[Ffmpeg-devel-irc] ffmpeg-devel.log.20150227

burek burek021 at gmail.com
Sat Feb 28 02:05:03 CET 2015

[03:34] <cone-657> ffmpeg.git 03Michael Niedermayer 07master:007498fc1a63: swscale/utils: clear formatConvBuffer on allocation
[09:39] <cone-443> ffmpeg.git 03Paul B Mahol 07master:5440076303df: avcodec/hqx: fix typo
[10:31] <cone-443> ffmpeg.git 03Paul B Mahol 07master:aa7a2fa58fdb: avcodec/hqx: use init_get_bits8()
[13:57] <cone-443> ffmpeg.git 03Michael Niedermayer 07master:629a03a9128b: avformat/rtpdec_vp9: Use ffio_free_dyn_buf()
[13:57] <cone-443> ffmpeg.git 03Michael Niedermayer 07master:d490b26fcb1b: avformat/ffmenc: Use ffio_free_dyn_buf()
[14:27] <cone-443> ffmpeg.git 03Clément BSsch 07master:3e0ae19f863f: avfilter/palettegen: fix frame mem leak
[14:34] <ubitux> it's interesting that asan detects some stuff valgrind doesn't seem to detect
[14:34] <ubitux> http://fate.ffmpeg.org/report.cgi?time=20150227021550&slot=x86_64-archlinux-gcc-asan
[15:06] <cone-443> ffmpeg.git 03Derek Buitenhuis 07master:a22032281c10: movenc: Write 'colr' box correctly for MP4
[15:14] <ubitux> Assertion av_fifo_size(stream->fifo) == 0 failed at libavformat/mpegenc.c:1185
[15:14] <ubitux> meh
[15:16] <Daemon404> \o/
[15:17] <ubitux> i was trying to reproduce a crash from someone
[15:17] <ubitux> i got an assert instead
[15:17] <Daemon404> to a user, an assert is the same as a crash
[15:18] <nevcairiel> app terminates, so yeah
[15:18] <ubitux> Daemon404: no it was a segfault
[15:19] <ubitux> and the assert is a level0 one, not something i would trigger because of my assert level
[15:19] <ubitux> anyway...
[15:49] <wm4> build_libs/lib/libavfilter.so.5: undefined reference to `pp_postprocess at LIBPOSTPROC_53'
[15:49] <wm4> why does this shit always happen when I'm in a hurry
[15:49] <wm4> and it's always useless crap deps
[15:49] <wm4> like libpostproc
[15:49] <wm4> or opencl
[15:50] <kierank> usual bullshit
[16:35] <cone-076> ffmpeg 03Michael Niedermayer 07master:78cd9c5671d8: postproc: mark alloc/free context functions as av_cold
[16:41] <Daemon404> anyone have much experience with the mp4toannexb bsf?
[16:41] <Daemon404> i was trying to only do it for the first packet (so i only get one sps/pps for the whole stream)
[16:41] <Daemon404> but that creates broken output and warnings
[16:42] <kurosu_> it's also called to filter extradata in addition to any in-band parameter set
[16:42] <Daemon404> source is mp4
[16:42] <Daemon404> there is only extradata
[16:42] <Daemon404> no in-band sets
[16:42] <Daemon404> thus it should only be needed once, no?
[16:43] <kurosu_> I think that was also one of the options of the changesps filter I posted once, but the code is so horrendous I dropped submitting it
[16:43] <kurosu_> I don't know, most probably
[16:43] <Daemon404> yeah
[16:44] <Daemon404> but if i only use it on packet 1, i get: [mpegts @ 0x186e060] H.264 bitstream error, startcode missing, size 10870 data 00002A72
[16:44] <Daemon404> on some packets
[16:44] <Daemon404> (i transmuxing mp4->ts)
[16:45] <kurosu_> from the code, it seems it checks for both 4 bytes and 3 bytes markers
[16:46] <kurosu_> so something must go horribly wrong for that to happen
[16:46] <Daemon404> this is a straight passthrough of packets
[16:46] <Daemon404> with bsf applied to first
[16:46] <kurosu_> well, can't help more here
[16:47] <Daemon404> perhaps mp4 does not much NALU headers?
[16:47] <Daemon404> mux*
[16:48] <Daemon404> yeah
[16:48] <Daemon404> https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/h264_mp4toannexb_bsf.c#L56
[16:48] <Daemon404> it looks like it prepends the 0001 startcode
[16:49] <Daemon404> yep mp4 has no start codes
[16:54] <cone-076> ffmpeg 03Michael Niedermayer 07master:f8b743b9e61f: avcodec/mjpegenc: Mark ff_mjpeg_encode_close() as av_cold
[16:54] <cone-076> ffmpeg 03Michael Niedermayer 07master:22a80272aea7: avcodec/diracdsp: Mark ff_diracdsp_init() as av_cold
