On Wed, Dec 18, 2002 at 03:11:59AM +0000, Bruce Gray wrote: > 2) PTR_SIZE == INTVAL_SIZE == LONG_SIZE (x86 Linux and Win32) > Intermediate casts to (unsigned long) removed, i.e. > (any)(unsigned long)(d) > becomes > (any)(d)
> Under MS VC++ for .Net, this patch changes the results of > none of the tests; it only stops the "UINTVAL_SIZE is not > defined as a preprocessor macro" warnings that are given > with the -Wall option. x86 is never the best platform to test any of this on. Crays are best, failing that any vendor compiler on a big endian 64 bit platform should be fairly effective [1]. However, they're all too costly for me. On Tue, Dec 17, 2002 at 08:31:05PM -0800, Steve Fink wrote: > That all sounded sensible to me. (And tests still pass here.) Thanks, > applied. Is the tendra compiler on one of the tinderboxes choking on removing the intermediate (unsigned long) cast? I think it's needed for the hack of mapping a function pointer to a data pointer and back. Is the tendra compiler free? I've tried to find it before, but couldn't manage to get a functional download URL. Nicholas Clark 1: except HP's FORTRAN 77 compiler sets all unitialised values to 0, which doesn't make it the best thing for writing your portable code on. IIRC -- Cat herding better than perl? http://www.perl.org/advocacy/spoofathon/
