<@sgallagh:fedora.im>
16:00:20
!startmeeting ELN SIG (2024-08-30)
<@meetbot:fedora.im>
16:00:22
Meeting started at 2024-08-30 16:00:20 UTC
<@meetbot:fedora.im>
16:00:22
The Meeting name is 'ELN SIG (2024-08-30)'
<@sgallagh:fedora.im>
16:00:25
!meetingname eln
<@meetbot:fedora.im>
16:00:26
The Meeting Name is now eln
<@sgallagh:fedora.im>
16:00:36
!topic Init Process
<@sgallagh:fedora.im>
16:00:38
!hi
<@zodbot:fedora.im>
16:00:39
Stephen Gallagher (sgallagh) - he / him / his
<@salimma:fedora.im>
16:01:04
!hi
<@zodbot:fedora.im>
16:01:05
Michel Lind (salimma) - he / him / his
<@davide:cavalca.name>
16:01:21
!hi
<@zodbot:fedora.im>
16:01:23
Davide Cavalca (dcavalca) - he / him / his
<@tdawson:fedora.im>
16:02:10
!hi
<@zodbot:fedora.im>
16:02:11
Troy Dawson (tdawson)
<@conan_kudo:matrix.org>
16:02:41
!hi
<@zodbot:fedora.im>
16:02:43
Neal Gompa (ngompa) - he / him / his
<@sgallagh:fedora.im>
16:03:04
Welcome, folks!
<@sgallagh:fedora.im>
16:03:08
!topic Agenda
<@sgallagh:fedora.im>
16:03:16
We have two items currently listed for the agenda:
<@jrichardson:matrix.org>
16:03:20
!hi
<@zodbot:fedora.im>
16:03:22
James Richardson (jrichardson)
<@sgallagh:fedora.im>
16:03:27
!info Agenda Item: Status update on composes
<@sgallagh:fedora.im>
16:04:02
!info Agenda Item: How to get Fedora packagers more involved with ELN
<@sgallagh:fedora.im>
16:04:17
Does anyone else have a topic they want to discuss today?
<@yselkowitz:fedora.im>
16:04:45
!hi
<@zodbot:fedora.im>
16:04:46
Yaakov Selkowitz (yselkowitz)
<@sgallagh:fedora.im>
16:04:49
Also, just as a point of order, we WILL be having a meeting next Friday as well. Today's meeting was shifted due to conflicts.
<@davide:cavalca.name>
16:04:50
did we reach a conclusion from the ELN Extras / EPEL 11 discussion at Flock?
<@salimma:fedora.im>
16:05:09
^ Davide said it first
<@sgallagh:fedora.im>
16:05:22
!info Agenda Item: ELN Extras vs. EPEL 11
<@sgallagh:fedora.im>
16:06:15
Alright, let's get started. I'll open with the compose status which is short, then we can decide as a group which of the other topics to hit second.
<@sgallagh:fedora.im>
16:06:31
!topic Status update on composes
<@sgallagh:fedora.im>
16:07:06
So, I was *hoping* to be able to share with you that the composes have been successfully migrated off of ODCS today.
<@sgallagh:fedora.im>
16:07:18
Unfortunately, I hit a few snags on that front.
<@sgallagh:fedora.im>
16:07:33
However, with the *extremely patient* help of Conan Kudo, I think I'm almost there.
<@conan_kudo:matrix.org>
16:07:48
if we're lucky we'll have it today
<@sgallagh:fedora.im>
16:07:58
I have what I hope will be a successful converted compose running right this minute.
<@sgallagh:fedora.im>
16:08:09
With a little luck, we'll know by the end of the meeting if it's solid.
<@sgallagh:fedora.im>
16:08:52
!info Fedora ELN cloud and container images have been converted to using Kiwi to generate them and the MR is under review.
<@zodbot:fedora.im>
16:08:56
salimma has already given cookies to sgallagh during the F40 timeframe
<@zodbot:fedora.im>
16:09:02
salimma gave a cookie to ngompa. They now have 137 cookies, 24 of which were obtained in the Fedora 40 release cycle
<@zodbot:fedora.im>
16:09:05
yselkowitz has already given cookies to ngompa during the F40 timeframe
<@sgallagh:fedora.im>
16:09:08
!link https://pagure.io/pungi-fedora/pull-request/1324
<@sgallagh:fedora.im>
16:09:57
!info Fedora ELN composes are nearly ready to migrate off of ODCS. Work-in-progress merge request is available to review.
