This is the mail archive of the
sid@sources.redhat.com
mailing list for the SID project.
Re: Possible remote.c patch for Z-packet breakpoints + Harvard + SID
- To: nsd at redhat dot com (Nick Duffek)
- Subject: Re: Possible remote.c patch for Z-packet breakpoints + Harvard + SID
- From: fche at redhat dot com (Frank Ch. Eigler)
- Date: 16 Jul 2001 18:05:43 -0400
- Cc: cagney at redhat dot com, jtc at redback dot com, sid at sources dot redhat dot com, gdb-patches at cygnus dot com
- References: <o5pub0sqqy.fsf@toenail.toronto.redhat.com><200107161931.f6GJVV719853.cygnus.project.sid@rtl.cygnus.com>
nsd wrote:
: [...]
: I see your point, though I think it's debatable. For one thing, Z-packet
: breakpoints are tagged as "Z_PACKET_SOFTWARE_BP" in gdb/remote.c, so it
: looks like they're intended to be a generic protocol speedup rather than a
: hardware breakpoint mechanism.
: [...]
That's true - it makes me rethink my earlier point.
: I proposed a patch in January to call BREAKPOINT_FROM_PC() before rather
: than after remote.c stores hardware breakpoint addresses in outgoing
: packets, but it was decided that such a change should be at a higher
: level, e.g. in target_insert_breakpoint().
:
: This problem can't be fixed in target_insert_breakpoint(), because the
: distinction between "hardware" (Z-packet) and software breakpoints is made
: in remote.c.
: [...]
If this gdb-side approach is not deemed acceptable to gdb folks, we
may be able to make complementary changes on the sid side without too
much littering.
One possibility: map hardware breakpoints to mask/value-type sid
"triggerpoints" on the PC register instead of plain value-type ones;
parametrize the gdb-interface-component with the desired mask for a
target. This sort of thing may be useful for PC-range breakpoints
too.
- FChE