Charles Oliver Nutter
Charles Oliver Nutter
Do you think you can turn this into a small reproduction that we can run locally? My first instinct would be that a class previously getting imported as a side...
Original reporter has disappeared. If anyone runs into this again please file a new issue.
At this point in execution I think it's probably impossible to have a null `rescueIPCs` because even simple methods will have one rescue wrapping the whole method.
Probably the case that we never have a null `unboxedVar` and also do not want to `createNew`.
The updated net-http must be depending on win32 more than previous versions. You may be able to specify a previous version in a Gemfile but I am not sure how...
Something must have changed how we handle case-preserving filesystems, at least when running on Windows.
The PR in #8542 could be a suspect, since it deals with globbing in case-preserving filesystems, but I'm not sure if that code is used by require.
JRuby 10 is not affected due to the updated stdlib no longer requiring `win32api` here: https://github.com/jruby/jruby/commit/ed2679c2d6f5023bcdddc18acb4ece6babcefb3d#diff-214de3cfd07c836f1acab856b5613dad84497ba4324e1356edda1c46ef67c626L3-R220
It appears to work ok on Darwin (MacOS) even though the file case doesn't match: ``` $ jruby -e 'require "Win32API"' LoadError: Win32API only supported on win32 at /Users/headius/work/jruby94/lib/ruby/stdlib/win32api.rb:2 ```
Much of the filesystem searching done here should not be any different on Windows. It falls back mostly on JDK classes to check the existence of the file. I'm wondering...