michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 7.1.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
<kasper93> lmoa
rvalue has quit [Remote host closed the connection]
iive has quit [Quit: They came for me...]
rvalue has joined #ffmpeg-devel
minimal has quit [Quit: Leaving]
_whitelogger has joined #ffmpeg-devel
cone-469 has quit [Quit: transmission timeout]
_whitelogger has joined #ffmpeg-devel
jamrial has quit []
indecisiveturtle has quit [Quit: indecisiveturtle]
TheVibeCoder has joined #ffmpeg-devel
<fflogger> [editedticket] v0lt: Ticket #8349 ([avcodec] Dolby AC-4 Support) updated https://trac.ffmpeg.org/ticket/8349#comment:99
<fflogger> [editedticket] MasterQuestionable: Ticket #11671 ([undetermined] enable TCP keepalive socket options) updated https://trac.ffmpeg.org/ticket/11671#comment:4
ngaullier has joined #ffmpeg-devel
zsoltiv__ has quit [Read error: Connection reset by peer]
zsoltiv_ has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
<fflogger> [editedticket] MasterQuestionable: Ticket #11668 ([ffmpeg] "-shortest" malfunction with audio much longer than video?) updated https://trac.ffmpeg.org/ticket/11668#comment:5
pelotron has quit [Ping timeout: 252 seconds]
pelotron has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 252 seconds]
quietvoid has quit [Read error: Connection reset by peer]
quietvoid has joined #ffmpeg-devel
<haasn> Oh no
System_Error has quit [Ping timeout: 244 seconds]
Kei_N has joined #ffmpeg-devel
Kei_N has quit [Client Quit]
Kei_N has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
jamrial has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
kurosu has joined #ffmpeg-devel
ccawley2011_ has joined #ffmpeg-devel
ccawley2011 has quit [Ping timeout: 252 seconds]
<fflogger> [editedticket] RavenWorks: Ticket #11668 ([ffmpeg] "-shortest" malfunction with audio much longer than video?) updated https://trac.ffmpeg.org/ticket/11668#comment:6
Traneptora has joined #ffmpeg-devel
Yalda[m] has joined #ffmpeg-devel
Yalda has quit [Ping timeout: 248 seconds]
<compn> BtbN, signed up
<TheVibeCoder> until when ffmpeg switches to fogejo 100%?
<Lynne> TheVibeCoder: have you figured out where the qscale/quantizer is read?
<Lynne> I can see where all the sizes for the components are parsed, but not where the qscale gets parsed
<TheVibeCoder> no idea
<TheVibeCoder> Lynne: ask kostya
<Lynne> I'll... look again
minimal has joined #ffmpeg-devel
<TheVibeCoder> until when ffmpeg switches to forgejo 100%?
<averne> Lynne: FYI I just completed initial support for my gsoc project: https://github.com/averne/FFmpeg/tree/vk-proresdec. This correctly decodes my test sample, though the iDCT part isn't bitexact compared to the software decoder.
<Lynne> averne: awesome
<Lynne> you should look into using the idct from the prores raw decoder I wrote, its super fast
cone-247 has joined #ffmpeg-devel
<cone-247> ffmpeg Niklas Haas master:8b647b3f8adb: avfilter/vf_colordetect: add x86 SIMD implementation
<cone-247> ffmpeg Niklas Haas master:545f721b443c: avfilter/vf_colordetect: add new color range detection filter
<cone-247> ffmpeg Niklas Haas master:f944a70fccde: tests/checkasm: add check for vf_colordetect
<averne> Yes, right now I'm just using an iDCT from the cuda samples repo. I saw yours on the ML and it looked a bit more efficient (fewer mults). I'd like to experiment with various stuff so I think I'll put together a small app that runs the accuracy test described in the SMPTE doc.
<averne> Also swscale on ffmpeg master refuses to convert yuv444p12 to yuv420p (or anything else I tried). I believe it landed a refactor recently, are there new requirements or something? Same command works fine on ffmpeg 7.1.1
<BtbN> compn: gave you permissions
<BtbN> compn: I just sent a test-mail to nut-devel (the only migrated list so far). Not sure if you have access to that in the old mailman interface, but could you double check that it didn't end up there? mailman3 is supposed to take priority.
<BtbN> you can also see it in the queue on Mailman3, maybe don't approve it :D
<jamrial> averne: probably unsupported color metadata
<averne> It does say "prim:reserved trc:reserved" in the error message. Should I provide this info somehow?
