00:17 <ramiro> haasn: how many faces do your dice have? :)

04:33 <fflogger> [editedticket] Gyan: Ticket #11610 ([swscale] scale with reference prevents video to be scaled with a second scale filter instance) updated https://trac.ffmpeg.org/ticket/11610#comment:1

07:07 <Lynne> michaelni: -level 4 is broken right now because of CRC ref issues in the extradata

07:07 <Lynne> the parser has not been updated

07:41 <haasn> ramiro: I had a left over d18446744073709551615 lying around from my pathfinder days

11:31 <michaelni> Lynne, "-vcodec ffv1 -strict -2 -level 4" works here, also theres a patch from andreas on the ML to fix a crcref issue not sure thats what you refer to

11:54 <microchip_> ffmpeg errors out when there's a : in the file name :/

11:55 <Lynne> michaelni: it fails here

11:56 <Lynne> git master, right?

12:05 <cone-561> ffmpeg Andreas Rheinhardt master:a6bcc773b5aa: avcodec/ffv1enc: Fix explicitly set -slicecrc 2

12:05 <cone-561> ffmpeg Andreas Rheinhardt master:1c78db8ffc6f: avcodec/vc2enc: Use LUT to avoid repeated av_log2()

12:05 <cone-561> ffmpeg Andreas Rheinhardt master:54c865fbec6f: swscale/utils: Fix potential race when initializing xyz tables

12:07 <mkver> Lynne: Did you set the slicecrc option explicitly (to 2)?

12:08 <mkver> If so, then your issue should have been fixed in a6bcc773b5aa3ad which I just applied.

12:38 <Lynne> michaelni: mkver: make -j3 && ./ffmpeg_g -i test.nut -level 4 -strict -2 -c:v ffv1 -y output.mkv && ./ffmpeg_g -i output.mkv -f null -

12:38 <Lynne> this fails here

12:39 <Lynne> adding -slicecrc 2 also fails

12:40 <mkver> Lynne: I think you run into the issue the second patch from https://ffmpeg.org/pipermail/ffmpeg-devel/2025-May/344097.html fixes.

12:41 <Lynne> oh, that does fix it

12:41 <Lynne> what does Video for Windows have to do with this, though?

12:42 <mkver> Lynne: Namely, if the extradata size before being muxed into Matroska is odd, the current demuxer will export it with an additional byte at the end, breaking checksums.

12:42 <mkver> Lynne: We do not store Matroska in its native, V_FFV1 form, but as VfW. Because using the latter might have created files incompatible with older demuxers.

12:43 <mkver> Changing this is the third patch of said patchset.

12:43 <Lynne> ah, and all other codecs don't care about the extradata total size, just that its large enough

12:43 <mkver> Seems so.

12:46 <mkver> For FFV1 version 3 the CRC is zero, which is not affected by appending a zero byte.

12:46 <mkver> Btw: There is a discussion about whether we should switch to V_FFV1.

12:51 <cone-561> ffmpeg Lynne master:bf6d3dc339b6: ffv1enc_vulkan: allow slicecrc=2

13:26 <fflogger> [editedticket] Balling: Ticket #11514 ([avcodec] Cuvid decoder problem with deinterlacing) updated https://trac.ffmpeg.org/ticket/11514#comment:7

13:37 <TheVibeCoder> something changed with bsf changing output codec id

13:38 <TheVibeCoder> does media100 decoding still works within ffmpeg?

13:38 <TheVibeCoder> its not covered by fate

13:39 <mkver> TheVibeCoder: Obvious culprit is 1c170613975d3cbcbb5aaa469b4a3cf0df5d4c2b.

13:39 <TheVibeCoder> samples are here https://trac.ffmpeg.org/ticket/959

13:40 <mkver> I think Zhao Zhili did not deem it possible that a BSF could change output codec id.

