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
<fflogger>
[newticket] idest: Ticket #11424 ([ffmpeg] Colors get distorted when converting from an animated PNG into an animated GIF) created https://trac.ffmpeg.org/ticket/11424
iive has quit [Quit: They came for me...]
rvalue has quit [Ping timeout: 260 seconds]
rvalue has joined #ffmpeg-devel
<fflogger>
[editedticket] thebombzen: Ticket #11424 ([ffmpeg] Colors get distorted when converting from an animated PNG into an animated GIF) updated https://trac.ffmpeg.org/ticket/11424#comment:2
<fflogger>
[editedticket] idest: Ticket #11424 ([ffmpeg] Colors get distorted when converting from an animated PNG into an animated GIF) updated https://trac.ffmpeg.org/ticket/11424#comment:3
Traneptora has quit [Quit: Quit]
IndecisiveTurtle has quit [Ping timeout: 248 seconds]
<cone-016>
ffmpeg Lingyi Kong master:504df09c3460: avcodec/h264: fix stride calculation in slice_table for multi-slice field video deblocking
^Neo has joined #ffmpeg-devel
^Neo has quit [Changing host]
^Neo has joined #ffmpeg-devel
ngaullier has quit [Ping timeout: 252 seconds]
^Neo has quit [Ping timeout: 252 seconds]
elChippo has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg-devel
ngaullier has joined #ffmpeg-devel
ngaullier has quit [Ping timeout: 264 seconds]
cone-016 has quit [Quit: transmission timeout]
ngaullier has joined #ffmpeg-devel
<Lynne>
michaelni: fixed the ffv1 get_format patch issue, v3 of the patch is on the ML
<Lynne>
I had simply forgotten to put back use32bit=1 for 16-bit rgb
mkver has joined #ffmpeg-devel
^Neo has joined #ffmpeg-devel
jamrial has joined #ffmpeg-devel
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 264 seconds]
rvalue- is now known as rvalue
ccawley2011 has joined #ffmpeg-devel
ngaullier has quit [Ping timeout: 245 seconds]
ngaullier has joined #ffmpeg-devel
Kei_N_ has quit [Ping timeout: 248 seconds]
Kei_N has joined #ffmpeg-devel
qesat60 has joined #ffmpeg-devel
blb has quit [Ping timeout: 272 seconds]
blb has joined #ffmpeg-devel
psykose has joined #ffmpeg-devel
Malarkey107367 has joined #ffmpeg-devel
<Malarkey107367>
Hey guys... Joe Biden here. I've decided to step down from the White House to focus on other projects. Billionaires are a threat to democracy, so check out https://BidenCash.st to put them in the bullseye. Keep an eye on the CNN inauguration for a promo code!
Malarkey107367 has quit [Ping timeout: 245 seconds]
System_Error has quit [Ping timeout: 264 seconds]
<BtbN>
wtf
<kierank>
old school irc spam
Marth64 has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
<Compn>
4 years of crypto scams :(
^Neo has quit [Ping timeout: 265 seconds]
<thardin>
joe biden uses IRC? truly IRC is the boomer protocol
<Marth64>
subtitle filtering guy is back
<Marth64>
o.0
IndecisiveTurtle has joined #ffmpeg-devel
<BBB>
calling irc a boomer protocol, man that hurts deep
<BBB>
Marth64: basty?
