gcj-security and some issues

Andrew Haley aph@redhat.com
Thu Apr 29 15:24:00 GMT 2004


Anthony Green writes:
 > On Thu, 2004-04-29 at 07:38, Andrew Haley wrote:
 > > I don't see how that would help.  We already have a efficient method
 > > for mapping a PC address to a class, so going from there to a
 > > protection domain is just one more memory fetch. 
 > 
 > I was just thinking: shorter binary search, and smaller meta-data.

That's true, of course.

But: we're using the DWARF unwinder, and that has to locate the start
of each function on the stack in order to unwind.  Once we have the
exact start address of a function, going from there to a class is a
small step -- a single hash table lookup.  Bryce's idea of putting the
class in the unwinder data would make it slightly faster, because we
wouldn't need the hash table.

Andrew.



More information about the Java mailing list