This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: RFC: don't set the pspace on ordinary breakpoints
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: tromey at redhat dot com (Tom Tromey)
- Cc: pedro at codesourcery dot com (Pedro Alves), gdb-patches at sourceware dot org
- Date: Thu, 10 Nov 2011 18:57:36 +0100 (CET)
- Subject: Re: RFC: don't set the pspace on ordinary breakpoints
Tom Tromey wrote:
> diff --git a/gdb/breakpoint.c b/gdb/breakpoint.c
> index 8c98bef..ec45335 100644
> --- a/gdb/breakpoint.c
> +++ b/gdb/breakpoint.c
> @@ -1568,6 +1568,9 @@ should_be_inserted (struct bp_location *bl)
> if (!bl->enabled || bl->shlib_disabled || bl->duplicate)
> return 0;
>
> + if (user_breakpoint_p (bl->owner) && bl->pspace->executing_startup)
> + return 0;
> +
> /* This is set for example, when we're attached to the parent of a
> vfork, and have detached from the child. The child is running
> free, and we expect it to do an exec or exit, at which point the
> @@ -5313,8 +5316,7 @@ describe_other_breakpoints (struct gdbarch *gdbarch,
> printf_filtered (" (thread %d)", b->thread);
> printf_filtered ("%s%s ",
> ((b->enable_state == bp_disabled
> - || b->enable_state == bp_call_disabled
> - || b->enable_state == bp_startup_disabled)
> + || b->enable_state == bp_call_disabled)
> ? " (disabled)"
> : b->enable_state == bp_permanent
> ? " (permanent)"
> @@ -6933,48 +6935,48 @@ enable_watchpoints_after_interactive_call_stop (void)
> void
> disable_breakpoints_before_startup (void)
> {
> - struct breakpoint *b;
> + struct bp_location *loc, **locp_tmp;
> int found = 0;
>
> - ALL_BREAKPOINTS (b)
> + current_program_space->executing_startup = 1;
> +
> + ALL_BP_LOCATIONS (loc, locp_tmp)
> {
> - if (b->pspace != current_program_space)
> + if (loc->pspace != current_program_space)
> continue;
>
> - if ((b->type == bp_breakpoint
> - || b->type == bp_hardware_breakpoint)
> - && breakpoint_enabled (b))
> + if ((loc->owner->type == bp_breakpoint
> + || loc->owner->type == bp_hardware_breakpoint)
Should this now use the same condition as above,
i.e. user_breakpoint_p (b) ?
> + && breakpoint_enabled (loc->owner))
> {
> - b->enable_state = bp_startup_disabled;
> found = 1;
> + break;
> }
> }
>
> if (found)
> update_global_location_list (0);
Or maybe rather, since the sole remaining function of that loop
is to sometimes skip the update_global_location_list, which doesn't
appear to have any reason except performance tuning (which is
really unnecessary here as this routine is called extremely rarely)
-- why not just remove the whole loop and just call
update_global_location_list unconditionally?
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
Ulrich.Weigand@de.ibm.com