Re: [PATCH] Fix -Os related -Werror failures.

[Jeff Law <law at redhat dot com>]

On 10/28/2016 02:10 PM, Paul Eggert wrote:
> On 10/28/2016 01:12 AM, Arnd Bergmann wrote:
> > I found that most often when gcc is confused about whether a variable
> > is uninitialized or not, the source code tends to be confusing to a
> > human reader as well and rewriting it differently results in better
> > readability and better object code while avoiding the warning.
>
> I'm afraid my experience has not been so good. Maybe 1/3 of the time
> rewriting is better, but otherwise rewriting doesn't help or even
> confuses the code. When that happens with -Wmaybe-uninitialized, in
> Emacs we typically use C declarations like this:
>
>       ptrdiff_t offset2 UNINIT; /* The UNINIT works around GCC bug
> 78081.  */
And I would echo that markup indicating that the initializer is to work 
around a GCC false positive is something I wish we had strictly enforced 
within GCC itself when it was made Wuninitialized clean.

GCC has made significant strides in its jump threading and predicate 
analysis to significantly such false positives and many of these 
initializers could probably be removed at this point.

> where UNINIT is defined something like this:
>
>   #ifdef GCC_LINT
>   # define UNINIT = {0,}
>   #else
>   # define UNINIT /* empty */
>   #endif
>
> and configuring with --enable-gcc-warnings compiles with something like
> 'gcc -Wall -Werror -DGCC_LINT'.\
But I would caution against blindly using 0 as an initializer.  Each 
case has to be examined to determine what a safe value would be.  Often 
0 is appropriate, but there are certainly cases where it is not the safe 
initializer to use.

Jeff