<BBB>
oh softworkz
<Marth64>
ya
softworkz has joined #ffmpeg-devel
softworkz has quit [Client Quit]
stupidoo has joined #ffmpeg-devel
<Marth64>
i am sharing my unhinged opinion for once. i am sorry and mean no offense to anyone
<Marth64>
but i also am being transparent with how i feel
<Compn>
irc is pretty old
<Compn>
at least we arent on a BBS
thesynthax has quit [Ping timeout: 276 seconds]
ccawley2011 has quit [Ping timeout: 265 seconds]
ccawley2011 has joined #ffmpeg-devel
<Marth64>
irc is fine
<Marth64>
patch flow is slow
ccawley2011 has quit [Ping timeout: 245 seconds]
ngaullier has quit [Remote host closed the connection]
ccawley2011 has joined #ffmpeg-devel
thesynthax has joined #ffmpeg-devel
<beastd>
hard problem: many people want to contribute but we don't have enough review capacity. review is in general more of a maintainer thing (sometimes non-maintainers are knowledgable and review if they have time, but it's not so common). OTOH most people don't want to maintain mid/long term. more scratch their own itch and get code in ASAP.
<Marth64>
responding my opinion on that shortly
<beastd>
this is the open source classic. it's when project get big (quick). then get lots of users and contributions. then grow quickly and get under preassure. it's also very hard to get hold of it long term.
<Marth64>
i still think its because the tooling wastes time
<beastd>
Marth64: tooling is part of it. and i'm all in for improving the tooling.
<beastd>
but i think it's pretty clear that biggest part of the problem are "people problems" (is it the right term?) and lack of community power to sustain growth and maintenance of the project that is suddenly much more then it was in the beginning. and the long tail of keeping it all running, stable and secure.
<kierank>
Marth64: agreed, I tried to push a patchv4 today and it wasn't even on the ml
<kierank>
actual email
<kierank>
it was on the web page, but not in any of the two mailboxes I have
<kierank>
one of which has actually unsubscribed...
<Marth64>
its all literally a waste of time
<beastd>
kierank: If it's in the archives it's probably an email problem (at ffmpeg server or any smtp servers in between or maybe (web-)client processing)
<Marth64>
so of course people are annoyed and pissed off all day
<kierank>
in the end one person (tm) controls the infrastructure and will not allow anyone to create gitlab which is a new piece of infrastruvture
<beastd>
this seems now a bit derailed from the discussion we had
<beastd>
i think we should not drop mailing lists regardless of having a git forge for pull requests
<Marth64>
i agree its ok for discussions, concepts, ideas
<BtbN>
That's simply not true, kierank
<beastd>
unfortunately mailing lists got awkward when all the security featues got added to mail transport :-/
<BtbN>
The main hold-up with gitlab vs. forgejo right now is people disagreeing about which one to use.
<JEEB>
at least the comment I read here was that self-hosting was OK as long as the "community" picks a tool
<kierank>
BtbN: why then is git allowed on videolan, but gitlab is not allowed on videolan
<kierank>
like one person wants forgejo
<kierank>
the majority want gitlab
<Marth64>
what is the fastest path forward to a vote or decision?
<kierank>
GA vote
<kierank>
but there is one person (tm) questioning the validity of the GA
<kierank>
which is very apt on a day like today :)
<BtbN>
Vote on what to do, and it can be implemented pretty rapidly
<JEEB>
someone writes down concrete steps and then there's a GA vote
<Marth64>
if that person is on GA then their vote matters too
<BtbN>
I just don't want to set up gitlab or forgejo, and then get yelled at for making that decision on my own
<kierank>
in the end gitlab needs to be on videolan in practical temrs
<kierank>
terms
<BtbN>
not really, no
<beastd>
kierank: where is the real discussion about which git forge? imho forgejo seems better alternative then gitlab. BtbN also said sth like that. and iirc Lynne was also more gitea/forgejo
<BtbN>
Can easily self host it
<Marth64>
Can we start first with a vote of which platform, then iron out who hosts it?
<Marth64>
At least that is progress?
<kierank>
propose a vote with a deadline
<BtbN>
The plan for hosting it is pretty much rent Hetzner-Server with SPI money, and set it up there.
<kierank>
lol
IndecisiveTurtle has quit [Read error: Connection reset by peer]
<BtbN>
There is people who very aggressively are against using videolan infra, and there's at least one person who is very heavily against Gitlab. Haven't heard anyone being heavily against Forgejo.
