Update to rails 7.2.0

[tomhughes]

No description provided.

[tomhughes]

Looks like this is a no-go until we can move to ruby 3.1 then...

[mmd-osm]

I did a smoke test on Ubuntu 22.04 + ruby 3.3.2 (2024-05-30 revision e5a195edf6), which I installed manually at one point. Tests are passing without issue:

Finished in 66.145466s, 26.2301 runs/s, 15520.4439 assertions/s.
1735 runs, 1026607 assertions, 0 failures, 0 errors, 0 skips

A few tests are reporting some warning "Test is missing assertions":

• test_3rd_party_token: test/integration/client_applications_test.rb:73
• test_body_class: test/helpers/application_helper_test.rb:92
• test_header_nav_link_class: test/helpers/application_helper_test.rb:94
• test_lost_password_recovery_success: test/integration/user_creation_test.rb:174
• test_show_no_friends: test/controllers/dashboards_controller_test.rb:13

[tomhughes]

Not sure why you're getting those messages because I don't see them.

I mean I can see what the problem is - people have written stubs for tests that don't do anything yet. It's just a bit odd that you get warned about them and I don't.

[tomhughes]

In some cases they're not even stubs - a test of an action with no assertions still checks that a given action will run without throwing any exceptions.

[mmd-osm]

The messages also show up in GitHub action runs, such as: https://github.com/openstreetmap/openstreetmap-website/actions/runs/10394204557/job/28783478253?pr=5070 around line 70:

Test is missing assertions: `test_lost_password_recovery_success` /home/runner/work/openstreetmap-website/openstreetmap-website/test/integration/user_creation_test.rb:174
Test is missing assertions: `test_body_class` /home/runner/work/openstreetmap-website/openstreetmap-website/test/helpers/application_helper_test.rb:92
Test is missing assertions: `test_header_nav_link_class` /home/runner/work/openstreetmap-website/openstreetmap-website/test/helpers/application_helper_test.rb:94

[mmd-osm]

According to the docs, https://api.rubyonrails.org/v7.1.3.4/classes/ActiveSupport/Testing/Assertions.html#method-i-assert_nothing_raised should handle this case. Probably a bit redundant, as exceptions would cause the test to fail anyway.

[tomhughes]

Yes I figured it out... I had RUBYOPT set to suppress warnings in my environment. I think at some point there were some annoying warnings we couldn't get rid of and rails doesn't normally use actual ruby warnings like that when it wants to warn about something.

[gravitystorm]

I think it's a shame that we have to drop 22.04 support, since I expect some developers won't have yet upgraded their Ubuntu to 24.04. (I haven't, but I used rbenv - we don't want to rely on that for new developers).

It seems like an unfortunate combination of release dates and support periods:

• ruby 3.0 was released in Dec 2020
• ruby 3.1 was released in Dec 2021, which was in time for Ubuntu 22.04 but wasn't the chosen version
• ruby 3.0 support ceased in Apr 2024
• rails drops 3.0 support (and therefore Ubuntu 22.04 support) in Aug 2024

So we end up with a period of only 4 months where we have two versions of Ubuntu available. We'll face a similar problem in two year's time, because Ubuntu 24.04 shipped with ruby 3.2, which will be EOL in Mar 2026 i.e. just before Ubuntu 26.04 LTS. I wish either ruby would extend their support a little, or Ubuntu would use the newest ruby, or rails would be more tolerant of slightly EOL rubies, all the make this developer experience a bit smoother.

Nevertheless, I'm happy to move forward. I would say that we should:

• Drop 22.04 from the testing matrix. If we can't run on real Ubuntu 22.04, there's little point in testing on Github Flavoured Ubuntu
• Adjust the Vagrantfile to use ubuntu/noble64 or generic/ubuntu2404 as appropriate
• There's a failure in building the docker image in CI

[tomhughes]

There's an extra complication which is that we can't (yet) use the GitHub 24.04 image to run tests as it's still beta and is missing firefox (planned to be fixed before it leaves beta) so for now we probably have to stick with 22.04 for GHA.

[tomhughes]

Also we're waiting on jumph4x/canonical-rails#92 before we can go to 7.2.1 which is already out.

[tomhughes]

I've updated this as follows:

• Reverted changes to ubuntu versions in testing matrix for now pending resolution of Firefox WebDriver not working on Ubuntu 24.04 actions/runner-images#10374
• Updated vagrantfile
• Fixed "no assertion" warnings in tests

The docker issue is rubyjs/mini_racer#305 / rubyjs/libv8-node#36 which stops mini_racer installing on 24.04 currently.

[tomhughes]

Oh just to add that all the production servers, and the dev server, are now on Debian 12 and hence ruby 3.1.

[mmd-osm]

#5010 (comment) proposes some changes to use Debian 12 for Docker. Maybe we should try to resolve the remaining issue there, to keep everything a bit more in sync.

[gravitystorm]

Looks good to me, thanks!

[tomhughes]

Oh I was going to wait until we could do 7.2.x updates in case there security issues that couldn't wait...

[tomhughes]

It looks like 7.2.1 is OK at least as there don't look to be any security fixes.

[gravitystorm]

Oh I was going to wait until we could do 7.2.x updates in case there security issues that couldn't wait...

Ah, I didn't realise it was blocked for any reason. I saw that the canonical-rails PR had been merged already, but I guess the problem is that it hasn't been included in a release.

I can make a PR that moves to canonical-rails from git master, if that helps? That way we can get point releases for rails 7.2.x

[tomhughes]

I don't think we need to worry for now but it's a way out if we do need to do a rails update urgently.

[gravitystorm]

Ah, I created #5246 before I saw your comment, so the work is done if you want it. I've no idea whether rails 7.2.2 or canonical-rails will be the first to be released.