http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54351
--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-08-22 18:43:47 UTC --- Hmm, the behaviour was probalby correct prior to fixing PR 43183, as the old implementation of reset() did exactly what is required of the destructor. However, the lifetime of the unique_ptr ends when the destructor starts ([basic.life]/1) so I don't believe it is legitimate to access it at that point.
