This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: [rfc;rfa:breakpoint] Pass full breakpoint/watchpoint count totarget
- From: Andrew Cagney <ac131313 at redhat dot com>
- To: Eli Zaretskii <eliz at is dot elta dot co dot il>
- Cc: gdb-patches at sources dot redhat dot com
- Date: Mon, 30 Sep 2002 12:59:57 -0400
- Subject: Re: [rfc;rfa:breakpoint] Pass full breakpoint/watchpoint count totarget
- References: <Pine.SUN.3.91.1020930073939.25050K-100000@is>
On Sun, 29 Sep 2002, Andrew Cagney wrote:
The attached patch changes the target vector so that a count of each
type of hardware watchpoint/breakpoint is passed to the
target_can_use_hardware_breakpoints() function. (Anyone got a better
name for the function?).
This lets targets make a better guess at how many watchpoints are going
to be needed.
Sorry, I don't see the large-scale picture: how will this facility be
used in routine GDB operations by the application-level GDB code?
Also, on a i386, for example, if you don't pass the size and alignment of
each address to be watched, the guess of the amount of required resources
can be very wrong. This is even more exacerbated due to debug register
sharing implemented for i386.
Yes, see my reply to the other e-mail. I think we can do slightly
better with the current model. However, to make things optimal we'll
need to totally overhaul things.
The intent is for the Z? packet to pass these totals down
to the target.
Do you mean to say this is only for remote targets?
Other architectures can use the info. Its just that remote will be first.
Andrew