On 04/01/11 15:52, Karel Zak wrote: > On Tue, Jan 04, 2011 at 03:04:14PM +0000, PÃdraig Brady wrote: >> On 04/01/11 14:37, Karel Zak wrote: >>> On Tue, Jan 04, 2011 at 11:53:21AM +0000, PÃdraig Brady wrote: >>>> On 01/01/11 22:56, Karel Zak wrote: >>>>> On Sat, Jan 01, 2011 at 08:59:02PM +0000, Jon Grant wrote: >>>>>> Just building latest code and saw the compiler generated some >>>>>> warnings. Just thinking, one idea is to check the return value, and >>>>> >>>>> the warn_unused_result for write() in glibc is pretty idiotic... >>>>> >>>>>> use it as best possible (even just for trace) or (void)bytes to make >>>>>> it go quiet when there really is no use. >>>>> >>>>> (void) foo() is mess and it has no effect for warn_unused_result >>>>> >>>>>> bytes = write(1, "\n", 1); >>>>> >>>>> hmm.. possible solution is to call error() (but I'm not sure if this >>>>> is appropriate solution for unimportant things like /etc/issue) or >>>>> really ignore the return value. It would be nice to have a macro that >>>>> allows to ignore the return values in some cases >>>>> >>>>> ignore_result( write(1, "\n", 1) ); >>>> >>>> We use something like that in coreutils: >>>> http://git.sv.gnu.org/gitweb/?p=gnulib.git;a=blob;f=lib/ignore-value.h;hb=HEAD >>> >>> Nice, but it does not work for all types, what about: >>> >>> #define ignore(x) ({ typeof(x) __rc = (x); }) >> >> That looks right for util-linux but >> is probably not portable enough for gnulib. > > Good point. > >> Also I'd be slightly worried that clang-analyzer >> would start complaining about dead stores with that. > > It seems that clang is not so pedantic, but gcc -Wall prints warning > about unused variable â__fooâ, so it's necessary to use the variable > somehow... > > $ cat a.c > #include <stdio.h> > > static int __attribute__((warn_unused_result)) foo(void) { return 0; }; > > #define ignore(x) ({ typeof(x) __foo = (x); (void) __foo; }) > > int main(int argc, char **argv) > { > ignore( foo() ); > return 0; > } > > > $ scan-build gcc -c a.c > ANALYZE: a.c foo > ANALYZE: a.c main > scan-build: Removing directory '/tmp/scan-build-2011-01-04-1' because it contains no reports. Thanks for confirming that. The above should work for aggregate types also, so looks good for util-linux >> Perhaps this might work? >> >> static inline void _ignore (int i) { (void) i; } >> #define ignore(x) _ignore ((int)x) > > I don't think so, it will produce warnings about cast from pointer to > integer if the function (foo() in my example) returns a pointer. Ah right, there would be a warning when the pointer size is different to the int. So this would be better: #include <stdint.h> static inline void _ignore (intptr_t p) { (void) p; } #define ignore(x) _ignore ((intptr_t)x) But that still doesn't work for aggregate types. cheers, PÃdraig. -- To unsubscribe from this list: send the line "unsubscribe util-linux" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
