Re: RFD: infrun.c: No bpstat_stop_status call after proceed over break ?

[Jim Kingdon <kingdon at redhat dot com>]

> I am currently trying to fix a GDB bug with missing watchpoint triggers
> after proceeding over a breakpoint on x86 targets.
> 
> Here is an example, using gdb.c++/annota2:

Yeah, I ran into the same bug (the testsuite hits it on x86 Linux).

My fix was similar to what you suggest (identical, unless I'm missing
something).  I don't remember it as creating testsuite regressions but
I'm not sure I checked - it was a while ago that I looked into this.

Index: infrun.c
===================================================================
RCS file: /cvs/gdb/gdb/gdb/infrun.c,v
retrieving revision 1.1.1.27
diff -u -r1.1.1.27 infrun.c
--- infrun.c	2000/02/05 07:29:43	1.1.1.27
+++ infrun.c	2000/02/07 13:22:02
@@ -2076,6 +2076,18 @@
 	    return;
 	  }
 
+#if 0
+	/* We might not want to think about breakpoints, but what
+	   about hardware watchpoints?  We yanked out the hardware
+	   (debug register or whatever) when we stepped over the breakpoint,
+	   so we need to check manually the watchpoint condition (mildly
+	   painful because it could get expensive, but I don't see how
+	   using the hardware could work without lots of spaghetti trying
+	   to keep track of what is inserted and what isn't (and how they
+	   interact)).
+
+	   TODO: Need to think more about this, test it, &c.  Right now it
+	   is kind of an idea more so than a well-thought-out solution.  */
 	/* Don't even think about breakpoints
 	   if just proceeded over a breakpoint.
 
@@ -2087,6 +2099,7 @@
 	    && through_sigtramp_breakpoint == NULL)
 	  bpstat_clear (&stop_bpstat);
 	else
+#endif /* 0 */
 	  {
 	    /* See if there is a breakpoint at the current PC.  */
 	    stop_bpstat = bpstat_stop_status