[RFA] improved handling of shared lib calls (reverse 2/2)

Hui Zhu teawater@gmail.com
Wed Jul 15 12:53:00 GMT 2009


I try this patch with cvs-head and program:
void foo() {
    return;
}

int main() {

	foo(); // 1
	foo(); // 2
	foo(); // 3

	return 0;
}
in ubuntu i386.
And test with reverse testsuite.
Everything is OK.

Thanks,
Hui


On Wed, Jul 15, 2009 at 09:08, Michael Snyder<msnyder@vmware.com> wrote:
> And now this one?
>
> Michael Snyder wrote:
>>
>> Now how about this one?
>>
>> Michael Snyder wrote:
>>>
>>> This change depends on the previous one (epilogue unwinder for i386,
>>> http://sourceware.org/ml/gdb-patches/2009-07/msg00099.html).
>>>
>>> With the frame unwinder working properly, we can remove the
>>> special case code in the test for "check for subroutine call"
>>> in handle_inferior_event.
>>>
>>> Of course that reveals some hidden problems with shared library
>>> calls, so this patch also addresses those.
>>>
>>> Michael
>>>
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> 2009-07-03  Michael Snyder  <msnyder@vmware.com>
>>>
>>>        * infrun.c (handle_inferior_event): Remove an execution_direction
>>>        check in the "check for subroutine calls" test, and add some
>>>  additional handling for reverse-stepping through shared libraries.
>>>
>>> Index: infrun.c
>>> ===================================================================
>>> RCS file: /cvs/src/src/gdb/infrun.c,v
>>> retrieving revision 1.400
>>> diff -u -p -r1.400 infrun.c
>>> --- infrun.c    2 Jul 2009 17:21:06 -0000       1.400
>>> +++ infrun.c    4 Jul 2009 00:27:53 -0000
>>> @@ -3776,9 +3776,8 @@ infrun: not switching back to stepped th
>>>      previous frame must have valid frame IDs.  */
>>>   if (!frame_id_eq (get_stack_frame_id (frame),
>>>                    ecs->event_thread->step_stack_frame_id)
>>> -      && (frame_id_eq (frame_unwind_caller_id (frame),
>>> -                      ecs->event_thread->step_stack_frame_id)
>>> -         || execution_direction == EXEC_REVERSE))
>>> +      && frame_id_eq (frame_unwind_caller_id (frame),
>>> +                     ecs->event_thread->step_stack_frame_id))
>>>     {
>>>       CORE_ADDR real_stop_pc;
>>>  @@ -3806,6 +3805,7 @@ infrun: not switching back to stepped th
>>>       /* Reverse stepping through solib trampolines.  */
>>>         if (execution_direction == EXEC_REVERSE
>>> +         && ecs->event_thread->step_over_calls != STEP_OVER_NONE
>>>          && (gdbarch_skip_trampoline_code (gdbarch, frame, stop_pc)
>>>              || (ecs->stop_func_start == 0
>>>                  && in_solib_dynsym_resolve_code (stop_pc))))
>>> @@ -3923,6 +3923,38 @@ infrun: not switching back to stepped th
>>>       return;
>>>     }
>>>  +  /* Reverse stepping through solib trampolines.  */
>>> +
>>> +  if (execution_direction == EXEC_REVERSE
>>> +      && ecs->event_thread->step_over_calls != STEP_OVER_NONE)
>>> +    {
>>> +      if (gdbarch_skip_trampoline_code (gdbarch, frame, stop_pc)
>>> +         || (ecs->stop_func_start == 0
>>> +             && in_solib_dynsym_resolve_code (stop_pc)))
>>> +       {
>>> +         /* Any solib trampoline code can be handled in reverse
>>> +            by simply continuing to single-step.  We have already
>>> +            executed the solib function (backwards), and a few +
>>>    steps will take us back through the trampoline to the
>>> +            caller.  */
>>> +         keep_going (ecs);
>>> +         return;
>>> +       }
>>> +      else if (in_solib_dynsym_resolve_code (stop_pc))
>>> +       {
>>> +         /* Stepped backward into the solib dynsym resolver.
>>> +            Set a breakpoint at its start and continue, then
>>> +            one more step will take us out.  */
>>> +         struct symtab_and_line sr_sal;
>>> +         init_sal (&sr_sal);
>>> +         sr_sal.pc = ecs->stop_func_start;
>>> +         insert_step_resume_breakpoint_at_sal (gdbarch, +
>>>                                 sr_sal, null_frame_id);
>>> +         keep_going (ecs);
>>> +         return;
>>> +       }
>>> +    }
>>> +
>>>   /* If we're in the return path from a shared library trampoline,
>>>      we want to proceed through the trampoline when stepping.  */
>>>   if (gdbarch_in_solib_return_trampoline (gdbarch,
>>
>>
>
>



More information about the Gdb-patches mailing list