[rfc] Don't thwack frame->pc in infun.c (DECR_PC_AFTER_BREAK)

[Andrew Cagney <ac131313 at redhat dot com>]

Hello,

The file "infrun.c" currently contains the `cute' trick:

   /* Make sure that the current_frame's pc is correct.  This
      is a correction for setting up the frame info before doing
      DECR_PC_AFTER_BREAK */
-  if (target_has_execution && get_current_frame ())
-    (get_current_frame ())->pc = read_pc ();

Come the revolution, and when all DECR_PC_AFTER_BREAK has been slain, 
this hack will just go away.

In the mean time, I'm trying to figure out a cleaner way of doing it ...

First thing to note is that the get_current_frame() [!= NULL] part of 
the `if' conditional is redundant.  If there isn't a current frame, 
get_current_frame() will throw an error.  Consequently the test can be 
dropped (with the side effect of forcing a frame create).

Second thing to note is that by changing the frame's PC the code has 
invalidated anything cached in that frame.  Consequently, the code 
should flush the cache, rather than just tweak the PC.  An unnecessary 
flush can be avoided by checking to see if the PC changed.

The attached patch implements both of these changes.  It only flushes 
the cache when the frame's PC has, some how, become invalid.

Unfortunatly, for the i386 (and alpha), this has consequences.  It 
guarentees the sequence:

- create frame (parse prologue)
- flush frames
- create frame (parse prologue)

Where parse prologue involves target reads and, hence, is considered 
expensive.   (I don't feel too guilty since the i386 has its [er, 
deprecated] codestream which should lessen the damange).

So anyway, the above needs to be changed.  An alternative might be to:

	if (target_has_execution)
	  deprecated_resync_current_frame_pc_hack (read_pc ());

thoughts?
Andrew
2002-12-06  Andrew Cagney  <ac131313@redhat.com>

	* infrun.c (normal_stop): If the PC and frame's PC are
	out-of-sync, re-initialize the frame cache.

Index: infrun.c
===================================================================
RCS file: /cvs/src/src/gdb/infrun.c,v
retrieving revision 1.81
diff -u -r1.81 infrun.c
--- infrun.c	4 Dec 2002 00:05:53 -0000	1.81
+++ infrun.c	6 Dec 2002 17:58:34 -0000
@@ -3281,8 +3281,15 @@
   /* Make sure that the current_frame's pc is correct.  This
      is a correction for setting up the frame info before doing
      DECR_PC_AFTER_BREAK */
-  if (target_has_execution && get_current_frame ())
-    (get_current_frame ())->pc = read_pc ();
+  if (target_has_execution
+      && get_frame_pc (get_current_frame ()) != read_pc ())
+    /* NOTE: cagney/2002-12-06: Ok, things are really messed up.  This
+       architecture has DECR_PC_AFTER_BREAKed the program counter
+       causing the frame's PC to be out-of-sync.  Rebuild the frame
+       cache.  This probably causes a re-parse of the function's
+       prologue.  Sigh!  This doesn't use reinit_frame_cache() since,
+       slightly further down the selected frame is explicitly set.  */
+    flush_cached_frames ();
 
   if (target_has_execution && breakpoints_inserted)
     {