Re: Aliasing brokenness (Was: Re: [patch RFC] SH: Use FRAME_GROWS_DOWNWARD)

[Kenneth Zadeck <zadeck at naturalbridge dot com>]

Mark Mitchell wrote:
> Andreas Schwab wrote:
> > Gabriel Dos Reis <gdr@integrable-solutions.net> writes:
> >
> >
> > > Is the conclusion that a presence of "potentially unused" union that
> > > contains two structures with the same prefix destroys non-aliasing
> > > assumptions?
> >
> >
> > IMHO the third example in 6.5.2.3 gives a strong hint at that.  AFAIU 
> > the
> > second fragment becomes valid as soon as declaration of the union is 
> > moved
> > before function f.
>
> Exactly.
>
> I think it's hard to read 6.5.2.3 to say anything other than that the 
> mere visibilty implies that sturctures contained in the union with 
> common initial sequences may alias.  I don't particularly like that 
> rule, but I think that's what it says.
And can we assume that visibility does NOT mean in some other 
compilation unit.