Chad Condon

> why even use a global here? I agree, but I think the use of `volatile` may still be necessary to prevent the loops from being optimized out.

So, rather than removing volatile, I suggest replacing `f *= f;` with `f = f * f;`.

OK, I'm going to try to translate whatever is missing from the old travis config into GitHub actions.

I suspect the automatic fenv support check isn't working as intended here. I'll get a windows/gcc build going in CI and see if I can repro.

Oddly, I can reproduce this in MSYS2 with [autotools](https://github.com/cpputest/cpputest/runs/8120653269?check_suite_focus=true), but not [CMake](https://github.com/cpputest/cpputest/runs/8120652956?check_suite_focus=true). #1623

> I recompiled on MSYS2 mingw64 Was this an incremental rebuild _after_ pulling CppUTest updates? You may need to delete your build directory (/home/jacosta/cpputest/cpputest_build) and reconfigure from scratch.

Which of `_WIN32` and `MINGW_HAS_SECURE_API` is not defined in your toolchain?

Would you mind posting the output of your CMake configure step? (`-- The C compiler identification is ...`, etc.)

Yeah, I haven't found a fix yet for this. Good find on #1381. From the limited investigation I've done so far, it looks like static initialization for the tests is...

I just noticed that this also affects `ENV`: ```dockerfile ENV \ # comment VAR=value ```