[ tree-ssa] PRE problem

[Daniel Berlin]

>> We do, however, insert these expressions later if their safety has 
>> been
>> verified
> And thus you have to make sure the annotations are updated.  Otherwise
> you can get incorrect code, segfaults, etc.  It is absolutely critical
> that the annotations be updated.
Well, these specific annotations didn't exist when the code was 
written. :)
In fact, neither did may_propagate_copy or propagate_copy, for that 
matter.
So whoever introduced the annotations (I think they came from Diego's 
flow sensitive aliasing stuff) should have added the code to update 
them in PRE.

I'll work up a patch to do this in a moment.  But your case still 
shouldn't abort when you add those lines, or at least, not because the 
memory tags are different, because those memory tags should *never* be 
different.