<wellsakus>
If so, I would appreciate your comment
<wellsakus>
I would like to move forward with the EVC codec topic
<haasn>
agh, I forgot there was a commit that was unfinished :( going through all AVCodecs and figuring out which ones accept limited range only
<j-b>
Could we finish EVC for 7 ? :D
<Dariusz>
It will be good to finish all EVC related issues before 7.0 relase. We can implement all required things but we need review :)
<jamrial>
yes, i'll look at it this week
<j-b>
not that I care about EVC
<j-b>
but I care a bit.
<wellsakus>
great thx
<Dariusz>
parts of EVC is already implemented so it will be good to finsh it
<j-b>
exactly
<j-b>
then VVC
<j-b>
then lc-evc
<j-b>
then we kill MPEG for having:
<j-b>
- HEVC
<j-b>
- EVC
<j-b>
- VVC
<j-b>
- LC-EVC
<j-b>
worst naming ever
<jamrial>
also AVC
<haasn>
YUVJ removal RFC on ML
<j-b>
jamrial: it has an A, at least.
<jamrial>
haasn: nice, christmas came early :p
<j-b>
haasn: NIIIIICE
<haasn>
probably it will be until christmas for it to pass through review
<j-b>
that's OK
<haasn>
oh derp, I probably don't need to add the ff_ API changes to APIchangelog
<jamrial>
nope
<durandal_1707>
this does not work correctly at all
<APic>
Thanks for the Highlight ♥
quietvoid has quit [Ping timeout: 240 seconds]
quietvoid has joined #ffmpeg-devel
<haasn>
durandal_1707: can you present test case where it does not work as expected?
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
noonien8 has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
<durandal_1707>
haasn: talk with NG
MrZeus has joined #ffmpeg-devel
quietvoid has quit [Ping timeout: 258 seconds]
Krowl has joined #ffmpeg-devel
quietvoid has joined #ffmpeg-devel
<durandal_1707>
AV1EVC
<durandal_1707>
who cares about FFmpeg - its no longer relevant at all.
<kierank>
durandal_1707: do not talk down project
<kierank>
just because you are not community leader
<durandal_1707>
kierank: get a grip and fork
<aaabbb>
is this some fork drama? i thought that drama was long gone
<JEEB>
it's paul being paul
<durandal_1707>
nobody using FFmpeg any more, its being deprecated.
<durandal_1707>
JEEB never was true to self.
AbleBacon has joined #ffmpeg-devel
cone-162 has quit [Quit: transmission timeout]
<kierank>
durandal_1707: use gstreamer
Krowl has quit [Read error: Connection reset by peer]
* gnafu
wonders if durandal_1707 should be less ffmpeg-deprecating and more self-deprecating.
quietvoid has quit [Ping timeout: 255 seconds]
quietvoid has joined #ffmpeg-devel
psykose has quit [Remote host closed the connection]
philipl has quit [Quit: leaving]
jarthur has joined #ffmpeg-devel
psykose has joined #ffmpeg-devel
Krowl has joined #ffmpeg-devel
<durandal_1707>
gnafu is typical russina censor
BetweenUs has joined #ffmpeg-devel
<gnafu>
XD
quietvoid has quit [Ping timeout: 240 seconds]
quietvoid has joined #ffmpeg-devel
BetweenUs has quit [Quit: Leaving]
<courmisch>
ah so now I have to pay to use Zuckerbook. How about no
<j-b>
Don't :D
<JEEB>
-_- is there something that can easily show available videotoolbox decoders in the system?
<JEEB>
as in, including software ones
<courmisch>
yes. Don't have macOS, so you can get the list with the "true" command
<JEEB>
I thought of hacking on the AV1 videotoolbox decoder, but I see that I can't even invoke the VP9 decoder (even if I use the code path to allow for swdec like with HEVC).
<elenril>
JEEB: reminder that if you don't show any interest in keeping -fix_sub_duration_heartbeat working, then I won't either
<courmisch>
my TDD sensor is sounding the alarm
<JEEB>
there's a FATE test for it
<JEEB>
of course if he just moves things to threads and it breaks due to different timings with non-live
<JEEB>
elenril: anyways if there's something needed to be poked, consider me poked about that. I just saw the patch set and there were fix_sub_duration things so I thought "oh OK, so he figured something out for it"
<durandal_1707>
ah so now I have to pay to use FFmpegbook
<courmisch>
do you?
