This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [patch] Warn on constant value watchpoints
On Sun, 08 Jun 2008 20:49:25 +0200, Eli Zaretskii wrote:
> > Date: Sun, 8 Jun 2008 14:09:09 -0400
> > From: Daniel Jacobowitz <drow@false.org>
> > Cc: Jan Kratochvil <jan.kratochvil@redhat.com>, gdb-patches@sourceware.org
> >
> > On Sun, Jun 08, 2008 at 09:04:43PM +0300, Eli Zaretskii wrote:
> > > > (gdb) watch 0x4343548
> > > > Watchpoint 1: 70530376
> > >
> > > Should we allow such watchpoints? under what circumstances are they
> > > useful?
> >
> > In my opinion, we should not allow such watchpoints.
>
> I agree. Does anyone disagree? If not, Jan, could you rework your
> patch accordingly?
I hope there is no longer any disagreement but requesting an approval.
Thanks,
Jan
2008-06-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* breakpoint.c (watch_command_1): New variable VAL_RESULT. Fill in
VAL_RESULT by the existing FETCH_WATCHPOINT_VALUE call. Refuse
constant VAL_RESULT list watchpoints.
2008-06-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* gdb.texinfo (Set Watchpoints): Document constant value watchpoints.
2008-06-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* gdb.base/watchpoint.exp: New test for constant value watchpoints.
--- gdb/breakpoint.c 6 Jun 2008 20:58:08 -0000 1.324
+++ gdb/breakpoint.c 8 Jun 2008 19:11:37 -0000
@@ -5818,7 +5818,7 @@ watch_command_1 (char *arg, int accessfl
struct symtab_and_line sal;
struct expression *exp;
struct block *exp_valid_block;
- struct value *val, *mark;
+ struct value *val, *mark, *val_result;
struct frame_info *frame;
struct frame_info *prev_frame = NULL;
char *exp_start = NULL;
@@ -5903,10 +5903,28 @@ watch_command_1 (char *arg, int accessfl
exp_end = arg;
exp_valid_block = innermost_block;
mark = value_mark ();
- fetch_watchpoint_value (exp, &val, NULL, NULL);
+ fetch_watchpoint_value (exp, &val, &val_result, NULL);
if (val != NULL)
release_value (val);
+ /* VAL may be unset for unreachable final values. */
+ while (val_result != NULL)
+ {
+ if (VALUE_LVAL (val_result) == lval_memory
+ || VALUE_LVAL (val_result) == lval_register)
+ break;
+ val_result = value_next (val_result);
+ }
+ if (val_result == NULL)
+ {
+ int len;
+
+ len = exp_end - exp_start;
+ while (len > 0 && isspace (exp_start[len - 1]))
+ len--;
+ error (_("Cannot watch constant value %.*s."), len, exp_start);
+ }
+
tok = arg;
while (*tok == ' ' || *tok == '\t')
tok++;
--- gdb/doc/gdb.texinfo 6 Jun 2008 20:58:08 -0000 1.503
+++ gdb/doc/gdb.texinfo 8 Jun 2008 19:12:15 -0000
@@ -3375,6 +3375,17 @@ This command prints a list of watchpoint
it is the same as @code{info break} (@pxref{Set Breaks}).
@end table
+If you watch for a change in a numerically entered address you need to
+dereference it as the address itself is just a constant number which will never
+change. @value{GDBN} refuses to create a never invokable watchpoint:
+
+@smallexample
+(@value{GDBP}) watch 0x600850
+Cannot watch constant value 0x600850.
+(@value{GDBP}) watch *(int *) 0x600850
+Watchpoint 1: *(int *) 6293584
+@end smallexample
+
@value{GDBN} sets a @dfn{hardware watchpoint} if possible. Hardware
watchpoints execute very quickly, and the debugger reports a change in
value at the exact instruction where the change occurs. If @value{GDBN}
--- gdb/testsuite/gdb.base/watchpoint.exp 15 Apr 2008 14:33:54 -0000 1.18
+++ gdb/testsuite/gdb.base/watchpoint.exp 8 Jun 2008 19:12:18 -0000
@@ -679,6 +679,17 @@ set prev_timeout $timeout
set timeout 600
verbose "Timeout now 600 sec.\n"
+# Test constant-value watchpoints.
+gdb_test "watch 123" "Cannot watch constant value 123." "constant watchpoint"
+gdb_test "watch 456 if 1 == 2" "Cannot watch constant value 456." \
+ "constant watchpoint with a condition"
+# For unsupported constant-value watchpoints catching we need to reset the
+# breakpoints counter.
+gdb_exit
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load $binfile
+
if [initialize] then {
test_simple_watchpoint