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 7.0 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
<frankplow>
How do people test things which require seeking? For example if I think there might be a bug which is triggered when a sequence is played to a certain point, then seeked to another point.
Krowl has quit [Read error: Connection reset by peer]
<JEEB>
frankplow: would be an API test most likely
<JEEB>
I think we already have a seek test somewhere since I've seen that listed in fate
Krowl has joined #ffmpeg-devel
<Lynne>
frankplow: I just use mpv
<Lynne>
scrubbing is the best test
<frankplow>
JEEB, Lynne: Thanks
<JEEB>
frankplow: but yea, `git grep "fate-.*seek"` should give you hints
<JEEB>
such as the seek_utils thing
ngaullier has joined #ffmpeg-devel
ngaullier has quit [Ping timeout: 260 seconds]
Krowl has quit [Read error: Connection reset by peer]
tufei has quit [Remote host closed the connection]
tufei has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
cone-976 has joined #ffmpeg-devel
<cone-976>
ffmpeg Geoff Hill master:6f6bd10531e4: avcodec/ac3: Implement ac3_exponent_min for aarch64 NEON
<cone-976>
ffmpeg Geoff Hill master:b69486ea18ed: avcodec/ac3: Implement float_to_fixed24 for aarch64 NEON
<cone-976>
ffmpeg Geoff Hill master:69cb34f8859e: avcodec/ac3: Implement ac3_extract_exponents for aarch64 NEON
<cone-976>
ffmpeg Geoff Hill master:42e88f18f30c: avcodec/ac3: Implement sum_square_butterfly_int32 for aarch64 NEON
<cone-976>
ffmpeg Geoff Hill master:ee1bc723de4c: avcodec/ac3: Implement sum_square_butterfly_float for aarch64 NEON
<haasn>
elenril: I still haven't figured out a clean way to make ost->enc_ctx reflect ost->encoder_opts, maybe you have an idea for that?
<haasn>
It seems ost->encoder_opts gets updated in a lot of places, at different times
<haasn>
do you think it would be too hacky to just av_opt_set_dict() inside ost_add()?
<haasn>
I worry that it might set a precedent where components down the line are not sure if enc_ctx is up-to-date or not w.r.t encoder_opts
mkver has joined #ffmpeg-devel
<haasn>
I also worry that av_opt_set_dict() mutates encoder_opts which may end up invalidating whatever downstream clients need this dict for
tmatth has quit [Ping timeout: 256 seconds]
jamrial has joined #ffmpeg-devel
tmatth has joined #ffmpeg-devel
<frankplow>
2d79ae3f8a3306d24afe43ba505693a8dbefd21b cut the outstanding fuzz ASAN crashes I have for the VVC decoder by over 70% and UBSAN errors by over 65%
lexano has joined #ffmpeg-devel
<cone-976>
ffmpeg James Almer master:9eca5b8369b8: avcodec/vvc/ps: reset sps_id_used on flush
<cone-976>
ffmpeg James Almer master:1f049d5627ce: avcodec/vvc/dec: fix typo
<haasn>
just doing av_opt_set_dict() seems to cause warnings about unconsumed options later on
<elenril>
haasn: encoder_opts should not be an API
<elenril>
IMO applying it as soon as possible is the right thing to do
<haasn>
I had a patch somewhere where I tried to remove encoder_opts
<haasn>
but there was one case where we needed it and did not have an enc_ctx
<haasn>
maybe for streamcopy
<elenril>
possibly
<haasn>
let me try again
<elenril>
streamcopy uses it in a weird way
<haasn>
there is also stuff like check_avoptions() and friends
<haasn>
which want to look at all unconsumed options
<haasn>
(also not clear how to handle "threads" here, since we need to distinguish between thread_count=1 due to default value and thread_count=1 due to -threads 1)
Krowl has quit [Read error: Connection reset by peer]
<kurosu>
kierank: little relevance, and maybe there's a pattern (me), but in all companies I've worked for, I've never seen ffmpeg or its libraries being used by any product in the areas I had access to (can't say about subsidiaries or very different activities)
<kierank>
I think you are the exception that proves the rule ;)
<kurosu>
(mentioning it not to contradict, but to clarify my situation)
<kurosu>
yes, I think few companies want to invest enough or have the "brains" for that type of work
<jdek>
also something to note, ffmpeg is used in other fields too like forensics, not just direct multimedia stuff
<cone-976>
ffmpeg Andreas Rheinhardt master:908a73df31ff: avcodec/mpegvideo_enc: Don't update current_picture unnecessarily
