[cfe-dev] 64 to 32 truncation warnings for ints and floats?

[John McCall]

On Nov 12, 2010, at 8:50 AM, Sean McBride wrote:
> In C & C++, Apple's gcc 4.2 produces 2 warnings for this code with -
> Wshorten-64-to-32:
> 
> int main(int,char**)
> {
> 	double d = 1.0;
> 	float f = d; // warning: 64 to 32
> 	
> 	long s1 = 10;
> 	long s2 = 20;
> 	int diff = s2 - s1; // warning: 64 to 32
> 	
> 	return 0;
> }
> 
> clang (trunk) only warns for the integer case.  Is there a way to get
> warnings for the double->float conversion too?

The generic "warn about any implicit conversions that might lose precision"
warning is -Wconversion.  Clang's implementation of that is similar to
gcc-4.3's, i.e. substantially better than gcc-4.2's.

-Wshorten-64-to-32 was never supposed to be more than a portability
warning, which is why I didn't warn about double->float in clang's
implementation.  Note that -Wshorten-64-to-32 is implied by -Wconversion.

John.