Made has_function survive optimization flags. Closes #1053.

[jpakkane]

No description provided.

[nirbheek]

I'm fairly certain we should fix this by changing the order in which we add the -O0 to CFLAGS.

[jpakkane]

The tests themselves need to be robust against any combination of optimization flags. Otherwise this will keep breaking in subtly different ways over and over again.

[nirbheek]

On the contrary, trying to fix tests by trying to be smarter than a compiler is an arms-race against compiler developers that we cannot win. Meson will magically/mysteriously keep breaking with newer versions of compilers, or with different compilers.

The sane way out is to force the compiler to not do any optimization at all, which is what we try right now but we pass the flags in the wrong order so it gets overriden.

[jpakkane]

We should do both, because defense in depth is the only way to go.

[nirbheek]

Fair enough! I'll be back tomorrow, and will make a PR for that + unit tests.