Richard Stallman <[EMAIL PROTECTED]> writes:
> In Emacs we have a macro that tends to produce the warning
>
> comparison is always false due to limited range of data type
>
> when compiling for 64-bit machines. EMACS_INT is 64 bits
> in that case. Here is the macro:
>
> #define FIXNUM_OVERFLOW_P(i) \
> ((EMACS_INT)(i) > MOST_POSITIVE_FIXNUM \
> || (EMACS_INT) (i) < MOST_NEGATIVE_FIXNUM)
>
> When this macro is used on an expression that has type int, it
> produces those warnings, and there is no way to change the macro to
> prevent the warnings. We used the workaround of copying the
> expression into an EMACS_INT in a previous statement, but that is
> cumbersome and ugly.
gcc has issued this warning for quite a while. It has been around at
least since this change:
Sun Apr 17 01:21:35 1988 Richard Stallman (rms at frosted-flakes.ai.mit.edu)
* typecheck.c (shorten_compare): Warn about constant result only in
cases like (char)x < 0x80, where old C compilers made it -0x80.
I'm not sure whether anything has changed in this area to cause the
warnings to appear more frequently today.
> Is there a way to write the macro so as to suppress this warning?
You can avoid it by using unsigned types. I think that something like
this will do the trick:
#define FIXNUM_OVERFLOW_P(i) \
((unsigned long long)(i) > MOST_POSITIVE_FIXNUM \
&& (unsigned long long)(i) < MOST_NEGATIVE_FIXNUM)
Ian