<courmisch>
planet.videolan.org is free to use
<courmisch>
of course, it's not seen a post in 6 months but
<JEEB>
elenril: I can attempt to take a look into it. the option is definitely useful with live subtitle input without defined subtitle durations
<durandal_1707>
exactly, its dead
<JEEB>
if you are doing live output, that is
<JEEB>
with file to file the latency of subtitle output isn't a thing one cares about
<elenril>
then I'd appreciate suggestions on how is it supposed to be changed
<JEEB>
sure, I'll try to poke at it
<courmisch>
durandal_1707: well, FFmpeg should switch to Rust, so the developers would feel compeled to spend 80% blogging about Rust and 20% actual coding
<JEEB>
elenril: also for the record funny enough if your input is that much further than output that you actually get your full durations calculated, it's actually good ^^;
<JEEB>
but of course it's not consistent
<JEEB>
also I think my initial attempt was to handle this on the muxer layer and skip fix_sub_duration, but then I found out that "undefined duration" and "zero duration" cannot be distinguished
<JEEB>
so then it ended up on ffmpeg.c layer :P
<JEEB>
elenril: anyways thanks for notifying that you just rm'd the feature instead of figuring it out
<JEEB>
since that was definitely not clear when I quickly saw your patch thread in my email client
<elenril>
:/ it says so very plainly
<elenril>
I also told you at VDD
<JEEB>
that I funny enough don't recall :D
<JEEB>
I recall you noting it months ago
<JEEB>
elenril: I did not have the mental time to look in detail, I just saw the string "fix_sub_duration" and assumed it was handled
<durandal_1707>
FUCK NG
<JEEB>
like, the ML has plenty of random stuff I would just get stress from if I read it
<elenril>
durandal_1707: be civil
<JEEB>
and I'm just trying to restart my patch set activity
<JEEB>
but in any case, I've been informed and I'll have to figure out what to do within TIME_RANGE
<JEEB>
elenril: just out of interest, did it break completely or it's just depending on the input being processed so it's no longer repeatable?
<elenril>
it's nondeterministic
<JEEB>
which is the latter
<JEEB>
OK
<JEEB>
that at least gives me hope that it compiles and I don't have to start with your branch and a trash fire
<JEEB>
also I'm not married to that specific solution, but due to the limitations in muxers wrt unknown and zero durations it doesn't seem feasible to handle it in the muxer and skip fix_sub_duration
<JEEB>
not sure where else to stick a "the output is going, so if you have any subtitles being calculated, give the current state kthx"
<JEEB>
also technically it should always be input in front of the rest of the flow, so if your output bits are slower it just means that you are getting more subtitles figured out. which while nondeterministic is useful.
<JEEB>
unless there's some other issue
<elenril>
well, the patch linked above removes it
<JEEB>
I'm not sure how I'm supposed to interpret that comment
<elenril>
the entire issue is that you need a deterministic algorithm for the option
<elenril>
because the muxer or encoder can be arbitrarily lagging behind the decoder
<JEEB>
yes, but actually lagging is generally OK. you have gained more information on the subtitle.
<elenril>
it's not ok if you want it to be deterministic
<JEEB>
sure
<elenril>
which you kinda do for testing
<elenril>
but then maybe you don't
<elenril>
you can adapt the tests and argue that it's fine that way
<elenril>
but there should be some kind of an argument behind that
<JEEB>
don't get me wrong, I'm not saying I don't want it to be deterministic.
<JEEB>
it's just that output being behind the input actually gives you the possibility of being more exact :D
<JEEB>
and as I said, if the functionality can be moved somewhere else than fix_sub_duration / ffmpeg I'm also open for that, but as I said > unknown|zero duration in muxers
<JEEB>
since the stuff I'm concerned with would mostly have unknown duration
<JEEB>
so if I can just pass that on in a manner that the muxer would understand that would be feasible, too
<JEEB>
or then some third place for such logic
<JEEB>
ideas are welcome :P
<JEEB>
> Bear requires openssl
<JEEB>
wtf
<JEEB>
or at least it attempts to bring in some BearSource thing
* JEEB
just wants to generate a compile commands file prkl
<thilo>
jdek: do you still have the list of the GA members your script produced back in 2020?
