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
<stevenliu>
I asked the student about the code of PEM process code, he said those code generate by AI, because that is better than write by himself
<stevenliu>
I admit that the code wasn't reviewed by you, but several RTC-related developers worked together to refine and review it. We posted it to ffmpeg-devel for two weeks and even reminded everyone, but it seems no one was interested—there wasn't even a single reply. For a student trying to contribute to FFmpeg development, this doesn't feel very welcoming.
<stevenliu>
I apologize—we didn't spot many issues ourselves, but after merging, everyone suddenly found so many problems. My thinking is straightforward: if there are issues, we'll fix them.
<stevenliu>
This was also translated by AI because, as you know, my English isn't very good.
<stevenliu>
BTW, Thanks to Timo and Marvin for dedicating their time to optimizing the TLS code during this period and providing the student with valuable guidance. The student is currently in their third year of university and still building foundational skills. They hope this GSoC opportunity will help improve their ability to contribute to FFmpeg development.
LainIwakura has quit [Quit: Client closed]
TheVibeCoder has quit [Ping timeout: 260 seconds]
<stevenliu>
kierank: don't worry, you have seen me offline, I'm not Jia Tan, the student maybe not Jia Tan too, i can double check about when i trival to his city :D
<kierank>
It's not about that, it's about AI bugs in encryption code
<kasper93>
I didn't want to make a drama about "ai use". I think we can all agree that those were big changes to security related code. It should get few iterations of review before merging.
<kasper93>
Hopefully with forgejo it will be easier to iterate on patches, with refinements. I always seen ML little bit intimidating to point out small suggestions.
<kasper93>
ai bug or just a normal bugs ;) After all one of the regressions was made in follow-up patches and not by ai.
<stevenliu>
Ok. I'll ping Jack Lau and ask him to rewrite the code for the TLS part he submitted before. To be honest, I reviewed quite a few times before, and then he submitted it to BtbN. Overall, sometimes I think AI-assisted code writing for him is more reliable than what he writes himself. It's just that it's a bit difficult to review.
<kasper93>
I think it got bit nervous because of ffmpeg 8 on the horizon
<BtbN>
That was all long before anyone talked about 8.0
<kasper93>
but now, on ML there is idea about reverting
<kasper93>
because of 8
<BtbN>
I don't see how reverting it would even be possible
<BtbN>
You'd have to revert several dozens of patches
<BtbN>
a lot of which are legitimate enhancements and bugfixes
<kasper93>
same, there was lots of patches, from you and others
<BtbN>
I don't think tls_openssl is in a dangerous state
<BtbN>
I have gone over it so many times now
<stevenliu>
I agreed with BtbN
<BtbN>
There's only that idx exhaustion, which is easily fixed
<kasper93>
I think those functions were iffy with memory managment and all
<stevenliu>
kasper93: i saw that before i login irc
<kasper93>
BtbN: I didn't see your comment before
<stevenliu>
then i think i must use VPN jump out the GFW wall say about the code is not full implement by AI, lot's of them written by us.Jack will explain the approach in detail via email to keep everyone informed. We'll do our best to resolve the issues.
<kasper93>
BtbN: I don't actually think BIO_TYPE_SOURCE_SINK is needed. I don't mind just restoring previous state, but looking at this topic NONE type should be enough if we don't use find() or related
<BtbN>
Are you sure? BIO_TYPE_SOURCE_SINK is not a normal type, it's flags OR'ed on top, telling OpenSSL how to interface with the BIO
<kierank>
I think on principle we shouldn't allow that code. It might be boilerplate but it sets a dangerous precedent
<BtbN>
Which code exactly? The entire whip.c?
<kierank>
No the dtls
<BtbN>
It's a bit too late for that now
<BtbN>
I think the openssl dtls code is fine. It's gone through a lot of refactoring now
<BtbN>
The whip.c stuff I never managed to get working...
<stevenliu>
Thanks BtbN and ePirat
<kierank>
I will let the TC decide
<stevenliu>
whip.c have lots modify continue, Jack is doing.
<kierank>
It's like Anton leaving was the breaking of a dam
<haasn>
can some of you elite coders please teach me how to use a web browser?
<haasn>
with my decades of industry experience I have trouble figuring out how to click on a hyperlink
<haasn>
until then, I reserve the right to publicly call everybody a poopyhead who does not babysit me
beastd has quit [Ping timeout: 260 seconds]
BradleyS has joined #ffmpeg-devel
cone-144 has quit [Quit: transmission timeout]
beastd has joined #ffmpeg-devel
<Lynne>
I launched a browser once
<Lynne>
the popups and marquee text scared me
<fflogger>
[newticket] baudneo: Ticket #11694 ([undetermined] Multi-GPU: NVDec cant find CUDA device when used inside a docker container and targeting specific GPU ID) created https://trac.ffmpeg.org/ticket/11694
DauntlessOne49 has joined #ffmpeg-devel
Warcop has joined #ffmpeg-devel
<pross>
"This website requires JavaScript."
nitroxis has quit []
nitroxis has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
jamrial has quit []
deeyes has joined #ffmpeg-devel
mkver has quit [Ping timeout: 248 seconds]
deeyes has quit [Ping timeout: 248 seconds]
deeyes has joined #ffmpeg-devel
<Yalda>
is 8.0 cut yet?
