#meeting-1:fedoraproject.org log

<@james:fedora.im>

17:00:04

!startmeeting fpc

<@james:fedora.im>

17:00:05

!topic Roll Call

<@meetbot:fedora.im>

17:00:05

Meeting started at 2025-01-09 17:00:04 UTC

<@meetbot:fedora.im>

17:00:05

The Meeting name is 'fpc'

<@salimma:fedora.im>

17:00:09

!hi

<@zodbot:fedora.im>

17:00:10

Michel Lind (salimma) - he / him / his

<@james:fedora.im>

17:00:44

!hi

<@zodbot:fedora.im>

17:00:45

James Antill (james)

<@limb:fedora.im>

17:01:01

!hi

<@zodbot:fedora.im>

17:01:03

Gwyn Ciesla (limb) - she / her / hers

<@james:fedora.im>

17:01:55

I know a few people who are still on vacation, so I'd guess we probably won't get five. But hi anyway ;)

<@limb:fedora.im>

17:03:23

🤷‍♀️

<@conan_kudo:matrix.org>

17:04:33

!hi

<@zodbot:fedora.im>

17:04:35

Neal Gompa (ngompa) - he / him / his

<@james:fedora.im>

17:10:20

!topic FPC PR#1429 New guidelines for Tree-sitter parsers

<@james:fedora.im>

17:10:27

!link https://pagure.io/packaging-committee/pull-request/1429

<@james:fedora.im>

17:10:52

I have no idea what this is ... but maybe just merge it?

<@salimma:fedora.im>

17:13:11

taking another pass to jog my memory

<@salimma:fedora.im>

17:14:01

This is something used by editors and diff tools IIRC. e.g. vim, or difftastic

<@salimma:fedora.im>

17:14:30

having a better way to package sounds nice, so no objection - I know difftastic currently bundles some tree parsers and it would be nice to be able to unbundle

<@james:fedora.im>

17:15:58

okay, I'm going to merge

<@james:fedora.im>

17:16:34

!topic FPC PR#1426 Create relative not absolute symlink to binary in Node.js example

<@james:fedora.im>

17:16:42

!link https://pagure.io/packaging-committee/pull-request/1426

<@james:fedora.im>

17:18:08

This seems ... good, although it doesn't merge. Relative symlinks being heavily prefered should be in the policy, right?

<@limb:fedora.im>

17:18:22

That was my memory....

<@decathorpe:fedora.im>

17:18:51

!hi

<@zodbot:fedora.im>

17:18:52

Fabio Valentini (decathorpe) - he / him / his

<@decathorpe:fedora.im>

17:18:58

sorry, missed my notifications.

<@salimma:fedora.im>

17:19:32

I remember being confused about whether to use relative or absolute for one package (then deciding on relative) and IIRC when I checked the policy did say relative

<@decathorpe:fedora.im>

17:20:15

rpmlint will also yell at you if you use absolute symlinks, IIRC

<@decathorpe:fedora.im>

17:20:35

ah, it's a warning from rpm itself too.

<@tibbs:fedora.im>

17:20:43

Hey, Looks like I'm super late.

<@tibbs:fedora.im>

17:20:46

Let me scroll back.

<@james:fedora.im>

17:22:48

!topic FPC PR#1412 Document one-to-many replacement

<@james:fedora.im>

17:22:53

!link https://pagure.io/packaging-committee/pull-request/1412

<@james:fedora.im>

17:23:34

I think I had a look at this towards the end of last year, and was confused we didn't already have similar text ... but I couldn't find it, if we do.

<@tibbs:fedora.im>

17:25:33

I think it's kind of a long way to say a simple thing but it is correct that we need to say it somewhere.

<@james:fedora.im>

17:25:54

I might be remembering someone trying to document 666 different reasons/ways why you'd use obsoletes and us rejecting that ... but this seems fine?

<@james:fedora.im>

17:26:26

Yeh, it seemed a bit long but eh.

<@salimma:fedora.im>

17:26:35

The example seems concise enough

<@salimma:fedora.im>

17:27:16

I'd prefer less vague language over the behavior under DNF 5, but as long as the recommended approach does the right thing on both, this looks fine

<@salimma:fedora.im>

17:28:06

nit: should we recommend that obsoletes be commented with the release at which point they can be removed? that might be orthogonal to this PR though

<@tibbs:fedora.im>

17:28:40

It's true and perhaps unfortunate that a bunch of our guidelines are really just codifying the rougher edges of RPM and now two versions of DNF.

<@decathorpe:fedora.im>

17:30:20

well .. if RPM (and upstream build tools) would only ever let you do the *one correct thing* (TM), we'd need zero guidelines :D

<@james:fedora.im>

17:30:42

Christmas just passed

<@decathorpe:fedora.im>

17:31:00

bummer 😄

<@tibbs:fedora.im>

17:31:13

Well we do have the capacity to do a good bit of buil-time enforcement....

<@james:fedora.im>

17:32:13

We've got better, for sure. Like the relative symlink thing.

<@salimma:fedora.im>

17:32:22

Dear Santa...

<@tibbs:fedora.im>

17:32:33

Anyway, I suggest merging this and then perhaps considering splitting the section out to a separate section.

<@salimma:fedora.im>

17:32:46

+1 on merging

<@decathorpe:fedora.im>

17:32:53

<@tibbs:fedora.im>

17:33:15

Sometimes I think it would be nice if the main guidelines page was just an introduction and links to individual sections.

