14:00:07 <kalev> #startmeeting Fedora Flatpak Packaging SIG
14:00:07 <zodbot> Meeting started Mon Jun 12 14:00:07 2023 UTC.
14:00:07 <zodbot> This meeting is logged and archived in a public location.
14:00:07 <zodbot> The chair is kalev. Information about MeetBot at https://fedoraproject.org/wiki/Zodbot#Meeting_Functions.
14:00:07 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
14:00:07 <zodbot> The meeting name has been set to 'fedora_flatpak_packaging_sig'
14:00:07 <kalev> #meetingname flatpak-sig
14:00:07 <kalev> #topic Init process
14:00:07 <zodbot> The meeting name has been set to 'flatpak-sig'
14:00:16 <kalev> who's around today?
14:00:44 <kalev> .hello kalev
14:00:45 <zodbot> kalev: kalev 'Kalev Lember' <klember@redhat.com>
14:01:01 <tpopela[m]> .hello tpopela
14:01:02 <zodbot> tpopela[m]: tpopela 'Tomas Popela' <tpopela@redhat.com>
14:01:18 <OwenTaylor[m]> .hello2
14:01:19 <zodbot> OwenTaylor[m]: Sorry, but user 'OwenTaylor [m]' does not exist
14:01:33 <OwenTaylor[m]> .hello otaylor
14:01:35 <zodbot> OwenTaylor[m]: otaylor 'Owen Taylor' <otaylor@redhat.com>
14:02:19 <kalev> #topic Announcements
14:02:27 <yselkowitz[m]> .hello yselkowitz
14:02:28 <zodbot> yselkowitz[m]: yselkowitz 'Yaakov Selkowitz' <yselkowi@redhat.com>
14:02:44 <kalev> nice, all the usual suspects are here today :)
14:03:06 <OwenTaylor[m]> I can do a status update on FlatpaksWithoutModules
14:03:28 <kalev> I haven't looked at flatpak stuff much lately. I wanted to do a round of updates but ran into https://pagure.io/fedora-infrastructure/issue/11368
14:04:11 <OwenTaylor[m]> The change proposal has not been approved yet, but the ticket is currently +8,0,0 so it should be RSN (tomorrow-ish)
14:04:17 <JanGrulich[m]> .hello jgrulich
14:04:18 <zodbot> JanGrulich[m]: jgrulich 'Jan Grulich' <jgrulich@redhat.com>
14:05:06 <kalev> #topic status update on FlatpaksWithoutModules
14:05:12 <kalev> the floor is all yours!
14:05:35 <OwenTaylor[m]> I've been working on the implementation side with the assumption that it will go through. I have code to build Flatpaks directly from packages locally, in OSBS v2 and as a Koji plugin
14:06:08 <kalev> ah, two different implementations?
14:06:31 <OwenTaylor[m]> It will be have to be determined for Feora whether we backport OSBS changes to OSBS v1 or use the Koji plugin
14:06:39 * matthiasc[m] waves
14:06:51 <OwenTaylor[m]> One implementation in flatpak-module-tools, but different glue code
14:06:52 <kalev> hi matthiasc[m]!
14:07:53 <OwenTaylor[m]> The Koji plugin has been a little harder than I expected - not because it's a lot of code, but because it was a bit hard to figure how to structure things, and once you start testing live with Koji builds on a laptop, cycle times are slow.
14:08:26 <OwenTaylor[m]> But I think it's fundamentally workable. There are a couple of hacks that could be removed with some simple changes to the Koji core code.
14:09:04 <kalev> nice, sounds like a good progress
14:09:24 <kalev> do you have the code anywhere public where I could take a look or is it too early for that still?
14:09:44 <tpopela[m]> yes, indeed! Thank you Owen!
14:10:37 <OwenTaylor[m]> Assuming that the proposal is approved, next steps are a) add the necessary tags/targets b) land changes to flatpak-module-tools and get a package of that into Fedora
14:11:40 <kalev> do you want to replace both app and runtime building at the same time, or stagger it somehow?