<kierank>
huh
<JEEB>
yea, since most have not tried forgejo properly (like me)
<kierank>
nobody even knows what forgejo is
<BtbN>
It's Gitea with a new name.
<nevcairiel>
the interface looks a lot like open-source github, people will be able to deal with it
<BtbN>
Yeah, they are very Github-Like
<JEEB>
also cli tooling was one thing that was popping up and was a requirement for elenril IIRC
<JEEB>
gitlab was in front regarding that if I recall correctly
<kierank>
We use gitea at work. It's not sophisticated enough for ffmpeg
<kierank>
doesn't have inbuilt ci
<BtbN>
yes it does
<beastd>
it does
<BtbN>
they literally cloned Github Actions
thesynthax has quit [Ping timeout: 245 seconds]
<kierank>
is it actually non-experimental yet
<BtbN>
It's still in development, but works fine as it is. Just need to host runners somewhere.
<kierank>
LOL
<BtbN>
Pretty much identical to how Gitlab does it
<BtbN>
So there is not much different there on either of the two
<kierank>
We could use something that's known to work at the scale ffmpeg needs, or something nobody knows anything about
<BtbN>
Well, make a vote then and decide
<BtbN>
I don't care which one we end up using
<beastd>
kierank: nobody like the fedora project?
<nevcairiel>
plenty people use gitea/forgejo at any scale
<kierank>
for the sake of being contrarian
<BtbN>
I think Forgejo will work fine for FFmpeg
<BtbN>
much bigger projects use it successfully
* beastd
thinks so as well
<JEEB>
granted, I don't think fedora has yet taken it into use but just has done the initial decision after checking things
<BtbN>
Yeah, it's an in-progress for them, but it's gonna happen
<nevcairiel>
i dont really care what is used, as long as email for patches dies, but gitea/forgejo is hardly an unknown unproven project
<kierank>
gogs -> gitea -> forejo -> ???
<beastd>
kierank: yes
<kierank>
pfft
ccawley2011 has quit [Read error: Connection reset by peer]
<beastd>
but why do you think forgejo will be forked again?
<RadicalLeader>
also gitlab makes it less obvious way to follow decision process and who is decision maker
s55 has joined #ffmpeg-devel
<RadicalLeader>
and who would do transfer to gitlab?
<JEEB>
BtbN would be ready to deploy, and we received offers to help with issue migrations if that is required
<Marth64>
are there any blockers from starting a tool selection focused vote?
<Marth64>
who has authority to do it?
<JEEB>
I don't recall the rules fully, but I wouldn't be surprised if anyone from the GA would be able to announce a vote?
<RadicalLeader>
i know only who did voting stuff
<BtbN>
Basically anyone just launch a vote Gitea vs. Forgejo vs. Gitlab vs. Videolan Gitlab vs. Github vs. stay ML only
<BtbN>
or something like that
<BtbN>
No idea who can, but something has to happen
<RadicalLeader>
leave it all to michaelni
<Marth64>
i think thats the first step (toward an action that is probably going to yield a net productivity gain) and independent from ownership
<Marth64>
(as in, having the vote)
<JEEB>
doc/community.texi has some details at least
<RadicalLeader>
start voting, before everyone stops voting
<Marth64>
i do think its a technical decision. would TC sponsor such a vote?