<@salimma:fedora.im>

17:33:42

that would make it easier to maintain I guess

<@tibbs:fedora.im>

17:33:44

Maybe we should just say that if you need subsections, you should consider just moving to a separate document.

<@james:fedora.im>

17:34:29

!topic Open Floor

<@decathorpe:fedora.im>

17:35:14

this PR seems to be stuck in limbo despite being tagged meeting: https://pagure.io/packaging-committee/pull-request/1333

<@decathorpe:fedora.im>

17:35:27

I promised to bring it up in the first meeting after the holidays :)

<@salimma:fedora.im>

17:36:37

oh fun

<@james:fedora.im>

17:37:12

I commented ... not sure if we talked about it in a meeting.

<@decathorpe:fedora.im>

17:37:45

I mean, given that guidelines don't apply retroactively, this *might* be fine, as an aspirational thing, but I don't think it should be required of packagers to migrate stuff, since that certainly isn't trivial

<@tibbs:fedora.im>

17:37:54

I'm kind of on the same side as Neil here; I just can't see banning this.

<@james:fedora.im>

17:37:56

But I don't think we can just say yes, even if we'd want to, without a few changes.

<@salimma:fedora.im>

17:38:00

Jason ティビツ: looks like you recommended this go through FESCo last time? 11 months ago

<@tibbs:fedora.im>

17:38:08

But I can see "recommending" it somehow.

<@salimma:fedora.im>

17:39:24

yeah

<@salimma:fedora.im>

17:40:09

maybe there needs to be a disclaimer that packagers are not expected to implement this, but should accept PRs to implement this if immutable folks provide such and can demonstrate they are safe to apply?

<@tibbs:fedora.im>

17:40:18

I think it just needs to be weak, as in a statement that "doing this will cause problems for some Fedora derivatives; you can do this instead, but are not required to do so".

<@salimma:fedora.im>

17:40:27

idk if such language should be part of packaging guidelines or not, but that seems to be the only concern we have left

<@salimma:fedora.im>

17:40:36

oh that sounds good

<@tibbs:fedora.im>

17:40:46

I'm sure calling ostree a derivative will anger someone, though.

<@decathorpe:fedora.im>

17:40:57

"deliverables" :)

<@tibbs:fedora.im>

17:40:58

It just feels like the tail is really trying to wag the dog pretty hard here.

<@tibbs:fedora.im>

17:41:32

On the other hand, it does mesh with the whole concept of scriptlets being undesirable for other reasons.

<@salimma:fedora.im>

17:42:15

hehe

<@salimma:fedora.im>

17:42:31

yeah, I'm in favor of minimizing the use of scriptlets in general

<@decathorpe:fedora.im>

17:42:34

yeah. it's just that in this case, this might often require quite extensive changes to avoid "the bad thing"

<@decathorpe:fedora.im>

17:43:02

and I don't think we can *require* that. we can recommend it though, if it's possible

<@tibbs:fedora.im>

17:45:36

require, recommend, suggest, mention.

<@tibbs:fedora.im>

17:47:11

Regardless, it doesn't hurt to mention the problem and then suggest an alternative, but even recommending that people do this when it could require significant work to get it right just seems too much.

<@tibbs:fedora.im>

17:47:52

But if we have it written down somewhere then the osfree folks who want to do the work can do so and then point to the guidelines section.

<@decathorpe:fedora.im>

17:48:44

so is that ... somewhere between SHOULD and MAY?

<@decathorpe:fedora.im>

17:50:40

hm. according to [RFC2119](https://www.rfc-editor.org/rfc/rfc2119), SHOULD is equivalent to RECOMMENDED

<@salimma:fedora.im>

17:51:31

we can also just reuse the keywords from RPM

<@james:fedora.im>

17:51:36

I guess I'm fine adding something saying "don't do this if you don't have to, and if you need help go speak to these people"

<@salimma:fedora.im>

17:51:39

recommend, then suggest as a weaker form

<@james:fedora.im>

17:51:47

But I'm not sure it really fixes the issue.

<@tibbs:fedora.im>

17:52:11

We can say "here is the problem. It can be avoided by doing this, but you are not required to do so".

<@tibbs:fedora.im>

17:52:29

That explains it and gets it documented, but doesn't force anything on anyone.

<@conan_kudo:matrix.org>

17:53:26

A practice as opposed to a rule is fine

<@conan_kudo:matrix.org>

17:53:32

but saying things aren't allowed is not

<@conan_kudo:matrix.org>

17:53:47

it's unreasonable and unworkable

<@tibbs:fedora.im>

17:54:28

We might get down to stricter rules around scriptlets in general in a number of years but I still don't think now is the time.

<@limb:fedora.im>

17:58:13

Sorry if I'm in and out. My ISP is being hot garbage today.

<@james:fedora.im>

18:00:07

We only have 1 minute left ... anyone want to try wording for 1333?

<@james:fedora.im>

18:00:15

0 minutes now ;)

<@soupcreator:matrix.org>

18:00:56

Cool

<@decathorpe:fedora.im>

18:01:29

I already left a comment saying basically that, I don't want to repeat myself

<@james:fedora.im>

18:01:57

Was meaning more like a PR ;)

<@james:fedora.im>

18:02:25

But I'll let everyone think on it for a week, and see if anyone wants to volunteer then.

<@james:fedora.im>

18:02:30

!endmeeting