Source location relative test

[bilgili]

We have found an issue with source_location where line numbers were reported wrong (https://developercommunity.visualstudio.com/t/std::source_location-started-reporting-w/10429484?q=source_location)
and I had a peek on the testing of the std::source_location.

It is as below:

const auto x = source_location::current();
    assert(x.line() == __LINE__ - 1);

It is relatively testing against the __LINE__ macro. Problem can arise if somehow implementation-wise __LINE__ is same as the __builtin_LINE() function. This will always report true. IMHO they should be static line numbers to test.

[cpplearner]

Agreed that static line numbers can be useful, but IMO they should be part of compiler tests. The library doesn't need to detect bugs with __LINE__ (or __builtin_LINE), and even if such bugs are detected, there's no way to fix them in the library.

[bilgili]

I dont have a very strong opinion but I could catch the issue in my unit test which was checking a static line number. At least an additional static_assert may be beneficial. But I also agree compiler tests should definitely catch any regression.

[StephanTLavavej]

The preprocessor lives at a different "level" than source_location, so I think that this is fine for a library test.