Re: [RFA] Reverse Debugging, 3/5

[Michael Snyder <msnyder@vmware.com>]

Pedro Alves wrote:
> Hi Michael,
> 
> Haven't read the other patches yet, but I'll go ahead and give
> some comments on this one.
> 
> On Wednesday 01 October 2008 20:18:35, Michael Snyder wrote:
>> Index: infrun.c
>> ===================================================================
>> RCS file: /cvs/src/src/gdb/infrun.c,v
>> retrieving revision 1.322
>> retrieving revision 1.322.2.2
>> diff -u -p -r1.322 -r1.322.2.2
>> --- infrun.c    22 Sep 2008 15:26:53 -0000      1.322
>> +++ infrun.c    30 Sep 2008 23:50:51 -0000      1.322.2.2
>> @@ -1193,11 +1193,17 @@ proceed (CORE_ADDR addr, enum target_sig
>>
> 
>>    if (addr == (CORE_ADDR) -1)
>>      {
>> -      if (pc == stop_pc && breakpoint_here_p (pc))
>> +      if (pc == stop_pc && breakpoint_here_p (pc)
>> +         && target_get_execution_direction () != EXEC_REVERSE)
> 
> Hmmm, so EXEC_ERROR is accepted here.  What exactly is
> EXEC_ERROR useful for?  Will there be a target that can't go
> either direction?  :-) 

No, silly...   ;-)

> Shouldn't failing to find ones
> direction always be an error (hence an error call from inside
> target_get_execution_direction, or something alike).

Targets that don't implement reverse return EXEC_ERROR,
rather than EXEC_FORWARD.  It was an early interface
design decision, and I'm not sure if I can remember the
justification after over 2 years, but I made it
consciously -- it seemed to simplify things.


>>  /* The PTID we'll do a target_wait on.*/
>> @@ -2141,6 +2149,12 @@ handle_inferior_event (struct execution_
>>        ecs->event_thread->stop_signal = ecs->ws.value.sig;
>>        break;
>>
>> +    case TARGET_WAITKIND_NO_HISTORY:
>> +      /* Reverse execution: target ran out of history info.  */
>> +      print_stop_reason (NO_HISTORY, 0);
>> +      stop_stepping (ecs);
>> +      return;
>> +
>>        /* We had an event in the inferior, but we are not interested
>>           in handling it at this level. The lower layers have already
>>           done what needs to be done, if anything.
>> @@ -2861,6 +2875,17 @@ infrun: BPSTAT_WHAT_SET_LONGJMP_RESUME (
>>             keep_going (ecs);
>>             return;
>>           }
> 
>> +       if (stop_pc == ecs->stop_func_start &&
>> +           target_get_execution_direction () == EXEC_REVERSE)
> 
> Split new line before the operator, not after:

OK

>>        case BPSTAT_WHAT_CHECK_SHLIBS:
>> @@ -3026,10 +3051,25 @@ infrun: BPSTAT_WHAT_SET_LONGJMP_RESUME (
>>        && stop_pc < ecs->event_thread->step_range_end)
>>      {
>>        if (debug_infrun)
>> -        fprintf_unfiltered (gdb_stdlog, "infrun: stepping inside range [0x%s-0x%s]\n",
>> +       fprintf_unfiltered (gdb_stdlog, "infrun: stepping inside range [0x%s-0x%s]\n",
>>                             paddr_nz (ecs->event_thread->step_range_start),
>>                             paddr_nz (ecs->event_thread->step_range_end));
>> -      keep_going (ecs);
>> +
>> +      /* When stepping backward, stop at beginning of line range
>> +        (unles it's the function entry point, in which case
> 
> unless

OK

>> +        keep going back to the call point).  */
>> +      if (stop_pc == ecs->event_thread->step_range_start &&
>> +         stop_pc != ecs->stop_func_start &&
>> +         target_get_execution_direction () == EXEC_REVERSE)
>> +       {
>> +         ecs->event_thread->stop_step = 1;
>> +         print_stop_reason (END_STEPPING_RANGE, 0);
>> +         stop_stepping (ecs);
>> +       }
> 
>> +      else
>> +       {
>> +         keep_going (ecs);
>> +       }
> 
> Unneeded braces.

Don't you think it's more readable if the if block
and the else block match?

