Fix SimpleModuleLoader on Windows

[hansl]

Basically I made a mistake with my previous PR and discovered it as I was building embed_module!.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 50.22%. Comparing base (6ddc2b4) to head (2609952).
Report is 123 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3795      +/-   ##
==========================================
+ Coverage   47.24%   50.22%   +2.98%     
==========================================
  Files         476      456      -20     
  Lines       46892    44823    -2069     
==========================================
+ Hits        22154    22514     +360     
+ Misses      24738    22309    -2429     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jedel1043]

To avoid the same error in the future, maybe the test should be changed to test the path components instead of the whole path string.

EDIT: Just realized that this change could also deduplicate the tests.

[hansl]

I don't think we would gain anything from checking the path components, but I'll make the change. There was a bug in the code and the tests were built to test the bug, so both were faulty. It happens.

(as proof that the bug was in the test AND the code, I didn't change the expected value, just the argument itself)

[jedel1043]

I don't think we would gain anything from checking the path components, but I'll make the change. There was a bug in the code and the tests were built to test the bug, so both were faulty. It happens.

But the bug passed the test because we were using separators on the test itself, right? If I'm correct, testing by components instead of using separators would have caught the bug.

[hansl]

But the bug passed the test because we were using separators on the test itself, right?

The bug was two fold:

1. in JavaScript you'd always use ./a/b/c in your import specifier, even on Windows. I was generating a path using the specifier, but didn't change the \\ to /, so if you were on Windows and used a/b/c it would have been a:\some\root\a/b/c which would be shown in components (as you said).
2. the test was actually wrong in that it passed ./a\\b\\c for specifier, and so Windows was returning the expected path but for the wrong reason. That's what I mean by I was "testing the bug". This meant that even doing a component check would still result in the right answer, because the test input was wrong. Changing the specifier in the test to be ./a/b/c then showed the bug, which I fixed in this PR as well.

I might be over explaining. We can sync offline tomorrow morning if you or I are missing something. Cheers!

[jedel1043]

That makes a lot of sense. Thank you for explaining!

[HalidOdat]

Looks good to me! :)