problems with "clobbered" warnings

[Zack Weinberg]

On Thu, Jun 22, 2000 at 05:19:00PM -0400, Michael Richardson wrote:
> 
>   I am attempting to get some code to compile with -Wall -Werror. I have just
> introduced a setjmp() to catch errors...
> 
>   I get warnings like:
> 
> paxas.c:211: warning: variable `infilename' might be clobbered by `longjmp' or `vfork'
...
> 
>   Is there someone who can offer a better description of when a variable may
> be considered to be clobbered, or a workaround that permits me to shut up
> just that test without giving up on the uninitialized warning?

The compiler generates those warnings for just about every variable in
a function that calls setjmp or vfork.

Your best option is not to use setjmp() at all.  Have a
cleanup_and_exit() that cleans up and calls exit; mark it noreturn;
and call it directly whenever you would have called longjmp.

If that's not feasible for some reason, you may be able to structure
your code like this:

jmp_buf env;

int
main(void)
{
  if(setjmp(env))
    {
      cleanup();
      return 1;
    }

  do_all_the_real_work();
  return 0;
}

- the key being to have no actual code in the routine that calls
setjmp, just function calls.

zw