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 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
abdu1 has quit [Quit: Client closed]
gnafu has quit [Ping timeout: 252 seconds]
gnafu has joined #ffmpeg-devel
Marth64 has joined #ffmpeg-devel
cone-054 has joined #ffmpeg-devel
<cone-054>
ffmpeg Michael Niedermayer master:33679f5325f5: avcodec/vvc/refs: fix negative pps_scaling_win offsets
Traneptora has quit [Quit: Quit]
Traneptora has joined #ffmpeg-devel
Marth64 has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
Mirarora has quit [Quit: Mirarora encountered a fatal error and needs to close]
aaabbb has quit [Ping timeout: 252 seconds]
aaabbb has joined #ffmpeg-devel
Mirarora has joined #ffmpeg-devel
aaabbb has quit [Ping timeout: 252 seconds]
aaabbb has joined #ffmpeg-devel
MisterMinister has joined #ffmpeg-devel
mindfreeze has quit [Quit: Connection closed for inactivity]
thilo has quit [Ping timeout: 260 seconds]
thilo has joined #ffmpeg-devel
jarthur_ has joined #ffmpeg-devel
jarthur has quit [Ping timeout: 272 seconds]
aaabbb has quit [Changing host]
aaabbb has joined #ffmpeg-devel
MyNetAz has quit [Remote host closed the connection]
<fflogger>
[newticket] berarma: Ticket #11454 ([undetermined] Ogg/Theora: Bad granulepos when GOP higher than 64) created https://trac.ffmpeg.org/ticket/11454
<fflogger>
[newticket] ValeryKot: Ticket #11455 ([undetermined] Degradation in Closed Captions decoding) created https://trac.ffmpeg.org/ticket/11455
karuzela has joined #ffmpeg-devel
<karuzela>
hello i have some questions about license.
<karuzela>
can someone help me?
Guest46 has joined #ffmpeg-devel
^Neo has quit [Ping timeout: 248 seconds]
System_Error has quit [Remote host closed the connection]
<thardin>
gah I haven't been keeping up with the ML
<Lynne>
jamrial: fix sent, thanks
<DeityLeader>
WoW
Marth64 has joined #ffmpeg-devel
<DeityLeader>
what happened with .pls playlists support in ffmpeg/mpv ?
electron_rotosco has joined #ffmpeg-devel
<electron_rotosco>
is there a best practices way to submit a request to add a new demuxer to ffmpeg? would that be best done via trac or something else?
<jamrial>
electron_rotosco: trac ticket, yes
<electron_rotosco>
thanks!
electron_rotosco has left #ffmpeg-devel [#ffmpeg-devel]
<cone-679>
ffmpeg Tomas Härdin master:0202c7cc2ec5: lavf/mxfenc: Make write_desc return int
<cone-679>
ffmpeg Tomas Härdin master:9729444c7d55: lavf/mxfenc: Return AVERROR(EINVAL) in mxf_write_jpeg2000_subdesc() is pixfmt not set
<Marth64>
thinking about ffpreset, the age of presets in presets/, and that we still support avconv style presets too ... wondering if people actually use the feature?
<Marth64>
rhetorical question, just it seems outdated
<Marth64>
(I like the concept personally but it doesn't seem to intuitively right now)
<Marth64>
work intuitively*
<wbs>
Lynne: I'm looking at the checkasm test for aacencdsp; have a look at quantize_bands in libavcodec/aacencdsp.c - is it intentional that the FFMIN() clipping is only applied for positive values? (i.e. we can end up with values < -maxval) this can happen in the checkasm test (after fixing it a little), does that happen in the real encoder?
