<@james:fedora.im>
16:00:35
!startmeeting fpc
<@james:fedora.im>
16:00:38
!topic Roll Call
<@meetbot:fedora.im>
16:00:38
Meeting started at 2024-09-19 16:00:35 UTC
<@meetbot:fedora.im>
16:00:38
The Meeting name is 'fpc'
<@limb:fedora.im>
16:00:46
!hi
<@zodbot:fedora.im>
16:00:48
Gwyn Ciesla (limb) - she / her / hers
<@conan_kudo:matrix.org>
16:01:10
!hi
<@james:fedora.im>
16:01:12
!hi
<@zodbot:fedora.im>
16:01:14
James Antill (james)
<@zodbot:fedora.im>
16:01:29
Neal Gompa (ngompa) - he / him / his
<@decathorpe:fedora.im>
16:01:32
!ji
<@decathorpe:fedora.im>
16:01:38
*sigh*
<@decathorpe:fedora.im>
16:01:41
!hi
<@zodbot:fedora.im>
16:01:44
Fabio Valentini (decathorpe) - he / him / his
<@james:fedora.im>
16:01:50
ja
<@carlwgeorge:matrix.org>
16:05:22
!hi
<@zodbot:fedora.im>
16:05:24
Carl George (carlwgeorge) - he / him / his
<@james:fedora.im>
16:05:34
And then there were 5!
<@salimma:fedora.im>
16:05:55
!hi
<@zodbot:fedora.im>
16:05:57
Michel Lind (salimma) - he / him / his
<@james:fedora.im>
16:11:54
!topic FPC#1049 https://pagure.io/packaging-committee/issue/1049
<@james:fedora.im>
16:12:16
Someone volunteered to comment on this last week ... I can't remember who though (hopefully it wasn't me).
<@decathorpe:fedora.im>
16:12:26
yeah I put writing guidelines for this onto my todo list last weke
<@decathorpe:fedora.im>
16:12:36
weke weke eh eh ...
<@james:fedora.im>
16:12:50
Ahh, okay. Just making sure it wasn't forgotten.
<@decathorpe:fedora.im>
16:13:00
no, officially on my list :)
<@carlwgeorge:matrix.org>
16:13:38
oh, the dreaded list
<@conan_kudo:matrix.org>
16:13:59
I volunteered to port over the generators to fedora
<@james:fedora.im>
16:14:07
Okay, going to pull up some different old things now...
<@conan_kudo:matrix.org>
16:14:19
honestly this has seriously irritated me for years, so I'm happy we want to solve this now :)
<@james:fedora.im>
16:14:20
!topic FPC PR#912 https://pagure.io/packaging-committee/pull-request/912
<@conan_kudo:matrix.org>
16:14:29
we're the only rpm distro without the generator
<@james:fedora.im>
16:15:46
This looks fine to me, but doesn't merge now :(
<@carlwgeorge:matrix.org>
16:17:29
at a glance, seems to be a semantic line break conflict, possible could be easy to rebase to resolve the conflicts
<@james:fedora.im>
16:18:18
Yeh, was hoping it'd be something like that.
<@carlwgeorge:matrix.org>
16:18:48
doesn't quite feel right to reply to a 5yo pr saying "hey can you rebase again", so perhaps we should just manually do it
<@carlwgeorge:matrix.org>
16:19:18
i noticed in fedpkg that you can do "Merges: <pr link>" in a commit message, and it will mark the pr as merged in pagure
<@salimma:fedora.im>
16:19:44
Oh nice
<@james:fedora.im>
16:19:47
Yeh, I know I can do it and push ... I'm less sure how to update the PR
<@james:fedora.im>
16:20:16
That's cool.
<@james:fedora.im>
16:22:11
!topic FPC PR#1097 https://pagure.io/packaging-committee/pull-request/1097
<@carlwgeorge:matrix.org>
16:22:19
i can take a stab at it, just need to move my versioning revamp aside temporarily
<@carlwgeorge:matrix.org>
16:23:08
i don't think pagure has "allow maintainers to push to my fork", so if i can resolve bex's pr are we ok just pushing it to main?
<@james:fedora.im>
16:23:12
Another usage of ^ in versions ... maybe? I'm less sure why it's better than what we did, so tempted to just close.
