Re: TUI + gdbserver broken?

[Pedro Alves <pedro_alves@portugalmail.pt>]

[-- Attachment #1: Type: text/plain, Size: 1382 bytes --]

Daniel Jacobowitz escreveu:
> On Mon, Mar 19, 2007 at 10:57:53PM +0000, Pedro Alves wrote:
>   
>> That has exactly the same effect as it was before the your select frame 
>> changes.
>> That is, only try to get the selected frame when there is one.  Maybe there 
>> should
>> still be a function like this in frame.c ?
>>     
>
>   
>> struct frame_info *
>> deprecated_get_selected_frame ()
>> {
>>  return selected_frame;
>> }
>>     
>
>   

Neither this, or your (level >= 0), worked - they both make tui enter
some kind of loop,
that I can't get a trace.  Perhaps in some other host it would be
easier (I'm on Cygwin).  Bummer.

>> (or perhaps call it get_selected_frame_if_any)
>>     
>
> The problem is that when we reinitialize the frame cache, the selected
> frame is indeterminate - the selected_frame global may randomly be
> initialized or NULL depending on what else has been called since we
> last did reinit_frame_cache.
>   

Maybe that has something to do with the hangs I see.

How about the attached?  Move the selected_frame creation logic
to select_frame.
It fixes both the remote debugging problem, and the repainting
problems - probably caused by the get_selected_frame calling
select_frame generating extra spurious events.

It would need a testsuite run and a bit of cleanup, since the
get_selected_frame would lose the parameter.

Cheers,
Pedro Alves



[-- Attachment #2: tui_frame3.diff --]
[-- Type: text/plain, Size: 1911 bytes --]

gdb/ChangeLog

	* frame.c (get_selected_frame): Move selected frame
	to current frame mapping to ...
	(select_frame): ... here.

---
 gdb/frame.c |   28 +++++++++++++++-------------
 1 file changed, 15 insertions(+), 13 deletions(-)

Index: src/gdb/frame.c
===================================================================
--- src.orig/gdb/frame.c	2007-03-18 23:22:10.000000000 +0000
+++ src/gdb/frame.c	2007-03-20 00:29:42.000000000 +0000
@@ -953,19 +953,6 @@ static struct frame_info *selected_frame
 struct frame_info *
 get_selected_frame (const char *message)
 {
-  if (selected_frame == NULL)
-    {
-      if (message != NULL && (!target_has_registers
-			      || !target_has_stack
-			      || !target_has_memory))
-	error (("%s"), message);
-      /* Hey!  Don't trust this.  It should really be re-finding the
-	 last selected frame of the currently selected thread.  This,
-	 though, is better than nothing.  */
-      select_frame (get_current_frame ());
-    }
-  /* There is always a frame.  */
-  gdb_assert (selected_frame != NULL);
   return selected_frame;
 }
 
@@ -994,6 +981,21 @@ select_frame (struct frame_info *fi)
   if (deprecated_selected_frame_level_changed_hook)
     deprecated_selected_frame_level_changed_hook (frame_relative_level (fi));
 
+  if (fi == NULL)
+    {
+      if (!target_has_registers
+	  || !target_has_stack
+	  || !target_has_memory)
+	return;
+
+      /* Hey!  Don't trust this.  It should really be re-finding the
+	 last selected frame of the currently selected thread.  This,
+	 though, is better than nothing.  */
+      selected_frame = get_current_frame ();
+    }
+  /* There is always a frame.  */
+  gdb_assert (selected_frame != NULL);
+
   /* FIXME: kseitz/2002-08-28: It would be nice to call
      selected_frame_level_changed_event() right here, but due to limitations
      in the current interfaces, we would end up flooding UIs with events