b3'
[11:19:32 CEST] <michaelni> ubitux, mateo` you need "./configure --enable-gpl --enable-libx264" at least to reproduce
[11:19:41 CEST] <michaelni> it doesnt happen with mpeg4
[11:21:52 CEST] <mateo`> michaelni: thanks
[11:22:17 CEST] <michaelni> i can repro it with codec copy too, can upload a file causing this with codec copy if needed ?
[11:22:52 CEST] <mateo`> michaelni: yes please, that can be helpful
[11:24:17 CEST] <michaelni> ./ffmpeg -i http://www.ffmpeg.org/~michael/ref.mp4  -c copy  -frag_size 20k fail.mp4
[11:24:39 CEST] <michaelni> both -an or -vn make it pass
[11:29:35 CEST] <michaelni> valgrind output is clean except he assert
[13:17:00 CEST] <ubitux> BtbN: http://sprunge.us/ijOa
[13:17:06 CEST] <ubitux> why nvenc is an exception?
[13:19:06 CEST] <BtbN> Historical reasons by now I guess.
[13:19:25 CEST] <BtbN> If it can be changed without breaking api/abi, I don't see why not.
[13:20:54 CEST] <ubitux> ee359c72ef8735122929da96006565e1558f1e55 was never merged??
[13:22:23 CEST] <ubitux> erf derek nooped them
[13:23:01 CEST] <ubitux> BtbN: so, mmh, can you do something about it?
[13:23:10 CEST] <ubitux> i have no mean to test it
[13:23:50 CEST] <BtbN> That breaks compatiblity with existing commandlines, so we decided against it.
[13:23:52 CEST] <andrey_turkin> we might as well follow libav here and deprecate nvenc_* in favor of *_nvenc
[13:24:03 CEST] <BtbN> The new name was added as an alias
[13:24:25 CEST] <andrey_turkin> right. libav also wraps old names such so they warn users about deprecation
[13:24:27 CEST] <nevcairiel> can always change it around to make the old name the alias
[13:24:49 CEST] <andrey_turkin> ffmpeg can do the same
[13:25:02 CEST] <andrey_turkin> and deprecate "nvenc" alias too
[13:25:47 CEST] <ubitux> can we do it before the release?
[13:26:14 CEST] <BtbN> The problem is, there isn't really the concept of an alias for an encoder name.
[13:26:24 CEST] <BtbN> so it's just a full duplication of the struct + class
[13:28:33 CEST] <ubitux> is it a problem?
[13:34:09 CEST] <BtbN> Well, can't tell one of them the alias. And it's a lot of duplication.
[13:44:05 CEST] <BtbN> But from the looks of it, that patch can almost be applied as-is. Only question is, what to do about the nvenc name, without h264 pre/suffix?
[13:44:35 CEST] <BtbN> I'd guess removing that would break at least a few scripts.
[14:44:39 CEST] <andrey_turkin> BtbN: add another alias with same deprecation code
[14:56:50 CEST] <cone-974> ffmpeg 03Petru Rares Sincraian 07master:84b252ba2a56: fate: add anequalizer test
[15:12:24 CEST] <BBB> ubitux: as for the finger print removal, I think you should leave ben littler in, hes still active (just not under his real name)
[15:28:00 CEST] <kierank> BBB: vp9 svc :(
[15:50:35 CEST] <ubitux> BBB: ok
[16:03:00 CEST] <mateo`> michaelni: side note, if I remove the assert next_dts >= 0 from the get_cluster_duration, the stream muxes fine (and playback works)
[16:03:31 CEST] <nevcairiel> well there gotta be a reason the assert existed
[16:11:02 CEST] <mateo`> nevcairiel: yes, the assert is perfectly valid
[16:11:29 CEST] <JEEB> you can check the resulting file's coded timestamps in L-SMASH's boxdumper
[16:12:18 CEST] <JEEB> because what I've noticed is that lavf for example ignores a lot of stuff when playing ISOBMFF. so you might have ended up with a weird value written somewhere
[16:16:16 CEST] <mateo`> what's happening is that the track_duration is adjusted (the value is actually lesser) and it does not seem to match anymore the value from the cluster entry leading to next_dts being negative
[16:26:12 CEST] <mateo`> JEEB: thanks for the reference, the output of the boxdumper is pretty useful
[16:26:43 CEST] <JEEB> yeah, esp. the box output can be real nice when debugging things
[16:26:50 CEST] <JEEB> and then it has CTS and DTS output for each sample
[16:26:55 CEST] <JEEB> in a separate mode
[16:43:51 CEST] <BBB> kierank: it exists :-p
[16:43:59 CEST] <BBB> kierank: btw my condoleances
[16:44:03 CEST] <kierank> yeah, sad times
[16:45:30 CEST] <nevcairiel> maybe your government is smart enough to just ignore the uninformed masses
[16:45:53 CEST] <ubitux> could be worse, you could be in the US having to choose between leprosy and cholera for your next president
[16:49:08 CEST] <BBB> omg and insults flew around
[16:49:12 CEST] <BBB> anyway
[16:50:25 CEST] <BBB> kierank: I feel terribly sorry for you (plural - the country), its a little unexpected but at a certain level I can understand the frustration of a large group of people with how things currently are
[16:51:06 CEST] <nevcairiel> the demographics of the poll paint an interesting picture though
[16:51:26 CEST] <BBB> link?