<@james:fedora.im>
16:23:24
I am
<@conan_kudo:matrix.org>
16:24:05
it does, but you have to enable it when you create the PR
<@carlwgeorge:matrix.org>
16:24:26
did it 5 yrs ago when this pr was opened? lol
<@conan_kudo:matrix.org>
16:24:30
no :)
<@carlwgeorge:matrix.org>
16:24:48
where's that time machine when you need it
<@conan_kudo:matrix.org>
16:25:16
if only :)
<@carlwgeorge:matrix.org>
16:25:19
alright i'm done derailing, back to 1097
<@limb:fedora.im>
16:27:04
Apologies, I have to bail. PIng me for ticket votes.
<@james:fedora.im>
16:27:54
No problem, see you next week.
<@nirik:matrix.scrye.com>
16:28:50
hey, I had a thing for open floor real quick...
<@james:fedora.im>
16:28:56
Keep Re-Reading the patch and comments and I'm not sure what it does better. Also seeing "foo < 1.0^" makes me want to hurt people.
<@james:fedora.im>
16:29:14
nirik: Sure
<@nirik:matrix.scrye.com>
16:29:42
wanted to make you all aware if you were not about https://pagure.io/fesco/issue/3269 ( Re-evaluate ban on pre-compiled CSS ) if you wanted to weigh in, please do on ticket.
<@nirik:matrix.scrye.com>
16:31:19
thats it.
<@carlwgeorge:matrix.org>
16:32:22
seems reasonable to me
<@james:fedora.im>
16:32:26
Cool. I wasn't aware of it, but it seems reasonable to not be more strict about CSS than JS ... and all the comments make me want to agree faster.
<@carlwgeorge:matrix.org>
16:34:03
i have a new thing too for open floor
<@james:fedora.im>
16:36:45
Cool ... I just commented and closed 1097 ... so go for it.
<@decathorpe:fedora.im>
16:36:56
!topic Open Floor
<@zodbot:fedora.im>
16:37:21
james gave a cookie to decathorpe. They now have 106 cookies, 18 of which were obtained in the Fedora 40 release cycle
<@carlwgeorge:matrix.org>
16:38:12
looking at rawhide, we have both llvm (version 18) and llvm18. this feels wrong, but i don't see any guidelines forbidding it. i think we need a consistent policy for this.
<@decathorpe:fedora.im>
16:39:09
we do? ah, the v19 builds are still in a side-tag
<@conan_kudo:matrix.org>
16:39:18
yeah v19 hasn't landed yet
<@carlwgeorge:matrix.org>
16:39:23
it feels like they were trying to emulate python's versioned srpms, but didn't finish retiring the unversioned package
<@conan_kudo:matrix.org>
16:39:47
the main llvm package is in the process of upgrading to v19
<@conan_kudo:matrix.org>
16:39:57
they pushed v19.1.0 this morning
<@carlwgeorge:matrix.org>
16:40:11
and then the problem repeats with llvm19
<@conan_kudo:matrix.org>
16:40:25
yes, that is generally how we expect compatibility transitions to work though
<@decathorpe:fedora.im>
16:40:37
well, ideally the compat package and the new version would land in the same side-tag update
<@conan_kudo:matrix.org>
16:40:37
I did this when we shifted to appstream v1.0 in F40
<@decathorpe:fedora.im>
16:40:53
but it doesn't hurt (much) if they don't
<@carlwgeorge:matrix.org>
16:40:54
then i would argue llvm18 should have also been in the side tag
<@conan_kudo:matrix.org>
16:41:05
sure i guess?
<@carlwgeorge:matrix.org>
16:41:34
f41 is in the same state, with llvm (18) and llvm18
<@decathorpe:fedora.im>
16:41:50
llvm 19 will be pushed to both
<@decathorpe:fedora.im>
16:41:57
it's an approved F41 Change
<@conan_kudo:matrix.org>
16:42:00
right
<@decathorpe:fedora.im>
16:42:03
and LLVM release schedule is *always* bad
<@conan_kudo:matrix.org>
16:42:12
and FESCo has also accepted that LLVM will always be late
<@james:fedora.im>
16:42:30
The idea being that if llvm19 doesn't work then people can buildreq llvm18 directly ... I guess I don't mind llvm18 landing early, so that people can premove their buildreqs if they know 19 is going to be a problem.