>>        return;
>>      }
>>
>> @@ -3116,10 +3156,28 @@ infrun: BPSTAT_WHAT_SET_LONGJMP_RESUME (
>>
>>        if (ecs->event_thread->step_over_calls == STEP_OVER_ALL)
>>         {
>> -         /* We're doing a "next", set a breakpoint at callee's return
>> -            address (the address at which the caller will
>> -            resume).  */
>> -         insert_step_resume_breakpoint_at_caller (get_current_frame ());
>> +         /* We're doing a "next".
>> +
>> +            Normal (forward) execution: set a breakpoint at the
>> +            callee's return address (the address at which the caller
>> +            will resume).
>> +
>> +            Reverse (backward) execution.  set the step-resume
>> +            breakpoint at the start of the function that we just
>> +            stepped into (backwards), and continue to there.  When we
>> +            get there, we'll need to single-step back to the
>> +            caller.  */
>> +
>> +         if (target_get_execution_direction () == EXEC_REVERSE)
>> +           {
>> +             struct symtab_and_line sr_sal;
>> +             init_sal (&sr_sal);
>> +             sr_sal.pc = ecs->stop_func_start;
>> +             insert_step_resume_breakpoint_at_sal (sr_sal, null_frame_id);
>> +           }
>> +         else
>> +           insert_step_resume_breakpoint_at_caller (get_current_frame ());
>> +
>>           keep_going (ecs);
>>           return;
>>         }
>> @@ -3176,9 +3234,21 @@ infrun: BPSTAT_WHAT_SET_LONGJMP_RESUME (
>>           return;
>>         }
>>
>> -      /* Set a breakpoint at callee's return address (the address at
>> -         which the caller will resume).  */
>> -      insert_step_resume_breakpoint_at_caller (get_current_frame ());
>> +      if (target_get_execution_direction () == EXEC_REVERSE)
>> +       {
>> +         /* Set a breakpoint at callee's start address.
>> +            From there we can step once and be back in the caller.  */
>> +         struct symtab_and_line sr_sal;
>> +         init_sal (&sr_sal);
>> +         sr_sal.pc = ecs->stop_func_start;
>> +         insert_step_resume_breakpoint_at_sal (sr_sal, null_frame_id);
>> +       }
>> +      else
>> +       {
>> +         /* Set a breakpoint at callee's return address (the address
>> +            at which the caller will resume).  */
>> +         insert_step_resume_breakpoint_at_caller (get_current_frame ());
>> +       }
> 
> Unneeded braces.

Oh come on -- I know they're syntactic null, but
they serve to keep the comment together with the
code it refers to.

>>        keep_going (ecs);
>>        return;
>>      }
>> @@ -3344,6 +3414,28 @@ step_into_function (struct execution_con
>>      ecs->stop_func_start = gdbarch_skip_prologue
>>                              (current_gdbarch, ecs->stop_func_start);
>>
>> +  if (target_get_execution_direction () == EXEC_REVERSE)
>> +    {
>> +      stop_func_sal = find_pc_line (stop_pc, 0);
>> +
>> +      /* OK, we're just gonna keep stepping here.  */
>> +      if (stop_func_sal.pc == stop_pc)
>> +       {
>> +         /* We're there already.  Just stop stepping now.  */
>> +         ecs->event_thread->stop_step = 1;
>> +         print_stop_reason (END_STEPPING_RANGE, 0);
>> +         stop_stepping (ecs);
>> +         return;
>> +       }
>> +      /* Else just reset the step range and keep going.
>> +        No step-resume breakpoint, they don't work for
>> +        epilogues, which can have multiple entry paths.  */
>> +      ecs->event_thread->step_range_start = stop_func_sal.pc;
>> +      ecs->event_thread->step_range_end   = stop_func_sal.end;
> 
> Somethings fishy with the whitespace.     ^

I just like things to line up when possible!
;-)


>> +      keep_going (ecs);
>> +      return;
>> +    }
>> +  /* else... */
>>    stop_func_sal = find_pc_line (ecs->stop_func_start, 0);
>>    /* Use the step_resume_break to step until the end of the prologue,
>>       even if that involves jumps (as it seems to on the vax under
>> @@ -3712,6 +3804,10 @@ print_stop_reason (enum inferior_stop_re
>>        annotate_signal_string_end ();
>>        ui_out_text (uiout, ".\n");
>>        break;
>> +    case NO_HISTORY:
>> +      /* Reverse execution: target ran out of history info.  */
>> +      ui_out_text (uiout, "\nNo more reverse-execution history.\n");
>> +      break;
>>      default:
>>        internal_error (__FILE__, __LINE__,
>>                       _("print_stop_reason: unrecognized enum value"));
> 
> Otherwise, I can't see anything wrong with it...

Thanks for reviewing.