<JEEB> so not undefined but reserved... I recall some dumb software was writing those CICP values but now we find it on this channel, too :P
<jamrial> those are invalid values, so it's rejecting them
<jamrial> JEEB: yeah, recent rework made it actually look at color values, and it's been a bit of a meltdown since then :p
<JEEB> there's the setparams video filter or whatever it was
<JEEB> and then you can also set them on the input side on ffmpeg cli possibly (input side being "before `-i INPUT`")
<JEEB> but I think I previously used the filter
<JEEB> I do wonder what the original reasoning was to have those value ranges as reserved
<JEEB> because if that reasoning already went past like 10 years ago, then it might make sense to interpret those as whatever certain applications are using them as
<JEEB> since CICP has had nice additions at the end of the listing just fine the recent XYZ years
<TheVibeCoder> fix mpv
<TheVibeCoder> still doesnt support ambisonic layouts
<TheVibeCoder> what a utter waste
<jamrial> does anything other than vlc with libspatialaudio support it?
<TheVibeCoder> yes, librempeg
<TheVibeCoder> but YT practically delted all ambisonic webms
<Lynne> lol, really?
<Lynne> guess apple's totally-not-vr thing is even deader now
BradleyS has quit [Read error: Connection reset by peer]
BradleyS has joined #ffmpeg-devel
<TheVibeCoder> Lynne: its only dead for normie webs
<Lynne> pro-webs exist?
<TheVibeCoder> nope
<Lynne> I know blackmagic released a new camera that targets the apple vr thing, but I read the specs and laughed
<TheVibeCoder> but its not 1999 and internet speeds allow to have at least 4 channels in audio
<TheVibeCoder> but no, we still have stereo only
<TheVibeCoder> total stagnation
<Lynne> fixed f6.0 aperture and fixed-focus at 1.5m, literally uless
<JEEB> TheVibeCoder: and even browsers support multichannel audio with stuff like AAC
<TheVibeCoder> and OPUS
<JEEB> yea
<Lynne> you can throw 7.1 aac on twitch and it happily deals with it
System_Error has quit [Remote host closed the connection]
<Lynne> ambisonics were a neat idea and I wish they had taken off, but consumer CPUs were barely fast enough to decode mp3s 30 years ago
jarthur has joined #ffmpeg-devel
MisterMinister has joined #ffmpeg-devel
ngaullier has quit [Remote host closed the connection]
Kei_N_ has joined #ffmpeg-devel
Kei_N has quit [Read error: Connection reset by peer]
System_Error has joined #ffmpeg-devel
s55 has quit [Quit: Bye]
ramiro has quit [Ping timeout: 260 seconds]
ramiro has joined #ffmpeg-devel
Kei_N has joined #ffmpeg-devel
Kei_N_ has quit [Read error: Connection reset by peer]
DauntlessOne49 has quit [Ping timeout: 240 seconds]
DauntlessOne49 has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 260 seconds]
ccawley2011_ has quit [Ping timeout: 260 seconds]
ccawley2011_ has joined #ffmpeg-devel
Warcop has joined #ffmpeg-devel
pelotron has quit [Ping timeout: 240 seconds]
pelotron has joined #ffmpeg-devel
<compn> TheVibeCoder, 2 channels, 2 ears. makes sense /s
<compn> BtbN, i dont see anything in pipermail archives https://lists.ffmpeg.org/pipermail/nut-devel/
<BtbN> compn: sweet, that means it must be working
<TheVibeCoder> compn: ears and head moves
<compn> BtbN, i dont see your mail to nut-devel in mm3 tho
<compn> so that may be issue
<BtbN> Well, I see it
<BtbN> Where are you looking? :D
<compn> oh held messages i see
<BtbN> I find the concept of unsubscription requests a bit strange
<compn> dont know if unsubscribe was ever standardized in rfc
<BtbN> I mean more that you have to/can request it, instead of just unsubbing yourself
<compn> my isp is still marking a lot of ffmpeg-devel as spam :\
<BtbN> Not much to be done about that, the concept of a mailing list is apparently too much for the "Modern Web"
<compn> TheVibeCoder, when looking at a screen on the wall? where is your ear moving lol
<TheVibeCoder> compn: search google
<TheVibeCoder> human move head all the time
<compn> click the motorcycle to continue your search of google
<TheVibeCoder> and thus true 3d hearing is possible
<TheVibeCoder> compn: click on QT test
<compn> click all squares with stairs in them.