<thilo>
elenril: do you remember when we add the "voted in for 2 years" rule for extra GA members? Did we have that in 2020 or is it a more recent consensus?
<elenril>
I don't think anybody touched that text since it was added until now
<JEEB>
*/33
<JEEB>
whoops
<ePirat>
JEEB, Bear?
<JEEB>
ePirat: a thing to run the compilation under so you get a clang compile command file
<thilo>
jdek: that is 50 people.... IIRC there was an issue, either in your script back then or with mailmap, that we had more than one output for s.o. with more mail addresses... maybe that's 50 in your list?
<jdek>
I mean yes people werent using mailmap properly at the time so there was no way to resolve mail addresses to the same person
<thilo>
jdek: aah... atomnuker was still in that list
<thilo>
jdek: so minus atomnuker thats 49 :)
philipl has joined #ffmpeg-devel
<jdek>
this list shouldnt be used
<jdek>
didnt we vote on updating it after the vote
<thilo>
however, there seems to be two mails for Nicolas as well...
<thilo>
that was the update exception from the current vote. I try to find the list of voters from 2020
<ePirat>
thilo, has CIVS no log of this?
<thilo>
unfortunately not back then. I patched it already, the next vote will log all authorized voters.
<ePirat>
thilo, where can I find the source code for CIVS btw?
<thilo>
google cornell.edu and CIVS... a link should also be on whatever page you look at for any vote we had
<ePirat>
thx, found it.
<JEEB>
OK, finally remembered to do the most obvious VT test
<JEEB>
to attempt to open a file in QT
<JEEB>
and AV1 doesn't work, so 13.6.1 doesn't even have swdec :)
<thilo>
ah no, actually not two instances of Nicolas in there, mixed him together with Clement.... so I guess that list of 49 people is as close to the thruth we can get
Krowl has quit [Read error: Connection reset by peer]
<courmisch>
the AC3 checkasm support has not appeared. Leadership is clearly at fault
<ePirat>
JEEB, thats not always an indicator VT cant do it though
b50d has joined #ffmpeg-devel
<JEEB>
ePirat: true. but VP9 or AV1 can't init whether I request hwdec or not on this 13.6.1 :D
<JEEB>
and I can't get a list of decoders, unlike encoders
<ePirat>
JEEB, which hardware do you have?
<JEEB>
2015 macbook
<JEEB>
pro
<JEEB>
:D
<JEEB>
I don't expect hwdec but I did expect swdec
<JEEB>
if I get my 14.x install fixed (14.1 update borked it), then I should be able to test with it too
<ePirat>
Ah yeah I dont think neither VP9 nor AV1 work have software decoders
<ePirat>
*-work
<JEEB>
but there were dav1d symbols found? :D
<ePirat>
for avif afaik
<JEEB>
hmm
<JEEB>
if so, that sucks
<ePirat>
maybe Gramner or BBB know more though
<JEEB>
since it means that you can't write code until you have hwdec capable hardware
<JEEB>
┐(´д｀)┌
<JEEB>
or well, you can write code
<JEEB>
but you cannot validate it at all
<ePirat>
good luck writing code without testing for VT :D
<JEEB>
yea
<ePirat>
JEEB, we nearly got VT extensions this year :P
<durandal_1707>
BAN MAC users!
<ePirat>
JEEB, but eventually the docs vanished again and no one mentioned anything about it anymore :D
<BBB>
?
<BBB>
vt=video toolbox on mac?
<JEEB>
ye
<BBB>
there's no av1 sw decoder
<BBB>
they explicitly mentioned that
<another|>
JEEB: just because Apple has dav1d for avif, doesn't mean they wire it up to do av1
<BBB>
"I’ve got nothing to announce on that score regarding AV1, though. iOS 17 will support software AV1 decode, but only for still images (AVIF)."
<BBB>
from someone@apple.com
<JEEB>
another|: I was not in the know it was just for AVIF. I only knew of the "ooh they have dav1d symbols there"
<BBB>
vp9 I don't know, I thought they included sw decoding but I'm not sure
<BBB>
yes dav1d is avif only on mac in VT/safari
<durandal_1707>
Echo-chamber CC and TC committe
<BBB>
and yes that sucks
<JEEB>
BBB: well VP9 would have just been a verification that the current code that's in FFmpeg works, and that one could start extending it
<JEEB>
but if there is no swdec you need M3 macs to verify it
<JEEB>
so eh
<BBB>
jb jeeb needs a m3 can you give him moneyz?