14:14:09 <OwenTaylor[m]> Ideally we just build everything for F39 with the new tooling
14:15:01 <OwenTaylor[m]> We can leave the old tooling in place for F38, and eventually switch the f38 runtime over if we want.
14:15:34 <kalev> that would probably be a prudent thing to do
14:16:46 <kalev> grr, the irc-matrix bridge seems to be dropping some of my messages :(
14:17:03 <tpopela[m]> Sort of a related question - you took the resolve-deps code from fedmod and put it into flatpak-module-tools or you've done some improvements to it (i.e. somehow handling the dependencies resolving on different architectures - generating runtime/SDK for x86_64, ppc64, aarch64)? Or no changes were done, but it might be easier to do as the code is now part of the project?
14:18:22 <OwenTaylor[m]> It's not really different in that regard, no.
14:18:53 <kalev> do you have the code anywhere public where I could take a look or is it too early for that still?
14:19:43 <OwenTaylor[m]> We could do improvements there to depsolve on each architecture, if we were willing to make runtime generation slower. I can easily make 'flatpak-module-depchase resolve-packages' take an --arch argument if it doesn't already.
14:19:52 <travier> .hello siosm
14:19:53 <zodbot> travier: siosm 'Timothée Ravier' <travier@redhat.com>
14:20:42 <OwenTaylor[m]> Alternatively, we *could* drop the whole "must list exact packages in source control" thing for runtimes.
14:21:13 <KalevLember[m]> I would be fine with slower runtime generation. I think it would be a net win to have it automated instead of manually chasing changes between different arches
14:21:25 <tpopela[m]> Owen Taylor: if you could add that support, then it would be amazing!
14:21:33 <tpopela[m]> Kalev Lember: +1
14:21:43 <OwenTaylor[m]> And consider changes in the runtime package list something we just have to handle at the QE / automated CI level. But I do worry about how good we are at keeping on top of that.
14:22:49 <KalevLember[m]> do you have the new flatpak-module-tools code somewhere public or is it too early for that still?
14:22:54 <OwenTaylor[m]> The one difference here in the new system is that when building apps the runtime package list is retrieved from the package list of the built container rather than from the profiles in the runtime module metadata.
14:23:44 <KalevLember[m]> do you mean when building app flatpak container?
14:24:08 <OwenTaylor[m]> There's something at https://pagure.io/flatpak-module-tools/tree/without-modules (about a week old), but it's really too early for anyone to spend time looking at it.
14:25:12 <OwenTaylor[m]> (I'd say the code is 95% the way it needs to be - but there's some restructuring to do, README hasn't been updatd, etc.
14:25:48 <OwenTaylor[m]> Kalev Lember: Yeah. It's really just an implementation detail as long as we are saying that we're strictly controlling the runtime package list out of source control
14:26:20 <KalevLember[m]> makes sense
14:28:30 <KalevLember[m]> awesome, I'll take a quick look at the changes :) I have some local flatpak-module-tools commits for building extensions that I wanted to send to you, but dropped the ball on this. It's probably a good time to do that now since you are hacking on that code.
14:29:25 <OwenTaylor[m]> Kalev Lember: cool.
14:29:31 <tpopela[m]> Kalev Lember: for the Firefox and openh264? :)
14:29:47 <KalevLember[m]> yes
14:30:58 <KalevLember[m]> awesome, I'm looking forward to a world without modules :) thanks a lot for working on this, Owen!
14:31:34 <KalevLember[m]> is there anything you need help with here?
14:32:50 <OwenTaylor[m]> Not so much at the moment. I'm currently at a "too much code on my laptop, too much stuff in my head" state :-) ... need to start unwinding that
14:33:08 <KalevLember[m]> ok
14:33:19 <KalevLember[m]> #info Owen has been working on the implementation and we're planning for landing it for F39.
14:34:29 <OwenTaylor[m]> I'll definitely need some beta testing on the flatpak-module cli once we have the tags/targets in Fedora so things don't have to be hacked as much.
