Re: Display of read/access watchpoints when HAVE_NONSTEPPABLE_WATCHPOINT

[Andrew Cagney <cagney at gnu dot org>]

> > Date: Mon, 03 May 2004 11:05:18 -0400
> > > From: Andrew Cagney <cagney@gnu.org>
> > >
> > > Check these bug/threads threads for background info.
> > >
> > > http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gdb&pr=967
> > > http://sources.redhat.com/ml/gdb-patches/2002-09/msg00739.html
>
>
> Thanks for the pointers.

Also:
http://sources.redhat.com/ml/gdb-patches/2002-09/msg00738.html

> Having read them, I wonder: what targets can benefit from this change?
> As similar discussuions in the past indicated, targets such as the
> i386, which have a small number of debug registers, can nevertheless
> support an infinite number of watchpoints if they watch the same
> memory region.  So, in general, for such a machinery to be useful on
> x86 at least, we need to pass more info about the watchpoints to get
> meaningful results.

I believe that was the conclusion last time.

> In any case, I think this particular issue if whether a target could
> accomodate all the watchpoints that GDB wants to insert is not a very
> important one.  Currently, GDB simply tries to insert them all, and if
> that fails, it barfs.  If the procedure to find out whether they could
> be all accomodated is going to be as complex as actually inserting
> them, nothing is gained from making TARGET_CAN_USE_HARDWARE_WATCHPOINT
> more accurate than it is now.

Andrew