00:50 <jamrial> haasn: that patch doesn't do any syncing for frame threading

00:53 <jamrial> wtf, AVFilmGrainAFGS1Params uses sizeof(AVFilmGrainParams)

07:49 <kurosu> I wasn't very happy with the patch, and while I think Michael's comment couldn't be addressed (that's the way it is), I guessed mine was irrelevant if everyone else was happy

07:51 <kurosu> Ah jamrial left

14:26 <jamrial> haasn, elenril: are the first three patches in my set ok? last two can be dropped i think

14:41 <elenril> jamrial: I don't think they can be dropped

14:42 <elenril> the code seems wrong both before and after

14:42 <elenril> seems to me we should revert it entirely and reimplement from scratch

14:43 <elenril> but I didn't test with tsan yet to verify it's definitely a rac

14:43 <elenril> e

15:16 <jamrial> elenril: full revert and applying patches 1-3 (so h274db is not dynamically allocated at all)?

15:17 <elenril> do 1 and 3 make sense after full revert?

15:17 <jamrial> yes, because hevcdec is using sizeof(AVFilmGrainParams)

15:19 <elenril> wasn't that added in this patch?

15:20 <jamrial> no, see AVFilmGrainAFGS1Params in avcodec/aom_film_grain.h

15:21 <elenril> ah, I thought it was hevcdec using it directly

15:27 <elenril> do we have tests for this?

15:30 <jamrial> http://coverage.ffmpeg.org/index.h2645_sei.c.9e026698b854471f1933d292b21d2bd0.html http://coverage.ffmpeg.org/index.aom_film_grain.c.7af06c7a07be61251ec3705362a2c9fb.html

15:30 <jamrial> looks like no

15:30 <elenril> yeah, seems so

15:30 <elenril> I addded abort() to that code and fate passes

15:30 <jamrial> haasn: do you have a sample for aom film grain in ITU T.35 hevc?

15:52 <jamrial> elenril: https://github.com/jamrial/FFmpeg/commits/film_grain/

15:52 <jamrial> last patch is optional. the important one is the one after the revert

17:51 <elenril> would be nice to have a test

17:51 <elenril> and check that tsan is happy

18:15 <jamrial> agree

18:32 <BBB> is everyone excited about going to korea next week?

18:54 <ePirat> BBB, not excited about the 14h flight :D

18:59 <BBB> yes that part sucks

19:05 <Marth64> enjoy the foods

19:11 <IndecisiveTurtle> First time going to VDD, flight time looks most tiring

19:13 <elenril> it's a nice opportunity to think

19:13 <elenril> reconsider your life choices and stuff

19:18 <IndecisiveTurtle> Last time had a trip that long was 10 hours by ship and it was bad, this will prob be worse :/

19:57 <IndecisiveTurtle> Anyone else have issues with setting debug=2 due to VK_KHR_shader_relaxed_extended_instruction being added unconditionally? It seems this was enabled very recently in mesa and isn't exposed on older 24.2.x radv that arch packages. Seems on nvidia prop driver too. Maybe it should be checked until it rolls out more broadly?

20:47 <cone-089> ffmpeg Michael Niedermayer master:14f5d67be38d: swscale/rgb2rgb_template: Fix ff_rgb24toyv12_c() with odd height

20:47 <cone-089> ffmpeg Michael Niedermayer master:23088a5ff2b5: avformat/mxfdec: Fix overflow in midpoint computation

20:47 <cone-089> ffmpeg Michael Niedermayer master:3fe301440549: swscale/output: used unsigned for bit accumulation

20:47 <cone-089> ffmpeg Michael Niedermayer master:db7b4fc89fb1: avformat/mpegts: Initialize predefined_SLConfigDescriptor_seen