On Fri, Feb 4, 2011 at 12:32 PM, Paul Pluzhnikov <ppluzhnikov@google.com> wrote:

>> Maybe the cache could point to the symbols themselves.  I think that
>> would fix the problem and be both safe and efficient.
>
> Ah, that also nicely takes care of the problematic use of -1 as invalid
> core address -- we *know* that 0 and -1 are *not* valid symbol values.

Here is the updated patch (applies on top of the [1/2] cosmetic patch).
Re-tested on linux/x86_64, no regressions.

Thanks,
-- 
Paul Pluzhnikov


2010-02-04  Paul Pluzhnikov  <ppluzhnikov@google.com>

	* breakpoint.c (longjmp_names): New variable.
	(struct breakpoint_objfile_data): New type.
	(breakpoint_objfile_key): New variable.
	(msym_not_found): New variable.
	(msym_not_found_p): New predicate.
	(get_breakpoint_objfile_data): New function.
	(create_overlay_event_breakpoint): Check per-objfile cache for
	symbols first.
	(create_longjmp_master_breakpoint): Likewise.
	(create_std_terminate_master_breakpoint): Likewise.
	(create_exception_master_breakpoint): Likewise.
	(_initialize_breakpoint): Register per-objfile data key.