Re: Reminder: Illegal Package-Private Accesses in libgcj

[Andrew Haley <aph at redhat dot com>]

Bryce McKinlay writes:
 > 
 > On Nov 4, 2003, at 11:15 AM, Andrew Haley wrote:
 > 
 > > Bryce McKinlay writes:
 > >>
 > >> On Nov 4, 2003, at 11:01 AM, Andrew Haley wrote:
 > >>
 > >>> We look it up in a hash table.  We sure don't call external programs!
 > >>
 > >> But you can't map a backtrace IP directly to a Class in the hashtable,
 > >> you have to find the function entry point first. I don't know of a way
 > >> to do that apart from using _Unwind_FindEnclosingFunction.
 > >
 > > Yes, that's why I wrote _Unwind_FindEnclosingFunction.
 > 
 > So what are we arguing about? ;-)

At last, I see where the misunderstanding occurs.

There are two separate issues here.

1.  How we walk the stack.
2.  How we turn an IP into an entry point.

When you wrote about using the unwinder, I thought you were writing
about 1, but it seems you were writing about 2.  I guess I don't think
of _Unwind_FindEnclosingFunction as part of the unwinder, despite its
name.  :-)

We might want to continue to use backtrace() on some targets.  Maybe
not, I don't know -- it might be better to use the unwinder
everywhere.  We should probably use the unwinder on AMD 64, but at the
moment we force no-omit-frame-pointer.

Andrew.