Re: [Bug-gnulib] Re: xsize.h needs SIZE_MAX defined on Windows

[Mark D. Baushke]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Paul Eggert <address@hidden> writes:

> Derek Robert Price <address@hidden> writes:
> 
> > Mark D. Baushke from the CVS development team suggested that ~0 might be
> > a better value than -1 as it is more obviously positive.
> 
> But ~0 == 0 on hosts with ones'-complement arithmetic; and on such
> hosts, (size_t) ~0 == 0.  There are C compilers for some
> ones'-complement hosts (though admittedly they're a bit unusual as
> porting targets for GNU code these days :-).

On such machines, I would expect that ((size_t) -1) is also a
problematic assignment (I must admit that I have mostly used
twos-complement machines).

> Anyway, I don't see why the fix is needed at all.  The gl_XSIZE
> autoconf macro invokes gl_SIZE_MAX, which should arrange for SIZE_MAX
> to be defined.  The advantage of this approach is that the resulting
> SIZE_MAX can be used in C preprocessor expressions.

Yes, that would be a good thing.

        -- Mark
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (FreeBSD)

iD8DBQFAULuo3x41pRYZE/gRAgRMAJ4yjukMW7ct58fi48EOxbyQKu04XgCgkp/L
2Po8qKpZLLnhmQg48FPb1nQ=
=eBmQ
-----END PGP SIGNATURE-----