This will be discussed during today's FESCo meeting (in ~40 minutes in #fedora-meeting-2 on libera.chat).

We discussed this during today's meeting (thanks to @adamwill for answering a few questions). We decided to wait with making any decisions until we can get a more concrete updated timeline, or until next week's FESCo meeting - whichever happens first.

Meeting logs: https://meetbot.fedoraproject.org/fedora-meeting-2/2023-08-10/fesco.2023-08-10-17.01.log.html

Sorry for not connecting. I totally missed the meeting + had different work in my real life during that time :(.

Thanks a lot @adamwill for answering the questions and please feel free to ask for any help if you need to. Ideally you can find us on Slack in #team-anaconda.

Hi, just want to notify you that GIS is progressing well from today: https://pagure.io/fedora-workstation/issue/362#comment-868920

Not sure what the progress is here, the discussion on the linked ticket is still ongoing. Can we get an update here please? We cancelled today's FESCo meeting, so it would be good to know whether we should 1. start an in-ticket vote to invoke the contingency mechanism, or 2. wait even longer ... though I'm getting uncomfortable with option 2 given that this is already quite late.

I'd ask you to wait until the next FESCo meeting before deciding to invoke the contingency mechanism. This change is unfortunately still not easily testable yet as it still requires building a custom image, but it's been progressing rapidly this week and should hopefully be ready in time for your next meeting.

Honestly, right now to execute the contingency plan is to do nothing. Web UI is not used in the image by default and the current GTK UI should work as before.
The question is really not much about executing the contingency plan but more about if we want to reject the change and for that it would be great to see what we have before deciding.

right now to execute the contingency plan is to do nothing

That is good news.

The beta freeze starts tomorrow. I think it's too late to switch the installer to a different implementation at this point.

Does that mean our change is postponed to next Fedora?

We need to know this to merge the GIS changes today or not, to get it there before freeze.

Landing this mere hours before the beta freeze goes into effect is, in my opinion, way too late for a big change like this. Enabling it for the Beta and reverting it again in time for the Final release is also a non-starter. So I'd say that the boat has sailed for F39 - but that's just my opinion, so let's make this a formal proposal:

The Change to default to the new Anaconda WebUI for Fedora Workstation is postponed to Fedora 40.

I am +1.

+1 to deferring to F40.

i mean we have the whole beta period to work out any issues that crop up. It's not like we're asking to get this in after beta... I don't understand the impetus for deferring, tbh.

the "whole beta period" is just over two weeks to the first go/no-go meeting, which isn't exactly acres of time.

Well, the impetus for deferring is that this should have landed in F39 weeks ago in a testable state ...

Well we've had test isos out since early last week (which granted was tardy and already discussed in last fesco meeting)... My expectation following https://meetbot.fedoraproject.org/fedora-meeting-2/2023-08-10/fesco.2023-08-10-17.01.log.html was that if we landed by today we'd be okay, but maybe I made too much of a mental leap there.

You guys are probably right in being prudent. Anyway, we should still land stuff into rawhide today, regardless.

If we going to deffer this to F40 (which I'm not strictly opposed to). I have another proposal for this.

Would it be feasible for FESCO to have this on the https://get.fedoraproject.org/ page as technology preview or similar solution for Fedora 39?

My point is, that I believe that it will be prepared for launch on F39 GA and lot of effort was put into that. So this way we could still promote it and find what users thinks about the new approaches.

I totally get your point that there is not enough time for support from Fedora QE and similar but I think it would be great to promote this as non-official (try it?) release.

Putting it on alt.fedoraproject.org might be possible?

FWIW, I'm -1 to deferring. It's limited in scope to the Workstation Live images and it is landing before the Contingency Deadline (even if only barely).

If things go disastrously wrong, we can revert before GA.

I tried to figure out what would need to merged before tomorrow, but it's hard. There's a bunch of tickets, many of them resolved already, but no comprehensive list that I can see.

i mean we have the whole beta period to work out any issues that crop up.

This is not how I understand the beta period. In beta freeze which lasts about two weeks, we are supposed to concentrate on fixing bugs. Only changes that fix known bugs are supposed to go in. And after the beta has been released, we can do version updates and such, but no breaking changes are expected. Together this means that the beta period and later are not suitable for general development. Merging a bunch of features which might be OK, but might require various adjustments that are more than bug fixes, right before the beta freeze is problematic. It'd be OK to merge even a big change right before the freeze, but only when we expect it to be "ready", maybe with some tweaks here and there. But in this case, we have a bunch of pull requests in flight, both in Anaconda itself and in other places, and it sounds like a bunch of additional development might be needed.

