Option to make GCC conform to C99 re:floating-point

The existence of this option to make GCC conform to C99 when compiling for the x87 floating-point instruction set is informative. The commentary that comes with the patch is quite interesting, if you are into this sort of thing.

If you are not that much into the minutiae of the implementation of C99's floating-point on x87 hardware, on the other hand, don't follow this link. The short story not contained there is that you can always use the SSE2 instruction set to get even more standard floating-point semantics and that you can always use Frama-C's value analysis option -all-rounding-modes to analyze x87 floating-point programs for all possible compilation choices even if the program is compiled without the option described there.

(Both these statements are true to the best of my understanding but I wouldn't sit under a piano hanging on them)