Disable all floating point functions for libc 2.1.0 on GCC12

[salkinium]

modm is switching to avr-gcc 12.2.0 but the compiler now generates infinite loops when calling sinf or sin. There is no warning or error emitted, so this is a very dangerous bug.

The libc 2.1.0 has proper function prototypes for the floating point functions, so we don't need to specify them here anymore.

I'm still not sure why this didn't occur with avr-gcc 11.2 or avr-gcc 10.3.

cc @ckormanyos @chris-durand

[ckormanyos]

Great catch. The recursive call surprises me also. I guess it really does make sense to simply plow ahead before digging through the myriad of defines and inline functions with varying and different <cmath> and ´<math.h>` headers.

Nice correction!

[chris-durand]

I know what happened. avr-libc 2.1.0 defines the double math functions as aliases for the float variant in math.h:

__ATTR_CONST__ extern float sinf (float __x);
__ATTR_CONST__ extern double sin (double __x) __ASM_ALIAS(sinf);        /**< The alias for sinf().  */

Thus, our duplicate definition of sinf calling sin effectively ended up calling itself recursively.

[chris-durand]

I'm still not sure why this didn't occur with avr-gcc 11.2 or avr-gcc 10.3.

Are you sure the compiler was built against avr-libc 2.1.0? Assuming you tried this on your Mac OS machine it is likely the compilers had 2.0.0.

[salkinium]

Ah, yes that's probably why!