Remove unneeded static_cast for xexp

[AreaZR]

We can just check for 0 since -0 is just 0

[StephanTLavavej]

Thanks!

• I agree that this change is correct. xexp is a local variable and it isn't mentioned after this if-statement (or address-taken), so conditionally performing the assignment is unobservable.
• The cast was presumably attempting to silence truncation warnings, because with short xexp, the expression -xexp triggers integer promotion, then assigning int to short may emit a truncation warning. However, this is in stl/src, and if the STL builds warning-free, there is no reason to have the cast here. (This may have been compiler-dependent, or the compiler may have gotten less aggressive about warning for expressions like val = -val;.)
• I believe that this change is desirable primarily because it gets the assignment out of the condition, which is an overly complicated practice that we avoid in new code.

[AreaZR]

I was actually considering making xexp an integer. All the tests pass if we make xexp an integer in this case and would certainly be faster to work on for RISC processors while also avoiding the problems of potentially needing casts.

However, I'm worried about tiny edge cases as I can't prove that at this point in execution, xexp is in range of a short, which would thereby change behavior, or if said behavior does change, whether such a change would have any adverse effect on the runtime.

So I decided to play it safe.

[StephanTLavavej]

So I decided to play it safe.

Agreed, thanks.

I've pushed an additional commit, as I remembered to check for additional affected locations (the float/double codepaths are often similar like this). Double-checked that the change is still safe and desirable here.

[AraHaan]

Is there other files with similar code as well?

[StephanTLavavej]

I searched for other files and I think we've got them all now, although I could have missed something.

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

[StephanTLavavej]

Thanks for this code cleanup! 😸 😸