<Yalda>
or branched
<Yalda>
nvm
<Yalda>
i want to get the BD demuxer in but there is one stupid issue left that i'm still trying to fix
LainIwakura has joined #ffmpeg-devel
<fjlogger>
[FFmpeg/FFmpeg] New comment on pull request #20024 rewrite/add mcc [de]muxer, add AV_CODEC_ID_SMPTE_436M_ANC with mxf support, and add conversions <-> eia-608/708 (https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20024) by programmerjake
<fjlogger>
[FFmpeg/FFmpeg] New comment on pull request #20024 rewrite/add mcc [de]muxer, add AV_CODEC_ID_SMPTE_436M_ANC with mxf support, and add conversions <-> eia-608/708 (https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20024) by programmerjake
<fjlogger>
[FFmpeg/FFmpeg] New comment on pull request #20024 rewrite/add mcc [de]muxer, add AV_CODEC_ID_SMPTE_436M_ANC with mxf support, and add conversions <-> eia-608/708 (https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20024) by programmerjake
<fjlogger>
[FFmpeg/FFmpeg] New comment on pull request #20024 rewrite/add mcc [de]muxer, add AV_CODEC_ID_SMPTE_436M_ANC with mxf support, and add conversions <-> eia-608/708 (https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20024) by programmerjake
DVedaa3 has quit [Read error: Connection reset by peer]
deeyes has quit [Ping timeout: 276 seconds]
deeyes has joined #ffmpeg-devel
bwu25 has joined #ffmpeg-devel
HarshK23 has quit [Quit: Connection closed for inactivity]
<fjlogger>
[FFmpeg/FFmpeg] New comment on pull request #20038 avfilter: add vf_scale_d3d11 filter and mfenc support for d3d11 frames (https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20038) by dashsantosh-mcw
<fjlogger>
[FFmpeg/FFmpeg] Comment edited on pull request #20038 avfilter: add vf_scale_d3d11 filter and mfenc support for d3d11 frames (https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20038) by dashsantosh-mcw
<haasn>
kierank: fwiw, I try to always attach review comments to the actual diff (i.e. on the affected line); that way it creates a separate thread which can be marked as resolved/unresolved, and which notifies you if the corresponding diff section changed as result of an update
<haasn>
makes keeping track of review comments slightly easier, I find
<BtbN>
Lynne: waiting for videolan admins to do something, regarding the merge button
<BtbN>
I'd love to enable the merge button, but since I have no admin on g.v.o, I can't setup forwarding there, and depend on thresh to set it up, and thresh is busy
<BtbN>
So our only option to do it ourselves would be to fully get rid of g.v.o, and only leave it as a mirror. We can do that without help, but it'd break everyones push setup
ccawley2011__ has joined #ffmpeg-devel
HarshK23 has joined #ffmpeg-devel
ccawley2011_ has quit [Ping timeout: 248 seconds]
LainIwakura has quit [Ping timeout: 272 seconds]
uau has joined #ffmpeg-devel
ccawley2011__ has quit [Ping timeout: 245 seconds]
<BtbN>
What is Nicolas problem now? The commit message explains the reasoning for the revert, and he goes "This should never heave reached the commit message". Huh?
ccawley2011__ has quit [Read error: Connection reset by peer]
<BtbN>
Does he think it's auto-generated by Forgejo, and just because of that hates it?
<TheVibeCoder>
ban NG
<jamrial>
BtbN: i guess he doesn't like what's essentially an exchange as commit message
<BtbN>
It's not an exchange though. It's quoting the documentation
<jamrial>
ah
<kasper93>
Sorry, I'm not good with commit descriptions, but I wanted to give some context. Also ESL, so I tend to word things in certain ways
<BtbN>
There is nothing wrong with the commit message imo
<BtbN>
Maybe he doesn't like the casual style of not being sure? No clue
<BtbN>
I suspect he thinks it's auto-generated by Forgejo
bwu25 has quit [Quit: bwu25]
bwu25 has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
minimal has quit [Remote host closed the connection]
minimal has joined #ffmpeg-devel
ccawley2011 has quit [Ping timeout: 244 seconds]
<haasn>
I suspect he doesn't think
<haasn>
just sees something he doesn't like and lashes out against it
<haasn>
it reminds me very much of my own behavior when I was a teenager and hated having to use GUI tools
<TheVibeCoder>
lol
<kasper93>
He is traying really hard to discard forgejo. Anything related with it is bad.
<JEEB>
any git forge thing
<kasper93>
I get it, if you are used to some tools the change of it may be challenging. But that's not the reason to discard it out right.
