c99/c++ localised variable definition

[Paul Richards]

On Mon, Jan 31, 2005 at 05:31:17PM +0100, Ulrich Spoerlein wrote:
> On Mon, 31.01.2005 at 13:36:09 +0100, Poul-Henning Kamp wrote:
> > >If you carelessly c++-ify a loop like:
> > >
> > >	for (int i = 0; i < N; i++)
> > >	{
> > >		if (some_condition(i)) break;
> > >	}
> > >	do_something_with(i);	/* use finishing index */
> > >
> > >you can miss the fact that the value of i is used outside of the
> > >loop.  The newly created scope for "i" shadows the presumably
> > >pre-existing definition of i at the top of the function, which
> > >is what do_something_with() gets to see.
> > 
> > I would _really_ hope we have the compiler warning about this
> > already ?
> 
> Doesn't look so:
> #include <stdlib.h>
> #include <stdio.h>
> 
> int
> main(int argc, char **argv) {
>   int N = 42;
>   int i;
>   for (int i = 0; i < N; i++)
>     if (i == 23)
>       break;
>   printf("%d\n", i);   /* use finishing index */
>   return (0);
> }
> 
> % cc -Wall -std=c99 test.c && ./a.out
> 1

gcc should be throwing an uninitialised warning here.

> % icc -Wall -std=c99 test.c && ./a.out
> test.c(12): remark #592: variable "i" is used before its value is set
>     printf("%d\n", i);   /* use finishing index */
>                    ^
> 0
> 
> But the ICC warning is bogus too, if you happen to set i before, the
> warning disappears.

icc looks correct to me since the i being printed out is the one
in the outer scope which is unitialised.

-- 
Paul Richards