<@conan_kudo:matrix.org>
16:42:39
right
<@conan_kudo:matrix.org>
16:42:55
we actually kind of lucked out this time, the schedule slippage meant that llvm19 is landing outside of a freeze window
<@conan_kudo:matrix.org>
16:43:18
normally it's kind of a mess because it lands right when final freeze kicks on
<@carlwgeorge:matrix.org>
16:43:40
in those cases i would recommend people use `llvm-devel(major) < 19`
<@carlwgeorge:matrix.org>
16:43:59
that would work before and after the llvm 19 sidetag is merged
<@james:fedora.im>
16:44:06
I guess llvm should provide llvm18 until it becomes llvm19 ... but, meh. I'm really don't care enough about rawhide, and I'm not sure I care if it leaks into a release either.
<@decathorpe:fedora.im>
16:44:38
I'm pretty sure we won't release F41 until llvm 19 has landed in it
<@carlwgeorge:matrix.org>
16:44:40
my main concern is it being exposed to users and them not knowing which one to install
<@conan_kudo:matrix.org>
16:44:46
I'm not sure it's really worth it to be overly prescriptive here either
<@carlwgeorge:matrix.org>
16:45:01
it helps that the intent is to have the unversioned one be the latest, and versioned older ones
<@decathorpe:fedora.im>
16:45:04
true, but still, this *is* beta and / or development branch.
<@conan_kudo:matrix.org>
16:45:16
for all the faults around llvm upgrades, they do a relatively decent job of doing the transition
<@decathorpe:fedora.im>
16:46:02
I mean, it won't hurt to ask them to push the compat packages in the same side-tag that they use for the main upgrade ...
<@carlwgeorge:matrix.org>
16:46:07
ok, without focusing too much on llvm specifics, do we want a policy about not doing two variants of the same version like this?
<@decathorpe:fedora.im>
16:46:08
(for next time, I mean)
<@james:fedora.im>
16:46:15
Carl George: Can you ping them privately and be like "hey, it'd be nice if you did it this way" and see what they say?
<@conan_kudo:matrix.org>
16:46:34
I don't think so.
<@carlwgeorge:matrix.org>
16:46:46
i don't know any of the maintainers directly, but i can send an email
<@conan_kudo:matrix.org>
16:46:47
It makes transitions too hard if they need to be done in multiple phases.
<@decathorpe:fedora.im>
16:46:58
yeah, I don't think we need to forbid any harmless temporary state
<@carlwgeorge:matrix.org>
16:47:07
that seems fine with a SHOULD
<@james:fedora.im>
16:47:13
I think if it stays in rawhide (is temporary) I'm happy to accept it if it's easier for the maintainers
<@conan_kudo:matrix.org>
16:47:37
if it's in a development branch and won't be part of a frozen GA repo, I'm fine with it
<@carlwgeorge:matrix.org>
16:48:31
let me give you a hypothetical. let's say some reckless rhel maintainers wanted to have llvm18 in fedora as just a build test for rhel, not just in dev branches, and wanted fedora users to ignore it and use the unversioned llvm. would we allow them to co-exist?
<@carlwgeorge:matrix.org>
16:48:58
let me give you a hypothetical. let's say some reckless rhel maintainers wanted to have llvm19 in fedora as just a build test for rhel, not just in dev branches, and wanted fedora users to ignore it and use the unversioned llvm. would we allow them to co-exist?
<@decathorpe:fedora.im>
16:49:35
well, that's something entirely different then
<@carlwgeorge:matrix.org>
16:49:53
honestly that's what this looked like to me at first glance
<@carlwgeorge:matrix.org>
16:50:09
i wasn't up to speed on the llvm schedule or the change proposal, so that context helps, but still
<@decathorpe:fedora.im>
16:50:11
well, it's not, though?
<@carlwgeorge:matrix.org>
16:50:30
to a user not following the change proposals, it does look like that
<@decathorpe:fedora.im>
16:50:34
is the only problem that the compat package and the update didn't land at the same time / in the same bodhi update?
<@carlwgeorge:matrix.org>
16:51:15
i do think that would solve it, yeah. the problem is the confusion about two llvm packages that are version 18, and landing them together would solve that.
<@carlwgeorge:matrix.org>
16:53:22
it feels like an obvious thing to not have doubles (outside of transition periods), but i'm thinking a small clarification about it in https://docs.fedoraproject.org/en-US/packaging-guidelines/Naming/#multiple would be worthwhile
<@james:fedora.im>
16:56:06
I guess I'm fine saying you should build foo<compat> in the side tag with foo-<new>, and change them at the same time to avoid confusion.
<@james:fedora.im>
16:56:56
Or better words than that, that mean that.
