Re: Some 64bit issues.

[Marius Vollmer]

Rob Browning <address@hidden> writes:

> In order to get the code to compile without warnings, I had to hack a
> few occurences of SCM_FIXABLE(x) and replace them with "1" because
> otherwise the compiler complains about the test always being true due
> to the limited range of the data type.
> 
> Marius suggested a fix for this a while back, but I didn't quite
> understand it at the time.  Marius, do you still know what you want
> done here?  If it's not too hard I'd be happy to fix that too, right
> now.

Hmm, the fix I have now applied to num2integral.c.i is quite involved.
Not really tricky but verbose.  If this is needed in more places, I
think we should provide SCM_FIXABLE_SHORT, SCM_FIXABLE_INT,
SCM_FIXABLE_LONG etc and use that in the appropriate places.  These
macros could be defined to do the right thing depending on the size of
the types and our fixnums.

> Compiling with --enable-error-on-warning pointed out most of the
> bugs on the alpha, so I'd really like to get our code "warning clean".

Yes, that's a good goal.