Re: Organization of breakpoint locations

[Daniel Jacobowitz <drow@false.org>]

On Mon, Feb 19, 2007 at 01:54:27PM +0100, Thomas Neumann wrote:
> > these?  Your key would not be the address, but the address plus the
> > breakpoint sequence number.  That handles one and two.  Then you can
> >   
> good idea. I thought that a breakpoint could have more than one
> bp_location associated with it, preventing such a scheme (but even then
> a breakpoint would probably not have more than one location with the
> same address).

It can't yet have more than one bp_location, but the separation is
there for a reason - I think the one to many mapping will arrive
sometime this year.  But I think we can add a sequence number to
bp_location's just like breakpoints have.

> > To be honest, I don't think this will be a small change either way.
> >   
> I will give the libiberty splay a try, that should be easy. But if
> resetting breakpoints is indeed as slow as you indicated, by patch is
> probably pointless. What a shame. That would have been a very portable
> and nice way to automatically trace execution flow.

I'd like GDB to be able to manage huge breakpoint lists; I'm just
warning you that a lot more work will have to be done first ;-)

One nice thing is that if you kill or crash GDB today, it tends 
to leave breakpoints removed.  If it didn't do this huge massive
removal, it probably wouldn't - perhaps we should add a signal handler
that takes care of that if we change it.

-- 
Daniel Jacobowitz
CodeSourcery