<JEEB>
they would all be invalid and bad for him
<kasper93>
even if he is capable of providing good feedback, I feel it's just muddied by his strong opinions about things
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 248 seconds]
ccawley2011 has joined #ffmpeg-devel
rvalue- is now known as rvalue
<Yalda>
I also didn't understand the point of that complaint
bwu25 has quit [Quit: bwu25]
<cone-676>
ffmpeg Martin Storsjö master:a4a3b3bf544c: fate: Add a missing ffprobe dependency for fate-filter-select-ffprobe
<kasper93>
I woudln't think I will be scrutinized for this commit description so much. I agree I could make it more formal sounding, but at the same time I still don't feel like it's such a huge deal.
<wbs>
if those were quotes from openssl docs, a reference to where it's quoted from would make it clearer maybe
<haasn>
BREAKING: 30 times more expeirenced dev can't figure out a web browser
<TheVibeCoder>
no
kurosu has joined #ffmpeg-devel
<quietvoid>
i don't enjoy reading these emails, very gate-keepy
<haasn>
In the several years since I have started contributing towards FFmpeg the only thing I have consistently heard, from almost EVERYBODY, concerning Nicolas George, is that he's a pain to work with and has not contributed anything of legitimate value to the project in many years
<haasn>
I honestly do not understand why this behavior is tolerated at all, and I think FFmpeg is dying because of such voices
<haasn>
I would be much happier if we could just ignore his opinions outright but the fact that michaelni gives him an unrpecedented amount of power to arbitarily revert commits he does not agree with, which are then dragged into a beaurocratic nightmare because of the de-facto policy of "the reverter wins in the case of a dispute, until overruled by a TC", makes it so we can't just ignore his threats either
<kierank>
Because the leadership controls the project not the CC
<kierank>
And the leadership has actively blocked nicholas from being banned
<kierank>
It's that simple
<kierank>
But nobody speaks out apart from me
<kierank>
And a few others
<haasn>
Well
<kierank>
This project gets what it deserves
Kimapr_ has joined #ffmpeg-devel
Kimapr has quit [Ping timeout: 245 seconds]
<kierank>
It's amazing how quick the standards fell when Anton left
<kierank>
There's nobody to stop Nicolas from being insulting and frankly outright abusive
<kierank>
00:29:54 <Traneptora> not prolific enough of a contributor to feel like I want to parse a reply from someone typing misspelled lowercase incomplete sentences
<haasn>
BtbN: To me, it makes sense to mirror comments and comment edits to the ML but not IRC, yet it seems we log them only on IRC but not the ML
<BtbN>
The events that get logged are not very flexible sadly
<BtbN>
for mail, there is always EVERYTHING, every single push, it was way too much spam
<BtbN>
for the IRC bot you can select Events, so it's not super spammy, but they're still not granular enough
<haasn>
still less spam than posting every single version of every single patch to the ML (= status quo)
<BtbN>
?
<BtbN>
Nothing gets posted to the ML, cause of how spammy it is
<BtbN>
It was on for like half a day
<BtbN>
Or you mean the classic ML workflow?
<haasn>
I'm just pointing out that being notified on the ML that somebody posted a new version of a patchset is strictly less spammy than somebody posting a new version of a patchset to the ML directly
<BtbN>
Oh, yeah.
<BtbN>
I guess the problem is that "quickly another force push" happens much more frequently than sending a new set
<haasn>
To be frank, I am concerned about being accused of "pushing without review" if I stop reading the ML and communicate my intent to push exclusively on the forge
<haasn>
yeah
<haasn>
I was wondering it it's possible to batch / time-gate these events
<BtbN>
I do intend to get mail notifications back
<haasn>
Like, comments could only be sent if there have been no edits since X minutes
<haasn>
And series could be re-sent if there have been no force pushes since X hours
<BtbN>
but I need to write an actual tool that does it, since the native E-Mail notifications have zero configuration options wrt what it sent
<BtbN>
It's literally all or nothing
<haasn>
ask ChatGPT to write a script /s
<haasn>
no pressure from my side, just sharing my thoughts on what would be the ideal design
<haasn>
alternatively we could just formally vote already and get the obvious conclusion that everybody knows will be the result of such a vote
<haasn>
and then we can discontinue the ML instead of sinking resources into a dead platform
<haasn>
but that's just me
<BtbN>
I absolutely do want mail notifications back
<BtbN>
they're important to have an actual migration period
<BtbN>
But "just post all FJ notifications to the ML" was not great
<BtbN>
They need to be somehow filtered
<BtbN>
I'm also not fully happy with the IRC bot. It's just Forgejos own "Slack notifications", translated to IRC
<BtbN>
so I have very little control over them as well, a bit more than E-Mails though
bcheng has quit [Remote host closed the connection]
<haasn>
BtbN: just posting everything to the ML at least offers plausible deniability that nobody can claim people on the ML didn't get a chance to review patches
<BtbN>
It was intense spam though imo, completely drowning out any ML discussions
<haasn>
besides, ffmpeg-devel is already unreadable without automatic tagging
<BtbN>
I'd want to limit it to "PR opened" and "PR comment"
<haasn>
surely ML lovers with their highly superior alien mail clients from 1970 can configure them to make the spam manageable
<BtbN>
And maybe one force-push notification per... 6h?
bcheng has quit [Remote host closed the connection]