This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[patch/rfc] Use frame_type for sigtramp test in infrun.c


Hello,

The attached patch modifies infrun.c so that, for the non-legacy case, the test:
frame_type == SIGTRAMP frame
&& step_frame_id != stop_frame_id
is used to determine if a single-step caused the inferior to stumble into a signal trampoline (either via a signal, or by returning from the handler). This replaces two pc_in_sigtramp calls and a stack-pointer heuristic -- not the most robust of tests.


The new test:

- eliminates the call pc_in_sigtramp(step_pc)
(i.e., on the PC from before the single step) (this was an unexpected bonus :-)


- uses the caching call get_frame_type call
pc_in_sigtramp has to re-do all the computation each time it is called (typically duplicating the effort of get_frame_type)


- uses !frame_id_eq
Which is a 100% robust frame-changed test. The old test couldn't detect a signal trampoline calling (via a signal) a singnal trampoline.


thoughts?

I'll look to commit this in a few days (already committed to my trad-frame branch).

Andrew
2004-03-16  Andrew Cagney  <cagney@redhat.com>

	* infrun.c (handle_inferior_event): For non legacy frames, use the
	frame ID and frame type to identify a signal trampoline.  Update
	comments.

Index: infrun.c
===================================================================
RCS file: /cvs/src/src/gdb/infrun.c,v
retrieving revision 1.140
diff -u -r1.140 infrun.c
--- infrun.c	15 Mar 2004 17:12:50 -0000	1.140
+++ infrun.c	16 Mar 2004 18:25:45 -0000
@@ -2521,22 +2521,19 @@
      But we can update it every time we leave the step range.  */
   ecs->update_step_sp = 1;
 
-  /* Did we just take a signal?  */
-  if (pc_in_sigtramp (stop_pc)
-      && !pc_in_sigtramp (prev_pc)
-      && INNER_THAN (read_sp (), step_sp))
+  /* Did we just step into a singal trampoline (either by stepping out
+     of a handler, or by taking a signal)?  */
+  /* NOTE: cagney/2004-03-16: Replaced (except for legacy) a check for
+     "pc_in_sigtramp(stop_pc) != pc_in_sigtramp(step_pc)" with
+     frame_type == SIGTRAMP && !frame_id_eq.  The latter is far more
+     robust as it will correctly handle nested signal trampolines.  */
+  if (legacy_frame_p (current_gdbarch)
+      ? (pc_in_sigtramp (stop_pc)
+	 && !pc_in_sigtramp (prev_pc)
+	 && INNER_THAN (read_sp (), step_sp))
+      : (get_frame_type (get_current_frame ()) == SIGTRAMP_FRAME
+	 && !frame_id_eq (get_frame_id (get_current_frame ()), step_frame_id)))
     {
-      /* We've just taken a signal; go until we are back to
-         the point where we took it and one more.  */
-
-      /* Note: The test above succeeds not only when we stepped
-         into a signal handler, but also when we step past the last
-         statement of a signal handler and end up in the return stub
-         of the signal handler trampoline.  To distinguish between
-         these two cases, check that the frame is INNER_THAN the
-         previous one below. pai/1997-09-11 */
-
-
       {
 	struct frame_id current_frame = get_frame_id (get_current_frame ());
 

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]