Re: [PATCH] Fix a recent warning in reorg.c

[Jeff Law <law at redhat dot com>]

On 04/26/2016 07:08 AM, Jakub Jelinek wrote:
> Hi!
>
> I've noticed a warning during bootstrap:
> ../../gcc/reorg.c: In function âvoid try_merge_delay_insns(rtx_insn*, rtx_insn*)â:
> ../../gcc/reorg.c:1431:12: warning: name lookup of âiâ changed
>        for (i = 0; i < XVECLEN (PATTERN (insn), 0); i++)
>             ^
> ../../gcc/reorg.c:1263:7: warning:   matches this âiâ under ISO standard rules
>    int i, j;
>        ^
> ../../gcc/reorg.c:1413:25: warning:   matches this âiâ under old rules
>        for (unsigned int i = len - 1; i < len; i--)
>                          ^
> It is not fatal, but still ugly.  The problem is that the function has
>   int i;
> ...
>   for (i = 0; ...)
> ...
>   for (unsigned int i = ... )
> ...
>   for (i = 0; ...)
> This patch just declares the var in the only affected loop, so that the warning
> is not emitted, unless we start checking -Wshadow warnings, I think this is
> good enough.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
>
> 2016-04-26  Jakub Jelinek  <jakub@redhat.com>
>
> 	* reorg.c (try_merge_delay_insns): Declare i inside the last
> 	for loop to avoid warning.
I'd like to declare this in the "obviously OK" category for future 
changes of a similar nature.  But I don't think we can because in the 
general case the value from the loop may be used outside the loop.

Makes me wonder if a plugin could help identify the obvious cases where 
the value from the loop isn't used outside the loop, thus allowing 
someone to quickly convert many of these things with a small effort.

jeff