<TheVibeCoder> compn: click red-like icon in top right corrent of currently active window
<TheVibeCoder> *corner
s55 has joined #ffmpeg-devel
s55 has quit [Client Quit]
s55 has joined #ffmpeg-devel
s55 has quit [Client Quit]
<kasper93> > detect_range_8_avx2 (illegal instruction)
<kasper93> :(
s55 has joined #ffmpeg-devel
<kierank> kasper93: which instruction
DauntlessOne49 has quit [Ping timeout: 240 seconds]
DauntlessOne49 has joined #ffmpeg-devel
<kierank> vpbroadcastb
cone-247 has quit [Quit: transmission timeout]
s55 has quit [Read error: Connection reset by peer]
<TheVibeCoder> avx != avx2 != avx512 != your old cpu
s55 has joined #ffmpeg-devel
zsoltiv_ has quit [Ping timeout: 240 seconds]
zsoltiv_ has joined #ffmpeg-devel
DauntlessOne49 has quit [Ping timeout: 240 seconds]
s55 has quit [Quit: Bye]
s55 has joined #ffmpeg-devel
cone-194 has joined #ffmpeg-devel
<cone-194> ffmpeg James Almer master:14f44783549b: avfilter/x86/vf_colordetect: fix use of AVX512 instruction in AVX2 function on non Unix64 targets
<jamrial> kasper93, kierank: fixed
kurosu has quit [Quit: Connection closed for inactivity]
<kierank> cool
<fflogger> [editedticket] guest: Ticket #11671 ([undetermined] enable TCP keepalive socket options) updated https://trac.ffmpeg.org/ticket/11671#comment:5
<TheVibeCoder> :(
<TheVibeCoder> do not fix bugs in ffmpeg ever
<jamrial> the more broken the better, right?
<TheVibeCoder> yes
<cone-194> ffmpeg James Almer master:fdca209f1f10: avfilter/x86/vf_colordetect: don't use rax to return a 32bit integer
<cone-194> ffmpeg James Almer master:70fc4e590980: avfilter/x86/vf_colordetect_init: don't enable ASM functions on targets where it's known they will be slower
<kasper93> thanks
<kasper93> I'm bombarded with messages from $dayjob, cant even read irc in time ;p
kurosu has joined #ffmpeg-devel
<fflogger> [editedticket] Cigaes: Ticket #11671 ([undetermined] enable TCP keepalive socket options) updated https://trac.ffmpeg.org/ticket/11671#comment:6
<cone-194> ffmpeg James Almer master:550ec9b7e698: avfilter/version: bump version after vf_colordetect addition
<cone-194> ffmpeg James Almer master:b9287fde1096: Changelog: add entry for vf_colorfilter addition
<kierank> jamrial: I think should be AVX512ICL, no?
<kierank> We shouldn't downclock on older CPUs
<jamrial> kierank: true, it's a BW function
<jamrial> so not available on base avx512
<kierank> At least in my code I made avx512 not use zmm
<kierank> And avx512icl use zmm
<jamrial> sure, but vpbroadcast{b,w} from xmm or memory is an AVX512BW instruction
s55 has quit [Quit: Bye]
<jamrial> so it needs to be under our icl flag
s55 has joined #ffmpeg-devel
<kierank> (irrespective of the new features)
<kierank> Yes
<cone-194> ffmpeg James Almer master:c62813a057ed: avfilter/x86/vf_colordetect: make the AVX512 functions run only on ICL targets or newer
s55 has quit [Read error: Connection reset by peer]
s55 has joined #ffmpeg-devel
s55 has quit [Quit: Bye]
s55 has joined #ffmpeg-devel
s55 has quit [Quit: Bye]
<cone-194> ffmpeg James Almer master:a01dc3aa272b: avfilter/x86/vf_colordetect: add missing preprocessor checks
DodoGTA has quit [Ping timeout: 260 seconds]
s55 has joined #ffmpeg-devel
bilboed has quit [Quit: Ping timeout (120 seconds)]
bilboed has joined #ffmpeg-devel
s55 has quit [Quit: Bye]
s55 has joined #ffmpeg-devel
<kierank> jamrial: thx
mkver has quit [Ping timeout: 260 seconds]
mkver has joined #ffmpeg-devel
s55 has quit [Read error: Connection reset by peer]
s55 has joined #ffmpeg-devel
s55 has quit [Read error: Connection reset by peer]
iive has joined #ffmpeg-devel
mkver has quit [Ping timeout: 265 seconds]
s55 has joined #ffmpeg-devel
<fflogger> [editedticket] RavenWorks: Ticket #11668 ([ffmpeg] "-shortest" malfunction with audio much longer than video?) updated https://trac.ffmpeg.org/ticket/11668#comment:7
<kasper93> BtbN: for CI it would be better to have in-tree scripts to handle that and test.yml only dispatch things with correct parameters, incidencially, it could just call fate.sh
<BtbN> Well, the idea is to expand it to enable external dependencies that make sense
<BtbN> Just calling fate.sh would barr us from ever doing that.
