clang std::source_location::current() fail

[PaltryProgrammer]

greetings kind regards clang std::current_location::current() results in error as shown MSVC compiles w/o error kindly advise Thank You Kindly

[EugeneZelenko]

Could you please create reproducer on https://godbolt.org ?

[PaltryProgrammer]

i do not know what a "reproducer" is . however w/ clang 15.0.0 as i did not see a 17.0.0 it seems to work as you can see

[fsb4000]

It's expected.

Microsoft STL requires __cpp_consteval to be defined for working source_location https://github.com/microsoft/STL/blob/e7f52446c6f41fddbed31904f594a1f711f85601/stl/inc/source_location#L11

And clang 15 doesn't fully implement cpp_consteval so it doesn't define that macro.

https://godbolt.org/z/soxeEaxnT

Even clang trunk doesn't define that macro: https://godbolt.org/z/K7rMfdW9E

[fsb4000]

Actually it looks like compiler devs just forgot to define that macro: #57094

but maybe some bugs still exist and compiler devs don't want to define the macro.

[llvmbot]

@llvm/issue-subscribers-c-20

[PaltryProgrammer]

if __cpp_consteval is required but not provided why was the compile shown here successful ?

[fsb4000]

if __cpp_consteval is required but not provided why was the compile shown here successful ?

Microsoft STL implementation checks that macro but libc++ and libstdc++ don't check.

(Your example uses libstdc++ , if you add -stdlib=libc++ then your example will use libc++ and as far as I know the compiler explorer can't use Microsoft STL together with clang)

[fsb4000]

if __cpp_consteval is required

std::current_location::current() needs __cpp_consteval 100%.

Look: https://en.cppreference.com/w/cpp/utility/source_location/current

It's consteval

[fsb4000]

Hi @AaronBallman !

Could you give us an update about the __cpp_consteval macro?

[PaltryProgrammer]

i merely added the macro and it seems to have worked

[AaronBallman]

Hi @AaronBallman !

Could you give us an update about the __cpp_consteval macro?

I know we're at least missing support for https://wg21.link/P2564R3 in terms of setting the value to 202211L, but we might be at the point to define it to 201811L. I know we're certainly close these days, but a quick look through the bug database shows ~40 open issues related to consteval (not all of these are going to be issues holding us back from claiming conformance though).

One thing that would help is if someone was able to run an experiment defining the macro to the 2018 value and trying to compile a bunch of large C++ relatively modern projects to see if that points out bugs we're not aware of.

[fsb4000]

Well, I have pretty slow PC (AMD FX8300) but I could run some projects too.

Do you have a list of such projects?

Also, look what I've found: chromium/subspace@e18b443#diff-eb105f54978bac9a1111fa159977cd7cf6d3372ce522c53019c294c4d5142f8eR95

and marzer/muu@25ff4e5#diff-9fee50b957f0636d37f5d5ed34361016c89b77729fdd415316d06598e1530ed1R412-R418

[AaronBallman]

Do you have a list of such projects?

Chromium was one such project, but you found something out there! :-D The other projects I usually use for this sort of thing (uncertain of how much consteval use they have, though) are Qt, kokkos, boost, and Folly.