<@sgallagh:fedora.im>
16:10:06
!link https://pagure.io/pungi-fedora/pull-request/1304
<@sgallagh:fedora.im>
16:11:08
I think that's everything I have to say on that score, unless there are questions?
<@sgallagh:fedora.im>
16:12:07
OK, then do we want to discuss EPEL 11 or Fedora Packagers next?
<@salimma:fedora.im>
16:12:15
packagers first I think
<@salimma:fedora.im>
16:12:26
the EPEL thing blew up last time and it might take a long discussion :)
<@sgallagh:fedora.im>
16:12:47
Fair point. We can also take that discussion to the EPEL SIG/meeting if needed.
<@tdawson:fedora.im>
16:12:59
I agree, let's do Extas/EPEL11 last.
<@sgallagh:fedora.im>
16:13:08
!topic How to get Fedora packagers more involved with ELN
<@sgallagh:fedora.im>
16:13:27
Conan Kudo: This was your proposed topic, do you want to start it off?
<@conan_kudo:matrix.org>
16:14:41
Yeah, so Carl George and I had been talking about this off and on for a while. I think we need to think about a framing for ELN that makes it relevant and useful for Fedora contributors to participate, even if it's just around ELN Extras initially.
<@conan_kudo:matrix.org>
16:15:26
Something that came out of that discussion is that I am not sure what workflow each of us use for leveraging ELN information.
<@conan_kudo:matrix.org>
16:15:49
Mine is kind of basic: I just check it once every few weeks with my workloads and go deal with things when I see them.
<@conan_kudo:matrix.org>
16:16:48
I am not even aware if we have fedora-messaging stuff or anything that could make it align more with the interrupt-driven management of Fedora packagers.
<@conan_kudo:matrix.org>
16:17:06
I am not even aware if we have fedora-messaging stuff or anything that could make it align more with the interrupt-driven management of Fedora packages.
<@sgallagh:fedora.im>
16:17:15
One point there:
<@salimma:fedora.im>
16:17:34
I have a point after Stephen
<@sgallagh:fedora.im>
16:17:38
One side-effect of the conversion away from ODCS is that the nightly.sh approach DOES send messages and compose diffs.
<@sgallagh:fedora.im>
16:17:57
(You may have seen the accidental compose report that I sent yesterday)
<@conan_kudo:matrix.org>
16:18:07
Yeah that was awesome, actually.
<@sgallagh:fedora.im>
16:18:30
Agreed; though I have needed to patch it to handle the frequency of ELN composes better.
<@sgallagh:fedora.im>
16:18:44
(Sending eight of those per day to the list would be spam, not information)
<@yselkowitz:fedora.im>
16:19:07
are we still going to compose that often?
<@conan_kudo:matrix.org>
16:19:20
I have even thought about coming up with some way that we could run tests in Fedora on images with specific workloads relevant for EPEL (e.g. KDE on CentOS Stream)
<@sgallagh:fedora.im>
16:19:26
yselkowitz: That's an open question to be answered once we get things working
<@conan_kudo:matrix.org>
16:19:42
Since both Troy Dawson and I do care about this with our CentOS SIGs
<@sgallagh:fedora.im>
16:20:09
yselkowitz: The new composes are also much faster; approximately one hour vs two and a half when using ImageFactory.
<@yselkowitz:fedora.im>
16:20:21
nice!
<@conan_kudo:matrix.org>
16:21:03
Oh yeah, kiwi is _fast_
<@conan_kudo:matrix.org>
16:21:19
each image build takes 20^~40% of the time the old ImageFactory builds did
<@salimma:fedora.im>
16:21:20
I like tools that focus on doing one thing well :)
<@conan_kudo:matrix.org>
16:21:24
each image build takes 20%~40% of the time the old ImageFactory builds did
<@sgallagh:fedora.im>
16:21:38
Anyway, that's a tangent.
<@salimma:fedora.im>
16:21:54
my point is related I think
<@sgallagh:fedora.im>
16:22:01
Could you talk more about how you want to organize tests?
<@sgallagh:fedora.im>
16:22:11
Sorry, Michel Lind 🎩. Please go ahead.
<@salimma:fedora.im>
16:22:13
people often get surprised when their package get pulled into ELN as a dependency, so anything we can do to get more visibility helps i guess
<@salimma:fedora.im>
16:22:34
but let's do Stephen's question first
<@sgallagh:fedora.im>
16:24:18
Conan Kudo: What sort of tests are you envisioning, and how would the results get reported?
