Re: using the interfaces in ctype.h

["Greg A. Woods; Planix, Inc." <woods%planix.ca@localhost>]

On 22-Apr-08, at 3:51 AM, Alan Barrett wrote:
> Can anybody confirm the existence of compilers such that
> this code:
>
>         #include <ctype.h>
>         int f(void)
>         {
>             signed char c = -1;
>             return isupper((unsigned char)c);
>         }
>
> produces a warning which can be quieted by an additional cast to int,
> thus:
>
>         return isupper((int)(unsigned char)c);
>
> ?

Potentially some compiler might complain about the unsigned char to  
signed int promotion without the explicit (int) cast, though  
apparently GCC doesn't (at least not so long as UCHAR_MAX <= INT_MAX).

I'll try to get access to IRIX and Solaris machines again to try  
these.  IIRC it was one of these I was seeing warnings from, and most  
likely it was IRIX, perhaps with whatever sensible warning options  
their compiler supports(*).  Unfortunately that's the one I'm least  
likely to be able to get access to again so anyone else who has access  
to IRIX systems should try it.

(*) In Smail-3 I used the following on IRIX-6.x last time I tested it  
there:

        CFLAGS="-Xa -g3 -O3 -fullwarn -diag_suppress 1184,1209,3201,3496"
        LDFLAGS="-g3 -O3"


BTW, can anyone say with any degree of certainty which, if any, C  
implementations (hosted) will extend the sign when the value of a  
"char *" pointer is assigned to a plain "int"?  I.e. in "int i = *s;"

DMR says (in the second edition of "The C Programming Language", p. 
44):  "On some machines a char whose leftmost bit is 1 will be  
converted to a negative integer ("sign extension")."  In the first  
edition he also said "(PDP11, for instance)".  Harbison and Steele, in  
"C  A Reference Manual" say much the same too of course, though never  
in exactly the context of using the result of indirecting a "char *"  
pointer.  DMR uses the "int c; char *s; while ((c = *s)) ..." idiom at  
least once (p.156), though perhaps that was a mistake.

--
                                        Greg A. Woods; Planix, Inc.
                                        <woods%planix.ca@localhost>