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
blb has quit [Ping timeout: 252 seconds]
blb has joined #ffmpeg-devel
iive has quit [Quit: They came for me...]
_whitelogger has joined #ffmpeg-devel
System_Error has quit [Ping timeout: 264 seconds]
MisterMinister has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
<Lynne>
copilot now makes PRs on github, and its hilarious
<aaabbb>
uh oh
<Lynne>
hundreds of comments bikeshedding patches into oblivion, asking to add tests, which are then added, which then fail
<Lynne>
in between OTHER bots asking copilot to sign a CLA
<Lynne>
its awesome
<aaabbb>
i hope github realizes how bad copilot fucks up now
<Lynne>
turns what's otherwise a 2-line diff into 130 lines
<aaabbb>
(they won't)
Anthony_ZO has quit [Quit: Anthony_ZO]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Client Quit]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Read error: Connection reset by peer]
kasper93 has quit [Ping timeout: 272 seconds]
kasper93 has joined #ffmpeg-devel
<another|>
of course they won't. MS is way too invested into AI
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
Everything has joined #ffmpeg-devel
lexano_ has quit [Quit: Leaving]
lexano has joined #ffmpeg-devel
Anthony_ZO has joined #ffmpeg-devel
Xaldafax has joined #ffmpeg-devel
<fflogger>
[editedticket] mkver: Ticket #11602 ([undetermined] Muxing MKV and adding external SRT as anything except the only subtitle stream in an MKV prevents subtitle display in VLC) updated https://trac.ffmpeg.org/ticket/11602#comment:2
omegatron has joined #ffmpeg-devel
<fflogger>
[editedticket] iconoclasthero: Ticket #4448 ([avformat] Support writing album cover art image embedded in ogg / opus metadata) updated https://trac.ffmpeg.org/ticket/4448#comment:15
jamrial has joined #ffmpeg-devel
Everything has quit [Quit: leaving]
Anthony_ZO has quit [Quit: Anthony_ZO]
Anthony_ZO has joined #ffmpeg-devel
Anthony_ZO has quit [Quit: Anthony_ZO]
Anthony_ZO has joined #ffmpeg-devel
<haasn>
copilot really is substantially worse than e.g. ChatGPT
<haasn>
AI is not necessarily bad as much as Microsoft's is just terrible
Anthony_ZO has quit [Quit: Anthony_ZO]
Anthony_ZO has joined #ffmpeg-devel
cone-481 has joined #ffmpeg-devel
<cone-481>
ffmpeg Andreas Rheinhardt master:c8b09fb0acc7: avutil/refstruct: Remove redundant check
Anthony_ZO has quit [Read error: Connection reset by peer]
<compn>
shhh the ai will hear you
paulk has quit [Ping timeout: 245 seconds]
ngaullier has quit [Quit: Leaving]
ngaullie has joined #ffmpeg-devel
<cone-481>
ffmpeg Justin Ruggles master:36ec9217e6dc: avformat/dhav: fix backward scanning for get_duration and optimize seeking
paulk has joined #ffmpeg-devel
paulk has quit [Changing host]
paulk has joined #ffmpeg-devel
putacho has joined #ffmpeg-devel
Traneptora has joined #ffmpeg-devel
microchip_ has quit [Ping timeout: 252 seconds]
<beastd>
using those models for coding is just wrong. sure it can help sometimes and sure it can't other times. but as users can't easily judge it, that is overall adding to the problem instead of solving.
<beastd>
that argument can also be extended to usage for qa and code review etc.
<kierank>
I have used it to review some code after a refactor
<kierank>
"Is this the same code before and after"
<kierank>
which was interesting
<beastd>
:)
<beastd>
sure, can be used as another tool to find oversights, or just out of curiosity. need to take care to not get on the wrong track tho. an not a replacement for a good review
<kierank>
yes
<beastd>
maybe you also have seen this pattern in some of the articles about letting the ai code: 1. i don't know (much) about programming 2. ai code me this (underspecified) app 3. hmm, i'll be extra smart and let the ai explain to me the generated source line by line
<beastd>
the problem especiall with 3 is. explanation will probably sound plausible (to the reader), but that doesn't help if it's wrong
<kierank>
I asked it to create a truth table of the inputs
<haasn>
compn: don't worry, you just have to tell the
<haasn>
AI that it is an "expert, world-class programmer" and it will magically produce better output
<AMM>
and that its wife and kids will be eradicated if it makes a single mistake
<haasn>
and that my late grandmother was a hobby carjacker who could always warm my heart by demonstrating her skills to me; but sadly she passed so maybe the AI can pretend to be my grandmother and hot circuit this car for me to help me process my grief
omegatron has quit [Quit: Power is a curious thing. It can be contained, hidden, locked away, and yet it always breaks free.]