So… if people who understand the extent of this change and feel confident merging pull requests can do it before tomorrow, then OK. Otherwise, I think this should be postponed to F40.

the main things that needed to merge were:

anaconda-39.30-1.fc39
livesys-scripts-0.5.0-1.fc39
gnome-initial-setup-45~beta-3.fc39

They're all merged and built now. There's also https://pagure.io/fedora-kickstarts/c/ce5b31416f0bef6ff70acc3363088a95e2ee5983?branch=main which would need to be cherry-picked to the f39 branch.

If we get enough -1s to force contingency, i'll do another gnome-initial-setup build to revert the vendor.conf changes.

All the pending upstream commits are are about fit-and-finish issues (titlebar aesthetics, focus improvements, that sort of thing)

About the Anaconda changes. We will probably still be delivering webui changes into the Fedora 39 because the changes are part of the package which is not now added to the Live media so it's safe thing to do.

From the code completion part - we should have all the code changes merged and bugs are tracked here:
https://bugzilla.redhat.com/show_bug.cgi?id=2231339

I suggest proceed with landing the kickstarts change in F39 ASAP to get it in before freeze so we don't foreclose the option of shipping the new installer before FESCo has made its decision. But of course we should expect that FESCo may require that we revert it (via a freeze exception).

Honestly, it'd be really nice to ship the new installer for beta even if we wind up needing to use the old installer for the final release, since using it for beta allows for significantly more user feedback than we would otherwise receive.

As said for previous change proposals: Shipping something for Beta with the express intent to not include it in the final release is an anti-pattern. Things should only be part of the beta if they are expected to be ready for final release as well.

fwiw, the kickstart changes are here:

https://pagure.io/fedora-kickstarts/pull-request/978

so my recommendation is we cherry-pick that today, get a few nights of the new installer, and discuss possible deferral at the next fesco meeting on thursday (where if deferred we'd revert that and the gnome-initial-setup changes)

OK, so I think we should proceed with this for F39. If we build the beta and it turns out that the installer is significantly broken, we can revert and rebuild.

I'll follow @sgallagh's example, and vote -1 to deferring for now.

I'll go with @zbyszek and @sgallagh and change to -1 too.

-1 to deferring

I've landed the PR for now, if we decided to back it out, we can later.

okay okay, consider my proposal withdrawn.

I still don't like that this is landing so late, but meh.

this isn't landing very smoothly. anaconda and g-i-s updates both failed gating, and the updates were sent separately (shouldn't they have been combined?) and apparently before the kickstarts change landed (because the live image openQA built for the anaconda update got oldUI, not webUI).

I can adjust openQA to understand the new gnome-initial-setup, and retrigger the tests for the anaconda update so we hopefully actually get webUI, and rewrite the openQA tests to handle webUI (that is going to take at least a couple of hours), but we still have the problem of anaconda prompting for authentication on startup on Workstation...the tests also failed on KDE apparently because the install process took a long time and the screensaver kicked in, I didn't look at that one too hard yet...

update: I worked on getting the tests to pass with the webUI changes, but couldn't quite finish before I had to go out, and there's also an issue with the KDE tests failing with the anaconda update because the screensaver kicks in. We decided to undo everything and try again tomorrow. For now, the anaconda update is gated for both f39 and f40, and the g-i-s updates have been untagged. The fedora-kickstarts change has been reverted. Tomorrow I can do more testing with openQA hacked up to apply the kickstarts change and pull in the g-i-s build when testing the anaconda update, and if we can get everything working, we can attempt again to land the changes.

OK, we now have an anaconda build that doesn't break on KDE, and I have the openQA tests modified to pass on webUI.

But, the Beta freeze has kicked in. Assuming FESCo still wants to land this, we need to approve https://bugzilla.redhat.com/show_bug.cgi?id=2233683 . It should get a ticket at https://pagure.io/fedora-qa/blocker-review/issues fairly soon.

FE voting ticket: https://pagure.io/fedora-qa/blocker-review/issue/1183

I'm fine with us trying to land it now. The fact it's so easy for us to back it out makes me more willing to do this. And realistically, if we don't have it in now, it won't be present for the Beta release, which is extremely important for getting broader feedback.

So we kind of voted to not defer this already? but I suppose that was it landing just before freeze? anyhow +1 to land today and if it proved untenable we can back out before beta...

yeah, I was reading that vote as applying to it landing before the freeze. It didn't seem right to just go ahead and grant an FE and land it the day after the freeze without getting FESCo's approval for that. But if we're gonna land it we should do it soon, so since three FESCo members have said 'go ahead' already, I'll wait for a couple more hours then go ahead and grant the FE (it has sufficient votes) and land the changes if nobody objects.

FE granted, changes landed, next F39 compose will have webUI on Workstation live.

The implementation of the feature went in, so let's close this.