<@conan_kudo:matrix.org>
16:24:51
So I was thinking more along the lines of the kind of tests we do in Fedora packages, including so-called "critpath" validation
<@conan_kudo:matrix.org>
16:25:13
which includes smoke testing and even running some subset of our integration tests for the package sets
<@conan_kudo:matrix.org>
16:25:39
I'm pretty sure we can swing extra resources from AWS through davdunc for this purpose, as it would be highly beneficial and useful to make sure everything actually works when we get to branch time
<@conan_kudo:matrix.org>
16:26:11
because I don't want to "steal away" compute resources from Fedora (which is plenty starved as it is)
<@sgallagh:fedora.im>
16:27:06
Do you see this as gating package tagging or composes in some way?
<@conan_kudo:matrix.org>
16:27:27
not initially, but maybe down the road, yes
<@conan_kudo:matrix.org>
16:27:45
because my thought is that we want to essentially implement Rawhide is Alpha for ELN
<@conan_kudo:matrix.org>
16:28:10
not saying that ELN is alpha, but that we have gates and quality checks that indicate some level of assurance long before we branch
<@conan_kudo:matrix.org>
16:28:37
this can also feed into finally fixing some fundamental quality check issues that we have with EPEL itself
<@conan_kudo:matrix.org>
16:28:51
as ELN is morphing to look more like a split Fedora than a merged one
<@conan_kudo:matrix.org>
16:28:58
so a lot of the same problems would apply
<@conan_kudo:matrix.org>
16:30:30
context of rawhide is alpha statement: https://fedoraproject.org/wiki/Changes/NoMoreAlpha
<@conan_kudo:matrix.org>
16:30:32
!link https://fedoraproject.org/wiki/Changes/NoMoreAlpha
<@sgallagh:fedora.im>
16:31:18
Ah, that makes a bit more sense. I was reading that as Rawhide is "Alpha for ELN". Not "Rawhide is Alpha" for ELN.
<@conan_kudo:matrix.org>
16:31:27
sorry about that :)
<@sgallagh:fedora.im>
16:31:36
No problem. Words is hard.
<@sgallagh:fedora.im>
16:32:26
So as a first pass, what does that look like? OpenQA results for ELN?
<@conan_kudo:matrix.org>
16:32:33
yeah
<@sgallagh:fedora.im>
16:33:08
I think we may actually have some tests running there, but I don't actually know how to see the results. I don't think they're posted anywhere [that I am subscribed]
<@conan_kudo:matrix.org>
16:33:11
we can start with the cloud and container images
<@conan_kudo:matrix.org>
16:33:24
and then we can engage people for more workloads
<@sgallagh:fedora.im>
16:33:58
Maybe start specifically with the container image first, since that's easiest to replicate on personal systesm
<@sgallagh:fedora.im>
16:34:03
Maybe start specifically with the container image first, since that's easiest to replicate on personal systems
<@conan_kudo:matrix.org>
16:34:40
yup
<@conan_kudo:matrix.org>
16:34:49
and it's the easiest to add tests for
<@conan_kudo:matrix.org>
16:35:15
but the ultimate idea is to make it possible for people to take community enterprise validation to the next level
<@sgallagh:fedora.im>
16:36:17
Bingo!
<@sgallagh:fedora.im>
16:36:19
(sorry)
<@sgallagh:fedora.im>
16:36:49
But I like where this is going. How do we get there?
<@conan_kudo:matrix.org>
16:38:52
I've been mentally thinking about it through the KDE workload lens, and I think the first step is getting the standard composes going so that we can at least manually test them
<@conan_kudo:matrix.org>
16:39:15
then we can start looking at getting a pool of openqa workers and see if we can just run the same tests Fedora runs
<@conan_kudo:matrix.org>
16:39:17
and see how that goes
<@conan_kudo:matrix.org>
16:39:37
I hesitate to go down the road of Bodhi activation, because Bodhi is... special
<@sgallagh:fedora.im>
16:40:00
Bodhi activation in what sense?
<@sgallagh:fedora.im>
16:40:17
We already behave like Rawhide in Bodhi; it's possible to set up blocking tests for stable approbal
<@sgallagh:fedora.im>
16:40:23
We already behave like Rawhide in Bodhi; it's possible to set up blocking tests for stable approval
<@conan_kudo:matrix.org>
16:40:30
but we could think about kicking off things like rpminspect and whatnot through it
<@conan_kudo:matrix.org>
16:40:55
or other kinds of things like warnings about core crossing to extras in the dep graph
<@conan_kudo:matrix.org>
16:41:21
since ELN Core should not cross into extras land
<@yselkowitz:fedora.im>
16:42:05
repoclosure should handle that?
