[sergio-catch-syscall] gdb.base/reread.exp regression

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Fri, 09 Jan 2009 02:37:13 +0100, SÃrgio Durigan JÃnior wrote:
> On Fri, 2009-01-09 at 01:01 +0100, Jan Kratochvil wrote:
> > #gdb
> > (2009-01-09 00:39:23) jankratochvil: sergiodj: Could you please verify - I have a regression in archer-sergio-catch-syscall on both x86 and x86_64 for: gdb.base/reread.exp: second pass: run to foo()
> > (2009-01-09 00:39:26) jankratochvil: Starting program: /home/jkratoch/redhat/archer-sergio-catch-syscall/gdb/testsuite/gdb.base/reread 
> > (2009-01-09 00:39:26) jankratochvil: Program received signal SIGSEGV, Segmentation fault.
> > (2009-01-09 00:39:26) jankratochvil: 0xd50482f4 in ?? ()
[...]
> Unfortunately, I couldn't reproduce the bug using a x86 architecture

Attached a possible fix - the bp_entry_breakpoint address got stale on an
inferior reload.

Sure there should be more a callback during breakpoint_re_set() than this way.
I just chose a proven method of existing remove_thread_event_breakpoints() but
I guess the proper way would be to use a catchpoint
with new `struct breakpoint_ops' specific for bp_entry_breakpoint.


> (I don't have a x86_64 available here).

This gdb.base/reread.exp regression I had reproducible on both x86 and x86_64
(particularly x86 running as a -m32 build on x86_64 but it should not matter).


Regards,
Jan

diff --git a/gdb/breakpoint.c b/gdb/breakpoint.c
index 143d4b0..09b3171 100644
--- a/gdb/breakpoint.c
+++ b/gdb/breakpoint.c
@@ -4403,11 +4403,13 @@ disable_overlay_breakpoints (void)
 }
 
 int
-create_entry_breakpoint ()
+create_entry_breakpoint (void)
 {
   CORE_ADDR taddr, entry_addr;
   struct breakpoint *b;
 
+  delete_entry_breakpoint ();
+
   taddr = entry_point_address ();
   /* Make certain that the address points at real code, and not a
      function descriptor.  */
@@ -4427,6 +4429,20 @@ create_entry_breakpoint ()
   return 1;
 }
 
+/* Remove the breakpoint from create_entry_breakpoint.  To be called in the
+   same moments as remove_thread_event_breakpoints (or its caller
+   disable_thread_event_reporting) are being called.  */
+
+void
+delete_entry_breakpoint (void)
+{
+  struct breakpoint *b, *temp;
+
+  ALL_BREAKPOINTS_SAFE (b, temp)
+    if (b->type == bp_entry_breakpoint)
+      delete_breakpoint (b);
+}
+
 struct breakpoint *
 create_thread_event_breakpoint (CORE_ADDR address)
 {
diff --git a/gdb/breakpoint.h b/gdb/breakpoint.h
index 384a156..4ef3c78 100644
--- a/gdb/breakpoint.h
+++ b/gdb/breakpoint.h
@@ -913,6 +913,9 @@ extern int catching_syscall_number (int syscall_number);
    or 1 if everything went OK.  */
 extern int create_entry_breakpoint (void);
 
+/* Remove the breakpoint from create_entry_breakpoint.  */
+extern void delete_entry_breakpoint (void);
+
 /* Tell a breakpoint to be quiet.  */
 extern void make_breakpoint_silent (struct breakpoint *);
 
diff --git a/gdb/linux-nat.c b/gdb/linux-nat.c
index 4b317f6..20a85d1 100644
--- a/gdb/linux-nat.c
+++ b/gdb/linux-nat.c
@@ -1681,6 +1681,8 @@ linux_nat_detach (struct target_ops *ops, char *args, int from_tty)
   int status;
   enum target_signal sig;
 
+  delete_entry_breakpoint ();
+
   if (target_can_async_p ())
     linux_nat_async (NULL, 0);
 
@@ -2896,6 +2898,9 @@ linux_nat_wait (ptid_t ptid, struct target_waitstatus *ourstatus)
   if (debug_linux_nat_async)
     fprintf_unfiltered (gdb_stdlog, "LLW: enter\n");
 
+  if (ourstatus->kind == TARGET_WAITKIND_EXECD)
+    delete_entry_breakpoint ();
+
   /* The first time we get here after starting a new inferior, we may
      not have added it to the LWP list yet - this is the earliest
      moment at which we know its PID.  */
@@ -3330,6 +3335,9 @@ linux_nat_mourn_inferior (struct target_ops *ops)
        there are other viable forks to debug.  Delete the exiting
        one and context-switch to the first available.  */
     linux_fork_mourn_inferior ();
+
+  /* It would fail with a warning before the inherited mourn calls above.  */
+  delete_entry_breakpoint ();
 }
 
 static LONGEST