[PATCH v3 6/6] Fix reverse-step and reverse-next over undebuggable solib code

[Omair Javaid <omair dot javaid at linaro dot org>]

This patch fixes failures to reverse-step or reverse-next over solib functions in absence of line information.
The problem is fixed by making sure that in solib functions we keep doing reverse single stepping in absence of line information.

Tested with no regressions on arm, aarch64 and x86_64.

gdb:

2014-08-13  Omair Javaid  <omair.javaid@linaro.org>

	* infrun.c (process_event_stop_test): Updated.

---
 gdb/infrun.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/gdb/infrun.c b/gdb/infrun.c
index c18267f..db8f15b 100644
--- a/gdb/infrun.c
+++ b/gdb/infrun.c
@@ -4905,12 +4905,15 @@ process_event_stop_test (struct execution_control_state *ecs)
       return;
     }
 
+  stop_pc_sal = find_pc_line (stop_pc, 0);
+
   /* Reverse stepping through solib trampolines.  */
 
   if (execution_direction == EXEC_REVERSE
       && ecs->event_thread->control.step_over_calls != STEP_OVER_NONE)
     {
       if (gdbarch_skip_trampoline_code (gdbarch, frame, stop_pc)
+	  || stop_pc_sal.line == 0
 	  || (ecs->stop_func_start == 0
 	      && in_solib_dynsym_resolve_code (stop_pc)))
 	{
@@ -4939,8 +4942,6 @@ process_event_stop_test (struct execution_control_state *ecs)
 	}
     }
 
-  stop_pc_sal = find_pc_line (stop_pc, 0);
-
   /* NOTE: tausq/2004-05-24: This if block used to be done before all
      the trampoline processing logic, however, there are some trampolines 
      that have no names, so we should do trampoline handling first.  */
-- 
1.9.1