[16:54] <cone-076> ffmpeg 03Michael Niedermayer 07master:0e699676f9a8: avcodec/snow: mark dwt init as av_cold
[17:43] <cone-076> ffmpeg 03Christophe Gisquet 07master:60bdc868b9c2: lavf/mov: correct printf specifier
[18:59] <cone-076> ffmpeg 03Michael Niedermayer 07master:b37c3396cd36: avcodec/h264: Add ff_ prefix to the shared h264_init_dequant_tables() function
[19:33] <cone-076> ffmpeg 03Michael Niedermayer 07master:23a90768a874: avcodec/v210dec: Add ff prefix to v210_x86_init()
[19:33] <cone-076> ffmpeg 03Michael Niedermayer 07master:bf6102044d47: avcodec/dvdec: Make dv_iweight_720_c static
[19:33] <cone-076> ffmpeg 03Michael Niedermayer 07master:4d00860ac797: swresample: Add prefix to soxr_resampler
[19:33] <cone-076> ffmpeg 03Michael Niedermayer 07master:ae20682f6b0e: swscale: Add prefix to updateMMXDitherTables()
[19:33] <cone-076> ffmpeg 03Michael Niedermayer 07master:0a23129bfc6f: avformat: Add prefix to get_codec_guid()
[20:12] <ubitux> wm4: i'm also interested about the mkv things, so yeah please elaborate :)
[20:13] <wm4> ubitux: I made a full list of stuff (whether it's useful or not) https://github.com/mpv-player/mpv/wiki/libavformat-mkv-checklist
[20:13] <wm4> but I'll reply
[20:13] <ubitux> thanks :)
[20:15] <ubitux> nice list
[20:15] <wm4> just added another thing to it
[20:24] <haasn> https://www.ffmpeg.org/doxygen/trunk/pixfmt_8h.html#ad4791ea14975f098b649db7fcd731ce6 seems to be missing an entry for something like AVCOL_TRC_BT1886 (which is defined by ITU-R BT.1886 as an idealized pure power gamma curve of 2.40)
[20:29] <haasn> Also, I want to extend pngenc.c to respect the color_primaries and color_trc fields when encoding an AVFrame
[20:30] <haasn> But (apart from sRGB) the PNG specification only allows pure power curves. What I want to do is use the closest pure power approximation to each TRC, eg. gamma 1.961 for BT.709
[20:30] <haasn> Is there a built-in function to translate a TRC to a pure power approximation?
[20:32] <kierank> no
[20:32] <kierank> you're probably overestimating what swscale can do
[20:37] <haasn> To clarify: I'm not interested in any form of conversion of the image, just appropriate tagging
[20:37] <JEEB> that should be possible
[20:38] <JEEB> poking swscale is on the other hand iä iä cthulhu f'thagn territory
[20:38] <haasn> Writing the tags to PNG is extremely easy; the hard part is converting AVColorPrimaries to a list of x/y chromaticity coordinates, and also AVColorTransferCharacteristic to a gamma value
[20:39] <haasn> This should be some sort of helper function that can be used by more than just the PNG library
[20:39] <haasn> s/library/encoder/
[23:34] <OceanSpray> so has there been an effort to make ffmpeg build in VS2015 yet?
[23:35] <nevcairiel> does it not work out of the box?
[23:35] <OceanSpray> well no
[23:35] <nevcairiel> it builds in 2013 without modifications
[23:35] <OceanSpray> I mean, it took quite a few modifications
[23:35] <nevcairiel> might just need a configure update to recongize the compiler i reckon
[23:36] <OceanSpray> because of the new unified CRT that implements snprintf already
[23:37] <nevcairiel> easiest would be to just not use it =P
[23:37] <OceanSpray> yeah I had to comment out all the lines that insert references to compat/msvcrt/snprintf.h
[23:42] <OceanSpray> make check also required some fiddling, since Windows doesn't have unistd.h and filtering_video.c needs an implementation of usleep
[23:42] <OceanSpray> I might just send you guys a pull request on github
[23:51] <nevcairiel> unistd.h isnt present on any previous versions either, and like i said, it builds without modifications on 2013 right now
[23:55] <OceanSpray> by "it", do you mean ffmpeg or the tests and samples?
[23:57] <nevcairiel> dunno if all samples are cross-platform, they are meant to demonstrate the API usage, not be cluttered by portable hacks
[23:57] <nevcairiel> but ffmpeg itself and make fate (the test suite)
[00:00] --- Sat Feb 28 2015

More information about the Ffmpeg-devel-irc mailing list