free_stmt_vec_info was releasing the LHS of a statement from the *wrong* 
statement.  Thankfully, the LHS it was actually releasing was just 
released by some earlier code, so nothing bad happened (the name manager 
explicitly allows multiple releases of the same node -- something we may 
want to change).

As with prior patches in this series, I have a reduced testcase that 
we'll try to utilize in some way.

Bootstrapped and regression tested on x86_64-linux-gnu.  Installed on 
the trunk.

Jeff