PATCH RFA: PR 29286: Handle placement new aliasing issues

[Ian Lance Taylor]

Eric Botcazou <ebotcazou@adacore.com> writes:

> > As far as I can see that SMT would have to alias every other SMT in
> > order to correctly implement CHANGE_DYNAMIC_TYPE_EXPR, so I don't
> > think this would actually pessimize less.
> 
> I don't think so.  If you say that CHANGE_DYNAMIC_TYPE_EXPR is a special 
> escape site and trace back which variables escape at it (like for variables 
> that escape in the call-clobbered sense), you could only include those 
> variables in the SMT.

Seems plausible, but I'm not sure it's significantly simpler than what
I implemented.

Ian