[Ffmpeg-devel-irc] ffmpeg-devel.log.20161118

[burek]

[00:06:56 CET] <wm4> jamrial_: what are those about? gapless stuff?
[00:08:44 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:1af7ddecda33: apng: use side data to pass extradata to muxer
[00:08:45 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:e622d7723bcf: fate: add apng encoding/muxing test
[00:08:46 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:e92f585bd90e: configure: make sure LTO does not optimize out the test functions
[00:08:47 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:346fa70bb88a: doc: fix spelling errors
[00:08:48 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:c90d521f161b: interplayacm: check for too large b
[00:08:49 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:266cf258cc93: interplayacm: validate number of channels
[00:08:50 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:6e5ccabbe846: interplayacm: increase bitstream buffer size by AV_INPUT_BUFFER_PADDING_SIZE
[00:08:51 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:d95568f9a2d3: mov: immediately return from mov_fix_index without old index entries
[00:08:52 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:3e33685892f9: apngdec: use side data to pass extradata to the decoder
[00:08:53 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:35db873534a8: fate: add streamcopy test for apng
[00:08:54 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:dc2942bbc889: avformat: close parser if codec changed
[00:08:55 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:35cb0c47bc89: ppc: pixblockdsp: do unaligned block accesses correctly again
[00:08:56 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:08b1fd6afb3d: diracdec: use correct buffer for slice_params_buf realloc
[00:08:57 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:b9a24cee3ba0: diracdec: clear slice_params_num_buf on allocation failure
[00:08:58 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:31cebfe789ff: diracdec: check return code of get_buffer_with_edge
[00:08:59 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:c72ac9ffd017: lzf: update pointer p after realloc
[00:09:00 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:c11fd9de7653: mxfdec: fix NULL pointer dereference
[00:09:01 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:1e4979f780a8: mpegaudio_parser: don't return AVERROR_PATCHWELCOME
[00:09:02 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:fa24e3780b77: matroskadec: fix NULL pointer dereference in webm_dash_manifest_read_header
[00:09:03 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:9dee25fbc796: mpegts: prevent division by zero
[00:09:04 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:72a2d6ff56ce: escape124: reject codebook size 0
[00:09:05 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:1ed4b527321f: proresdec_lgpl: explicitly check coff[3] against slice_data_size
[00:09:06 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:581cce0cca80: dvbsubdec: fix division by zero in compute_default_clut
[00:09:07 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:41359d381a2b: icodec: fix leaking pkt on error
[00:09:08 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:1615d83dcf6e: icodec: correctly check avio_read return value
[00:09:09 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:d8affeea82e5: smvjpegdec: make sure cur_frame is not negative
[00:09:10 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:039a3e6db892: pnmdec: make sure v is capped by maxval
[00:09:11 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:b526958ca4a1: filmstripdec: correctly check image dimensions
[00:09:13 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:c8f5154fc1ca: softfloat: handle -INT_MAX correctly
[00:09:14 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:a2c7840a6b69: mlz: limit next_code to data buffer size
[00:09:14 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:598016b85f29: dds: limit 4 bpp handling to AV_PIX_FMT_PAL8
[00:09:15 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:2b863d4e9b10: libschroedingerdec: don't produce empty frames
[00:09:17 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:cbc9d46066b6: libschroedingerdec: fix leaking of framewithpts
[00:09:18 CET] <cone-780> ffmpeg 03Andreas Cadhalpun 07release/3.2:c7d38efbc26b: exr: fix out-of-bounds read
[00:09:19 CET] <cone-780> ffmpeg 03Martin Vignali 07release/3.2:08f26d99b52f: libavcodec/exr : fix channel size calculation for uint32 channel
[00:09:19 CET] <cone-780> ffmpeg 03Stefano Sabatini 07release/3.2:31c9c7ad826e: ffprobe: fix crash in case -of is specified with an empty string
[00:13:58 CET] <Chloe> jeez. that's a lot of commits
[00:21:11 CET] <BBB> michaelni: tnx for the fix
[00:23:23 CET] <atomnuker> michaelni: is cehoyos' commit still needed after those 2 commits?
[00:36:05 CET] <michaelni> it would affect cases for which we do not have a explicit pixel format (we cannot represent these without extra lsbs, i think theres no way to encode these atm either), future ffmpeg versions could support encoding these though and old ffmpeg could decode it by adding lsbs.