<@sgallagh:fedora.im>
16:42:08
(Side-note: I don't like the phrase "ELN Core", and not just because it sounds like a videogame mid-boss)
<@conan_kudo:matrix.org>
16:42:12
lol
<@conan_kudo:matrix.org>
16:43:10
it depends on what you're looking for, but yeah mostly
<@salimma:fedora.im>
16:43:38
if we can automate filing a bug against the ELN Extras workload maintainer for packages that should be removed, that'd be great
<@conan_kudo:matrix.org>
16:43:50
indeed
<@salimma:fedora.im>
16:44:02
or even just getting a report that is posted, and the workloads and their owners are tagged
<@sgallagh:fedora.im>
16:44:07
Michel Lind 🎩: "should be removed" for what reasons?
<@conan_kudo:matrix.org>
16:44:20
ultimately, I want us to start having reports to maintainers about the stuff they care about
<@salimma:fedora.im>
16:44:30
if package P gets added to "ELN Core", it should not be listed in ELN Extras right?
<@yselkowitz:fedora.im>
16:44:44
that's complicated
<@conan_kudo:matrix.org>
16:44:44
yup
<@salimma:fedora.im>
16:44:48
though... yeah it's tricky
<@salimma:fedora.im>
16:45:08
wearing my company hat I might want to signal "we really need this package, we dont' care if it's in EL or in EPEL"
<@conan_kudo:matrix.org>
16:45:16
but the point is, I think to make it relevant, we need workflows that align better with Fedora maintainers
<@salimma:fedora.im>
16:45:21
the current yaml does not really let us make that distinction
<@carlwgeorge:matrix.org>
16:46:09
i'm pretty sure that was a throwback to the old fedora core/extras split, which we've basically reinvented
<@conan_kudo:matrix.org>
16:46:16
:D
<@sgallagh:fedora.im>
16:46:24
Yes, I'm aware. It still sounds like a videogame enemy :)
<@conan_kudo:matrix.org>
16:46:44
well, this is kind of a boss battle :P
<@sgallagh:fedora.im>
16:48:02
Just to take a moment to preen: the compose that I kicked off just before this meeting completed [almost] successfully!
<@tdawson:fedora.im>
16:48:11
I'm sorta thinking we're blurring the lines between the two topics, meaning we're really crossing into what would make ELN Extras better for epel11.
<@zodbot:fedora.im>
16:48:17
yselkowitz has already given cookies to sgallagh during the F40 timeframe
<@sgallagh:fedora.im>
16:48:33
All of the important things finished, but it looks like I broke something in the reporting pieces.
<@zodbot:fedora.im>
16:48:42
tdawson gave a cookie to sgallagh. They now have 239 cookies, 14 of which were obtained in the Fedora 40 release cycle
<@conan_kudo:matrix.org>
16:48:45
well it's also useful to make it more relevant for Fedora maintainers who have packages going into ELN Core
<@tdawson:fedora.im>
16:49:01
Very true
<@conan_kudo:matrix.org>
16:49:15
I have packages as a Fedora maintainer that go into RHEL, both simple and complex
<@sgallagh:fedora.im>
16:49:17
https://kojipkgs.fedoraproject.org/compose/eln/latest-Fedora-ELN/
<@conan_kudo:matrix.org>
16:49:21
e.g. redhat-fonts and libseccomp :)
<@tdawson:fedora.im>
16:49:38
I wasn't meaning what I said as a bad thing, I was just saying that getting that sort of thing in place will help Extras
<@conan_kudo:matrix.org>
16:49:53
I want it to help both, because what we're doing now can't scale
<@yselkowitz:fedora.im>
16:50:27
so what are the action items here?
<@conan_kudo:matrix.org>
16:50:54
I'd like Stephen Gallagher to tell me what he thinks it would be
<@sgallagh:fedora.im>
16:52:36
Honestly, I think better reporting needs to be the first step. People need to know 1) that their package is part of the ELN or ELN Extras package set and 2) whether it is functional there (FTBFS and FTI are easy first steps, better testing is the necessary follow-in)
<@sgallagh:fedora.im>
16:52:57
Honestly, I think better reporting needs to be the first step. People need to know 1) that their package is part of the ELN or ELN Extras package set and 2) whether it is functional there (FTBFS and FTI are easy first steps, better testing is the necessary follow-on)
<@carlwgeorge:matrix.org>
16:53:49
is there any kind of plan for https://github.com/fedora-eln/eln/issues/194 (cr tracking buildreqs) to get resolved?
