[Bug c++/12963] Wrong and misleading warning encourages writing non-portable code

[bagnara at cs dot unipr dot it]

------- Additional Comments From bagnara at cs dot unipr dot it  2003-11-08 22:44 -------
(In reply to comment #1)
> The warning is correct for this implementation and since you are compiling
with this 
> implementation, you can recieve warnings (yes it would be better if it was
enabled by a -
> W* but I think it is good warning no matter what).  To disable the warning use
-w (notice 
> the lower case w).

If I understand correctly, -w inhibits all warning messages, which is
certainly not what I call for.  The fact that, when producing code for
platforms where CHAR_BIT == 8, the test is redundant is a matter for
the optimizer (who can omit the test), not something the user should
be warned about.  In general, I believe target-dependent warnings
should default to disabled.  By the way, are there other target-dependent
warnings in GCC?  I mean that this is the first time I meet one (unless
I am missing something).

Now try with

void*
foo(void* c) {
  if (c <= (void*) 0xFFFFFFFF)
    return 0;
  else
    return c;
}

and you will see that (notice -W -Wall)

g++ -W -Wall -c t.cc -O3 -fomit-frame-pointer -S

does not produce any warning.  Yet, the produced code is,

.globl _Z3fooPv
	.type	_Z3fooPv, @function
_Z3fooPv:
.LFB3:
	xorl	%eax, %eax
	ret

as expected.
Would you advocate a warning also in this case?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12963