14:35:04 * OwenTaylor[m] wonders about renaming flatpak-module-tools - but to what?
14:35:50 <KalevLember[m]> I can imaging how hacky it must be to test it locally without having the right tags in koji :)
14:36:15 <tpopela[m]> flatpak-toolbox ? But I think that there was something like that in the past..
14:36:27 <KalevLember[m]> flatpak-building-tools? fedora-flatpak-tools?
14:36:40 <OwenTaylor[m]> fedora-flatpak-tools would be fine, but a bit weird, since it's rpm+dnf+Koji specific not Fedora specific. 'flatpak-rpm build-container' seems yucky. Plan is to hide the CLI behind fepdkg anywayss
14:37:29 <yselkowitz[m]> what happens to local builds?  fedmod rpm2flatpak?
14:38:07 <OwenTaylor[m]> Local build work much better than before - they will be faster and less confusing.
14:39:28 <OwenTaylor[m]> rpm2flatpak is no longer needed since you list only leaf packages in container.yaml. You'll be able to do something like 'fepdkg build-flatpak-rpms' that will depchase your dependencies, check that they have been rebuilt in f39-flatpak-app, and fire off builds as necessary.
14:40:13 <OwenTaylor[m]> (And fedpkg build-flatpak-rpms-local to do the same but in mock putting results in a local repository)
14:41:08 <KalevLember[m]> it going to be awesome to be able to do a simple koji scratch build against f39-flatpak-app to test that things build, instead of having to go through all the modular builds glue
14:44:01 <KalevLember[m]> anything more on this? should we move on?
14:44:13 <yselkowitz[m]> move on
14:44:33 <kalev> #topic flatpak container builds broken
14:44:59 <KalevLember[m]> ok, just a quick note that we can't do any builds right now due to https://pagure.io/fedora-infrastructure/issue/11368
14:46:05 <KalevLember[m]> Owen, do you understand where the sources are fetched from and how the filters should be set up exactly?
14:48:17 <OwenTaylor[m]> I haven't really looked at the koji policy engine stuff at all, I'm only familiar with setting allowed_scms out of hte koji config. But I can take a look.
14:49:17 <KalevLember[m]> Thanks - I'd appreciate that
14:50:08 <OwenTaylor[m]> I'm not sure about the 'fedpkg sources' stuff, but the main issue seems to be a src.fedoraproject.org vs pkgs.fedoraproject.org confusion - the koji configs used to alow either
14:51:00 <KalevLember[m]> I guess alternatively we could change it on the flatpak building side to use pkgs.fedoraproject.org?
14:52:24 <OwenTaylor[m]> Hmm, no, it seems that kevin just quoted part of the hub policy. So it's not exactly that. But I'll investigate
14:52:40 <KalevLember[m]> Thanks!
14:53:14 <KalevLember[m]> Anything else to discuss or should we wrap this up? I believe we are almost out of time
14:53:22 <yselkowitz[m]> https://src.fedoraproject.org/flatpaks/stellarium/pull-request/2
14:53:41 <yselkowitz[m]> can we merge this already?
14:54:25 <KalevLember[m]> Hm, I thought we agreed to add the KDE runtimes only for building KDE apps and keep the rest on the Fedora runtime
14:54:56 <yselkowitz[m]> not afaik, I certainly am against that
14:55:42 <yselkowitz[m]> mediawriter was an exception in order to be included in workstation/silverblue
14:55:49 <yselkowitz[m]> even that isn't great tbh
14:56:23 <yselkowitz[m]> but otherwise apps should NOT include qt when we already have runtimes which provide it
14:56:42 <KalevLember[m]> I think we should schedule this discussion for the next meeting so we have a bit more time,
14:57:00 <yselkowitz[m]> it's already been two months...
14:58:28 <KalevLember[m]> Right, we definitely need to discuss it - next time
14:58:58 <KalevLember[m]> see you all in two weeks! we are out of time now
14:59:05 <kalev> #endmeeting