<@james:fedora.im>
16:00:14
!startmeeting fpc
<@meetbot:fedora.im>
16:00:15
Meeting started at 2025-04-24 16:00:14 UTC
<@meetbot:fedora.im>
16:00:15
The Meeting name is 'fpc'
<@james:fedora.im>
16:00:17
!topic Roll Call
<@decathorpe:fedora.im>
16:00:32
!hi
<@zodbot:fedora.im>
16:00:33
Fabio Valentini (decathorpe) - he / him / his
<@limb:fedora.im>
16:00:36
!hi
<@zodbot:fedora.im>
16:00:37
Gwyn Ciesla (limb) - she / her / hers
<@tibbs:fedora.im>
16:00:40
Hello.
<@james:fedora.im>
16:00:52
!hi
<@zodbot:fedora.im>
16:00:53
James Antill (james)
<@conan_kudo:matrix.org>
16:01:06
!hi
<@zodbot:fedora.im>
16:01:08
Neal Gompa (ngompa) - he / him / his
<@salimma:fedora.im>
16:03:06
!hi
<@zodbot:fedora.im>
16:03:07
Michel Lind (salimma) - he / him / his
<@james:fedora.im>
16:07:58
Okay, today looks to be pretty unhinged. So hold on...
<@james:fedora.im>
16:08:09
!topic FPC PR#1454 https://pagure.io/packaging-committee/pull-request/1454
<@james:fedora.im>
16:08:59
So ... not much for us to do right now, I think. But it's pretty funny that Panu responded and it's basically "Yeh, don't do what we did before or what you are proposing."
<@decathorpe:fedora.im>
16:09:02
meh.
<@decathorpe:fedora.im>
16:09:38
surprise surprise, turns out the solution is "use what fits your case best" ...
<@salimma:fedora.im>
16:09:52
oh no
<@decathorpe:fedora.im>
16:09:52
it would help if there were better documentation for all this
<@tibbs:fedora.im>
16:10:16
Yes, I think when RPM documentation is lacking, people look to the packaging guidelines.
<@salimma:fedora.im>
16:10:48
right
<@tibbs:fedora.im>
16:10:50
I only vaguely remember when we initially put this "global > define" thing in there.
<@conan_kudo:matrix.org>
16:11:04
it would help if there were documentation about this
<@salimma:fedora.im>
16:11:07
if the upstream would just say "this is what global means, this is what define means, and these are where it makes sense to use each"
<@salimma:fedora.im>
16:11:28
and if the semantics change between RPM versions it would be really nice to know too
<@james:fedora.im>
16:11:39
Yeh, long time ago ... and I'm guessing it was around the time we started having python variables for the python version, which subshell'd out to get it.
<@decathorpe:fedora.im>
16:11:45
yes, this is essentially what I think we need.
<@tibbs:fedora.im>
16:12:14
But the real right place for this is in RPM's documentation, not our guidelines. Unless we want to condense it and make a recommendation.
<@salimma:fedora.im>
16:12:43
definitely
<@decathorpe:fedora.im>
16:12:54
I still don't entirely understand why `%global` is suddenly considered "bad" - it might be more strict than actually necessary, but it doesn't *break* things?
<@salimma:fedora.im>
16:13:17
if they don't have docs ... should we ... submit what we think is the proper way to use them to the RPM docs, and let them nitpick that PR?
<@salimma:fedora.im>
16:13:40
yeah, I don't understand either, I thought we should use global by default and use define only if global does not suffice
<@tibbs:fedora.im>
16:13:40
I think if it broke things we wouldn't have much of a distro. The problem really is that we make a recommendation when we shouldn't be making any recommendation.
<@james:fedora.im>
16:13:43
I think it's just a response to the previous recommendation of "always use global" and that meaning that people were doing the "global + %{expand: ... }
<@tibbs:fedora.im>
16:14:07
I think we're better off saying nothing at all instead of having what we have now.
<@conan_kudo:matrix.org>
16:14:17
that probably makes sense to consider
<@james:fedora.im>
16:14:21
And now maybe people aren't subshell'ing out much, so "use define most of the time" is probably fine.
<@tibbs:fedora.im>
16:14:54
You could be right. I really can't remember who drove the initial change, but RPM was a black box then.
<@decathorpe:fedora.im>
16:15:07
I don't get why `"global + %{expand: ... }` is bad either?
