How to debug ssa optimizer

[Zack Weinberg]

kenner@vlsi1.ultra.nyu.edu (Richard Kenner) writes:

>     Ah yes, the "Fold *&B to B" case?  Sure, I'd be happy with a check
>     added there that we don't return the wrong type.
>
> Yup.  But then if it falls through, verify_stmt will fail since the
> operand isn't valid.  So, as I said, there are still some holes here.

Maybe the canonical form for "char *foo, bar[10]; foo = bar;"
should be 
<modify_expr <var_decl foo>  <nop_expr type <char *> <var_decl bar>>
instead of using addr_expr?

zw