[Ffmpeg-devel-irc] ffmpeg-devel.log.20180922

[burek]

[02:00:25 CEST] <cone-282> ffmpeg 03Michael Niedermayer 07master:fe315feab59f: avcodec/eac3dec: Check that channel_map does not contain more than EAC3_MAX_CHANNELS
[02:00:26 CEST] <cone-282> ffmpeg 03Michael Niedermayer 07master:148a21611d85: avcodec/ac3dec: Fix shift signedness in mask creation
[10:36:11 CEST] <durandal_1707> no video, no audio, no live streaming, nothing, future is here!
[11:17:29 CEST] <atomnuker> its a bit dead this year tbh, room's less than 3/4 full
[11:17:50 CEST] <atomnuker> everyone undershot their allocated time by quite a lot, including me
[11:19:20 CEST] <atomnuker> I think I'll try to go back to my room and work on vulkan
[11:21:03 CEST] <atomnuker> after I get lunch which will hopefully not contain insects like last time
[11:21:37 CEST] <durandal_1707> lol
[11:22:37 CEST] <atomnuker> I hope it'll be pizza, if so it'll be the first actual food I've had in days
[11:23:24 CEST] <atomnuker> you can't find anything in paris after 8 or 9, I do mean anything
[11:39:41 CEST] <TD-Linux> atomnuker, BBB actually overshot his time
[11:40:39 CEST] <atomnuker> well its all about averages, and the way we're going we'll be waiting for theoretical pizzas for a while
[14:13:07 CEST] <kierank> how do I make a mpeg4video fate test
[14:13:23 CEST] <JEEB> depends on what you want to test and if it requires an API client?
[14:13:38 CEST] <kierank> doesn't need an api client
[14:13:40 CEST] <kierank> just ffmpeg.c
[14:14:25 CEST] <JEEB> ok, for decoding it's basically doing the checksum claculation per frame or so usually (framemd5 IIRC), there's quite a few examples of that in the tests IIRC
[14:14:52 CEST] <JEEB> for encoding you could hash the output compressed frames, or something possibly better
[14:15:47 CEST] <kierank> michaelni: so are you going to revert your get_unary change
[14:33:40 CEST] <kierank> JEEB: https://git.videolan.org/?p=ffmpeg.git;a=blob;f=tests/fate/mpeg4.mak;h=ed6a2fac2003f5992c72c14271847ac0a990a9f6;hb=HEAD
[14:33:45 CEST] <kierank> so what do I add to this?
[14:54:11 CEST] <michaelni> kierank, you mean push the doc improvment ? not sure what you refer to by revert 
[14:54:49 CEST] <kierank> michaelni: change the mpeg4 decoder to use the correct length
[14:54:58 CEST] <kierank> instead of checking for the case that's wrong
[14:56:49 CEST] <michaelni> kierank, the way i remember the spec/draft is that theres a invalid case, so a check is needed.
[14:57:45 CEST] <kierank> michaelni: nope, this is entirely because of the misleading get_unary docs
[14:59:14 CEST] <michaelni> the case of 12 zero bits is invalid according to what i read here. So no matter what get_unary() does a check and "breakout" is needed
[14:59:26 CEST] <kierank> yes so you set the "length" field to 10
[14:59:55 CEST] <kierank> oh I see what you mean
[14:59:59 CEST] <kierank> wow get_unary is stupid
[15:00:20 CEST] <kierank> actually no it's fine
[15:00:22 CEST] <kierank> you'll get 11
[15:00:23 CEST] <kierank> and that's it
[15:00:33 CEST] <kierank> so yes the mpeg4 decoder needs changing
[15:01:44 CEST] <kierank> wow the docs are still confusing
[15:02:22 CEST] <michaelni> if get_unary() is used with 11 instead of 12 then  the error case is folded into the escape and we would need to read an additional bit and we would need to deal with end of input exclpicitly or check the extra read bit
[15:02:31 CEST] <michaelni> none of that seems simpler
[15:02:36 CEST] <kierank> yes but the docs don't make this clear at all
[15:02:37 CEST] <kierank> "Unary length - 1 or length for the stop code (all 1 or all 0 of len)."
[15:02:47 CEST] <kierank> does the unary length include the 1 or not?
[15:02:56 CEST] <kierank> if there is a 1
[15:04:06 CEST] <michaelni> kierank, are you refering to the docs in git or the patch i posted 2 days ago ?
