Reducing the cost of exception handling

[Jeffrey A Law]

  In message < u94t34423j.fsf@yorick.cygnus.com >you write:
  > >>>>> Bruno Haible <haible@ilog.fr> writes:
  > 
  > >  - How do debuggers react on the fact that a function consists of two
  > >    totally disjoint address ranges?
  > 
  > Badly, I'm afraid.  In particular, DWARF 2 assumes that a function is one
  > address range, and so unwinding will break with this patch.
Or more correctly, they don't handle it at all.

jeff