02:28 <cone-571> ffmpeg Michael Niedermayer master:c250a903a793: doc/encoders: very basic FFv1 documentation

02:28 <cone-571> ffmpeg Michael Niedermayer master:4cc1495aca45: avformat/iamf_parse: reject ambisonics mode > 1

02:28 <cone-571> ffmpeg Sean McGovern master:b9eaf6e05c2c: swscale/ppc: disable YUV2RGB AltiVec acceleration

02:28 <cone-571> ffmpeg Michael Niedermayer master:c314a68d0462: avcodec/rangecoder: Do not loop renormalization

02:28 <cone-571> ffmpeg Michael Niedermayer master:360e7cafd0e6: avcodec/h2645_parse: Ignore NAL with nuh_layer_id == 63

02:28 <cone-571> ffmpeg Michael Niedermayer master:832649986c2b: avcodec/rangecoder: Move refill check out of refill() function

02:28 <cone-571> ffmpeg Michael Niedermayer master:70fa8bd38cdb: avcodec/rangecoder: Avoid checking for the first byte on every renormalization

02:28 <cone-571> ffmpeg Michael Niedermayer master:18922f4ae3d4: avcodec/rangecoder: Remove unneeded outstanding byte mask

02:28 <cone-571> ffmpeg Michael Niedermayer master:b2da4c33e31f: avcodec/rangecoder: eliminate main branch from renorm_encoder()

02:28 <cone-571> ffmpeg Michael Niedermayer master:aeec39f3c1be: swscale/slice: clear allocated memory in alloc_lines()

02:28 <cone-571> ffmpeg Michael Niedermayer master:7e020f214132: avformat/dxa: check bpc

02:28 <cone-571> ffmpeg Michael Niedermayer master:cdce6a8b97c3: avcodec/eatgq: move array to where it is used

02:28 <cone-571> ffmpeg Michael Niedermayer master:4949e34d96cc: avcodec/eatgq: Check bytestream2_get_buffer() for failure

02:28 <cone-571> ffmpeg Michael Niedermayer master:f52b9d05837c: avformat/qcp: Check for read failure in header

02:28 <cone-571> ffmpeg Michael Niedermayer master:4fc43e45f9e7: avcodec/hevc/hevcdec: initialize qp_y_tab

03:09 <fflogger> [editedticket] Balling: Ticket #9428 ([avcodec] avcodec: truehd parser drops data but does not adjust initial timestamps) updated https://trac.ffmpeg.org/ticket/9428#comment:4

08:30 <kurosu> Does anyone know if the MSVC breakage on alignas is tied to particular versions of the compiler and/or IDE? That sounds like something that should/would be reported, as it's not a small breakage

08:32 <kurosu> (but little comfort for ffmpeg, having to detect that version or that the builtin/... is broken)

09:04 <fflogger> [editedticket] nafmo: Ticket #10732 ([undetermined] avcodec_flush_buffers() not resetting E-AC-3 decoder) updated https://trac.ffmpeg.org/ticket/10732#comment:6

11:21 <haasn> is there a way to easily build ffmpeg + all dependencies with MSAN enabled?

12:06 <haasn> michaelni: what a shocker, https://trac.ffmpeg.org/ticket/11265 is because of the half chroma input "optimization"

12:07 <haasn> it might be possible to fix it properly by either plumbing in the true width somehow and then clamping in the interpolation, or alternatively pre-padding the input (which may require a memcpy)

12:07 <haasn> but I think in all honesty this flag and optimizations like it are on their way out

12:07 <haasn> so I propose to just disable this optimization for odd input sizes

12:39 <cone-703> ffmpeg Zhao Zhili master:6e49b8699657: avutil/mem_internal: Don't use alignas for MSVC

12:47 <JEEB> grmbl, so finally got a stable set of perf data, and it seems like d119ae2fd82a494d9430ff4d4fc262961a68c598 is where ~5% additional resource usage for the same task happened

12:48 <michaelni> "so I propose to just disable this optimization for odd input sizes" <-- this sounds reasonable to me

12:48 <michaelni> haasn, ^

12:49 <nevcairiel> kurosu: havent had time to investigate the msvc breakage much yet, but i do know that its broken with the very latest at least, tracing it back is a bit of a chore

12:50 <kurosu> np and thanks for the reply, I just hoped it would be fixed by now in MSVC

12:51 <kurosu> I'm asking because I remember such a breakage but only for >=32 alignment, and somehow I think not even gcc can do that under Windows, as if some ABI think

12:51 <kurosu> *thing

12:51 <nevcairiel> i did look at their bug tracker but couldnt find any mention of alignas problems

12:52 <nevcairiel> but it may also not be that popular of a feature

12:54 <nevcairiel> when i get some time i may try to setup a simple test case ... if a simple case can reproduce it

12:55 <JEEB> hopefully

12:57 <nevcairiel> it would seem a bit weird that alignment works in general, but only with the proprietary instruction, one would think they would eventually reach the same instruction path

12:57 <nevcairiel> its not like we dont align on msvc, we just use __declspec(align(x)) instead

12:58 <nevcairiel> .. unless we use it on things that alignas is not actually meant to work on

13:58 <haasn> jamrial: did you see my fix for the memleak issue?

13:58 <haasn> I'd appreciate a short LGTM so I can push it

14:01 <jamrial> haasn: no, i missed it

14:01 <jamrial> what's the subject?

14:17 <haasn> jamrial: https://ffmpeg.org//pipermail/ffmpeg-devel/2024-December/336964.html

14:25 <fflogger> [editedticket] haasn: Ticket #11265 ([undetermined] Chroma Artifacts on the right edge of some odd width 4:2:0 videos in ffplay) updated https://trac.ffmpeg.org/ticket/11265#comment:5

15:54 <cone-546> ffmpeg James Almer master:d2c90abab74a: doc/filters: add an entry for vf_lcevc

15:54 <cone-546> ffmpeg James Almer master:cb27e478f721: doc/codecs: add missing enhancements value for export_side_data

15:58 <KillerWasp> hello. avcodec_receive_packet don't have sense, why they give EAGAIN() where they already receive all the datas what he needed? i'm only build a video file mp4 with only 1 stream of video, by each frame receive always the same image for all the 10 seconds that I'm try save, the video are 30fps, with 300f for save, but lost 41f in EAGAIN all in the begin and is reduced to 8.63s.

16:01 <KillerWasp> I understand that it may be handling everything in an internal buffer before producing output, but why does it have to lose all those frames even after calling av_write_trailer for closure?

16:01 <Compn> KillerWasp, you're going to have to post your code

16:01 <Compn> we dont know what you are doing

16:02 <nevcairiel> sounds like you are encoding a video with delay and didnt send a flush packet

16:02 <nevcairiel> but for help, you want #ffmpeg

16:03 <KillerWasp> nevcairiel: yes, have a delay of 1/30fps for each frame. And i don't know about of flush packet, you mean with av_packet_unref? yes, it's already have in the end of each frame.

16:05 <KillerWasp> in the end of each frame and even with EAGAIN, but still**

16:09 <JEEB> KillerWasp: I responded to you on #ffmpeg with https://lists.ffmpeg.org/doxygen/trunk/group__lavc__encdec.html , and please for API usage questions stay on the user channel :P

16:24 <KillerWasp> sorry, but is hard know it. Because handle the codes of ffmpeg is also development. 😛

16:27 <JEEB> the topic of the channel specifically notes about it

18:24 <JEEB> Traneptora: so you posted the PNG patch. nice

18:51 <Traneptora> I did, ye