I wonder if https://phabricator.wikimedia.org/T197726 could be somehow also related (non-main namespace; mobile). Probably not.

It's not unbreak-now, I suppose, but it's unbreak-before-Tuesday, which is when this will start going out to production wikis.

If we can't figure out a proper fix, we can always just revert the change easily.

This may be a little unrelated, but I am very confused by the editor situation. It seems that we did not really think about the new cases that were introduced when Minerva and MobileFrontend was split. I feel like this is part of the reason for this bug.

The mobile wikitext editor is always used on Minerva skin, in both mobile and desktop mode, even though the desktop editor is much more useful on desktop. Even weirder is that the mobile visual editor is only used on mobile.

The desktop editor is always used on all other skins, even in mobile mode, even though it seems to me that the mobile editor would work fine with other skins, especially with single-column layouts like narrow Timeless. Maybe it wouldn't fit design-wise with MonoBook or something, but the code should work fine.

Can we instead change this behavior, so that the mobile editor is used on mobile regardless of skin, and the desktop editor is used on desktop regardless of skin?

(I can split this to a separate task if you want to discuss.)

It's not unbreak-now, I suppose, but it's unbreak-before-Tuesday

Agreed

This may be a little unrelated, but I am very confused by the editor situation.

Me too now :)

The mobile wikitext editor is always used on Minerva skin, in both mobile and desktop mode, even though the desktop editor is much more useful on desktop.

Agree with this too.

so that the mobile editor is used on mobile regardless of skin

I like this in principle but is likely going to require some skinStyles.

and the desktop editor is used on desktop regardless of skin?

I don't think there is any blockers to this and at least what I was imagining we were working towards.

The problem from my POV is the definition of mobile mode. Historically, we tend to rely on the ResourceLoader targets system or skin when really we should be relying on screen resolution to determine that.

It sounds like we all might benefit from syncing about what we want to achieve long term and understanding this code a bit more given there have been quite a few patches/regressions now - T185729 > T196150 > T196915 . Right now let's address the broken state. Personally I worry about more patches given the experience so far and would favour a revert and try again approach.

I do agree with matmarex's suggested approach. The issue here seems to stem from a lack of clarity (on my part, on reviewer's part, on the code's part, etc) about exactly when things are providing a mobile experience. Wiping that clean and getting a simple rule for it would make sense.

I'd also be happy to scope everything that has changed so far behind some "only apply to mobile" wall, since that was mostly its intent, but as Joe says we don't have a clear way of specifying that currently. (Well, beyond the whole "people deliberately using Minerva on desktop is uncommon" aspect.)

The issue here seems to stem from a lack of clarity (on my part, on reviewer's part, on the code's part, etc) about exactly when things are providing a mobile experience. Wiping that clean and getting a simple rule for it would make sense.

👍. Some browser tests in Minerva documenting how things should behave would probably also be a good idea.

I'd also be happy to scope everything that has changed so far behind some "only apply to mobile" wall, since that was mostly its intent, but as Joe says we don't have a clear way of specifying that currently

Note, the mobile grade C browser workflow is still broken (unusable) with all these changes so scoping to mobile may not be enough here. I don't know enough about the mobile editing redirect code to know what's going on there and how we might fix that.

Given next train is monday and I'm off tomorrow what's the plan here? Revert them or add additional logic and commit to more vigorous testing?

Change 441569 had a related patch set uploaded (by Bartosz Dziewoński; owner: Bartosz Dziewoński):
[mediawiki/skins/MinervaNeue@master] Revert "Suppress display of wikitext editor on action=edit"

https://gerrit.wikimedia.org/r/441569

Yeah. Let's not leave this broken over the weekend. I personally will not have time to investigate before Tuesday and I assume y'all won't either.

Change 441569 merged by jenkins-bot:
[mediawiki/skins/MinervaNeue@master] Revert "Suppress display of wikitext editor on action=edit"

https://gerrit.wikimedia.org/r/441569

Thanks for dealing with the immediate problem @matmarex.

I think we can resolve this task? https://en.m.wikipedia.beta.wmflabs.org/wiki/Module:T173949 is editable again now so the bug as reported is fixed. Should we continue discussion on T197834 ?

The issue here seems to stem from a lack of clarity (on my part, on reviewer's part, on the code's part, etc) about exactly when things are providing a mobile experience

Happy to chat through the code/editor this week. Should we set something up?