<BBB>
michaelni: same question for spi
<durandal_1707>
yes, j-b give moneys to JEEB !
<BBB>
please give jeeb a m3 for science etc
<BBB>
also I need a new iphone 15 pro
<BBB>
pleaz?
<durandal_1707>
sure
<JEEB>
xD
<BBB>
also for science
<durandal_1707>
FFmpeg "science"
<courmisch>
NG did say that FFmpeg should go back to doing experimentaton
<courmisch>
without the shackles of business
<durandal_1707>
kierank approves above message
<courmisch>
but
<JEEB>
I actually thought whether I should in a year+ buy 15 pro or something similar, since being able to video capture onto a USB-C SSD is <3
<courmisch>
I do not condone giving monies to the world's richest corporation
<courmisch>
even if I probably indirectly old fractions of Apple stock
<courmisch>
hold*
<kierank>
durandal_1707: agreed
<gnafu>
JEEB: You could just get a Sigma fp ;-).
<gnafu>
(For recording video to an SSD.)
Krowl has joined #ffmpeg-devel
Krowl has quit [Client Quit]
Krowl has joined #ffmpeg-devel
rvalue has quit [Read error: Connection reset by peer]
rvalue has joined #ffmpeg-devel
whatever_nicknam has joined #ffmpeg-devel
b50d has quit [Remote host closed the connection]
<durandal_1707>
FFmpeg is closing doors.
quietvoid has quit [Ping timeout: 258 seconds]
<j-b>
BBB: no.
<j-b>
BBB: he should ask SPI.
<BBB>
?
<j-b>
I don't see why I should be the only one sponsoring FFmpeg devs.
<j-b>
and being shit on.
<BBB>
I didn't shit on you
<j-b>
BBB: 20:07 <@BBB> jb jeeb needs a m3 can you give him moneyz?
<BBB>
I asked nicely
<j-b>
BBB: of course not, you are one of the nicests :)
<BBB>
I also asked for a m15 and was denied, I'll ask SPI next
<BBB>
not m15. I mean iphone 15 pro
<BBB>
crap
<durandal_1707>
Lynne: how much fast one can get when doing 3x3 convolution? By doing different kernel so scalar product can be used.
<BBB>
youknowhwatImean
<j-b>
no I don't
<BBB>
michaelni: can I get a iphone 15 pro for science and av1 hw testing please? (from spi)
<j-b>
I now think you are an Apple Corporate Shill
<j-b>
Who has access to prototypes
<j-b>
and is here to make FFmpeg slower and dav1d slower outside of Apple platforms.
<durandal_1707>
BAN MAC users and wanna be MAC users!!!
<BBB>
but then why am I x86 optimizing?
<BBB>
makes no sense if apple is arm centered
<j-b>
fake news
<j-b>
prove that you are not
<j-b>
Prove the non-existence of something that does not exist
<durandal_1707>
BBB: what was your last commit to FFmpeg repo?
<BBB>
17 years ago, the matroska demuxer
<durandal_1707>
j-b: give him what he wants!
<BBB>
Date: Mon Apr 17 10:52:10 2023 -0400
<BBB>
yuv4mpegenc: add bitdepth multiplier after rounding width.
<BBB>
my last commit, I think
<courmisch>
17 years ago of what astral body?
<courmisch>
that's not even 17 moons ago
<Lynne>
durandal_1707: simd-wise, you'll have a horizontal add, which is never fast, the rest of the multiplies are all fine
<BBB>
Date: Thu Mar 11 16:26:42 2004 +0000
<BBB>
matroska demuxer by ("Ronald S. Bultje" <R dot S dot Bultje at students dot uu dot nl>)
<BBB>
that's my first commit
quietvoid has joined #ffmpeg-devel
<BBB>
19.5 years ago... I feel so old now
lemourin5 has joined #ffmpeg-devel
lemourin is now known as Guest2984
Guest2984 has quit [Killed (copper.libera.chat (Nickname regained by services))]
<BBB>
j-b: true. I can't disprove it
<durandal_1707>
prove FFmpeg exist!