<@carlwgeorge:matrix.org>
16:57:17
i'll work up something as a pr, and we can nit pick the wording then
<@conan_kudo:matrix.org>
16:57:21
sure
<@carlwgeorge:matrix.org>
16:57:50
speaking of prs i owe y'all, i'm not done with the snapshot one, but i have a preview if y'all want to see it
<@carlwgeorge:matrix.org>
16:57:54
https://carlwgeorge.fedorapeople.org/docs/packaging-committee/packaging-guidelines/Versioning/#_snapshots
<@carlwgeorge:matrix.org>
16:58:25
still need to fixup the examples, and move the traditional/legacy stuff to an appendix page
<@conan_kudo:matrix.org>
16:58:38
right off the bat, I'm firmly against dropping the VCS identifier from the snapshot guidelines
<@conan_kudo:matrix.org>
16:59:13
the rest of it, I'll look at later, but I really don't think it's a good idea to drop the VCS
<@carlwgeorge:matrix.org>
16:59:17
and we've discussed that in https://pagure.io/packaging-committee/pull-request/1380
<@conan_kudo:matrix.org>
16:59:34
and that was not concluded
<@carlwgeorge:matrix.org>
16:59:56
do we even know how many packages in fedora have upstreams that aren't git?
<@conan_kudo:matrix.org>
17:00:32
that is not to point
<@carlwgeorge:matrix.org>
17:00:34
having extra characters for the 0.01% edge case is not really appealing to me
<@conan_kudo:matrix.org>
17:00:35
that is not the point
<@conan_kudo:matrix.org>
17:00:50
if we're going to be pedantic about snapshot versioning, this is a hill that I really do care about
<@carlwgeorge:matrix.org>
17:01:22
please don't belittle this as pedantry, it's just the nature of writing guidelines
<@conan_kudo:matrix.org>
17:01:40
this whole thing came out of pedantry
<@decathorpe:fedora.im>
17:01:47
fwiw, I would also prefer the `git` / `svn` / `hg` string to stay
<@carlwgeorge:matrix.org>
17:02:33
no, it came out of the fact that we officially allow like four different formats, and there are others in the wild
<@conan_kudo:matrix.org>
17:03:18
you complained about my format that I used because I wanted the same convention across packages I maintain across 4 distributions
<@decathorpe:fedora.im>
17:03:33
then we just ban the ones without SCM identifier? 😉
<@conan_kudo:matrix.org>
17:03:33
and that we used to have in our guidance too
<@carlwgeorge:matrix.org>
17:04:11
the format you wanted to use isn't allowed by our current guidelines, so yes i pointed out that fact
<@carlwgeorge:matrix.org>
17:04:44
keeping things consistent with other distributions is not, and imo should not be, a goal for fedora guidelines
<@james:fedora.im>
17:04:44
Maybe say only use the VCS identifier for non-git upstreams?
<@carlwgeorge:matrix.org>
17:04:58
i don't mind that as a compromise
<@conan_kudo:matrix.org>
17:05:04
I don't think it makes sense to omit it ever
<@conan_kudo:matrix.org>
17:05:12
even for git
<@conan_kudo:matrix.org>
17:05:25
it makes it clear at a glance that it's a git snapshot
<@carlwgeorge:matrix.org>
17:06:03
so does `^` or `~` without a label like `rc`, `beta`, etc
<@carlwgeorge:matrix.org>
17:06:35
in fact, that's the only MUST in our current guidelines about snapshots is to use the `^`
<@decathorpe:fedora.im>
17:06:55
yeah, and that's likely a mistake from when they were merged ...
<@carlwgeorge:matrix.org>
17:07:31
no, the mistake was not allowing `~`, but the point is that scm labels are not MUST
<@decathorpe:fedora.im>
17:07:57
they're not, but any reasonable person would add them
<@carlwgeorge:matrix.org>
17:08:12
so i guess i'm not a reasonable person
<@decathorpe:fedora.im>
17:08:42
I don't understand why this is the hill you want to defend
<@carlwgeorge:matrix.org>
17:09:15
i need a break, and we're over time anyways. let's put a pin in this.
<@james:fedora.im>
17:09:22
Given the proportion of the usage of git, I'm fine assuming it's git unless told otherwise.
<@james:fedora.im>
17:09:22
But I don't want to die on either hill.
<@james:fedora.im>
17:09:22
Also fine assuming no label means something like "beta" or "rc" depending on where the package is.
<@james:fedora.im>
17:09:39
Oh, yeh, didn't notice it was past the hour. My bad.
<@james:fedora.im>
17:10:03
See you all next week.
<@james:fedora.im>
17:10:08
!endmeeting