Re: [PATCH] pthread_cleanup_push macro generates warning when -Wclobbered is set

[Florian Weimer <fweimer at redhat dot com>]

On 11/21/2017 06:35 AM, Paul Eggert wrote:
> Joseph Myers wrote:
> > the warning
> > seems justified to me, in that the second return of __sigsetjmp might
> > result in undefined behavior from variables being accessed after the
> > containing block has exited, which is what the warning is about.
>
> I'm with Florian on this. I don't see the warning as being justified.
>
> Basically the warning is saying "Watch out! Since these variables are 
> not declared to be volatile, they might be trashed by longjmp!" You can 
> work around such problems by declaring the variables to be volatile.
>
> Here, though, declaring the variables to be volatile would not fix the 
> problem that you describe. Because a longjmp makes them go out of scope, 
> they become uninitialized regardless of whether they're declared to be 
> volatile. So the warning is not appropriate here.

I think the standard assumes that storage for all local variables is 
allocated when the function is entered (or when a scope is entered with 
which contains a variable of variably modified type).  This is certainly 
an odd requirement.

Thanks,
Florian