<BBB>
it's not a legal entity. it's just a thing in the mind of its users
<BBB>
ffmpeg can't vote or make political donations in the US
<BBB>
thus, it does not exist
<j-b>
durandal_1707: fair enough: the day we meet in real life in HRAVA
<durandal_1707>
big code censor
<durandal_1707>
and fflabs exist, its legal entity.
<another|>
´´´´´´´´´´´´ßßßßßßßß099üääüäü+
<another|>
oops. sorry
<durandal_1707>
what happened to you?
<another|>
this was definitely not a covert message
<durandal_1707>
it IS covert message.
<j-b>
CLEARLY
<j-b>
another is suspicious
<another|>
just cleaning my keyboard
<durandal_1707>
all clear!
<durandal_1707>
j-b: we will never meet in real life!
<durandal_1707>
kierank: same!
<kierank>
Mi6 already met
<j-b>
durandal_1707: we already met. You just don't know about it :D
<durandal_1707>
j-b: we never met, you are on drugs, and delusional.
<j-b>
and?
<courmisch>
I thought I was going mad that my code was working when it clearly should not
<courmisch>
well, madder
<durandal_1707>
kierank: when? where? how?
<courmisch>
can anybody check my last checkasm patch (file last touched by wbs, jamrial and elenril but not giving any real names)
<wbs>
courmisch: quick reaction; content-wise, it's clearly "if you say so", I have no idea about that function right now. it mostly comes down to, does the test still pass with the set of assembly + architectures we have in fate? it not, our asm is either broken, or the test happens to test stuff outside of the expected range
<wbs>
I can give the patch a testrun on x86, x86_64, arm and aarch64
<courmisch>
it works on my x86_64, but I don't have the fanciest AVXes
<wbs>
neither do I, but I guess we don't have avx512 simd for sbrdsp either
<courmisch>
though AFAICT if it breaks, it's a bug in the tested code, not the test
<wbs>
mostly depends on whether the random values match what parameters can be passed by the actual decoders in practice, or if it tests cases that assembly rightfully might have decided not to care about. (in best cases these details would be documented in the foodsp.h header, but they seldom are)
<Lynne>
michaelni: ping
<michaelni>
BBB, if you want anything from SPI for work related to ffmpeg, post to ffmpeg-devel. see spi.txt
<michaelni>
Lynne, yes ?
<BBB>
that's way too complicated for me, sorry
<michaelni>
BBB, why ?
<michaelni>
just a mail with [REFUND-REQUEST] in the subject
vincento has joined #ffmpeg-devel
<courmisch>
kierank: you got competition for the FOSDEM track?
whatever_nicknam has quit [Ping timeout: 250 seconds]
Krowl has quit [Quit: Krowl]
<ePirat>
JEEB, VP9 hwdec is not on M3 only btw
<JEEB>
yea, but that's already implemented :)
<ePirat>
JEEB, even my intel mac has it apparently
<JEEB>
I was just testing with VP9 to attempt to verify that everything already existing in the code base worked
<ePirat>
JEEB, yep, theres an issue with it though iirc
<JEEB>
or well, the two most recent formats
<JEEB>
HEVC and VP9
<JEEB>
HEVC worked, VP9 not. and then I hacked AV1 value and that failed too (which is now 100% understandable with the new quotes)
<ePirat>
some VP9 files can cause it to hang
<courmisch>
sounds secure
<thilo>
elenril: I added the two year rule into doc/communit in 6709e468 myself, I guess after some meeting 0:-)
<thilo>
doc/community.texi
<elenril>
no, it is present in the original 4ccb68dc670bbc98b4dede6f21615343dd46561e
<thilo>
right, I think I moved it just around with the commit... I shall not write onIRC so quickly
<thilo>
well the 4ccb68dc670bbc98b4dede6f21615343dd46561e is also from me, dated oct 2020.... so I guess it is valid then for current votes
<Lynne>
michaelni: release?
uau has quit [Server closed connection]
uau has joined #ffmpeg-devel
<michaelni>
Lynne, ill try to do the 6.0.1 and 5.1.4 ASAP and then look at 6.1 but iam a bit behind due to vote investigation and all the discussions related
MikhailAMD has quit [Quit: Leaving]
<Lynne>
no one cares about 5.1.4
<Lynne>
we've had users waiting for more than 8 months for 6.1
<elenril>
it does not fix any tickets, it fixed the patches that follow it
<JEEB>
I think this was it
<mkver>
JEEB: rcombs patch was different IIRC: elenril's patch will lead to more buffering, because a subtitle with a timestamp that is far away from the current timestamp will no longer lead to immediate output; rcomb went the other way IIRC.
