[PATCH][alias-improvements] Fix PR39506 cause

[Daniel Berlin]

On Fri, Mar 20, 2009 at 6:16 PM, Richard Guenther <rguenther@suse.de> wrote:
> On Fri, 20 Mar 2009, Daniel Berlin wrote:
>
>> I still strongly disagree that we should expect 39506's testcase to
>> work or consider it a bug.
>> I don't have a problem if it happens to work, but i wouldn't go
>> specifically trying to make it work
>
> This is why I didn't include the testcase in the patch and closed
> the bug as invalid.

>
> Nevertheless I believe this particluar disambiguation has noting to
> do in TBAA pruning code nor does it (or at least should it) add any
> disambiguation in the first place.

Sure.

>
> So better remove it, also on the technical grounds that when we
> compute points-to sets for SSA_NAME pointers in a program in
> SSA form a rule about decl pointers looks strange.  We could
> side-step this by not rewriting these pointers into SSA form, but
> that would of course be stupid as well.

Agreed.
:)