The Safe module did not properly restrict the code of implicitly called
methods (such as DESTROY and AUTOLOAD) on implicitly blessed objects
returned as a result of unsafe code evaluation. These methods could have
been executed unrestricted by Safe when such objects were accessed or
destroyed. A specially-crafted Perl script executed inside of a Safe
compartment could use this flaw to bypass intended Safe module
restrictions.
