[Ffmpeg-devel-irc] ffmpeg-devel.log.20181113

[burek]

[08:17:28 CET] <cone-337> ffmpeg 03Tobias Rapp 07master:7992c2867efc: avcodec/libopenh264enc: fix class_name
[11:40:01 CET] <cone-428> ffmpeg 03Martin Vignali 07master:13ee107afd3d: avcodec/prores_metadata_bsf : fix colorspace check
[12:42:29 CET] <thardin> that sscanf() solution sure is something
[12:55:03 CET] <thardin> there has to be a better solution that reimplementing it or fork()ing
[12:56:16 CET] <durandal_1707> no
[12:59:31 CET] <thardin> fflibc when?
[13:02:29 CET] <thardin> my inner formal verifier cries a little looking at code like this
[13:30:05 CET] <BtbN> locale issues with printf/scanf are a huge pain, and there really does not seem to be a proper solution
[13:30:26 CET] <BtbN> You'd wish there'd be something like disable_locale() reenable_locale() you can just slap around it
[13:35:00 CET] <thardin> that wouldn't be reentrant
[13:35:17 CET] <BtbN> depends on if it's set globally or thread local
[13:35:26 CET] <thardin> right
[13:36:09 CET] <thardin> this reminds me of errno, and the efforts to try and make it less annoying
[13:36:37 CET] <thardin> a libified libc (liblibc if you will) would be useful
[13:36:37 CET] <BtbN> errno just seems like a design flaw that somehow dragged on to me
[13:38:06 CET] <durandal_1707> tell him that: https://github.com/wm4/libinsanity
[13:41:01 CET] <thardin> gpl might be a hindrance (not for me though)
[16:33:54 CET] <orbea> Hi, I am having a segfault in the ffmpeg master (as of a few days ago) and older releases when running a single alephone scenerio (game). Here is a backtrace https://pastebin.com/YZ1qSnCC and the alephone issue. https://github.com/Aleph-One-Marathon/alephone/issues/121 My question, is this an ffempg bug or an alephone bug? maybe it could be both?
[16:35:01 CET] <orbea> my limited understanding its crashing on an audio frame
[16:44:37 CET] <jamrial> orbea: https://github.com/Aleph-One-Marathon/alephone/blob/master/Source_Files/FFmpeg/SDL_ffmpeg.c#L2060
[16:44:40 CET] <jamrial> sizeof(AVFrame) is not a part of the public ABI
[16:45:11 CET] <jamrial> they should allocate a frame with av_frame_alloc(), and not in stack
[16:47:09 CET] <jamrial> it's crashing because that frame on stack is uninitialized
[16:50:40 CET] <orbea> ah, thanks!
[16:50:54 CET] <orbea> I'll make sure that is conveyed to the alephone devs
[16:51:25 CET] <orbea> would you mind if I quoted irc for that issue?
[16:58:23 CET] <atomnuker> its written in libavutil/frame.h, line 219 in current git master
[16:59:31 CET] <orbea> oh, thanks. *looks for the comments*
[17:17:20 CET] <JEEB> durandal_1707: finally got some time to test things, and seems like it isn't framesync2 since if I take upstream + revert my first attempt at fixing the sub2video issue (e760c12aeef608aa8b416664687b9aca3a2c6f68) then the performance seems good
[17:17:55 CET] <JEEB> it's most likely one of hte sub2video things I've posted since on the ML
[17:19:10 CET] <JEEB> I was going to say that it's funny because the affected input had no subtitles to begin with, but oh yes it did :P
[17:24:33 CET] <JEEB> it's not completely derped though, since another thing has now run N inputs and two with subtitle overlay
[17:24:54 CET] <JEEB> but yea, will be doing more does it vs does it not testing
[18:43:15 CET] <durandal_1707> anybody have dot-crawl samples to share?
[18:43:59 CET] <atomnuker> yeah, JEEB had one
[18:45:39 CET] <atomnuker> https://fushizen.eu/u/jeeb/derp/00002.MPS
[18:46:31 CET] <JEEB> ah yes, that's why it popped up during the PSP video discussions
[18:46:41 CET] <atomnuker> you'll need to deinterlace, but the dot crawl is monsterous
[18:46:56 CET] <JEEB> mistagged full range, dot crawl, ATRAC3+ in MPEG-PS
[18:50:44 CET] <atomnuker> eh? the file has audio?
[18:51:41 CET] <JEEB> yes
[18:51:50 CET] <JEEB> if you apply mistydemeo's patch it probably sounds
[18:52:10 CET] <JEEB> although last I checked the parser or something barfs if you try to seek
[18:53:16 CET] <atomnuker> where did you get that, can't be from a dvd
[18:53:45 CET] <JEEB> UMD Video
[18:53:55 CET] <atomnuker> didn't know that was a thing
[18:54:06 CET] <JEEB> oh yes, I also have Spiderman on it
[18:54:14 CET] <JEEB> since Sony gave them out if you registered your PSP
[18:54:14 CET] <JEEB> :D
[18:59:04 CET] <gnafu> Classic.
[19:24:02 CET] <BBB> michaelni: do you remember what the ffmpeg flag is called to not scale frames? e.g. if the input is scalable, and I want the output frames to be variably scaled also, you mentioned an option a few years back to not scale to one standard size in the output stream, what was that called again?
[20:23:51 CET] <michaelni> BBB,i think you refer to "-reinit_filter 0" the intend of that flag IIRC was to send input changes into the filter graph so the filter graph could handle them in any way choosen without loosing state. If the filter graph didnt remove the change then it would be passed on the encoder this was more a sideeffect and possibly usefull but not why the flag was there. It seems the flag doenst work in master anymore
[20:26:51 CET] <BBB> ok, thanks, little bit unfortunate but Ill find some other way then
[20:26:55 CET] <BBB> thanks for answering
[20:50:43 CET] <durandal_1707> anybody against ffvhuff patch? i gonna push it now!
[21:43:12 CET] <kierank> durandal_1707: if you are bored you can finish this patch https://trac.ffmpeg.org/ticket/7163#comment:7
[21:54:11 CET] <durandal_1707> kierank: that patch is incomplete
[21:54:25 CET] <kierank> "Here is unfinished patch" 
[21:54:27 CET] <kierank> no shit
[21:55:04 CET] <durandal_1707> kierank: unfinished patch is incomplete 
[21:55:19 CET] <durandal_1707> partial
[21:55:28 CET] <durandal_1707> missing some code i think?
[21:55:37 CET] <kierank> 12-bit alpha
[21:55:52 CET] <durandal_1707> not that
[21:57:27 CET] <durandal_1707> do we have 12 bit idct
[21:58:47 CET] <kierank> we do
[21:58:52 CET] <kierank> that patch produces clean output
[22:03:52 CET] <durandal_1707> kierank: write simd code
[22:04:10 CET] <kierank> maybe at christmas
[22:15:34 CET] <cone-428> ffmpeg 03Andreas Rheinhardt 07master:959521b58de9: h264_levels, h264_metadata_bsf: Fix levels typo
[22:15:35 CET] <cone-428> ffmpeg 03Andreas Rheinhardt 07master:3c9c9b1568ef: h264_metadata: Don't use inferred value of buffering frames
[23:01:38 CET] <cone-428> ffmpeg 03Michael Niedermayer 07master:6b4c9854daba: avcodec/golomb: Speed up long ur_golomb codes
[23:01:39 CET] <cone-428> ffmpeg 03Michael Niedermayer 07master:2f888771cd1c: avcodec/shorten: Fix integer overflow with offset
[00:00:00 CET] --- Wed Nov 14 2018