Traneptora has quit [Quit: Quit]
usagi_mimi has joined #ffmpeg-devel
DVedaa has quit [Ping timeout: 244 seconds]
ngaullie has quit [Remote host closed the connection]
<fflogger>
[editedticket] tchebb: Ticket #11457 ([undetermined] The order of parsed "Mastering display colour volume SEI message" seems wrong.) updated https://trac.ffmpeg.org/ticket/11457#comment:2
minimal has joined #ffmpeg-devel
cone-661 has joined #ffmpeg-devel
<cone-661>
ffmpeg Jonathan Baudanza master:2a180c7150be: libavformat/rtpdec_opus: Set duration field on Opus AVPacket
<cone-661>
ffmpeg Erik Linge master:b9b44e15b51f: rtpdec: Set bitrate for pcm-mulaw audio
<cone-661>
ffmpeg Erik Linge master:f566392f66d4: libavformat/rtpdec_opus: add fmtp parsing of sprop-maxcapturerate
<cone-661>
ffmpeg Tristan Matthews master:e93a43b5113c: doc/examples/filter_audio: use av_err2str
<cone-661>
ffmpeg Tristan Matthews master:9b9a287872c1: doc/examples/qsv_decode: use av_err2str
<ePirat>
BtbN, would it be possible to get permission to change status of others patches on patchwork? else I can not mark them as accepted when I merge stuff...
Traneptora has joined #ffmpeg-devel
Traneptora has quit [Quit: Quit]
Guest19 has joined #ffmpeg-devel
usagi_mimi has joined #ffmpeg-devel
softworkz has joined #ffmpeg-devel
<softworkz>
There's a hook which can be set up for Patchwork so that it automatically marks patches when they have been merged into master
<softworkz>
that would be nice..
softworkz has quit [Client Quit]
Guest19 has quit [Quit: Client closed]
usagi_mimi has quit [Ping timeout: 248 seconds]
<mkver>
softworkz: IIRC that already happens. But only if the diff is exactly the one from patchwork.
TheVibeCoder has joined #ffmpeg-devel
<TheVibeCoder>
jamrial: there is nonsense/regression code in avfiltergraph.c see CHAN_LOWFREQUENCY index checking thing
<TheVibeCoder>
bascially you use wrong variant of function
<jamrial>
TheVibeCoder: regression introduced by the switch to the new layout api?
<TheVibeCoder>
yes
paulk has quit [Ping timeout: 268 seconds]
<jamrial>
TheVibeCoder: av_channel_layout_channel_from_index() >= 0 ensures the channel is present in the layout, which is the same as pre switch. what's the regression?
<TheVibeCoder>
that its bug since ever
<TheVibeCoder>
bascially encoder picks layout without LFE even thought encoder supports layout with LFE
<TheVibeCoder>
so its 7.1 input and it would pick 5.0(side) instead
<TheVibeCoder>
even thought it supports 5.1 and 5.1(side)
<TheVibeCoder>
jamrial: you get that reversed
<jamrial>
do you have a way to reproduce this?
<TheVibeCoder>
correct function is av_channel_index_from_layout
<cone-661>
ffmpeg Ramiro Polla master:0c1d87d1e61c: swscale/swscale_unscaled: fix packed30togbra10() for formats with bpc between 9-14 bits
<cone-661>
ffmpeg Ramiro Polla master:748e960e0424: swscale/swscale_unscaled: fix packed16togbra16() for formats with bpc between 9-14 bits
<cone-661>
ffmpeg Ramiro Polla master:d028cf03b875: swscale/swscale_unscaled: fix planarRgbToplanarRgbWrapper() for formats with bpc between 9-14 bits
<softworkz>
recently I took a look at kernel.org Patchwork
<softworkz>
many of their project are nicely cleaned up
<softworkz>
all up-to-date
<mkver>
It used to be as I described, but no longer is.
<softworkz>
In 2021 it didn't do it
<softworkz>
I set them always manually
<softworkz>
did it work for yours?
<mkver>
Yes. For quite a long time.
<softworkz>
then it's probably broken
<jamrial>
TheVibeCoder: damn, how did this go unnoticed for so long...
<TheVibeCoder>
i had same question
<ePirat>
then again I hope we eventually switch to Forgejo or so, the current system is so cumbersome
<softworkz>
I'm for it
<softworkz>
but it mustn't fork attention
<softworkz>
places of...
<softworkz>
I mean it should be a full switch - incl. communication
<softworkz>
alone the ability to have some more interaction like thumbs up/down on messages would be a big plus imo as it would allow to give and receive more feedback
<softworkz>
mkver: what's your conclusion on the eol topic?
