Hey.

!hi

!hi

hello 👋

!hi

Ohh, meetbot is v. slow today ... thought it was dead for a minute

yeah, message bus seems unhappy due to a short network outage

!hi

Gwyn Ciesla (limb) - she / her / hers

!hi

Michel Lind (salimma) - he / him / his

We talked about this last week and Fabio Valentini kind of volunteered ... anything else to add?

there's a PR that would implement this: https://pagure.io/packaging-committee/pull-request/1436

but after comments from Panu, latest status seems to be "let's put this on hold" https://pagure.io/packaging-committee/pull-request/1436#comment-214320

maybe dropping the table and referring to the setup package can be split off?

yeh, hopefully

Holding documentation stuff like that up on when rpm changes occur would be a very slow pace.

yup - left a comment: https://pagure.io/packaging-committee/pull-request/1436#comment-214412

Cool

Anyone have anything they'd like to discuss?

anyone going to fosdem?

Me

Conan Kudo: I thought you were going?

me too ... busy week 😬

same... 🫂

It's next week, right?

damn those C compilers, all of a sudden enforcing more type safety! ✊️

Maybe want to cancel the meeting?

yeah, I will most be at CentOS connect next week at this time

CentOS connect Thu Fri, FOSDEM sat sun

I'm running things on Thursday at Connect

so there's no way I'm attending any fedora meetings that day 😅

so many FTBFS

How many? I have 29 🫠

you can me both

On par with last FOSDEM iirc

you and me both (can't type today)

50

The spring releases are brutal

Oy vey

yeah

I'm drowning in Go failures _and_ GCC 15 failures

It's a sign from the universe to drop packages that don't spark joy anymore 🫠

Ugh, what broke in go?

I kind of wish there was an easier way to pass -std=gnu11 to the compiler

Golang 1.24 i think

apparently everything, because most of my failures are Go based'

> Go fish!

Go 1.24 adds more format string checking logic

Yeah. I have a new vendored golang package and that broke too

(anyone going to fosdem is allowed to slap me for that one)

I have very few GCC 15 ones, mostly a ton of Go 1.24 ones

on a positive note, the caddy ftbfs inspired me to dig into why i kept having to skip flaky tests, and found that upstream runs their tests with `-short` which took care of all my problems and let me drop a patch

yay

Ahh

Want to pull all new commit from os repository

Which OS repository is appropriate to build and assemble new releases to test rotation deployment of new releases and stable release of okd and OpenStack ?

Weird, I'd seen the relnotes for 1.24 and it seemed new features and perf. improvements.

So you were getting timeout problems that even upstream doesn't care about huh

Not actually tried it yet though

Um... Wrong channel?

Sorry for that

James Antill: https://tip.golang.org/doc/go1.24#vet under "printf analyzer"

Since this is FPC... can we encourage compiler changes to be tested like python tests new releases?

they *should* have been

Not mandate it, but point to the python copr testing as a gold standard

apparently they kind of assumed that the mass rebuild would be postponed and were surprised when it wasn't

(and file bugs against packages before landing the change in rawhide... Febian does this too)

that would be a good question for #devel:fedoraproject.org

(and file bugs against packages before landing the change in rawhide... debian does this too)

It was pushed a bit due to GCC update, IIRC ... just a day or two and it might have needed more like a week or two

But, yeh, golang should think about doing side tags and rebuilding everything if it's going to start failing builds on updates.

yeah. they said they did a mass rebuild in COPR but didn't have time to investigate stuff

(not go, but gcc)

I wonder if we need to add a cutoff for toolchain changes

to the change proposal deadlines

we added one for infra changes a few years ago

yeah. pushing major compiler update hours before mass rebuild seems ill-advised

no chance to fix any immediate fallout

I genuinely did not expect this much fallout from the Go compiler update

Hard to do things like that when one of our Fs is for First

Cutoff, plus encouraging testing pre releases in COPR and filing issues, seems like a reasonable improvement