<@james:fedora.im>
16:15:32
Conan Kudo: You do patches for dnf/etc. ... what do you think of submitting an rpm docs patch?
<@james:fedora.im>
16:16:00
Fabio Valentini: AIUI %global + %expand is just %define spelled in a really weird way.
<@james:fedora.im>
16:16:19
Or mostly %define, with some weird corner case differences.
<@decathorpe:fedora.im>
16:16:28
I only use %expand as syntactic sugar so I don't need to escape newlines
<@salimma:fedora.im>
16:16:56
same
<@conan_kudo:matrix.org>
16:17:05
I think that's a good idea. Are you asking me to write something without asking it outright? :P
<@conan_kudo:matrix.org>
16:17:39
there's actually a decent chunk of our guidelines that are better described as "rpm docs in fedora" too
<@conan_kudo:matrix.org>
16:17:53
it is worth considering whether some of that can be upstreamed
<@james:fedora.im>
16:17:55
Was mostly leaning on asking it outright tbf ... "what do you think about you writting/submitting an rpm docs patch?
<@james:fedora.im>
16:18:46
FWIW I think it'd be useful to have something in our docs ... but if that can be "here is a pointer to rpm docs on how this works", that'd be great.
<@james:fedora.im>
16:19:28
Then rpm can break their online docs again and we can get 666 tickets about it ;)
<@salimma:fedora.im>
16:20:22
666 seems apt
<@salimma:fedora.im>
16:20:42
a very yummy number... ok do not force me to continue this tortured analogy
<@james:fedora.im>
16:21:31
!info Neal "volunteered" to look at patching upstream rpm docs.
<@james:fedora.im>
16:21:57
!topic FPC#1456 https://pagure.io/packaging-committee/issue/1456
<@conan_kudo:matrix.org>
16:22:20
lol ok
<@james:fedora.im>
16:22:50
This issue makes me ¯\\_(ツ)_/¯
<@salimma:fedora.im>
16:22:51
voluntold :)
<@salimma:fedora.im>
16:22:54
sorry Neal
<@conan_kudo:matrix.org>
16:23:58
this rule change makes sense, but the examples don't
<@conan_kudo:matrix.org>
16:24:03
the hplip one is just weird and wrong
<@decathorpe:fedora.im>
16:24:38
yeah "no new things got installed" is just wrong
<@james:fedora.im>
16:24:43
The "people should be able to install plugins without the thing the plug into" argument makes me shake my head.
<@decathorpe:fedora.im>
16:24:50
this is not "no new things" 😆
<@decathorpe:fedora.im>
16:24:50
https://src.fedoraproject.org/rpms/hplip/blob/rawhide/f/hplip.spec#_936-969
<@decathorpe:fedora.im>
16:24:50
<@conan_kudo:matrix.org>
16:25:03
there relevant case would be dnfdragora, which installs the correct backend dependencies based on what toolkits are present, but it's already a graphical application
<@james:fedora.im>
16:25:05
I can kind of understand it from a QA POV, but for normal random users?
<@james:fedora.im>
16:25:31
And QA type people can just turn weak deps off, right?
<@decathorpe:fedora.im>
16:25:44
yeah, for QA you can use `setopt=install_weak_deps=False`
<@conan_kudo:matrix.org>
16:26:04
it's not weak, it's conditionally strong
<@james:fedora.im>
16:26:32
I thought all of these were conditional recommends?
<@conan_kudo:matrix.org>
16:26:50
there are cases of conditional requires, though I don't think any are in the ticket
<@conan_kudo:matrix.org>
16:27:13
libreoffice does this, for example
<@conan_kudo:matrix.org>
16:27:43
that package needs a hammer taken to it
<@james:fedora.im>
16:27:51
Yeh, I'd be happier to ban requires ... or at least say they should be recommends
<@conan_kudo:matrix.org>
16:28:33
well, conditional requires can make sense, it's really case by case
<@salimma:fedora.im>
16:28:59
we really should document some of the use cases in the documentation I guess
<@salimma:fedora.im>
16:29:13
it's a bit hard to reason about the use cases and make sure they all work, otherwise
<@conan_kudo:matrix.org>
16:29:24
but like, this rule basically means nothing for KDE because the whole stack is plugin based architectures
<@salimma:fedora.im>
16:29:27
or even, for some cases where the packaging is wrong and needs to be changed
<@conan_kudo:matrix.org>
16:29:43
so visible apps are dependent on other visible apps because internally everything is a plugin
<@james:fedora.im>
16:30:40
KDE doesn't break out the plugin bits from the main desktop/UI bits?