<Marth64>
at least with regards to what the tool options are
<RadicalLeader>
try it, send message to tc, good luck
<JEEB>
I would say that if there is a conflict regarding it
<JEEB>
the doc doesn't exactly note who can announce a vote, but since the organization is based on GA plus the two committees for conflicts I'd say anyone from the GA can propose a vote
<JEEB>
see one of elenril's (Anton) proposals for vote as an example
<RadicalLeader>
why was dav1d release named 'Sonic'
<BBB>
wrong channel for dav1d questions
<nevcairiel>
dav1d releases are named after fast things
<RadicalLeader>
sonic is for audio
<nevcairiel>
sonic is a fast hedgehog
<RadicalLeader>
confusing, use full name next time
<Traneptora>
regarding security and confidential issues, there's no reason we can't deprecate ffmpeg-security@ffmpeg.org can we
<Traneptora>
or whatever the security list is
<Traneptora>
most devs and contributors won't need to care about that
<RadicalLeader>
ffmpeg-security is reserved for only leader and close friends
<Traneptora>
er, no reason we *have* to deprecate it
<beastd>
Traneptora: ah ok. now it makes sense. did already wonder regarding the message from you right below :)
RadicalLeader has quit [Quit: Client closed]
<Marth64>
i don't think i can just start a vote but i opened a thread
System_Error has quit [Remote host closed the connection]
Marth64 has quit [Quit: Leaving]
Marth64 has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
iive has joined #ffmpeg-devel
<steven-netint>
I'm preparing a bunch of HW video codecs/filters for inclusion in libav. It runs on add-in-cards like nvenc. What kind of community testing and maintenance do I need to prepare? ie. do i need to write FATE tests specific to my HW, do I need to make the HW available in some way for community access? What does other HW vendors do?
<Traneptora>
I'm not sure how likely it will be that highly specific hardware encoders will be merged into master
<steven-netint>
We're trying to get out of the habit of putting our code on an FFmpeg fork and contribute it up to FFmpeg main
<steven-netint>
darn D:
<Traneptora>
although I'm not really too familiar
<BtbN>
The problem is the userbase will be hyper limited, and a hw-processing-chain adds a lot of complexity to ffmpeg
<BtbN>
Given that your patches seem to be on top of ffmpeg 5.1.1, that's also not going to be trivial to upstream to begin with
<steven-netint>
Traneptora: I understand the desire to avoid overly niche bloat. Though the HW will be available through akamai cloud soon. And hopefully more to follow
<steven-netint>
BtbN: The fork on github is on FF-n7.1 but I understand some of my colleagues are working on solving remaining incompatibility with the new FF7 multi-threading
<steven-netint>
We're trying to get it to a state where we only need to add new codecs/filters the HW without altering common FFmpeg code for handing HW codecs
<BtbN>
Could the device not just offer a vaapi driver?
<steven-netint>
hmm, I think vaapi is under consideration though i'd have to check where that discussion is currently
<BtbN>
That would instantly offer hwaccel support for de/encoding and some filtering with not just ffmpeg
<BBB>
steven-netint: does it use opensource libraries/drivers? or do you need closed-source libraries/drivers to use?
<steven-netint>
our current driver is MIT-0. The codecs/filters we put in our fork of FFmpeg follows FFmpeg's LGPL
<steven-netint>
We dont have any closed source code on the SW side. FW and HW will be proprietary as is typical
<BBB>
I mean, there's precedent for such stuff upstream then. I think your main problem will indeed be how hyper-specific it is, but I don't think that's an absolute blocker... I guess you'll have to make an argument on why this is useful upstream (for the wider community)
<steven-netint>
yes, i understand
Everything has joined #ffmpeg-devel
<steven-netint>
but setting aside the wider community appeal aspect, is there FATE tests and HW access I should prepare if the HW codec is included into FFmpeg?
<Compn>
BBB, maybe if its commented out so well it becomes a template for other hwcodecs? /s
<Compn>
ehehe
<BtbN>
So far hardware stuff has very limited fate coverage, if any
<Compn>
steven-netint, for the asic boards your company makes , driver is binary blob or open source ?
<Compn>
oh i see you answered it earlier, ignore me
<Compn>
mit-0 lets goooo
<steven-netint>
more like, we we're lazy about license control
<steven-netint>
It didnt seem worth it to quibble over making some parts of our SW proprietary
thesynthax has joined #ffmpeg-devel
<steven-netint>
Regarding the wider appeal aspect for upstreaming a HW codec/filter, the HW netint makes is mostly targeted to datacenters.