<BtbN> fate.sh would also try to upload the results somewhere?
<kasper93> why?
<BtbN> Cause that's what it does
<kasper93> I was asking about external deps
<kasper93> what's the problem with fate.sh for external deps?
<BtbN> You want to start maintaining multiple configure invocations in fate.sh?
<BtbN> I really don't see the point of calling fate.sh from the workflow
<kasper93> you will start mainining this the same way inside the yml
<BtbN> Yes, where it makes sense cause it corresponds to the image the workflow uses or which depds it installs
indecisiveturtle has joined #ffmpeg-devel
<BtbN> Calling fate.sh would generally just obfuscate the CI job output, since you can't easily see anymore which stage fails
<BtbN> the script really doesn't do any rocket science either, it calls configure, make and make fate
<BtbN> And also tries to update the git repo, which would likely clash with the checkout action
<BtbN> Cause on CI you want to test the specific commit the CI was tasked to test, not HEAD of some branch
<BtbN> all in all, it really does not seem like a good fit
<kasper93> in the current form for sure not
<kasper93> but it is not set in stone
<kasper93> and it can be another script
<kasper93> but advantage of having script is that you can easily reporoduce the build locally
<kasper93> without unwraping all the things that out of tree config has done
<kasper93> but it's fine, it's mostly configure call that changes anyway
<BtbN> You can run the workflow locally as well
<kasper93> anyone knows why av_export_avutil exists?
<BtbN> It's more likely that there'll be a new shell script if the shellcode inside of the workflow grows too much
<BtbN> av_export_avutil is used exactly to dllimport avpriv_cga_font and avpriv_vga16_font in not-avutil
<kasper93> can't dllimport when linking statically
<BtbN> That's why it's only set to something when not linking statically
jarthur has quit [Quit: jarthur]
<kasper93> then why it's used when linking statically?
<kasper93> NOTE: a relevant symbol 'avpriv_cga_font' is available in libavutil/libavutil.a but cannot be used because it is not an import library.
<BtbN> What do you mean?
<BtbN> av_export_avutil just defines to nothing
<BtbN> in that case
<kasper93> clearly it does not
<BtbN> if CONFIG_SHARED is not 1, it'll always define to nothing
<BtbN> is it maybe missing the new second config.h include, and thus does not see CONFIG_SHARED?
ccawley2011_ has quit [Read error: Connection reset by peer]
<BtbN> config_components.h, not sure what exactly goes in there
<BtbN> nah, CONFIG_SHARED is in config.h
<kasper93> something borked
<kasper93> probably on my end
<BtbN> mixing up build artifacts between different builds?
TheVibeCoder has quit [Ping timeout: 244 seconds]
rvalue has quit [Remote host closed the connection]
<kasper93> I think it actually never worked on Windows shared build
<kasper93> checkasm is forcefully linked statically
<kasper93> but on windows shared and static doesn't got togheter in one build
<kasper93> because if you inject somewhere dllimport it won't take this symbol from static one
<BtbN> if that bypasses the normal config.h, yeah
<BtbN> it'd have to set its own special flag, and the headers need to check that as well
<kasper93> but if you change headers need to rebuild all object files, so it is whole another build
<BtbN> Yeah, I guess a shared checkasm build on Windows is simply not possible
<kasper93> I think there is at least one build on fate that works, but this is lld-link erroring, probably link.exe just eats static symbol even if dllimport, without complaining much.
indecisiveturtle has quit [Ping timeout: 240 seconds]
kurosu has quit [Quit: Connection closed for inactivity]
<kasper93> BtbN: where are ci runners hosted? It's the same machine as forgejo instance?
<BtbN> separate instances
<kasper93> I see, thanks
<BtbN> On a Ubuntu 24.04 instance
<kasper93> hmmmmm
<kasper93> I'm not confused, avpriv_cga_font seems to be regression after adding vf_colorfilter to checkasm
<kasper93> or colordetect
<cone-194> ffmpeg Jacob Lifshay master:157d3b007e93: lavu: add av_scanf_format() and use it on av_sscanf()
<cone-194> ffmpeg Bernat Arlandis master:3cd4fa9b895a: libtheoraenc: Add encoding speed level option
<kasper93> maybe we actually were missing config.h and it was working by accident before,strange
<BtbN> I'd expect compilation to just hard fail then
<BtbN> cause it includes it unconditionally after all
<BtbN> and if it's empty, it'd trip up cause trying to use an undefined macro
<kasper93> that's why it's weird
<kasper93> I stopped carying about it and now I want to know again... because it looks like real issue now
DodoGTA has joined #ffmpeg-devel