13:41 <TheVibeCoder> thing is if i set bsf to not change codec id decoders still work

13:41 <TheVibeCoder> because there are special decoders for it

13:41 <TheVibeCoder> so on that side its cargo culted non-sense

13:42 <TheVibeCoder> but regression is still there even though for media100 decoder and others there is workaround

13:43 <TheVibeCoder> what is correct way to do this? change bsf to not change codec id and fix this regression that mkver pointed out?

13:44 <fflogger> [editedticket] oldshev: Ticket #11610 ([swscale] scale with reference prevents video to be scaled with a second scale filter instance) updated https://trac.ffmpeg.org/ticket/11610#comment:2

13:45 <TheVibeCoder> or if regression is fixed than bsfs must really change codec id?

13:46 <mkver> TheVibeCoder: IMO we should not update the externally visible extradata at all; instead, decoders there should be a no-ownership pointer in AVCodecInternal pointing to the actual extradata (it is set to the initial, user-supplied extradata when avcodec_open2() and set to the output-extradata of the BSF if there is one. Decoders that may use a BSF changing extradata need to use AVCodecInternal.extradata; all the others can continue as now.

13:46 <mkver> 1c170613975d3cbcbb5aaa469b4a3cf0df5d4c2b would then be reverted.

13:47 <mkver> s/decoders there should be/there should be/

13:48 <TheVibeCoder> also please someone add media100 coverage to  FATE

13:48 <jamrial> do we really need a separate decoder for media100 if it's going to be converted to mpegb?

13:48 <TheVibeCoder> looks like yes

13:49 <jamrial> we could maybe try to make it so the bsf is only inserted if the input data is recognized as media100

13:49 <TheVibeCoder> because it triggers bsf in its spec

13:50 <TheVibeCoder> that is over-engineering, the normal data is properly signaled by fourcc

13:52 <TheVibeCoder> its just there to not duplicated decoders code, but still separate codec_id should be there because bitstream is not compatible even though its basically modified mjpeg-b

13:53 <TheVibeCoder> so media100, ahx (mp2 variant), wwvorbis (wwise vorbis variant) are currently bitten by this

13:55 <TheVibeCoder> - ahx is CRI mpv2 variant, those game formats are sometimes ridiculous

13:55 <TheVibeCoder> *mp2

13:56 <TheVibeCoder> and none of them modify extradata, vvc is so huge that I do not dare to peek at it

14:38 <fflogger> [editedticket] bugsan: Ticket #11582 ([ffmpeg] Please add an option to make the new "elapsed" stat optional) updated https://trac.ffmpeg.org/ticket/11582#comment:2

14:51 <fflogger> [newticket] gabriel: Ticket #11611 ([avformat] VVC in MKV container: IDRs with leading pictures not decoded correctly) created https://trac.ffmpeg.org/ticket/11611

15:25 <fflogger> [editedticket] Balling: Ticket #3643 ([swscale] XYZ to RGB color space) updated https://trac.ffmpeg.org/ticket/3643#comment:8

15:34 <fflogger> [editedticket] Balling: Ticket #11578 ([ffmpeg] Waveform discontinuity in decoded E-AC-3) updated https://trac.ffmpeg.org/ticket/11578#comment:8

16:36 <fflogger> [newticket] Andrew-R: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) created https://trac.ffmpeg.org/ticket/11612

17:00 <fflogger> [editedticket] Lynne: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) updated https://trac.ffmpeg.org/ticket/11612#comment:1

17:02 <fflogger> [newticket] Andrew-R: Ticket #11613 ([avcodec] ffv1_vulkan and level 4 does not work well) created https://trac.ffmpeg.org/ticket/11613

17:05 <fflogger> [editedticket] Lynne: Ticket #11613 ([avcodec] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:1

17:06 <fflogger> [editedticket] j7n: Ticket #9471 ([avformat] EAC3 native encoder is only gapless in the beginning, not in the end) updated https://trac.ffmpeg.org/ticket/9471#comment:9

17:09 <fflogger> [editedticket] Andrew-R: Ticket #11613 ([avcodec] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:2

17:11 <fflogger> [editedticket] Andrew-R: Ticket #11613 ([avcodec] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:3

