[libcxx-commits] [libcxx] [libcxx][test] Prevent compiler optimising out uses of 64-bit atomics (PR #172517)

[David Spickett via libcxx-commits]

DavidSpickett wrote:

gcc build failing with:
```
  # .---command stderr------------
  # | /home/gha/actions-runner/_work/llvm-project/llvm-project/libcxx/test/std/atomics/atomics.types.generic/atomics.types.float/wait.pass.cpp: In instantiation of ‘void test_impl() [with T = float; MaybeVolatile = std::__1::add_volatile_t]’:
  # | /home/gha/actions-runner/_work/llvm-project/llvm-project/libcxx/test/std/atomics/atomics.types.generic/atomics.types.float/wait.pass.cpp:113:38:   required from ‘void test() [with T = float]’
  # |   113 |     test_impl<T, std::add_volatile_t>();
  # |       |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~
  # | /home/gha/actions-runner/_work/llvm-project/llvm-project/libcxx/test/std/atomics/atomics.types.generic/atomics.types.float/wait.pass.cpp:118:14:   required from here
  # |   118 |   test<float>();
  # |       |   ~~~~~~~~~~~^~
  # | /home/gha/actions-runner/_work/llvm-project/llvm-project/libcxx/test/std/atomics/atomics.types.generic/atomics.types.float/wait.pass.cpp:41:5: error: implicit dereference will not access object of type ‘volatile std::__1::atomic<float>’ in statement [-Werror]
  # |    41 |     DoNotOptimize(a);
  # |       |     ^~~~~~~~~~~~~
```
So I'm not sure this code is reliable across compilers either.

https://github.com/llvm/llvm-project/pull/172517