[Ffmpeg-devel-irc] ffmpeg-devel.log.20161124

burek burek021 at gmail.com
Fri Nov 25 03:05:02 EET 2016


[00:56:28 CET] <cone-728> ffmpeg 03Moritz Barsnick 07master:0700d02a697e: lavfi/pan: allow negative gain parameters also for other inputs than the first named
[01:51:35 CET] <cone-728> ffmpeg 03Andreas Cadhalpun 07master:995512328ed8: pgssubdec: only set w/h/linesize when allocating data
[02:00:13 CET] <Chloe[m]> atomnuker: why wait?
[02:01:21 CET] <atomnuker> Chloe[m]: he told us that he needed "a few days" to move it to an external repo
[02:01:30 CET] <atomnuker> like, 3 weeks ago
[02:31:11 CET] <jya> what's the rationale behind http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=5702416c57afb2bb062eb69d60fc42c31c91b674 ? I can understand not attempting to seek if the stream isn't seekable. But duration and size can certainly be updated. This has broken mythtv flash streaming
[03:20:58 CET] <Timothy_Gu> michaelni: So I'm seeing that https://www.ffmpeg.org/ still uses a certificate from StartCom CA
[03:21:24 CET] <Timothy_Gu> But apparently their CA had some pretty serious messup Mozilla, Google, Apple, etc. are distrusting them
[03:21:31 CET] <Timothy_Gu> for future certs
[03:21:33 CET] <Timothy_Gu> https://blog.mozilla.org/security/2016/10/24/distrusting-new-wosign-and-startcom-certificates/
[03:21:54 CET] <Timothy_Gu> https://security.googleblog.com/2016/10/distrusting-wosign-and-startcom.html
[03:34:24 CET] <michaelni> Timothy_Gu, reimar is the one who does/did the cert stuff
[03:34:42 CET] <michaelni> please tell him
[10:47:18 CET] <Chloe> atomnuker: sure, well, the current FFserver is in git history. No reason why he couldn't do that even after it's removed
[10:56:01 CET] <Chloe> michaelni: how are you sending signed mail? My client seems to be sending it in a way which results in a bad signature
[11:00:25 CET] <Chloe> Ah, I think I was using PGP/MIME which I guess won't work
[11:00:29 CET] <wm4> merges are behind by 600 commits and 4 months
[11:04:34 CET] <wm4> that makes me even want to consider helping out with the merges, but the next commits to merge are already quite intrusive and complicated
[11:05:16 CET] <nevcairiel> i havent checked yet
[11:42:47 CET] <nevcairiel> philipl: have you tested what happens if you try to to decode 12-bit on older hardware? does it just refuse to create the decoder?
[12:49:17 CET] <matija> trying to build ffmpeg with libnpp, using latest cuda toolkit but ther eis no libnpp there only libnppc,s and i, tried 3.2 and latest git, same result - ERROR: libnpp not found
[12:49:23 CET] <matija> any ideas?
[13:15:16 CET] <superware> I'm using ffmpeg 3.1.4, avformat_open_input is calling my interrupt_callback, but when the callbak returns 1 (I verified that it's actually being called and returns 1) then avformat_open_input doesn't abort but rather keeps blocking, any idea?
[13:17:23 CET] <superware> (the url being opened is "udp://224.0.19.0:40004/"
[13:23:34 CET] <superware> anyone alive? :|
[13:28:34 CET] <superware> ok, this seems to be the same issue from 6 months ago https://trac.ffmpeg.org/ticket/5615
[14:44:46 CET] <wm4> superware: try to find out where it hangs
[14:46:38 CET] <superware> wm4: unfortunately it might be too complex for me to understand exactly where, keep in mind it can be stuck in a specific state and not a specific position/line.
[14:49:01 CET] <superware> wm4: it's very easy to reproduce, it can be any UDP url (eg. udp://127.0.0.1:8000) in fact it shouldn't even exists, because if it does exist the stream will be opened and not have a chance to hang :)