<cone-976>
ffmpeg Andreas Rheinhardt master:66923165d5b7: avcodec/mpegutils: Move definitions to better places
<cone-976>
ffmpeg Andreas Rheinhardt master:ca7b2c393a7c: avcodec/mpeg12: Remove always-false check
<cone-976>
ffmpeg Andreas Rheinhardt master:a9709b3c8382: avcodec/mpegvideo_dec, mpeg12dec: Move MPEG1/2 code to mpeg12dec.c
<cone-976>
ffmpeg Andreas Rheinhardt master:1a5452655aee: avcodec/mpeg12dec: Remove redundant check
<cone-976>
ffmpeg Andreas Rheinhardt master:518f79f2a83d: avcodec/mpeg12dec: Don't pretend MPEG-1/2 to support alpha
<cone-976>
ffmpeg Andreas Rheinhardt master:f48313526b5c: avcodec/mpegvideo_dec: Don't emit non-keyframe warning for H.261
<cone-976>
ffmpeg Andreas Rheinhardt master:00e48e6da824: avcodec/mpeg12dec: Remove unnecessary FFCodec.close
<jamrial>
wbs: looks like the new ac3 checkasm tests fail on win64
tonio has joined #ffmpeg-devel
Workl has joined #ffmpeg-devel
rajivharlalka6 has joined #ffmpeg-devel
psykose_ has joined #ffmpeg-devel
<jamrial>
mkver, wbs: sign extend as usual, at least for the tests i fixed so far
rix_ has joined #ffmpeg-devel
Martchus has joined #ffmpeg-devel
<jamrial>
mkver, wbs: all of them, lol
Coinflipper has quit [Ping timeout: 268 seconds]
kepstin has quit [Ping timeout: 268 seconds]
hbbs has quit [Ping timeout: 268 seconds]
Fenrir has quit [Ping timeout: 268 seconds]
rix has quit [Ping timeout: 268 seconds]
j-b has quit [Ping timeout: 268 seconds]
ramiro has quit [Ping timeout: 268 seconds]
Krowl has quit [Ping timeout: 268 seconds]
Pasternyk has quit [Ping timeout: 268 seconds]
lexano has quit [Ping timeout: 268 seconds]
psykose has quit [Ping timeout: 268 seconds]
Martchus_ has quit [Ping timeout: 268 seconds]
rajivharlalka has quit [Ping timeout: 268 seconds]
xvaclav has quit [Ping timeout: 268 seconds]
rix_ is now known as rix
rajivharlalka6 is now known as rajivharlalka
psykose_ is now known as psykose
Coinflip- has joined #ffmpeg-devel
Coinflip- is now known as Coinflipper
j-b has joined #ffmpeg-devel
ramiro has joined #ffmpeg-devel
kepstin has joined #ffmpeg-devel
mkver has quit [Ping timeout: 260 seconds]
Fenrir has joined #ffmpeg-devel
cone-344 has joined #ffmpeg-devel
<cone-344>
ffmpeg James Almer master:b0ef9e35e809: x86/ac3dsp: clear the upper 32 bits for input arguments where needed
<BBB>
I'm confused, do mpeg-1 and mpeg-2 not support odd resolutions? I was pretty sure that they did
<BBB>
michaelni: ^
<BBB>
they do, right? or am I crazy?
xvaclav has joined #ffmpeg-devel
<nevcairiel>
you can argue about subsampling being invalid with odd resolutions, but in practice such files exist anyway, no clue what the spec has to say on the matter
hbbs has joined #ffmpeg-devel
hbbs has quit [Changing host]
hbbs has joined #ffmpeg-devel
<BBB>
why would it be invalid? you just pad 1 pixel
<JEEB>
I think the VPx decoders did that
<JEEB>
not sure if anything else did?
lexano has joined #ffmpeg-devel
Traneptora has joined #ffmpeg-devel
<BtbN>
The xstack heap overflow due to odd resoltutions is also still around...
<BtbN>
I don't understand the filter well enough to fix it, and nobody cares about my botched patch for it
* michaelni
is firmly against discriminating against odd resolutions
<BtbN>
well, the various xstack segfault with a heap overflow if you make them stack odd heights
mkver has joined #ffmpeg-devel
lexano_ has joined #ffmpeg-devel
lexano_ has quit [Remote host closed the connection]
<BtbN>
Cause it rounds up the subsampled height. so if you stack a 3x3 picture, the subsampled size is rounded up to 2x2, and it then tries to write two lines into the receiving frame at offset 2, which obviously does not go well
<michaelni>
i meant for mpeg* and anything that could support it
<michaelni>
of course segfaulting or any other undefined behavior must not occur
mkver has quit [Remote host closed the connection]
mkver has joined #ffmpeg-devel
<BtbN>
I've thought about it a bit, and I have legitimately no idea how to correctly vertically stack odd resolutioned subsamples frames
<BtbN>
*d
<BtbN>
For a while I thought such frames would be plain invalid, but they seem to exist
<wbs>
jamrial: sorry, and thanks for the prompt fix!
