Add ruby 4.0 and HEAD to CI

[nevans]

This will provide early warning for changes, not only to ruby (the language), but also to important default stdlib gems, such as openssl.

For example, this will not pass until the following other PRs are merged:

• Relax dev dependencies #361
• Don't modify SSLContext::DEFAULT_PARAMS #362

It passes for me (locally) with those both merged.

[nevans]

@chrisarcand Any thoughts on this #361 and #362 (which are needed for ruby 4.0 compatibility) or this PR (which will help detect future compatibility issues a little bit sooner)?

If you don't have time before then, that's fine (and very relatable!). I'll update this PR after the 4.0 release to add 4.0 to the matrix, too.

[chrisarcand]

Sweet! Other PRs merged. Go ahead and add 4.0 and let's watch it pass :)

[nevans]

@chrisarcand Thanks! I had to update the GitHub Action for ruby/setup-ruby, but it's all passing now.

[nevans]

Actually, I'm going to rewind that last commit and just merge #355 and #372 into this PR. That'll automatically close all three PRs when this one is merged (and add all three to the release notes, if you use that).

[chrisarcand]

Looks like there's some conflicts to settle

[nevans]

@chrisarcand I'm not seeing any conflicts on GitHub... but I did rebase, force push, then rewind a commit, and force push that. Are you seeing local conflicts? Maybe try git remote update && git reset --hard @{upstream}?

[chrisarcand]

Weird! I'm seeing this:

I'll toy with it a bit.

[chrisarcand]

Yeah, needs some help:

[chrisarcand]

I can squash merge instead, that's fine.

[nevans]

Ah, that explains it. You were using rebase (for flat history, I presume). Because I did use a couple of merge commits for #355 and #372, and git's flat rebase throws away any information from the merge commits (which was needed to resolve the conflict).

[nevans]

I personally use a (slightly idiosyncratic) workflow (which I forget is idiosyncratic): I only "rebase merge" single commit PRs, I use "regular" merge commits for multi-commit PRs, and I usually only use "squash merge" on dependabot PRs (because of dependabot/dependabot-core/issues/4565 😉). So I usually forget that the rebase PR merge will be confused by merge commits. But squash should pretty much always get the job done.