[00:36:40 CET] <nevcairiel> but dont you agree that adding random bits into the format is anything but not lossless?
[00:37:20 CET] <michaelni> i belive the changed codepath is unused except such future (not explicitly supported formats)
[00:38:07 CET] <Chloe> if it's unused then there's no reason why it should be there?
[00:38:10 CET] <michaelni> nevcairiel, i dont know, adding 0 lsbs is not really lossless either
[00:38:35 CET] <Chloe> michaelni: why not?
[00:38:37 CET] <nevcairiel> well this only happens if you encode a format you cant decode to, doesnt it
[00:38:51 CET] <nevcairiel> obvious solution: make sure its implemented at both ends
[00:38:53 CET] <nevcairiel> no need for hackery
[00:40:43 CET] <nevcairiel> half-assed "support" for gray10 is what got us here, why not simply enforce adding both parts in the future, the decoder patch doesn't look  particularly complicated now
[00:41:56 CET] <michaelni> the ffv1 spec does not restrict things to 8,10,12,16 bits, our decoder should support as much as it can even if our encoder doesnt IMO
[00:43:10 CET] <nevcairiel> but scaling it up to some other bitdepth and adding bits that didnt exist from the start is just not lossless, is all
[00:43:34 CET] <nevcairiel> the decoder shouldnt decode a format it doesnt know to output unmodified, then
[00:44:28 CET] <nevcairiel> to some degree its probably nice to do that, but on a more stricter level it may cause issues with people assuming its properly lossless/bitexact when it isnt
[00:46:26 CET] <nevcairiel> but alas thanks for fixing the 10-bit case, i'll leave it to you to decide what to do about those "fallback" decoding modes
[00:58:21 CET] <jamrial_> i need to revert a patch, which will break a fate test, to apply another, which will make the test pass again
[00:58:26 CET] <jamrial_> should i instead revert and apply the new code all in one commit to avoid breaking fate for the sake of bisects?
[00:58:47 CET] <jamrial_> such a commit will look ugly and hard to read i guess
[00:59:46 CET] <jamrial_> maybe i could disable the test within the revert commit instead
[00:59:46 CET] <nevcairiel> you could put a fate change into the commit, but thats probably still ugly
[01:03:53 CET] <jamrial_> nevcairiel: afaik without either of the patches the test would fail, not just output wrong data
[01:04:03 CET] <nevcairiel> i see
[01:04:21 CET] <nevcairiel> well disabling it in the revert is probably the best choice, squashing the two will likely be a very ugly diff
[01:04:52 CET] <nevcairiel> you wouldnt be reverting if the two approaches were at all similar
[01:05:29 CET] <jamrial_> yeah, i'll do that
[03:38:57 CET] <kinetik> jamrial_: ping
[03:39:19 CET] <jamrial_> kinetik: pong
[03:40:33 CET] <kinetik> jamrial_: hi! just looking at fixing the comment you made on the FLAC MP4 muxer patch.  I cribbed the experimental flag check from the VP9 code (which has the same issue, it'll mux into mov without the strict flag)
[03:41:47 CET] <kinetik> jamrial_: are you saying that i should remove it from ff_codec_movaudio_tags or just make the strict flag check exclude non-MODE_MP4 entirely?