I have been prepared for GCC fallout, so that one didn't bother me

Asking for more testing seems much more doable though

pleasantly surprised to not have too much of that

I think "new GCC and Go major releases must land in rawhide one week before mass rebuild" wouldn't contradict that principle

we should probably generalize that to new toolchain major releases and then carve out an exception for llvm

since llvm will never ever make it

who's going to file the fesco ticket? :D

Snowflakes...

no, apples

llvm always lands the day of beta freeze or the day after

it's annoying as heck, but fesco is resigned to this... we tried :(

Roughly how many packages rely on llvm now?

the main critical ones are rustc and mesa

I owe too many documentation PRs already, not me (at least not soon)

I can do it

$ dnfraw repoquery --whatrequires llvm-libs --qf "%{source_name}\n" | wc -l

21

For llvm I feel like we should ask them to package it like python

we kind of did, they didn't want to

And not switch the default to the new one if it lands late. Just make it available as an option to people who need it

So maybe the tool chain deadline can be something that we can point them to, whether it helps or not

we should consider asking gcc to do the same too

Yeh, have the wording be something like "the default version of the toolchain, used to build packages"

fedora has had multiversioned gcc for a couple of releases now

it might make sense to just switch to that model like opensuse does

(it exists for cuda atm)

Yeah, they're already doing an older version for cuda. Should be easy to now say also make the newest version non default

Interesting ... it's been a few years since I've really done any work in C ... not sure if I want to go back ;)

> https://pagure.io/fesco/issue/3347

salimma gave a cookie to decathorpe. They now have 114 cookies, 4 of which were obtained in the Fedora 41 release cycle

ngompa gave a cookie to decathorpe. They now have 115 cookies, 5 of which were obtained in the Fedora 41 release cycle

I expect that we'd want a devel list discussion for this too, but it seems uncontroversial

carlwgeorge gave a cookie to decathorpe. They now have 116 cookies, 6 of which were obtained in the Fedora 41 release cycle

james gave a cookie to decathorpe. They now have 117 cookies, 7 of which were obtained in the Fedora 41 release cycle

That's probably a good idea. We never know what people might object to

I expect some objection from the toolchain folks, actually

that too

probably a mix of "7 days isn't enough" and "how dare you not let me push things in last minute"

*objection!*

us: *overruled!*

if the gcc freeze was November 1 instead of November 15, it probably could all be done on time

Yeah, so a devel discussion will probably have the packagers siding with fesco against the tool chain team

It's nice being on the popular side for once

yes, like what happened with the annobin thing during gcc10

ah, actually I guess that was gcc8

oh nope it's just happened multiple times 🙃

the most recent one was https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/FNOL635TXCZ2AUEJ7VS6GZHLMPTHMYWO/

that's what led to https://pagure.io/fesco/issue/2454

oh, familar faces

yup 😅

We should cite this in the new ticket too

I'm not shocked, but still amused.

To justify getting a generalized deadline

yeah

toolchain stuff is hard, and packagers are squeezed by the fallout when there isn't enough time

Slight segue - James Antill how do you script sending the meeting notifications as DMs to everyone?

Very complicatesd script ... I open https://fedoraproject.org/wiki/Packaging_Committee_Meeting_Process#On_the_day_of_the_meeting in another window and copy and paste each line

human script :o

On IRC I used to be able to paste the entire thing as one block, but that doesn't work in matrix (would love to see an option/fix for that).

I wonder if Quaternion has such a thing

it is a Matrix client that tries to behave like an IRC one

I had high hopes for the rust matrix client but it choked trying to log in to my fedora.im 🫠

I think Fedora SSO for matrix hits most of the new clients people want to try

So ... not shocked.

I think Quaternion supports ODIC/SSO login

On the upside, I think they all get it to work eventually

I think Fractal should work now

NeoChat does and they use the same backend

(I have used both NeoChat and Fractal with fedora.im before, but the former crashed on me and the latter is hideous)

Fractal does work, it's just missing some features

So I'm still on element, schildi and if I really have to, fluffy

I use Element and NeoChat currently

i've tried a handful of clients and not really happy with any of them. element web seems like the least bad.

close meeting? :D

seems to be very offtopic now

yeah I think we're pretty much done here

Yeh