[Bug objc++/49221] [4.7 Regression] Several ICEs in the obj-c++ test suite after revision 174307

["jason at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49221

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jason at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #8 from Jason Merrill <jason at gcc dot gnu.org> 2011-06-06 03:51:55 UTC ---
During EH lowering the back end wants to replace a TYPE with something else
that represents the type at runtime.  objc_eh_runtime_type generates such a
thing, and tries to generate a DECL, but we're in function scope at the time so
we try to add a DECL_EXPR for the DECL, which breaks.

We definitely shouldn't be trying to add a DECL_EXPR for a non-local variable. 
I suppose it's easy enough to work around this in cp_finish_decl, though there
might be other code that assumes that any new variable while we're in function
scope belongs to that scope.