[Ffmpeg-devel-irc] ffmpeg-devel.log.20170916

burek burek021 at gmail.com
Sun Sep 17 03:05:02 EEST 2017


[00:40:02 CEST] <testsin> Hi
[00:40:06 CEST] <testsin> @Holden
[01:08:03 CEST] <cone-379> ffmpeg 03Kaustubh Raste 07master:e5a650e14154: avcodec/mips: Improve avc lpf msa functions
[01:08:04 CEST] <cone-379> ffmpeg 03Kaustubh Raste 07master:0105ed551cb9: avcodec/mips: Improve avc mc copy msa functions
[01:08:05 CEST] <cone-379> ffmpeg 03Kaustubh Raste 07master:1a85fb7e1eb3: avcodec/mips: Improve hevc sao band filter msa functions
[01:08:06 CEST] <cone-379> ffmpeg 03Thomas Mundt 07master:a7f6bfdc185a: avfilter/interlace: prevent over-sharpening with the complex low-pass filter
[01:08:07 CEST] <cone-379> ffmpeg 03Thomas Mundt 07master:ed48e22748ee: avfilter/interlace: simplify code
[01:08:08 CEST] <cone-379> ffmpeg 03Thierry Foucu 07master:42a41c395697: vf_fps: Fix memory leak introduced by eea64ef4
[01:38:15 CEST] <atomnuker> jamrial: if postproc isn't enabled (which here it isn't enabled by default) libavfilter's pkg-config still indicates it depends on libpostproc
[01:38:51 CEST] <atomnuker> (so it fails if you look for libavfilter)
[01:44:38 CEST] <atomnuker> the hell, even with --disable-postproc it's still depending on libpostproc
[02:27:30 CEST] <tmm1> Anssi: i'd like to teach hls.c to take advantage of http keep-alives, to avoid making a new tcp connection for every segment. any thoughts on the matter?
[02:27:36 CEST] <jamrial> atomnuker: https://pastebin.com/raw/EBAXvrvy
[02:28:18 CEST] <tmm1> seems it would be easier if using URLContext with ff_http_do_new_request(), but hls.c was converted from using URLContext to AVIOContext
[02:35:23 CEST] <tmm1> well AVIOContext seems to wrap URLContext for http anyway, so it should be possible to create the context with multiple_requests=1, keep it open and then re-use it for subsequent requests
[02:35:57 CEST] <nevcairiel> URLContext should not be used outside of avio
[02:46:03 CEST] <tmm1> hmm ok
[02:46:32 CEST] <atomnuker> jamrial: pkg-config files get regenerated after a configure, right?
[02:47:41 CEST] <tmm1> so maybe io_open can be passed an existing AVIOContext along with something like AVIO_FLAG_REUSE and it will re-use if the underlying context is a URLContext and re-use is possible (http host matches existing connection)
[02:53:43 CEST] <jamrial> atomnuker: if you run configure without first doing make clean or make distclean it will not be regenerated. the build system only tracks the library's version.h file as dependency for the .pc files it seems
[02:55:08 CEST] <jamrial> sorry, distclean only. .pc is a distclean suffix
[02:58:18 CEST] <atomnuker> ok, that solves it then
[02:58:25 CEST] <atomnuker> I basically never run distclean
[02:59:10 CEST] <jamrial> i never do it either. i run "rm -rf *" on my out-of-tree build folder :p
[03:00:26 CEST] <jamrial> guess it should depend on something else. just the library's version.h is insufficient
[03:01:26 CEST] <jamrial> maybe config.h
[03:02:59 CEST] <jamrial> mmh, fuck, it should probably track version.h of every other library the library depends on as well, or something like that
[04:34:37 CEST] <cone-553> ffmpeg 03Aman Gupta 07master:73bf0f42e342: avformat/hlsenc: fix segfault when using -hls_segment_type fmp4 with -hls_segment_filename
[16:57:14 CEST] <cone-411> ffmpeg 03Reimar Döffinger 07master:a149fa97d950: avcodec/frame_thread_encoder: Fix AV_OPT_TYPE_STRING handling in avctx
[19:56:04 CEST] <cone-411> ffmpeg 03Michael Niedermayer 07master:0f5576a22b11: avutil/imgutils: Fix warning: missing braces around initializer
[19:56:05 CEST] <cone-411> ffmpeg 03Michael Niedermayer 07master:d76838c1adc2: avcodec/fits: Fix include type
[19:56:06 CEST] <cone-411> ffmpeg 03Michael Niedermayer 07master:08ec828de966: avcodec/vorbisenc: Fix mixed declaration and statements
[00:00:00 CEST] --- Sun Sep 17 2017


More information about the Ffmpeg-devel-irc mailing list