<Lynne>
I believe that is unintentional, yes
ccawley2011 has quit [Read error: Connection reset by peer]
<Lynne>
the quantization coding function seems to be happy enough to code without triggering decoder issues
ccawley2011 has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
<wbs>
ok, interesting, so there's potentially something to fix if we'd make it clip properly... but if I get the checkasm test fixed, we'd also need to sync a couple of asm implementations once/if we do that
thresh_ is now known as thresh
<wbs>
then secondly; is there someone who can read x86 simd assembly, who can help out? in trying to fix the aacencdsp checkasm test, I'm hitting differing outputs for quant_bands on x86; try running with https://github.com/mstorsjo/ffmpeg/commits/checkasm-aacenc, run e.g. "./tests/checkasm/checkasm --test=aacencdsp 0" and you'll have a bunch of differing values printed. can someone look at
<wbs>
libavcodec/x86/aacencdsp.asm and see what the function ...
thresh has quit [Changing host]
thresh has joined #ffmpeg-devel
<wbs>
... actually is doing; it doesn't seem to do the negating of negative values for is_signed==1
<Marth64>
xubuntu
<Marth64>
oops
<BBB>
mkver: I have a report from firefox that our vp9 decoder is having races since their last sync. I suspect this could be either because of general work on frame-threading, or it could be because of some of the vp9 specific patches. Both were done by you. Would you be interested in having a look? (I sadly don't have a lot of time to look into this.)
<mkver>
BBB: Yes.
<BBB>
can I cc your email on the relevant bug report in their bug tracker?
abdu has quit [Quit: Client closed]
<mkver>
yes. (But it would be better to just post the link here)
<mkver>
(Or is it private?)
<BBB>
it's a security bug so it's private, sadly
<BBB>
sorry
abdu has joined #ffmpeg-devel
abdu has quit [Client Quit]
abdu has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
<DeityLeader>
Lynne: does having variable frame size lossless audio codec have sense? having variable frame sizes/nb_samples is to auto-adjust to different auto-correlations - transients/noise/tones
<Lynne>
generally all lossy MDCT-based codecs I can think of use fixed sizes
<welder>
wbs: I hit the same error in aacencdsp with NEON assembly. It seems to me the culprit is that the C code first takes the minimum of two floats, casts it to int, and optionally negates this int, while assembly code first negates the float and then casts it to int.
<Lynne>
I can't think of one where that isn't the case
<Lynne>
oh, yeah, there it makes much more sense, you could choose which bands to predict or not, whether they would affect the current context
<DeityLeader>
Lynne: i do not plan to use bands/FFT just LPC with auto-correlations done via RDFT convolutions
<Lynne>
sure, though IMO an RDFT is sort of an FFT
<DeityLeader>
where autocorrelations are used to split audio to different segments depending on self-similarity
<wbs>
welder: it feels like a different issue; for the aarch64 NEON I did hit that for cases around INT_MIN, where negating the integer doesn't get it into a positive one. But for the x86 SIMD it fails to negate values like -1700
cone-679 has quit [Quit: transmission timeout]
Guest37 has joined #ffmpeg-devel
abdu has quit [Quit: Client closed]
abdu has joined #ffmpeg-devel
DeityLeader has quit [Quit: Client closed]
DeityLeader has joined #ffmpeg-devel
<DeityLeader>
pross: what is status of decoders?
Guest37 has quit [Quit: Client closed]
Guest23 has joined #ffmpeg-devel
abdu has quit [Quit: Client closed]
abdu has joined #ffmpeg-devel
Guest23 has quit [Quit: Client closed]
DeityLeader has quit [Quit: Client closed]
DeityLeader has joined #ffmpeg-devel
Guest3 has joined #ffmpeg-devel
aljazmc has joined #ffmpeg-devel
Coinflipper has quit [Quit: ]
Coinflipper has joined #ffmpeg-devel
Coinflipper has quit [Quit: ]
Coinflipper has joined #ffmpeg-devel
DeityLeader has quit [Quit: Client closed]
mkver has quit [Remote host closed the connection]
mkver has joined #ffmpeg-devel
aljazmc has quit [Quit: Leaving]
<BtbN>
What's happening today? Second issue independent people opened claiming "Thing XY isn't there.". I check, it's there. I tell them it's there. They go "Oh, yeah, it's there. *closes issue*"