[16:51:37 CEST] <BBB> (Ive heard summaries but didnt see actual tables or anything)
[16:51:38 CEST] <nevcairiel> young people get screwed over by the old people
[16:51:39 CEST] <RiCON> old people = leave; young people = remain
[16:51:47 CEST] <BBB> ah, ok
[16:51:48 CEST] <ubitux> isn't london asking for independency btw?
[16:51:54 CEST] <BBB> anything rural vs. city differences?
[16:52:09 CEST] <ubitux> http://www.standard.co.uk/news/london/london-referendum-results-londoners-call-for-capital-to-become-independent-state-after-brexit-vote-a3279801.html
[16:52:19 CEST] <ubitux> (no graphicz)
[16:52:20 CEST] <BBB> I saw some graph in NYT which suggested northern ireland, london and scotland were remain, and most of the rest of england + wales was leave
[16:52:29 CEST] <iive> i thought london city is already independent. ;)
[16:52:50 CEST] <atomnuker> what in the actual fuck is going on, neg tot_vq + sub tot_vq, 1 isn't the same as add tot_vq, 1
[16:53:16 CEST] <atomnuker> how the hell is it not the same, the value is unsigned
[16:54:53 CEST] <BBB> is the upper half initialized?
[16:55:06 CEST] <BBB> if its on stack, its often not initialized at all if the input was int instead of int64_t
[16:55:13 CEST] <BBB> (ptrdiff_t, intptr_t, etc.)
[16:55:40 CEST] <jsebechlebsky> Hi, does someone know if rw_timeout can be set up from ffmpeg command line? Documentation says so, but everytime I check the value of corresponding URLContext->rw_timeout field it's zero... 
[16:56:20 CEST] <jsebechlebsky> (even when I pass the rw_timeout argument with positive value)
[17:00:51 CEST] <atomnuker> BBB: not the issue here
[17:00:56 CEST] <BBB> ok
[17:01:35 CEST] <atomnuker> changing to ptrdiff_t doesn't make a difference
[17:01:38 CEST] <iive> atomnuker: is that simd/sse ?
[17:02:44 CEST] <atomnuker> yep, sse4
[17:08:03 CEST] <iive> so, you have ( ~a+(a-1))  ? 
[17:08:31 CEST] <iive> maybe i'm missing something, but this should give -2 
[17:14:21 CEST] <jamrial> atomnuker: you forgot to change jl into jg
[17:45:16 CEST] <ubitux> (there is a gsoc going on?)
[17:51:27 CEST] <mateo`> I struggle fixing the regression introduced by b1e1da52fe2e7320363fdaa0cdd276847e25bf53 (my mp4 knowledge is a bit rusty, what a shame :(), I probably won't have time to continue to work on it over the week-end, so ...
[17:51:48 CEST] <mateo`> 1) it can wait monday and i'll continue
[17:52:25 CEST] <mateo`> 2) the relevant patch/merge is disabled/reverted until the regression is fixed
[18:36:16 CEST] <mateo`> Regarding the regression, it looks like the packet that comes from ff_interleaved_peek as a dts lesser than the last pkt dts that is used by ff_mov_write_packet
[18:38:57 CEST] Action: mateo` off
[18:47:51 CEST] <cone-974> ffmpeg 03Michael Niedermayer 07master:f573acc42447: avformat/matroskadec: Call matroska_read_close() on header parsing error
[22:03:22 CEST] <omerjerk> can someone give an initial review on this patch please - https://github.com/omerjerk/FFmpeg/commit/7e15d5402d9ea4744896a641acac34a61287b0dd
[22:03:55 CEST] <omerjerk> I've incorporated all of the changes as suggested so far.
[22:09:24 CEST] <michaelni> omerjerk, minor nitpicks, there are tabs in it (should be spaces) also typo " #return normalized result" should be @return
[22:11:06 CEST] <omerjerk> tabs, where exactly ? :o
[22:11:46 CEST] <michaelni> moment git showed it in ugly red for me so i couldnt miss it 
[22:12:00 CEST] <michaelni> +       int dic_code_bit;
[22:12:28 CEST] <michaelni> also is av_mlz* intended to be used outside of the lib its defined in ?
[22:13:02 CEST] <michaelni> if not it should be ff_mlz*, OTOH if its intened / makes sense to use outside then it might fit better in libavutil
[22:13:51 CEST] <michaelni> when in doubt choose ff_mlz*
[22:18:13 CEST] <michaelni> also, you could test the code with a fuzzer to check for crashes and infinite loops
[22:19:47 CEST] <omerjerk> okay. 
[22:20:01 CEST] <omerjerk> Thilo suggested to put it in libavcodec only. 
[22:20:31 CEST] <omerjerk> so I guess ff_ make sense. 
[00:00:00 CEST] --- Sat Jun 25 2016