Question on scalar replacement

[Richard Kenner]

    > Jason, is a.f = foo() valid GIMPLE?  If not, then the bug is in the
    > gimplifier.  Otherwise, the patch I sent should be OK.

    It's valid for aggregate_type_p type.

Nevertheless, I think Diego's patch is worthwhile on the grounds of "be
conservative in a what you generate and liberal in what you accept".  If
we're going to disallow a scalar CALL_EXPR on the RHS of a MODIFY_EXPR
with a compound LHS, we need to test for it and ICE if we see it.