[15:04:12 CEST] <michaelni> the patch contain an example
[15:04:17 CEST] <michaelni> containS
[15:04:26 CEST] <kierank> the patch you posted
[15:04:47 CEST] <michaelni> ok, so we need to improve the pacth than if it makes no sense to you
[15:06:22 CEST] <kierank> it kind of makes sense
[15:15:35 CEST] <kierank> durandal_1707: I want to fix prores idct today
[15:17:12 CEST] <michaelni> kierank, is this better: https://pastebin.com/GxaxLz3w ?
[15:17:41 CEST] <kierank> michaelni: seems better
[15:18:42 CEST] <michaelni> ok ill post it to the ML and will apply later if noone suggests something else/better
[15:18:54 CEST] <durandal_1707> kierank: it doesn need fixing
[15:22:59 CEST] <cone-366> ffmpeg 03Martin Vignali 07master:59256dea146d: fate/mov : use do_md5sum for mov-faststart-4gb-overflow test
[15:24:24 CEST] <kierank> durandal_1707: wrong
[15:24:47 CEST] <kierank> It needs to avoid prores idct hacks
[15:24:52 CEST] <kierank> We just discussed
[15:26:17 CEST] <durandal_1707> kierank: which hacks?
[15:28:11 CEST] <kierank> Wrong idct
[15:33:24 CEST] <durandal_1707> kierank: you mean it is not 12 bit?
[15:34:04 CEST] <kierank> Correct
[15:40:15 CEST] <durandal_1707> kierank: that conflict with my work on adding lowres support
[15:40:34 CEST] <durandal_1707> so do it fast
[15:40:42 CEST] <kierank> We try to do it today
[15:41:02 CEST] <durandal_1707> we? you and Carl?
[15:45:25 CEST] <kierank> And BBB probably
[15:46:51 CEST] <durandal_1707> BBB is working on Dav1d
[16:19:07 CEST] <atomnuker> takes 30 minutes to get from hotel to dinner place, and I can't find any info that there's a restaurant at the dinner place
[16:27:45 CEST] <durandal_1707> atomnuker: don't get lost!
[16:36:12 CEST] <lede> atomnuker: i guess it will happen on a boat
[16:46:09 CEST] <durandal_1707> where is ffv2 source code?
[16:53:48 CEST] <lede> https://github.com/atomnuker/FFmpeg/tree/exp_ffv2_daala
[17:21:17 CEST] <durandal_1707> atomnuker: tell me when you gonna test afftdn and how
[17:39:29 CEST] <durandal_1707> dav1d code is full of warnings when compiled with clang
[17:39:59 CEST] <lede> even without clang
[17:42:35 CEST] <durandal_1707> 1.ivf is not an AV1 file [tag=av0180u|0x61763031]
[17:42:54 CEST] <durandal_1707> it doesnt supports .ivf created by ffmpeg
[17:43:04 CEST] <durandal_1707> neither mp4
[17:49:52 CEST] <lede> tools/input/ivf.c:64
[17:52:43 CEST] <jamrial> durandal_1707: where's the repo?
[18:00:51 CEST] <durandal_1707> jamrial: https://code.videolan.org/videolan/dav1d
[18:04:07 CEST] <durandal_1707> ./build/dav1d --framethreads 2 -i build/1.ivf --muxer yuv -o /dev/null
[18:04:13 CEST] <durandal_1707> ^ hangs
[18:06:05 CEST] <durandal_1707> and it doesnt support pipeing output
[18:09:22 CEST] <lede> framethreading is unstable
[18:10:36 CEST] <lede> it crash, hangs, segfault, double-free, fail assertions, you name it
[18:12:45 CEST] <durandal_1707> am i correct that there is no single asm at all?
[18:13:11 CEST] <lede> yes
[18:13:31 CEST] <lede> yet about as fast as libaom with simd
[18:13:45 CEST] <durandal_1707> yes, i noticed
[18:25:26 CEST] <Shiz> /w/w 16
[19:42:55 CEST] <durandal_1707> imho afftdn is much higher quality audio denoiser than that ladspa plugin
[19:52:43 CEST] <durandal_1707> Compn: add native prosumer decoder to mplayer too?
[21:21:22 CEST] <durandal_1707> anybody against applying ilbc decoder?
[00:00:00 CEST] --- Sun Sep 23 2018