Re: RFA: Fix signbit() for 16-bit targets

[Nicholas Clifton <nickc at redhat dot com>]

Hi Craig, Hi DJ,

> > > +  return (w & 0x80000000) != 0;
> > Also 0x80000000 -> 0x80000000UL ?
> Given that uint32_t is being used, UINT32_C(0x80000000) would seem most
> appropriate, although this might mean stdint.h would have to be added.

In theory no cast is needed.  According to the ISO C standard an integer 
constant has:

  6.4.4 (5)
    The type of an integer constant is the first of the corresponding
    list in which its value can be represented.

    Suffix   Decimal Constant
     none        int
                 long int
                 long long int

So for targets where "int" is 32-bits 0x80000000 will be an int, but for 
targets where "int" is 16-bits 0x80000000 will be a long int.  (The 
signed-ness of the type should not matter here as we are only performing 
a binary operation on the value).

This also matches the practice.  At least for the 16-bit RL78 toolchain 
compiled with GCC 4.8.  The code, without a cast, works there.

Cheers
  Nick