Re: [patch]: inform user that a watchpoint is hit

[Chandru <chandru at in dot ibm dot com>]

Jan Kratochvil wrote:
> This is not reproducible for me.
> * Which distribution?
> * Which exact NVR (name-version-release) of the distribution package?
> * The URL http://sourceware.org/ml/gdb-patches/2009-08/msg00254.html was
>   referring to 4 patches, do you have all of them applied?
>   (but these 4 patches have no effect on this testcase for me, shown as
>   "archer-jankratochvil-watchpoint" below)
> * Please provide both the testcase and complete GDB run screenshot in the same
>   mail.  Once you refer to "rawatch.c", second time "rwatch.c".
>   When you provide a screenshot refer to all the exact versions and possible
>   patches applied to it.
> * Provide the architecture in use (x86_64? i686? could it be ppc64?).
> * Provide even the compiler version used.
> * Provide the kernel version in use due to various ptrace syscall bugsin the
>   past and present (asking primarily for the possibility of ppc64 in use and
>   the watchpoints on it).
>
>
> Thanks,
> Jan
>
>   
Hello Jan and Doug,

Thanks for your mails. I today downloaded gdb-6.6 and gdb-6.8.50.20090824.
gdb-6.6 outputs the messages as follows when the program is restarted


Program exited normally.
(gdb) r
Starting program: /home/chandru/rawatch
Hardware read watchpoint 2: value1
Hardware access (read/write) watchpoint 3: value1
Hardware read watchpoint 2: value1
Hardware access (read/write) watchpoint 3: value1
Hardware read watchpoint 2: value1
Hardware access (read/write) watchpoint 3: value1

Breakpoint 1, main () at rawatch.c:20
20          value1 =3;
(gdb)



whereas gdb-6.8.50.20090824 doesn't.

Program exited normally.
(gdb) r
Starting program: /home/chandru/rawatch

Breakpoint 1, main () at rawatch.c:20
20          value1 =3;
(gdb)



It is not the case that the read or access watchpoints are not being hit.
Just the messages like 'Hardware read watchpoint 2: value1' and
'Hardware access (read/write) watchpoint 3: value1'  are not shown
on the screen when the program is restarted and the condition

if (breakpoint_enabled (b))
     mention(b);


only helped to see those messages. It is only the first occurrence of the
watchpoints that the messages don't get displayed for. We could see the
messages for later occurrences of the watchpoints on value1. Hence this
is just a cosmetic issue and doesn't look like a architecture or a kernel
specific issue.

Thanks,
Chandru