MisterMinister has quit [Remote host closed the connection]
<elenril>
oh
<JEEB>
mkver: gotcha :)
<elenril>
longer-term, I think we really want:
<elenril>
1) separating the interleaving queue into its own object, possibly as a public api
<elenril>
2) heartbeat packets in the api
<elenril>
mkver: but do you think a patch like the above is acceptable for now?
<mkver>
For now: Yeah.
<elenril>
good to hear
<michaelni>
JEEB, iam not sure distros (who are the main users of this) would appreciate this, they cannot update their packages at a weekly rate ...
<JEEB>
they would just update at their own rate just fine. there would just never be a large timediff between something entering a release branch, and it getting out in the wild :)
Swedaniel has joined #ffmpeg-devel
lemourin9 has joined #ffmpeg-devel
lemourin is now known as Guest2753
Guest2753 has quit [Killed (tantalum.libera.chat (Nickname regained by services))]
<Lynne>
michaelni: release notes sent
<Lynne>
the release is almost there anyway, all that's left is to tag it and sign it
<ePirat>
JEEB, what would be the use of such frequent releases
<JEEB>
if there are backports they'd automagically get pushed out :)
<JEEB>
also whatever the schedule, not having to manually poke releases
<BtbN>
Just call every single commit to a release branch a release, and call it a day
<wbs>
once a week is probably a bit overkill, but e.g. llvm has adopted a scheme where they do point releases for stable releases every 2 weeks, for maybe 5-6 releases after each new major version
<wbs>
doesn't need to be often as long as it happens somewhat regularly, I guess, is the point. so perhaps every 2 months?
l-boogie has joined #ffmpeg-devel
l-boogie has quit [Client Quit]
<ePirat>
JEEB, wbs, that sounds more reasonable
<ePirat>
also releases are signed iirc so some manual intervention to do that is still needed
vincento has quit [Ping timeout: 250 seconds]
lemourin is now known as Guest3930
lemourin has joined #ffmpeg-devel
Guest3930 has quit [Killed (iridium.libera.chat (Nickname regained by services))]
ccawley2011 has quit [Read error: Connection reset by peer]
quietvoid has quit [Ping timeout: 252 seconds]
lemourin has quit [Read error: Connection reset by peer]
lemourin has joined #ffmpeg-devel
quietvoid has joined #ffmpeg-devel
MrZeus_ has joined #ffmpeg-devel
MrZeus has quit [Ping timeout: 252 seconds]
lemourin has quit [Read error: Connection reset by peer]
lemourin has joined #ffmpeg-devel
MrZeus__ has joined #ffmpeg-devel
MrZeus_ has quit [Ping timeout: 255 seconds]
MrZeus_ has joined #ffmpeg-devel
cone-954 has joined #ffmpeg-devel
<cone-954>
ffmpeg Michael Niedermayer master:19fcf4313148: avformat/mov: Disallow FTYP after streams
<cone-954>
ffmpeg Michael Niedermayer master:bb0a684d93f7: avcodec/4xm: Check for cfrm exhaustion
<cone-954>
ffmpeg Michael Niedermayer master:ac4e3e188af8: avcodec/evc_parse: Check num_remaining_tiles_in_slice_minus1
lemourin has quit [Read error: Connection reset by peer]
MrZeus__ has quit [Ping timeout: 255 seconds]
lemourin has joined #ffmpeg-devel
<cone-954>
ffmpeg Michael Niedermayer release/5.1:b49de8dfe68a: avformat/mov: Disallow FTYP after streams
<cone-954>
ffmpeg Michael Niedermayer release/5.1:49beba036c83: avcodec/4xm: Check for cfrm exhaustion
<cone-954>
ffmpeg Michael Niedermayer release/5.1:4729204c17f7: Update for 5.1.4
MrZeus_ has quit [Ping timeout: 258 seconds]
<michaelni>
"<BtbN> Just call every single commit to a release branch a release, and call it a day" <-- you dont need release branches, just call every commit to git master a release :)
<cone-954>
ffmpeg Michael Niedermayer n5.1.4:HEAD: avcodec/evc_parse: Check num_remaining_tiles_in_slice_minus1