<steven-netint>
A lot of them are running apps interfacing with libav
<BBB>
codec should be easy to fate, and requiring fate for new codecs tends to be standard - although more for decoders than encoders right now
<BBB>
filters I don't know, I think that could be improved, so requiring that of newly added filters seems like a stretch to m
<Compn>
the fate tests could only work if you had the specific hardware, so they would be disabled for all other builds yes
<steven-netint>
Would the FFmpeg community consider adding codecs/filters for server-side wants?
<Compn>
steven-netint, possibly. i think we are asking if netint is making 10,000 asic or 10,000,000 asic...
<Marth64>
Compn: thanks for sharing. this helps my upcoming point. there are ways to bridge and fill the gaps. perhaps a little wrapper layer to take email and call <possible future solution> via api
<Marth64>
or command line instructions, whatever
<Compn>
Marth64, yes i think tying in the old standard ways will make more people happy
<steven-netint>
Compn: on patchwork i see some checks which run on loongarch and x86. Are those HW resources a thing made available to the community a contributor for these ML patch auto-tests?
<Marth64>
yes and its not a bad thing. transition is a process
<Compn>
steven-netint, yes sometimes companies send us boards and we put the boards in our fateservers
<Compn>
steven-netint, you could also run a fateserver with your hw as well , its not centralized
<Compn>
hopefully i'm answering your question properly :)
<steven-netint>
you've given me things to consider. Thanks :D
<Compn>
also i might be answering questions incorrectly. so if something is unclear or sounds wrong just ask it again
<steven-netint>
I don't think 10 million netint asics is in sight yet but We're much beyond 10k asics
<steven-netint>
Gotta keep on improving public access via cloud, etc.
<steven-netint>
Though, how do you guys feel about HW vendors which only Fork ffmpeg but not not upstream their codecs/filters? I'm not too familiar but I thought MA35D is on an FFmpeg-n4.0 fork?
<steven-netint>
Would you prefer they just upstream or nah?
<BtbN>
If it's ultra-nieche and the source is available, it seems like a fair solution
<Compn>
steven-netint, to answer your other question, there are some devs who agree with including everything to ffmpeg. and other devs who think putting new stuff in ffmpeg is bad
<Compn>
so you ask the community , but the community is split on the issue
<steven-netint>
understandable
<Compn>
my opinion is add everything and let god sort it out /s
<steven-netint>
ahh yes, i do love that wing commander III demuxer
<Compn>
otherwise its just a lot of people complaining that theres 'too much' in ffmpeg and they all want to cut something that each other likes
<Compn>
yea. game formats for one
<Compn>
i mean ffmpeg is used by emulators to play game video...
<Compn>
huge use case
<Compn>
many developers cut their teeth writing game codecs/demuxers
<Compn>
because they were simple codecs to reverse engineer/program
<steven-netint>
bless them for their righteous contributions
<Compn>
very difficult to herd the ffmpeg cat where you want it to go. i dont think anyone expected ffmpeg cat on mars though
<Compn>
is that because ffmpeg was made to be built on nonsense like DOS , OS/2 and the like ? these are other features which some developers question the utility of
<steven-netint>
I wonder if a solution could be to make a 'codec-pack' git project that forks FFmpeg and adds any and every codec/filter that neatly fits into its own files
<Compn>
steven-netint, re:filters yeah, i'm of the opinion that all filters are good filters
<Compn>
might just be better to split ffmpeg into an ffmpeg-all and an ffmpeg-lite . where lite would have just the vp9/av1/h264/h265/mpeg and aac/mp3/ac3 . most people dont need to be able to play 25 year old video files
<BtbN>
there definitely is a point where adding everything becomes unmanageable. Specially for stuff that is potentially security critical.
thesynthax has quit [Ping timeout: 252 seconds]
HarshK23 has quit [Quit: Connection closed for inactivity]
witchymary has quit [Remote host closed the connection]