[03:42:10 CET] <jamrial_> i tried removing the line from ff_codec_movaudio_tags but it broke demuxing
[03:44:58 CET] <jamrial_> maybe just return EINVAL with an error message saying flac is not supported if mode is not mp4
[03:45:20 CET] <jamrial_> i'd prefer the movenc maintainer to chime in, but he hasn't replied when i CCd him
[03:46:08 CET] <jamrial_> trying to mux to m4a (ipod muxer) correctly fails with your patch as is, at least
[03:46:42 CET] <kinetik> removing it doesn't seem to stop it being muxed into mov either
[03:48:41 CET] <jamrial_> it makes the muxer use "MS style audio codec tag" instead of your code
[03:49:06 CET] <jamrial_> which probably just creates a broken file
[03:49:06 CET] <kinetik> ah, i see
[04:00:11 CET] <kinetik> jamrial_: thanks, i'll post an updated patch shortly
[10:17:29 CET] <cone-106> ffmpeg 03Vittorio Giovara 07master:390b95b88b2b: fate: Add a mixed NAL coding sample
[10:17:30 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:9e1ddc08208d: Merge commit '390b95b88b2b896b63f257f69e434dfc0111e076'
[10:22:55 CET] <cone-106> ffmpeg 03Vittorio Giovara 07master:0b1bd1b2057d: lavd: Drop unneeded av_init_packet()s
[10:22:56 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:d7d6f9c78272: Merge commit '0b1bd1b2057d41fd0ccba7317911c484a50f9207'
[10:44:32 CET] <cone-106> ffmpeg 03Matthieu Bouron 07master:61cb9fac4749: mov: fix stream extradata_size allocation
[10:44:33 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:8a91452e83e7: Merge commit '61cb9fac47498a38dfe7623f66aa1f3696e9158c'
[10:44:46 CET] <cone-106> ffmpeg 03Martin Storsjö 07master:d0b1e6049b06: libopenh264dec: Fix cleanup if the init failed early
[10:44:47 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:6fb07c7d85c0: Merge commit 'd0b1e6049b06eeeeca146ece4d2f199c5dba1565'
[10:45:16 CET] <cone-106> ffmpeg 03Martin Storsjö 07master:36b380dcd52e: libopenh264dec: Simplify the init thanks to FF_CODEC_CAP_INIT_CLEANUP being set
[10:45:17 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:0bd76401d166: Merge commit '36b380dcd52ef47d7ba0559ed51192c88d82a9bd'
[10:46:37 CET] <cone-106> ffmpeg 03Martin Storsjö 07master:2d097c16b833: libopenh264enc: Return a more sensible error code in some init failure paths
[10:46:38 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:7e9474ca47fd: Merge commit '2d097c16b833c532ac974a7f1fd05c0a1f3b7675'
[10:47:37 CET] <cone-106> ffmpeg 03Martin Storsjö 07master:7a76371437f9: libopenh264enc: Simplify init by setting FF_CODEC_CAP_INIT_CLEANUP
[10:47:38 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:07502e473f0c: Merge commit '7a76371437f9562c3414f985523f883489e3936a'
[10:52:20 CET] <cone-106> ffmpeg 03Anton Khirnov 07master:d60c2d521693: mp3dec: read the initial/trailing padding from the LAME tag
[10:52:21 CET] <cone-106> ffmpeg 03Anton Khirnov 07master:46278ec90ac5: mp3enc: write trailing padding
[10:52:22 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:f4c3aa7ade13: Merge commit 'd60c2d5216930ef98c7d4d6837d6229b37e0dcb3'
[10:52:23 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:5e78126bbdd5: Merge commit '46278ec90ac5ad1dab5e85991f176afe49003fee'
[11:37:33 CET] <cone-106> ffmpeg 03Anton Khirnov 07master:8d07e941b04d: FATE: add a test of H.264 SEI recovery in an intra refresh stream
[11:37:34 CET] <cone-106> ffmpeg 03Hendrik Leppkes 07master:127cc6dd3d0b: Merge commit '8d07e941b04d63fc4443dd986e3dc7b69cdcca43'
[13:47:15 CET] <BBB> michaelni: can you add support for specifying color_range, colorspace, color_trc and color_matrix to ffv1?
[13:47:38 CET] <BBB> michaelni: or alternatively, can you add support for this being specified in the container (I believe webm/mp4 support that)?
[13:48:02 CET] <BBB> michaelni: maybe chroma location also
[13:49:04 CET] <wm4> is color_matrix = primaries?