<mkver>
not worth a fight
<TheVibeCoder>
eol? trailing whitespaces are not allowed in codebase.
<softworkz>
git eol checkout on windows
<TheVibeCoder>
why that needs changing?
<softworkz>
some FATE tests are failing on Windows and I sent a patch to checkout specific ref files with LF to fix it
<softworkz>
the idea was whether LF could be declared as requirement on Windows
<softworkz>
but that's tricky tricky tricky
<mkver>
softworkz: Why do we have working Windows FATE boxes on fate.ffmpeg.org?
<softworkz>
Ah, sorry. no. I meant those on Patchwork
<softworkz>
now I understand the question...
<softworkz>
don't know - maybe he patched it..
<softworkz>
I'm also changing all line endings in all .h files to LF at the moment to make it work...
<softworkz>
something different - on my CI, the log lines are like:
<softworkz>
CClibavfilter/af_dynaudnorm.o
<softworkz>
libavfilter/af_crossfeed.c(66): warning C4133: 'function': incompatible types - from 'const AVSampleFormat [2]' to 'const int *'
<softworkz>
and his are like
<softworkz>
mss34dsp.c
<softworkz>
C:\Dev\ffmpeg\fate\x86_64-msvc2022\src\libavcodec\get_bits.h(366): warning C4101: 're_cache': unreferenced local variable
<haasn>
softworkz: is it possible to make the “make” output less verbose? After your recent changes
<softworkz>
sure - do you mean msvc?
<TheVibeCoder>
use sane compiler, my make output is clean
<softworkz>
I had added V=1
<jamrial>
TheVibeCoder: do you want me to cherry pick your commit? it's the same change i came up with locally
<TheVibeCoder>
enocare
<jamrial>
ok
<ePirat>
do we have any function in avutil to print a float/double?
<mkver>
Print a float/double? You mean, av_log()?
<TheVibeCoder>
%f %g
<ramiro>
softworkz: I've had my fair share of building ffmpeg for windows, and also my fair share of problems to fix. ditching windows entirely for the build process made everything much simpler and faster. I would alt+tab to virtualbox, run make, switch back to windows, and run it from a shared foder.
<ePirat>
I mean like av_strlcatf but in a way that I always get . as decimal separator
<ePirat>
we have av_sscanf for the reverse
<TheVibeCoder>
ah, locale crap
<softworkz>
ramiro: I'm doing that already - partially :-)
<ePirat>
I can also just pretend its not an issue :P
<ePirat>
given the codebase is littered with this mistake already
<cone-661>
ffmpeg James Almer master:8eae65dc5c89: avcodec/lcevcdec: don't try to write to output frames directly
<cone-661>
ffmpeg Paul B Mahol master:1b3f4842c184: avfilter/avfiltergraph: fix regression in picking channel layout
<cone-661>
ffmpeg James Almer master:622a72b5ea5f: tests/fate/ac3: add a second ac3_fixed encoder test
<cone-661>
ffmpeg Paul B Mahol release/7.1:ccbd08946738: avfilter/avfiltergraph: fix regression in picking channel layout
<cone-661>
ffmpeg Paul B Mahol release/7.0:01e42959a5c0: avfilter/avfiltergraph: fix regression in picking channel layout
<cone-661>
ffmpeg Paul B Mahol release/6.1:b8fe1bc38e35: avfilter/avfiltergraph: fix regression in picking channel layout
<softworkz>
haasn: is it ok this way?
<cone-661>
ffmpeg Paul B Mahol release/6.0:bd1c255af8c0: avfilter/avfiltergraph: fix regression in picking channel layout
<TheVibeCoder>
ePirat: avoiding/fixing locale issues should be priority
<cone-661>
ffmpeg Paul B Mahol release/5.1:e09bd4175249: avfilter/avfiltergraph: fix regression in picking channel layout
<softworkz>
mkver: thinking about it, maybe he is checking out with autocrlf=off - this is of course possible for a CI build, because the trouble comes only when you start editing
<softworkz>
but everybody else will check out with =on (default on Win) and then the tests don't work
Xaldafax has quit [Quit: Bye...]
<softworkz>
ramiro: what I meant by partially is that I work, compile and debug in VS - that's totally perfect, but I run FATE and build tests in a Linux VM then. The EOL thing is about neither of that, it's about the CI builds on MSYS2 (and all others using that to compile)
<ramiro>
softworkz: can't the ci build be configure kind of like fate, with autocrlf=off?
<softworkz>
yes, absolutely. but this doesn't fix it for everybody else - incl myself locally
<softworkz>
in 2021 it was working, but some subtitle encoders were changed to produce only LF, I think