[Ffmpeg-devel-irc] ffmpeg-devel.log.20150527

[burek]

[00:03:15 CEST] <philipl> Compn: Yes, that's a typo. It doesn't change anything because the resolution of the calculation isn't affected.
[00:03:48 CEST] <philipl> I can fix it.
[00:04:38 CEST] <llogan> Compn "fixed" a bug.
[00:12:21 CEST] <Compn> i didnt fix it 
[00:13:24 CEST] <llogan> obliquely
[00:32:50 CEST] <Compn> theres way too many 'new' bugs :P
[00:32:56 CEST] <Compn> time to close a few.
[00:44:57 CEST] <Daemon404> ^ also latency si a thing
[00:45:02 CEST] <Daemon404> and http *can *seek.
[00:48:31 CEST] <Compn> yes http can seek
[00:48:37 CEST] <Compn> and you can seek range
[00:48:48 CEST] <Compn> BUT, av is interleaved
[00:48:54 CEST] <Compn> so you downloading video packets no matter what
[00:49:15 CEST] <Compn> if someone feels like benchmarking ffmpeg -i http://blah vs local file, then dooo it :P
[00:49:24 CEST] <Compn> er vs wget & ffmpeg local
[01:55:14 CEST] <cone-663> ffmpeg 03Michael Niedermayer 07master:023040ed0d26: avformat/movenc: Allow muxing mp3 with samplerate < 16khz if the user sets strict to -1
[02:54:29 CEST] <BBB> I dont know if you guys ever answered jessegs question, but grep for trap in configure and youll find it
[02:54:36 CEST] <BBB> $ grep trap ../configure
[02:54:42 CEST] <BBB> trap 'rm -f -- $TMPFILES' EXIT
[02:54:59 CEST] <Daemon404> ah.
[02:55:11 CEST] <Daemon404> that wouldnt fix his problem though.
[02:55:48 CEST] <BBB> he could sigint himself instead of exiting
[02:56:10 CEST] <BBB> or remove the trap of remove the file in question from TMPFILES
[02:56:16 CEST] <BBB> (dont know what he was after)
[03:06:03 CEST] <Compn> you can /q jesseg BBB
[03:06:09 CEST] <Compn> and tell him / her you self ;P
[03:07:25 CEST] <BBB> ok done
[03:07:26 CEST] <BBB> tyu
[03:07:41 CEST] <Compn> thanks to you for caring about user problems too
[03:16:52 CEST] <jamrial> https://git.videolan.org/?p=x264/x264-sandbox.git;a=commitdiff;h=9c7346e766cfd26e702f5ceb4f4b1cebcdab7afd i don't get this, we (supposedly) have had nasm support for a while without these changes to x86inc
[03:17:08 CEST] <jamrial> there are even a couple fate clients using nasm
[03:19:48 CEST] <BBB> jamrial: well our x86inc.asm files arent exactly in sync
[03:19:59 CEST] <BBB> and they didnt historically care about our nasm changes so maybe thats some of our differences
[03:22:28 CEST] <jamrial> ah, so we have nasm related changes on our side. i thought x86inc was mostly in sync, and that it was x86util the one that diverged a lot (because gpl vs lgpl)
[03:22:52 CEST] <BBB> daemon404 mentioned x86inc.asm also isnt exactly in sync
[03:28:02 CEST] <rcombs> x86inc should be in its own repo that people can add as a submodule
[03:32:08 CEST] <BBB> probably& but then again its just one file
[03:32:48 CEST] <rcombs> maybe it and x86util, if people reuse that too
[03:33:28 CEST] <rcombs> a whole repo does feel slightly "overkill", but it's not like it costs anyone anything, and I can't think of a better method
[03:36:26 CEST] <Compn> laurent was iirc porting x264 stuff back
[03:36:29 CEST] <cone-663> ffmpeg 03Philip Langdale 07master:7ae805db7097: avcodec/nvenc: Fix typo: 1204 -> 1024
[03:36:34 CEST] <Compn> but i havent seen em in a while
[03:36:53 CEST] <Compn> and d_s i havent seen in forever
[05:20:50 CEST] <cone-663> ffmpeg 03Michael Niedermayer 07master:4240e6a92117: avfilter/vf_hqdn3d: Initialize the whole LUT
[05:20:51 CEST] <cone-663> ffmpeg 03Michael Niedermayer 07master:52fc3e372f8e: avfilter/x86/vf_hqdn3d: Fix register types
[05:49:44 CEST] <philipl> Do we have a sample hevc file that uses long term references? All of the samples I've found so far do not.