17:13 <fflogger> [editedticket] Lynne: Ticket #11613 ([avcodec] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:4

17:42 <fflogger> [editedticket] Andrew-R: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) updated https://trac.ffmpeg.org/ticket/11612#comment:2

17:45 <fflogger> [editedticket] Lynne: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) updated https://trac.ffmpeg.org/ticket/11612#comment:3

17:49 <fflogger> [editedticket] Andrew-R: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) updated https://trac.ffmpeg.org/ticket/11612#comment:4

18:08 <fflogger> [editedticket] Andrew-R: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) updated https://trac.ffmpeg.org/ticket/11612#comment:5

19:21 <fflogger> [editedticket] Balling: Ticket #9471 ([avformat] EAC3 native encoder is only gapless in the beginning, not in the end) updated https://trac.ffmpeg.org/ticket/9471#comment:10

20:16 <cone-344> ffmpeg Andreas Rheinhardt master:92e310eb82e0: avformat/matroskadec: Fix VfW extradata size

20:18 <fflogger> [editedticket] Andrew-R: Ticket #11612 ([avcodec] Vulkan HEVC decode 1st frame broken) updated https://trac.ffmpeg.org/ticket/11612#comment:6

20:24 <fflogger> [editedticket] mkver: Ticket #11613 ([avformat] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:5

20:48 <thardin> almost 8000 unread emails on the ML

20:48 <thardin> so.. behind

20:49 <thardin> I almost feel guilty. luckily there's nothing about MXF

21:00 <fflogger> [editedticket] Andrew-R: Ticket #11613 ([avformat] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:6

21:10 <fflogger> [editedticket] Andrew-R: Ticket #11613 ([avformat] ffv1_vulkan and level 4 does not work well) updated https://trac.ffmpeg.org/ticket/11613#comment:7

21:55 <ramiro> wbs: last year I did a bunch of improvements on neon yuv2rgb, but in the end it became an overhaul of the neon code. I said to myself that some day I would split that into separate commits and submit for review, but now that I look back at the code, it seems like a lot of work to split into smaller functional commits. do you think I could submit the overhaul, or is it best to try and split it, for

21:55 <ramiro> commit history's sake and make review easier?

21:55 <ramiro> (here's the "final" commit: https://github.com/ramiropolla/ffmpeg/commit/f7284e12 )

22:12 <kierank> thardin: indeed unmanageable to read all that

22:16 <thardin> lots of traffic. which is good I guess. but also not LKML levels

22:18 <kierank> mostly noise imo

22:18 <kierank> maybe swscale is the exception

22:20 <thardin> haasn is love. haasn is life

22:22 <thardin> I had some ideas earlier today how to improve the throughput of the cinepak encoder, for the CD decoder usecase. because I know at least one group of users want that. you want near constant bitrate there, which has some implications

22:23 <thardin> I have some VQ patches that I haven't bothered sending to the ML

22:23 <thardin> which would speed up all VQ encoders

23:04 <cone-344> ffmpeg Timo Rothenpieler master:4cb42551d4ad: fftools/resources: add missing extensions to .gitignore

23:04 <cone-344> ffmpeg Timo Rothenpieler master:af943831248d: ffbuild: correctly silence and tag new css/html steps

23:04 <cone-344> ffmpeg softworkz master:aa38270630f5: fftools/makefile: Remove resources from ffprobe

23:04 <cone-344> ffmpeg softworkz master:75a5e0959f58: fftools/ffmpeg: Free print_graph option variables

23:04 <cone-344> ffmpeg softworkz master:a0ac89e46c70: fftools/graphprint: Fix memory leaks

23:04 <cone-344> ffmpeg softworkz master:b7d44dde3b89: fftools/tf_mermaid: Add missing uninit and fix leaks

23:58 <fflogger> [editedticket] mkver: Ticket #11164 ([avcodec] visual macroblock error with mpeg2video encoding) updated https://trac.ffmpeg.org/ticket/11164#comment:4