<@james:fedora.im>
16:31:10
Package wise, I mean.
<@conan_kudo:matrix.org>
16:32:50
not always
<@james:fedora.im>
16:33:03
After asking that, I'm not sure I want to know anymore ;)
<@james:fedora.im>
16:33:03
You could always comment in the issue and say "please don't break KDE", which would be amusing if nothing else.
<@decathorpe:fedora.im>
16:33:21
can't the whole paragraph be replaced with essentially "apply the principle of least surprise"?
<@conan_kudo:matrix.org>
16:33:25
A lot of it is subpackaged out, but some of the older ones haven't
<@salimma:fedora.im>
16:33:32
"KDE is an edition now, you can't break it"
<@limb:fedora.im>
16:34:02
s/can't/may not/g
<@conan_kudo:matrix.org>
16:34:04
same goes for GNOME, most of their packaging is ancient and pretty out of compliance with current guidelines
<@limb:fedora.im>
16:34:14
Because oh yes you can. ;)
<@james:fedora.im>
16:34:19
I'm pretty sure I'd +1 something like this and close the issue as fixed.
<@conan_kudo:matrix.org>
16:34:34
I think this is pretty fair
<@conan_kudo:matrix.org>
16:34:47
to be honest, we probably should say that for dependencies _in general_
<@decathorpe:fedora.im>
16:35:14
- Was I surprised that a normal system update pulled in HP Printer software GUI? Yes. :)
<@decathorpe:fedora.im>
16:35:14
- Would I be surprised if installing a gnome-shell extension installed the official app for managing extensions? Probably not.
<@conan_kudo:matrix.org>
16:35:36
our philosophy is already that, which is why we don't do openSUSE-style packaging of each dso as its own subpackage
<@conan_kudo:matrix.org>
16:35:57
we just don't spell it out :P
<@james:fedora.im>
16:35:57
I think it's the update part that is surprising there though ... if I installed all the hp printer stuff and also got a GUI for it, that doesn't seem surprising.
<@decathorpe:fedora.im>
16:36:21
yes, that would definitely be less surprising
<@decathorpe:fedora.im>
16:36:30
though why install the GUI only if Qt5 is installed then?
<@james:fedora.im>
16:36:58
I guess people want printers but to not have Qt installed?
<@tibbs:fedora.im>
16:37:21
I can't be the only person with a headless print server.
<@conan_kudo:matrix.org>
16:37:33
the thing is, Qt5 was default-installed on most Fedora variants until only a couple years ago
<@conan_kudo:matrix.org>
16:37:45
because Fedora Media Writer is Qt-based
<@james:fedora.im>
16:37:58
Wouldn't be shocked if you are the only person with a printer, TBH ;)
<@salimma:fedora.im>
16:38:19
I have a printer. I refuse to use HP :)
<@conan_kudo:matrix.org>
16:38:27
and yes refuse to use HP
<@conan_kudo:matrix.org>
16:38:33
I owned a Brother printer
<@conan_kudo:matrix.org>
16:38:37
an MFC even
<@conan_kudo:matrix.org>
16:39:04
it wasn't wireless and eventually broke, so I got rid of it and didn't need one ever since eFiling for taxes became possible
<@tibbs:fedora.im>
16:39:21
Everyone here wants their own printer in their office.
<@salimma:fedora.im>
16:39:29
I had a Brother laser printer, now a Lexmark MFC
<@conan_kudo:matrix.org>
16:39:47
Lexmark used to be great
<@salimma:fedora.im>
16:39:57
still quite decent really
<@conan_kudo:matrix.org>
16:40:00
I used to have a refurb office grade Lexmark MFC that was as tall as me
<@salimma:fedora.im>
16:40:08
my HOA and a lot of expense forms require printing and signing :P
<@decathorpe:fedora.im>
16:40:49
:gavel: :gavel: :gavel:
<@decathorpe:fedora.im>
16:40:56
ordaaaaar
<@conan_kudo:matrix.org>
16:41:02
🖨️
<@conan_kudo:matrix.org>
16:41:13
🗞️
<@decathorpe:fedora.im>
16:41:42
so what do we do about the ticket?
