Charles Oliver Nutter

@MSP-Greg Ok thanks for the job run... One of the two failures appears to be skipped normally on JRuby. I'm not sure why it is not skipped for the new...

Correction... the command lines use either `bundle exec` **or** whatever `Gem.ruby` produces, so the latter form could be implicated as well.

It reproduces for me locally on MacOS on M1... I'll see what I can see from this end.

I believe this is a bug in the restart logic. The server is properly initialized during the tests, but when the restart is triggered it's launching a new instance incorrectly.

I just stumbled onto the same file and I'm investigating whether it is necessary anymore. It's doing some argv-munging that's likely causing this issue, but I'm not clear what its...

I believe this workaround file was added to do two things: * **Actually** do a native chdir to the target directory. In order to avoid interfering with other threads or...

Well here's some "good" news: if I pull that `chdir_exec` function out and call it from CRuby, it errors the same way: ``` $ ruby33 chdirexec.rb SyntaxError: /Users/headius/work/jruby/bin/jruby:8: syntax error,...

This could be a bug in how FFI handles varargs on both CRuby and JRuby.

A minimal patch to get the restart working correctly for now: ```diff diff --git a/lib/puma/launcher.rb b/lib/puma/launcher.rb index 9f0a0a49..afd484c7 100644 --- a/lib/puma/launcher.rb +++ b/lib/puma/launcher.rb @@ -287,7 +287,9 @@ module Puma close_binder_listeners...

I still need to follow up with fixes for JRuby + FFI + Apple Silicon but I'm glad the workaround is good for now.