<@sgallagh:fedora.im>
16:53:53
The compose delta is probably a good start.
<@sgallagh:fedora.im>
16:54:23
Carl George: I'll point you to yselkowitz for that; he's taken ownership of Content Resolver.
<@yselkowitz:fedora.im>
16:55:32
right now, PTC
<@sgallagh:fedora.im>
16:55:39
Though if I could turn my boss upside down and shake the money from his pockets, I'd love to contract Conan Kudo to help with a major refactor there. He's one of like three people on the planet that knows how to use libsolv :)
<@conan_kudo:matrix.org>
16:56:00
😂
<@sgallagh:fedora.im>
16:56:00
PTC?
<@yselkowitz:fedora.im>
16:56:03
https://cygwin.com/acronyms/#PTC
<@sgallagh:fedora.im>
16:57:00
For the sake of the meeting record: PTC == "Patches Thoughtfully Considered".
<@carlwgeorge:matrix.org>
16:58:58
i'm concerned about the viability of starting epel11 early if cr buildreq support is in "patches welcome" mode
<@tdawson:fedora.im>
16:59:41
Well, CR does what it was originally written for. The problem is that Extras was not on the original manifest.
<@yselkowitz:fedora.im>
17:00:10
I do believe it's possible but it's not my codebase, so it's going to take time to figure out how to improve it
<@sgallagh:fedora.im>
17:00:24
I'm sure there's a good reason, but why aren't we just processing it as if it was EVERYTHING and then just subtracting the stuff that's in ELN *cough* core *cough* from the total result?
<@yselkowitz:fedora.im>
17:00:27
besides this kind of tooling work isn't really my specialty
<@yselkowitz:fedora.im>
17:00:46
runtime and build deps are handled separately
<@sgallagh:fedora.im>
17:00:48
I mean, it's not efficient, but it might be effective
<@carlwgeorge:matrix.org>
17:00:54
i think there is still the philosophical difference i brought up at the hackfest, that cr was designed around reducing content, and extras/epel is about expanding content
<@yselkowitz:fedora.im>
17:01:10
it's not just about reducing though
<@carlwgeorge:matrix.org>
17:01:32
but it was designed around that goal, hence the minimization name in so many places
<@sgallagh:fedora.im>
17:01:38
Hmm, I just noticed we're over time. I can continue, but if others have places to be we can put a pin in this here.
<@yselkowitz:fedora.im>
17:01:51
it's very good for that, but it's also quite useful in anything where dependency resolution is needed
<@carlwgeorge:matrix.org>
17:02:13
except build dependencies apparently
<@carlwgeorge:matrix.org>
17:02:21
does the cr codebase have an "owner"?
<@yselkowitz:fedora.im>
17:02:45
it's me now, for better or for worse
<@tdawson:fedora.im>
17:02:57
I'm also shared owner.
<@sgallagh:fedora.im>
17:03:06
Carl George: Builddeps work fine for primary stuff like "ELN" and "CentOS Stream"
<@sgallagh:fedora.im>
17:03:11
It's the "addons" that have some issues.
<@carlwgeorge:matrix.org>
17:03:19
ah, so some nuance
<@sgallagh:fedora.im>
17:04:07
Right, which is why I was wondering if we could just take the slow-but-effective route of just processing "ELN+Extras" as a "primary" set and then just subtracting from the results the set of things in "ELN" only.
<@salimma:fedora.im>
17:04:45
that will help when things are in flux I guess
<@conan_kudo:matrix.org>
17:05:29
which is pretty much all the time with ELN
<@conan_kudo:matrix.org>
17:05:50
the divisions between "core" and "extras" are fuzzy and shifting by design
<@conan_kudo:matrix.org>
17:06:42
after all, I expect ELN definitions to update as content changes in released RHEL too
<@sgallagh:fedora.im>
17:09:28
OK, we're over time. Do we have any specific actions to call out today or shall we resume this in the #eln:fedoraproject.org channel (or at the meeting next week)?
<@davide:cavalca.name>
17:10:14
my only request is to write down the conclusion in the docs once we've reached consensus
<@yselkowitz:fedora.im>
17:10:22
at next meeting please
<@conan_kudo:matrix.org>
17:10:25
absolutely
<@tdawson:fedora.im>
17:11:12
next meeting
<@sgallagh:fedora.im>
17:12:35
Works for me.
<@sgallagh:fedora.im>
17:12:43
OK, thank you for coming, everyone.
<@sgallagh:fedora.im>
17:12:46
!endmeeting