[06:12:53 CEST] <jamrial> maybe you can create one with x265, or nvenc
[06:27:30 CEST] <philipl> With nvenc, i can tell it it's allowed to use ltrs, but I can't force it.
[06:43:28 CEST] <philipl> can't make x265 do it either.
[08:36:53 CEST] <cone-307> ffmpeg 03James Almer 07master:c312bfac4c8a: x86/cpu: add AV_CPU_FLAG_AVXSLOW flag
[11:39:09 CEST] <cone-894> ffmpeg 03Dave Rice 07master:09ec31ce3aab: ffprobe: print stream_index with frame data
[12:07:41 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:073f4b93b270: fate: Update exif tests after 09ec31ce3aab41407965e7a95b7d145702434c26
[12:32:53 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:1cf87e168042: doc/build_system: Document GEN, text taken from the webpage
[13:37:29 CEST] <wm4> holy shit I hate RELATIVE_TS_BASE so much
[13:37:44 CEST] <wm4> such a shitty, confusing, nonsensical hack that got into my way more than once
[13:37:55 CEST] <wm4> I'm lucky to remember it, saved me some thorough confusion
[13:43:09 CEST] <nevcairiel> i still dont really know wtf its for
[14:17:33 CEST] <wm4> ok, sometimes utils.c calls av_get_audio_frame_duration() to somehow magically make up a mp3 frame size...
[14:18:01 CEST] <wm4> which of course goes wrong, for some reason
[14:32:13 CEST] <wm4> so the parser simply can't deal with junk
[15:03:20 CEST] <ubitux> philipl: please indicate the hash of the commit in the ticket you close
[15:03:41 CEST] <ubitux> it gives some insight to the reporter and later readers
[15:48:55 CEST] <cone-894> ffmpeg 03hSÇ 07master:e89e23e1bc1b: avcodec: loongson3 optimized h264dsp weighted mc with mmi
[15:48:56 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:2ae03968147b: avformat/mpsubdec: change multipler to int, it only stores 1 and 100
[16:23:28 CEST] <wm4> rcombs: done; you can make your changes on top of these
[16:23:37 CEST] <wm4> *this patch
[16:24:25 CEST] <rcombs> \o/
[16:26:36 CEST] <nevcairiel> i'm somewhat confused about CONFIG_TLS_PROTOCOL in tls.h, you already have a tls_protocol element in the configure script, why didnt you just make that emit a config.h entry
[16:27:12 CEST] <wm4> nevcairiel: because configure creates these entries automagically from parsing allformats.c
[16:27:20 CEST] <wm4> I guess it could be emitted manually
[16:28:02 CEST] <nevcairiel> you can add things to it by adding things to HAVE_LIST or what the var was called
[16:28:15 CEST] <nevcairiel> although the name might change then
[16:28:55 CEST] <nevcairiel> or CONFIG_LIST as it may be
[16:29:25 CEST] <nevcairiel> or CONFIG_EXTRA
[16:29:28 CEST] <nevcairiel> so many choices!
[16:30:08 CEST] <nevcairiel> although this concludes my level of caring
[16:30:16 CEST] Action: nevcairiel goes back to lurking
[16:30:44 CEST] <wm4> I've run configure too often lately, so I'll leave this problem to the next contributor
[17:12:06 CEST] <philipl> ubitux: done.
[17:12:13 CEST] <ubitux> thx
[17:13:43 CEST] <philipl> ubitux: my intern has some questions about using the ASS callbacks. He posted to the mailing list. Could you give him some guidance?
[17:13:55 CEST] <ubitux> ah yeah i wanted to answer
[17:13:57 CEST] <ubitux> give me a moment
[17:13:59 CEST] <philipl> cheers.
