Charles Oliver Nutter

@fxn Most of the other failures I fixed were due to us not calling `const_added` for autoloads. It seems likely this is another place where we are not properly triggering...

@rsim That is great news, thank you! We'll be looking to get JRuby 10.0.1.0 out soon with those fixes, but I'd like to try to get Zeitwerk green for a...

@fxn I tried running the failing tests in isolation, but they all pass (the Hotel ones, anyway). Being unable to fail in isolation means this is going to be a...

Here's a run with no failures other than the known path-canonicalization issue (#6735): ``` $ bundle exec rake /Users/headius/work/jruby/lib/ruby/gems/shared/gems/minitest-5.20.0/lib/minitest/test.rb:86: warning: previous definition of run was here # Running tests with...

I decided to just brute-force it and delete tests until I found a reproducible sequence. Got one! ``` $ bundle exec rake TESTOPT=-v SEED=47270 /Users/headius/work/jruby/lib/ruby/gems/shared/gems/minitest-5.20.0/lib/minitest/test.rb:86: warning: previous definition of run...

The problematic code seems to be this stub in the `TestRequireInteraction` test: https://github.com/fxn/zeitwerk/blob/ef59cc29de85897d603f295b5f6abf3ee08258ec/test/lib/zeitwerk/test_require_interaction.rb#L169-L176 Could be something about the way this stubs out `Dir.foreach` is breaking it for future calls. If...

Current minimal test to trigger the failure: ```ruby test "explicit namespaces are loaded correctly (directory first, Object)" do files = [["hotel.rb", "raise"]] with_files(files) do require './hotel' rescue nil end files...

@fxn You have any thoughts about what state might be getting screwed up by this example? It's like the Hotel constant is getting registered as raising an error and just...

@fxn Thank you, that's perfect! I was hoping you'd be better equipped to untangle the Zeitwerk parts than me! This is a very peculiar behavior, but it should be easy...

I believe I found the issue. When entering a require, we register a lock under the expanded form of that filename, so any other threads attempting to require at the...