[16:32:24 CET] <philipl> nevcairiel: i have not, but i assume so based on how it behaved with 12bit on the old driver which didn't support it.
[16:33:58 CET] <cone-784> ffmpeg 03Paul B Mahol 07master:8f5a2bed5eab: ffmpeg_filter: fix several logic failures
[17:43:34 CET] <jamrial> BtbN, philipl: is this ok? http://pastebin.com/raw/bjqMkQFm
[17:44:12 CET] <BtbN> "check_header cuda.h"?
[17:44:18 CET] <BtbN> There is no cuda.h used anymore
[17:44:37 CET] <jamrial> it's used by hwcontext_cuda.h
[17:44:50 CET] <jamrial> and it's making make checkheaders fail
[17:45:16 CET] <jamrial> http://fate.ffmpeg.org/log.cgi?time=20161124053746&log=compile&slot=x86_64-archlinux-gcc-checkheaders
[17:45:20 CET] <BtbN> oh right, the fallback.
[17:45:44 CET] <jamrial> that patch fixes it by making sure to not test that header if the Cuda Toolkit is not installed in the build system
[17:45:59 CET] <jamrial> the _internal one can be tested fine since it includes the compat header
[17:46:02 CET] <BtbN> it can test that header though, it just has to include the header from compat first
[17:46:38 CET] <nevcairiel> checkheaders is specifically meant to test if every header works standalone
[17:46:38 CET] <jamrial> that's what hwcontext_cuda_internal.h does, but hwcontext_cuda.h can't
[17:46:43 CET] <nevcairiel> so including something first goes against its entire idea
[17:47:02 CET] <nevcairiel> and if some headers depend on some external library, they need to be excluded
[17:47:14 CET] <BtbN> ah, so that's what that configure check does
[17:47:24 CET] <BtbN> Looked like it makes cuda depend on cuda.h again
[17:47:34 CET] <BtbN> Yeah, that's fine then
[17:47:35 CET] <jamrial> no :p
[17:47:43 CET] <jamrial> it just adds a check for a header if cuda is enabled
[17:49:03 CET] <jamrial> ok, thanks
[17:52:33 CET] <cone-784> ffmpeg 03James Almer 07master:d1de725beed3: cuda: check for cuda.h when enabled
[21:11:07 CET] <cone-784> ffmpeg 03Wan-Teh Chang 07master:dceac9a4a754: avfilter/tests/.gitignore: add integral
[21:11:08 CET] <cone-784> ffmpeg 03Wan-Teh Chang 07master:048b46b4e294: avutil/tests: add cpu_init to .gitignore and tests/fate
[21:16:27 CET] <michaelni> Timothy_Gu, http://fatebeta.ffmpeg.org/v3.2 doesnt seem to exist
[21:23:18 CET] <cone-784> ffmpeg 03Muhammad Faiz 07master:ebb4c783d025: fate/swresample: add resample exact_lin and exact_lin_async test
[21:23:19 CET] <cone-784> ffmpeg 03Muhammad Faiz 07master:06f94149c61f: swresample/resample: optimize exact_rational=on:linear_interp=on case
[23:00:57 CET] <wm4> nevcairiel: does dxva2 (or d3d11va) have a mechanism to check max. video size?
[23:01:57 CET] <nevcairiel> Try to create a decoder, see it fail?
[23:02:26 CET] <nevcairiel> I do thay
[23:02:48 CET] <nevcairiel> It needs no bitstream params or anything, just the dimensions
[23:03:00 CET] <wm4> well I heard that LAV has an option to enable UHD decoding
[23:03:28 CET] <nevcairiel> Some hardware just sucks despite supporting it
[23:03:48 CET] <wm4> slower than sw dec, or failure to decode correctly?
[23:03:59 CET] <nevcairiel> Especially with 4k it was very very slow on first gen
[23:04:32 CET] <nevcairiel> And AMD had broken 4k until they figured it out and blacklisted it in the driver
[23:04:55 CET] <nevcairiel> Half the image was green :)
[23:05:00 CET] <wm4> eh
[23:05:27 CET] <wm4> mpv and google chrome still have issues with normal d3d11va direct rendering on amd
[23:05:45 CET] <nevcairiel> In general it can autodetect the dimensions, but the options are workarounds for bugs
[23:05:57 CET] <nevcairiel> And AMD is just terrible on hwaccel more or less
[00:00:00 CET] --- Fri Nov 25 2016


More information about the Ffmpeg-devel-irc mailing list