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

[Michael Snyder <msnyder@vmware.com>]

Committing this change.

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,