Buggy debuginfo test suite, or just `function-names.rs`?

[workingjubilee]

In #125958 (a PR which changes the way consts are handled in the compiler, and thus alters details like their hashing, etc.) a few lines were trimmed out of a debuginfo test because it was unreasonable to get it to work, despite the function names seemingly being present.

It seems like there might be a bug in the way that the debuginfo test suite handles [...]? Or it might incorrectly treat certain checks as ordering-sensitive, when hashing can change orders? It's unclear, but in any case, undesirable.

This mostly affected the Windows variant of the test.

[workingjubilee]

cc @michaelwoerister

[michaelwoerister]

Thanks for the pointer, @workingjubilee! I'll take a look.

[workingjubilee]

To concretize this issue slightly further, I'd like to apply changes as close as possible to the other ones in this patch to the debuginfo test suite: #125963

[BoxyUwU]

cc #115786