05:47 <fflogger> [editedticket] RandomPerson: Ticket #11489 ([ffmpeg] MediaCodec code not working, likely outdated) updated https://trac.ffmpeg.org/ticket/11489#comment:1

07:52 <fflogger> [editedticket] JackLau: Ticket #10786 ([avformat] CODECS for HLS mux hard coded to 4) updated https://trac.ffmpeg.org/ticket/10786#comment:5

08:20 <cone-884> ffmpeg Nuo Mi master:dede00f003c5: lavc/vvcdec: fix undefined reference to ff_h2645_pixel_aspect

10:15 <fflogger> [editedticket] giannis-ff: Ticket #11426 ([avcodec] Memory leak in Intel QSV decoder) updated https://trac.ffmpeg.org/ticket/11426#comment:3

12:48 <cone-057> ffmpeg Frank Plowman master:600ad36949a7: lavc/vvc: Fix pps_single_slice_picture

12:49 <fflogger> [editedticket] MasterQuestionable: Ticket #11489 ([avcodec] MediaCodec code not working, likely outdated) updated https://trac.ffmpeg.org/ticket/11489#comment:2

13:32 <GuruLeader> michaelni: please disband TC and CC entities they are pointless

13:52 <fflogger> [editedticket] RandomPerson: Ticket #11489 ([avcodec] MediaCodec code not working, likely outdated) updated https://trac.ffmpeg.org/ticket/11489#comment:3

13:54 <fflogger> [editedticket] RandomPerson: Ticket #11489 ([avcodec] MediaCodec code not working, likely outdated) updated https://trac.ffmpeg.org/ticket/11489#comment:4

16:07 <cone-432> ffmpeg Michael Niedermayer release/7.1:68af2cc3feb8: Prepare for 7.1.1

16:07 <cone-432> ffmpeg softworkz release/7.1:a8a3fc84afa2: avformat/hls: Partially revert "reduce default max reload to 3"

16:36 <mkver> Gramner: Do you happen to know any trick to make compilers actually bake the offset into the relocation in https://godbolt.org/z/cTKvvodvT?

16:36 <mkver> (This affects VP9 dsp code.)

16:47 <Gramner> can't say I do. generally speaking if you rely on a compiler doing the right thing you're gonna have a bad time. if you use ptrdiff_t instead of int gcc manages to produce the following for the foo2 case: lea rax, [ff_filters_sse2+1920]; lea rdi, [rax+rdi-128] which I guess is the least terrible one

16:47 <Gramner> it almost got there

16:52 <mkver> Thanks for your time.

17:05 <BBB> mkver: is this to reduce the binary size of vp9dsp's c mc?

17:09 <mkver> BBB: GCC currently emits code that relies on there being a GOT entry for ff_filters_sse2 (like ff_foo3). Simply declaring the symbol as hidden makes GCC emit code like ff_foo: It has one instruction more and therefore causes an increase in .text (outstripping the savings from the one then unnecessary relocation). Modifying the macro to something would mitigate this, but changing code just because GCC doesn't like the actually superior form is bad.

17:13 <BBB> oh ok so it's something else, nevermind then

17:13 <BBB> at some point I'd like to do some work on reducing the binary size of vp9dsp's c mc code

17:14 <mkver> BBB: It is to avoid pessimizing code size in vp9dsp.

17:14 <BBB> ok, that's always good

17:24 <BtbN> Is it correct that the "host" library of a shared code file has it in both OBJS and STLIBOBJS?

17:24 <BtbN> That seems a bit redundant, OBJS should cover it, no?

17:25 <jamrial> example?

17:37 <BtbN> jamrial: https://code.ffmpeg.org/FFmpeg/FFmpeg/commit/7c338f470f869978efa3ec161c5f28fa83840c00 is what I'm looking at

17:37 <BtbN> after it, to this day, libavcodec has ffjni.o in both OBJS and STLIBOBJS, and I just emulated that in my patch that does something similar

17:38 <BtbN> but at least logically to me, just OBJS should be enough?

17:39 <jamrial> yeah, it's enabled with CONFIG_JNI, not a muxer or similar, so it seems pointless

17:40 <BtbN> OBJS should be SHLIBOBJS + STLIBOBJS

17:55 <cone-432> ffmpeg Timo Rothenpieler master:9d5d51bd129c: avutil/timecode: add ff_timecode_set_smpte

17:55 <cone-432> ffmpeg Timo Rothenpieler master:31ca6c1bfe5d: avcodec/utils: use new ff_timecode_set_smpte function

17:55 <cone-432> ffmpeg Timo Rothenpieler master:e567dca12f27: avcodec/nvenc: add time code writing for h264

17:55 <cone-432> ffmpeg Timo Rothenpieler master:5c5be37daff4: avcodec/nvenc: factor out mastering display data into its own function

21:17 <jamrial> BtbN: looks like mastering metadata as written by nvenc is correct now?

21:51 <BtbN> in the latest driver you mean?

21:51 <BtbN> Haven't re-checked