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

[Joseph Myers <joseph at codesourcery dot com>]

On Wed, 15 Nov 2017, Florian Weimer wrote:

> > As far as the compiler can see, the __sigsetjmp call makes the jmp_buf
> > contents escape, and at any subsequent point (before the function returns
> > or a scope with a variably modified type is left) the function might
> > return again - and if this is when the containing scope has reentered, the
> > warning is meant to warn, not deduce that in fact that case does not
> > occur.  Some more precise way of describing the possible times of a second
> > return would be needed to make it valid not to warn.
> 
> Still not convinced about this, and what's worse, I don't understand why you
> come to a different conclusion.  This shouldn't be a matter of opinion. 8-/

Given that the compiler knows nothing about the semantics of the functions 
in question, beyond the returns-twice property of __sigsetjmp, 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.  This is 
a matter of heuristics for what possible problematic returns should be 
considered by the warning.

-- 
Joseph S. Myers
joseph@codesourcery.com