`<source_location>`: Investigate EDG issues

[StephanTLavavej]

We have a workaround for internal VSO-1285779 "REPORTED: EDG rejects source_location intrinsics within consteval functions" in the test:

STL/tests/std/tests/P1208R6_source_location/test.cpp

Line 4 in 969b2e2

#if defined(__cpp_consteval) && !defined(__EDG__) // TRANSITION, VSO-1285779

This was resolved as fixed on 2021-09-07 by internal VS-PR-348924 in VS 2022 17.0. However, attempting to remove this workaround with VS 2022 17.3 Preview 3 fails with numerous compiler errors. We need to investigate and report these issues.

[cpplearner]

EDG seems to have trouble computing the correct location.

D:\test>type test-edg-srcloc.cpp
#include <source_location>

template<unsigned> struct A {};

A<std::source_location::current().line()> x = 0;

D:\test>cl /std:c++20 /BE test-edg-srcloc.cpp
Microsoft (R) C/C++ Optimizing Compiler Version 19.33.31627.1 for x64
Copyright (C) Microsoft Corporation.  All rights reserved.

test-edg-srcloc.cpp
"test-edg-srcloc.cpp", line 5: error: no suitable constructor exists to convert
          from "int" to "A<26U>"
  A<std::source_location::current().line()> x = 0;
                                                ^

Note that it thinks std::source_location::current().line() at line 5 evaluates to 26U, maybe because current is defined at line 26 in <source_location>.

[frederick-vs-ja]

It seems that this issue can be closed as the workaround was removed by #3866.

[StephanTLavavej]

Agreed. All remaining EDG checks in the test are either column number variation (which we don't care about) or are actually guarding an MSVC bug. Thanks!