[13:53:46 CET] <BtbN> it should be, yes. At least from what I remember from VAAPI
[14:06:56 CET] <nevcairiel> matrix is color space
[14:07:33 CET] <nevcairiel> (ie. the most important of them all)
[14:08:23 CET] <nevcairiel> to avoid confusion just use the terms the api uses, range, primaries, trc, space
[14:10:02 CET] <nevcairiel> BBB: matroskaenc (and as such webm) can already write thsoe elements, but the spec for those is not formally finalized yet, so it requires a -strict flag
[14:10:30 CET] <BBB> with matrix, I meant primaries, yes
[14:10:43 CET] <BBB> ok, strict might be useful
[14:10:54 CET] <nevcairiel> -f_strict -1
[14:10:55 CET] <nevcairiel> or so
[14:10:59 CET] <BBB> cool, ok
[14:11:03 CET] <BBB> tnx
[14:26:45 CET] <wm4> this should probably include some HDR stuff as well then
[14:44:47 CET] <nevcairiel> which mkv also has
[14:45:18 CET] <nevcairiel> its included in the specs here now https://www.matroska.org/technical/specs/index.html .. i dont know if that means its official now
[14:45:38 CET] <nevcairiel> the development of mkv has splintered onto a second mailing list which i dont follow
[14:45:40 CET] <nevcairiel> so :p
[16:58:58 CET] <cone-306> ffmpeg 03James Almer 07master:16c429166ddf: Revert "apngdec: use side data to pass extradata to the decoder"
[16:58:59 CET] <cone-306> ffmpeg 03James Almer 07master:50e0e96f3f8e: avformat/apngenc: use the stream parameters extradata if available
[16:59:00 CET] <cone-306> ffmpeg 03James Almer 07release/3.2:cf655d164334: Revert "apngdec: use side data to pass extradata to the decoder"
[16:59:01 CET] <cone-306> ffmpeg 03James Almer 07release/3.2:f97bee9ad565: avformat/apngenc: use the stream parameters extradata if available
[17:24:47 CET] <Chloe> Why is there yet another HEVC library...
[17:29:57 CET] <Chloe> Oh hmm, I was under the impression there were quite a few libraries, looks like there's only a few. (was probably thinking of h264)
[17:33:29 CET] <RiCON> there's only two non-hardware for h264, same as hevc
[17:37:15 CET] <Chloe> RiCON: I see that now too.
[17:45:44 CET] <atomnuker> reynaldo: what's happening with ffsever?
[17:53:16 CET] <Chloe> atomnuker: I believe it was being continued in a separate repo, so dropping it in ffmpeg was going ahead as planned
[17:54:16 CET] <atomnuker> I know that, but reynaldo said it wouldn't take him long but just a couple of days a couple of weeks ago
[18:19:00 CET] <cone-306> ffmpeg 03Michael Niedermayer 07master:d7a3bb208867: avcodec/ffv1dec: Support gray 10/12/16 explicitly avoid shifts
[18:19:01 CET] <cone-306> ffmpeg 03Michael Niedermayer 07master:d1d18de6ade3: avcodec/ffv1dec: Set packed_at_lsb for 16bit YUV
[19:14:53 CET] <ubitux> nevcairiel: "Our H264 decoder drops 3 frames from the beginning of the stream" why?
[19:15:14 CET] <nevcairiel> if you find out, you can tell me
[19:15:30 CET] <ubitux> i thought you knew, ok :)
[19:34:54 CET] <jamrial> that libturing library needs boost
[19:37:41 CET] <RiCON> it comes with boost in the repo too, in case you don't want or have it installed
[19:37:50 CET] <jamrial> it also has the weirdest method to send parameters to an encoder i've seen so far
[19:38:21 CET] <atomnuker> it reconstructs a command line
[19:38:33 CET] <Chloe> RiCON: yeah, I think I cried at that
[19:38:34 CET] <atomnuker> complete with the executable name in the front
[19:40:18 CET] <llogan> boost is the best
[19:40:50 CET] <nevcairiel> boost isnt too bad if you're a c++ project, but linking C++ libs to C projects is always a bit terrible =p
[19:41:14 CET] <nevcairiel> but as long as these shenangians are hidden through pkg_config or something, not hardcoded in configure..
[19:43:11 CET] <llogan> plus you get to install a million billion files for boost
[19:43:22 CET] <BtbN> Did they turn a command-line encoder into a library, or what the hell is that interface?
[19:46:02 CET] <nevcairiel> probably
[20:12:31 CET] <wm4> the overflow checks seem to be severely broken in that patch too
[20:15:10 CET] <jamrial> kinetik: i'm trying files created with your patch on Firefox 51 beta 1, and i get no video if the file also has a flac stream
[20:15:19 CET] <jamrial> the audio plays, though
[20:35:07 CET] <ubitux> BBB: libavcodec/x86/vp9itxfm.asm:2151: error: instruction expected after label
[20:35:15 CET] <ubitux> (yasm 0.8.0)
[20:35:23 CET] <BBB> add HAVE_EXTERNAL_AVX2 around the function
[20:35:26 CET] <nevcairiel> update that shit =p
[20:35:27 CET] <ubitux> (doesn't matter much but i have a fate instance with that as requested)
[20:35:28 CET] <BBB> someone else reported the same bug yesterday
[20:35:38 CET] <ubitux> probably Carl reading FATE
[20:35:40 CET] <ubitux> ?