<@james:fedora.im>
16:42:44
I think this is as close to agreement as we've got with any proposed wording.
<@decathorpe:fedora.im>
16:43:23
so you're saying that I should make this a PR?
<@tibbs:fedora.im>
16:43:51
My perspective is that things which can be installed without a GUI shouldn't magically get a GUI through package dependencies. Just because servers and headless systems exist.
<@james:fedora.im>
16:44:30
I would add "if weak dependencies are turned off"
<@tibbs:fedora.im>
16:44:47
And I thought that was the point of the current guideline, but I guess there is a difference between "GUI" and "package containing a desktop file" which someone has decided means the rule doesn't apply to what they are doing.
<@salimma:fedora.im>
16:45:58
should the distinction be between apps meant to just configure something vs the main apps?
<@conan_kudo:matrix.org>
16:46:16
unfortunately nowadays there's a lot of abuse of the desktop file format for non application things
<@salimma:fedora.im>
16:46:24
not sure how to phrase it, but you know, GNOME Extensions is something to configure extensions (should really be in control center, but... that's another can of worms), hplip to configure printers, etc.
<@conan_kudo:matrix.org>
16:46:59
to be honest, if we had hardware supplements, that would make sense to use here
<@conan_kudo:matrix.org>
16:47:08
that's what the suse folks do
<@james:fedora.im>
16:47:51
I think the main difference is between "I know what I'm doing and want to install X to solve Y" and "I'm a user and want to try X" ... the former shouldn't install extra things unless they are really needed, the later probably wants to install a bunch of stuff so the user doesn't have to hunt around for X-gui or whatever it's called.
<@decathorpe:fedora.im>
16:48:09
RPM would know whether I have an HP printer in my network or not ???
<@conan_kudo:matrix.org>
16:48:37
well... we _could_ do that 🤣
<@decathorpe:fedora.im>
16:48:46
🙅🙅🙅🙅🙅🙅
<@conan_kudo:matrix.org>
16:48:48
easy enough with the right plugin
<@limb:fedora.im>
16:48:54
That sounds like a job for systemd
<@tibbs:fedora.im>
16:49:03
Could we just say what we are trying to accomplish and ask people to try to achieve that?
<@conan_kudo:matrix.org>
16:49:29
but but but automagic plug and play
<@james:fedora.im>
16:49:34
Yeh, if you know how to word that.
<@conan_kudo:matrix.org>
16:50:11
but in all seriousness, driverless networked printers probably need hplip less since they all can be configured through a web endpoint
<@conan_kudo:matrix.org>
16:50:34
and cups-ipp is supposed to be able to work with that
<@decathorpe:fedora.im>
16:51:53
ok, let's agree that hplip is a bad example?
<@limb:fedora.im>
16:52:06
I have a networked HP MFC and forgot hplip existed.
<@limb:fedora.im>
16:52:26
coreutils Requires: scribus is always wrong.
<@conan_kudo:matrix.org>
16:53:29
wtf who would consider such insanity?
<@limb:fedora.im>
16:53:41
Exactly.
<@limb:fedora.im>
16:54:11
Sometimes reducto ad absurdum helps find a helpful boundary.
<@james:fedora.im>
16:54:35
I think that was an intentionally wrong example nobody would ever try, except Neal has PTSD and assumed it was attempted.
<@limb:fedora.im>
16:55:06
And I'll offer Neal a hug if we meet IRL.
<@conan_kudo:matrix.org>
16:55:52
Indeed.
<@conan_kudo:matrix.org>
16:56:05
aww :)
<@limb:fedora.im>
16:56:22
And now Neal will never merge a PR from me without reading it.
<@james:fedora.im>
16:56:30
haha
<@conan_kudo:matrix.org>
16:57:59
😰
<@salimma:fedora.im>
16:58:05
is that your intent :P
<@limb:fedora.im>
16:58:33
I mean that should already have been true TBH
<@james:fedora.im>
16:58:45
Well, we are almost at time ... I think the preface warning that this meeting would be unhinged was valid and successful.
<@james:fedora.im>
16:58:56
Any final words?
<@limb:fedora.im>
16:58:57
Happy to help.
<@james:fedora.im>
16:59:52
If nobody else comments on 1456, I'll probably post something later.
<@james:fedora.im>
17:00:17
See you next week.
<@james:fedora.im>
17:00:20
!endmeeting