Don't install the bootstrap gem unless they are using sassc

[jcoyne]

No description provided.

[jrochkind]

I would find this more readable if it were

unless used_bootstrap_css?
   generate_with_sassc_rails
   return

(Unless you actually want to use the return value of generate_with_sassc_rails).

This code is getting pretty convoluted and branching, keeping it readable seems good, what do you think?

[jcoyne]

Sure, done.

[jrochkind]

This does not fix the test-order-dependent Rails 8 failure in #3361, but i guess it was not intended to?

The logic does seem right.

The logic is getting pretty convoluted and hard to follow, and kind of starts to seem "accidental" to me; I think it could use refactoring these asset generators into more clear use cases, instead of just making tests green.

I could be interested in submitting a PR for such a refactor, but resolving the test-order-dependent failures in Rails 8 seems like a pre-requisite, to be confident green on such a PR was actually a signal the refactor works out.

[jrochkind]

I think this makes sense (and is similar to what I ended up doing in a version in blacklight_range_limit investigating this stuff). I am interested in if it breaks #3360 if we merge that first.

At some point I'd be interested in adding a switch so the file does not NEED to be named application.bootstrap.scss -- there are cases where you are using a package.json-based solution but didn't actually generate with --css=bootstrap, and you have an application.[s]css that you'd still want the @import 'blacklight' added to.

But that's a separate thing, this PR doesn't change it. I'll approve, but what if we merge #3360 first and rebase this one to make sure they don't conflict?

[jcoyne]

Rebased