[17:18:56 CEST] <cone-894> ffmpeg 03wm4 07master:2b3e9bbfb529: avformat/mp3: skip junk at the beginning of mp3 files
[17:18:57 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:f722009ad997: avformat/mp3dec: Ensure the buffer is large enough to avoid seeks from the first frame search
[17:24:49 CEST] <wm4> michaelni: maybe this should have been before the loop, with 65*1024+4 as parameter
[17:34:02 CEST] <michaelni> wm4, i was thinking of that as well but thought it might end up reading 64k in and possibly causing startup delay at low bitrates, though that shouldnt happen. if you prefer, i can change it, either should be fine
[17:37:40 CEST] <wm4> I don't care particularly much
[17:37:51 CEST] <wm4> it would matter for the failure case only
[17:45:45 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:07fc47909f6f: avformat/mpsubdec: Use double instead of float for timestamp calculations
[17:50:14 CEST] <michaelni> wm4, either should be fine, even for the failure case or am i missing something ?
[17:50:53 CEST] <wm4> michaelni: ah, you're right, I misunderstood the code
[17:58:26 CEST] <philipl> A repeat from yesterday: Do we have a sample hevc file that uses long term references? All of the samples I've found so far do not.
[17:58:33 CEST] <philipl> and I can't convince nvenc or x265 to use them.
[18:03:01 CEST] <nevcairiel> i'm somewhat surprised that feature is even still with us, i dont think h264 used it much either
[18:03:34 CEST] <BtbN> What is it even usefull for?
[18:04:08 CEST] <Daemon404> RPS_D and RPS_E have long-term refs
[18:04:14 CEST] <Daemon404> according to the conformance docs
[18:04:17 CEST] <BtbN> Not needing an I-frame for a scene change, when the same/a similar scene was used much earlier?
[18:04:47 CEST] <nevcairiel> i have no idea what its useful for, i suppose there is a limit on how long a frame can live in the shortterm ref buffer otherwise?
[18:05:08 CEST] <Daemon404> http://pastie.org/10210361
[18:06:17 CEST] <Daemon404> http://wftp3.itu.int/av-arch/jctvc-site/bitstream_exchange/draft_conformance/RPS_D_ericsson_6.zip
[18:06:21 CEST] <philipl> cheers.
[18:06:25 CEST] <Daemon404> http://wftp3.itu.int/av-arch/jctvc-site/bitstream_exchange/draft_conformance/RPS_E_qualcomm_5.zip
[18:06:28 CEST] <Daemon404> enjoy,.
[18:06:38 CEST] <nevcairiel> chances are we even have those in fate
[18:06:46 CEST] <Daemon404> probably.
[18:06:55 CEST] <philipl> I couldn't find an hevc or h265 section on the samples site.
[18:07:05 CEST] <Daemon404> fate, not samples
[18:07:09 CEST] <nevcairiel> yeah both of those are tested in fate
[18:07:12 CEST] <philipl> good point.
[18:07:16 CEST] <nevcairiel> i think most ref samples are
[18:07:18 CEST] <Daemon404> http://fate-suite.ffmpeg.org/hevc-conformance/
[18:08:23 CEST] <philipl> My long term ref handling doesn't work yet. :-)
[18:08:41 CEST] <nevcairiel> i dont think i ran the hevc conformance  through dxva2
[18:08:44 CEST] <nevcairiel> i should do that some day
[22:12:02 CEST] <cone-894> ffmpeg 03wm4 07master:4a006b9eb7e7: lavf: split tls.c
[22:23:44 CEST] <cone-894> ffmpeg 03wm4 07master:d8ffb2055f0e: lavf: split tls.c
[22:23:45 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:fbf9583f9f88: Merge commit 'd8ffb2055f0e0fcb5d025bab72eb19c2a886c125'
[22:31:11 CEST] <cone-894> ffmpeg 03wm4 07master:a9f1d584e53f: lavf: move TLS-related ifdeffery to library specific files
[22:34:28 CEST] <wm4> michaelni: don't merge the schannel patch yet
[22:34:35 CEST] <wm4> rcombs probably wants to make amends
[22:44:22 CEST] <rcombs> yeah, I'll probably tweak at least a few things
[22:46:30 CEST] <cone-894> ffmpeg 03wm4 07master:57cde2b180fc: lavf: move TLS-related ifdeffery to library specific files
[22:46:31 CEST] <cone-894> ffmpeg 03Michael Niedermayer 07master:eff2ed2fde8d: Merge commit '57cde2b180fcec0eaf60aad65f436ab6420546f5'
[00:00:00 CEST] --- Thu May 28 2015