<michaelni>
BtbN, if you want to tile a bunch of frames with funny chroma subsampling, it can be complicated. The reason is the chroma sample locations differ for a frame starting at even pixel than a odd pixel. So you would have to resample the chroma of some images and also merge the border columns and rows
<michaelni>
that is for odd resolutions of course
<BtbN>
So the correct fix for now would probably just be to have the xstack family of filters fail with an error if they encounter odd resolutions with subsamples frames
<BtbN>
Wouldn't break any existing functionality, but replace a segfault with an error
<michaelni>
yes probably or print a warning and do some hack like copy one line / column less
<michaelni>
long term doing the correct resampling and merging should be done. Its not complex in terms of number of lines of code needed. Just maybe complex to understand
<BtbN>
I'm also not sure if the width is affected at all
Workl has quit [Read error: Connection reset by peer]
<BtbN>
The overwrite there would at worst be 1 pixel, if I understand it correctly, which is probably always available as padding
<JEEB>
wow, x265 actually decided to make a release
<JEEB>
thought they'd given up on doing such things
dykai has quit [Ping timeout: 240 seconds]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 256 seconds]
<Daemon404>
i saw it when reviewing a downstream set that includes a workaround ;)
<JEEB>
at least FATE passes, which means that we generally don't test intra refresh :D
Krowl has joined #ffmpeg-devel
<Daemon404>
FATE doesnt real test seeking either
<JEEB>
we do have the seeking API test thingy, although not sure what it does internally
<mkver>
JEEB: It only does demuxing.
Livio has joined #ffmpeg-devel
<JEEB>
ah
iive has joined #ffmpeg-devel
IndecisiveTurtle has joined #ffmpeg-devel
Livio has quit [Ping timeout: 264 seconds]
Warcop has joined #ffmpeg-devel
cone-344 has quit [Quit: transmission timeout]
Livio has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
<beastd>
BBB: I have just recently come across odd res mpeg1 video. Though somehow I remember when I started digging into multimedia around 2000, most tools would not allow me to create odd res videos.
<JEEB>
yea, most of my recollection was regarding On2 formats
<JEEB>
for some reason
<BBB>
Daemon404: presumably michaelni?
<rajivharlalka>
Curious to know if ffmpeg has some annual conference kinda setup? Have seen about some orgs on youtube but wasn’t able to find one for ffmpeg
<kierank>
Videolan Dev Days is the defacto ffmpeg conference
ccawley2011_ has joined #ffmpeg-devel
<rajivharlalka>
Quick followup: how is videolan related to ffmpeg? Not on a product usage basis but anything other than that. Saw it coming a few places in the docs too
<j-b>
kierank: Seoul! Seoul!
<kierank>
videolan is a user of ffmpeg
<kierank>
probably the most well known open source one
<beastd>
Though videolan dev days is not at all about videolan topics only. Kind of a broader multimedia meet up with presentations and participation from all kinds of multimedia developers.
<j-b>
It lacks trolls
mkver has quit [Ping timeout: 240 seconds]
Raz- has quit [Ping timeout: 240 seconds]
* kierank
there
Livio has quit [Ping timeout: 256 seconds]
<andrewrk>
j-b: Seoulがかっこいいだが日本はどう思う？
scat117 has quit [Ping timeout: 256 seconds]
kasper93 has quit [Ping timeout: 252 seconds]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
kasper93 has joined #ffmpeg-devel
IndecisiveTurtle has quit [Remote host closed the connection]
tonio has quit [Read error: Connection reset by peer]
scat117 has joined #ffmpeg-devel
philipl has quit [Quit: leaving]
Traneptora has quit [Quit: Quit]
philipl has joined #ffmpeg-devel
Traneptora has joined #ffmpeg-devel
elenril has quit [Ping timeout: 260 seconds]
elenril has joined #ffmpeg-devel
dykai has joined #ffmpeg-devel
staceee has quit [Ping timeout: 240 seconds]
staceee has joined #ffmpeg-devel
Traneptora has quit [Quit: Quit]
dykai1 has joined #ffmpeg-devel
dykai has quit [Ping timeout: 255 seconds]
dykai1 is now known as dykai
kurosu has quit [Quit: Connection closed for inactivity]