This is the mail archive of the
gdb@sources.redhat.com
mailing list for the GDB project.
watchpoint fixes
- From: Paul Koning <pkoning at equallogic dot com>
- To: gdb at sources dot redhat dot com
- Date: Thu, 22 May 2003 16:04:27 -0400
- Subject: watchpoint fixes
Gentlepeople,
A few weeks ago I reported problems in the way hardware watchpoints
work in gdb -- or rather, don't until patched. I described a fix but
have heard no reaction. Should I submit that as a patch to the
patches list? Note that I'm just a gdb observer...
I've noticed some more problems along similar lines: if you set a
watchpoint, and then the program hits a break instruction that doesn't
match a breakpoint set by gdb, the program doesn't stop. The reason
is that it falsely matches the watchpoint but since the data address
doesn't match, the stop_status ends up "keep going". The solution is
simple: if the stop was not a watchpoint (there isn't a data address
from the target) then pretend watchpoints don't exist at all, rather
than marking them as "keep going".
I could submit a patch for that too if there's interest. Suggestions?
paul