[20:36:03 CET] <ubitux> i might be wrong but i think he's the one who asked me to add that instance
[20:36:12 CET] <ubitux> maybe because a stable distro was using this release
[20:36:32 CET] <ubitux> i must say i don't care much, i was just raising the problem
[20:36:52 CET] <nevcairiel> its an easy fix either way, but at some point we might as well bump the requirement
[20:37:11 CET] <BBB> can you guys add the right %if HAVE_EXTERNAL_AVX2 around the given functions?
[20:37:14 CET] <BBB> makes my life easy
[20:39:16 CET] <jamrial> we could just bump the minimum yasm version to 1.1.0. that's what debian oldstable uses
[20:39:22 CET] <cone-306> ffmpeg 03Clément BSsch 07master:b9c26aed4e3c: ffmpeg: remove dead code out of the video filter loop
[20:39:23 CET] <cone-306> ffmpeg 03Clément BSsch 07master:825f6fbfac6c: ffmpeg: factor out sending frame to filters
[20:39:36 CET] <jamrial> it will not address this, though, since avx2 was added in 1.2.0
[20:40:21 CET] <jamrial> but it would let us remove guards for other sets
[20:51:40 CET] <BBB> up to you guys
[20:51:58 CET] <BBB> I dont mind the %if HAVE_EXTERNAL_AVX2 if people think its useful
[20:52:06 CET] <BBB> but if youd rather remove it thats obviously fine with me also
[20:53:35 CET] <jamrial> BBB: i'll add the guards for now
[20:53:53 CET] <RiCON> BBB: it was wm4
[20:54:00 CET] <RiCON> ffmpeg-git was breaking mpv's travis build
[20:54:19 CET] <nevcairiel> if we ever get avx512 we'll need to switch to recommending nasm anyway, at which time we might as well require a more recent version
[20:54:33 CET] <RiCON> it was "fixed" by adding --disable-asm
[20:56:06 CET] <BBB> RiCON: omg
[20:56:25 CET] <BBB> RiCON: Im not sure that was the best solution
[20:56:27 CET] <RiCON> travis is still on ubuntu 12.04, so yasm 1.1.0
[20:56:49 CET] <BBB> I hope that thing isnt running cpu-intesive unit tests using ffmpeg
[20:57:12 CET] <RiCON> it's not, it's just building ffmpeg to build mpv, not running any test
[20:57:55 CET] <BBB> ok then
[20:57:57 CET] <RiCON> fwiw, libass also disables asm on its own
[20:57:58 CET] <BBB> sorry about the breakage
[20:59:17 CET] <llogan> nasm is dead, let's use yasm. yasm is dead, let's use nasm.
[21:00:33 CET] <BBB> thats pretty much the cycle of life, yes
[21:01:13 CET] <BBB> the sad thing is that for software to succeed, you pretty much have to give up your life and any other amibitions for it
[21:01:29 CET] <BBB> the yasm author got distracted; yasm dead
[21:03:25 CET] <cone-306> ffmpeg 03James Almer 07master:2de1c79b61b6: x86/vp9itxfm: add missing AVX2 guards
[21:07:52 CET] <BBB> jamrial: ty!
[21:09:01 CET] <wm4> mpv's travis build uses ffmpeg only to the extent of testing header files and linking against it, so --disable-asm is fine
[21:09:13 CET] <wm4> trying to use it with a shit-old nasm is only a side-effect
[22:45:04 CET] <cone-306> ffmpeg 03Andreas Cadhalpun 07master:7289aa2d71a8: options_table: limit codec parameters to sane values
[23:46:04 CET] <Chloe> Anyone got any AVC or RS422 hardware?
[23:46:34 CET] <Chloe> AVC being 'AV/C Digital Interface Command Set General Specification Version 4.1'
[00:00:00 CET] --- Sat Nov 19 2016