Re: [RFA/7.8] user breakpoint not inserted if software-single-step at same location

[Pedro Alves <palves at redhat dot com>]

On 05/30/2014 02:26 PM, Joel Brobecker wrote:
>>>>  - if there's still a non-sss breakpoint inserted at the
>>>>    same address, then don't actually remove the breakpoint
>>>>    off of the target, just wipe it from gdb's list.
>>>
>>> It seems to me that we'd need to merge your initial recommendation
>>> into your summary above, right?
>>
>> I admit I don't know what recommendation you're referring to.  :-)
> 
> Sorry! This one:
> 
>     | but we'll need to create/clone the location and its shadow buffer,
>     | and then still handle the issue in the "remove" path.

Ah.  Yes.

> I am wondering how to create that test, because it would be
> a little tricky. We need to set ourselves into a situation
> where we single-step out of a breakpoint with the second SSS
> breakpoint being at the same address as one of the user breakpoints,
> that second SSS not being the one that gets hit during that
> first single-step-out-of-breakpoint.  

Yeah.  Hmmm.  

*thinks*

I have a very simple idea, around "jump" + "always-inserted".

E.g., with, where b+ indicates a user breakpoint:

       00001 nop        <- PC
    b+ 00002 nop        

 - enable breakpoints always inserted mode
 - step to 00002
 - gdb removes the sss breakpoint.
 - due to always inserted mode, gdb does not remove b+, but
   due to the bug, it's actually no longer planted.
 - the b+ breakpoint should be reported to the user.
 - now do "jump $pc".
 - expected:
    The breakpoint should trigger immediately again.
 - what we get on sss targets:
    GDB loses control, and program runs to end.

Ah, I just went ahead and tried that against my by sss-on-x86
branch, and indeed it fails here, while it passes on
pristine mainline / hardware stepping.

8<----------
>From e13bf4d64bf299111193a1f27a0bbc194d9b34f4 Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Fri, 30 May 2014 16:52:36 +0100
Subject: [PATCH] test for sss breakpoints bug

---
 gdb/testsuite/gdb.base/sss-bp-on-user-bp.c   | 30 ++++++++++++++++
 gdb/testsuite/gdb.base/sss-bp-on-user-bp.exp | 51 ++++++++++++++++++++++++++++
 2 files changed, 81 insertions(+)
 create mode 100644 gdb/testsuite/gdb.base/sss-bp-on-user-bp.c
 create mode 100644 gdb/testsuite/gdb.base/sss-bp-on-user-bp.exp

diff --git a/gdb/testsuite/gdb.base/sss-bp-on-user-bp.c b/gdb/testsuite/gdb.base/sss-bp-on-user-bp.c
new file mode 100644
index 0000000..ff82051
--- /dev/null
+++ b/gdb/testsuite/gdb.base/sss-bp-on-user-bp.c
@@ -0,0 +1,30 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+   Copyright 2014 Free Software Foundation, Inc.
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+#include <signal.h>
+#include <unistd.h>
+
+int
+main (void)
+{
+  /* Assume writes to integers compile to a single instruction.  */
+  volatile int i = 0;
+
+  i = 1;     /* set foo break here */
+  i = 2;     /* set bar break here */
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.base/sss-bp-on-user-bp.exp b/gdb/testsuite/gdb.base/sss-bp-on-user-bp.exp
new file mode 100644
index 0000000..bb63d3f
--- /dev/null
+++ b/gdb/testsuite/gdb.base/sss-bp-on-user-bp.exp
@@ -0,0 +1,51 @@
+# Copyright 2014 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# Test that removing a single-step breakpoint that is placed at the
+# same address as another regular breakpoint leaves the regular
+# breakpoint inserted.
+
+standard_testfile
+set executable ${testfile}
+
+if {[prepare_for_testing "failed to prepare" $testfile $srcfile debug]} {
+    return -1
+}
+
+if ![runto_main] then {
+    fail "Can't run to main"
+    return 0
+}
+
+gdb_breakpoint [gdb_get_line_number "set foo break here"]
+gdb_continue_to_breakpoint "first breakpoint" ".* set foo break here .*"
+
+gdb_breakpoint [gdb_get_line_number "set bar break here"]
+
+# So that GDB doesn't try to remove the regular breakpoint when the
+# step finishes.
+gdb_test_no_output "set breakpoint always-inserted on"
+
+# On software single-step targets, this step will want to momentarily
+# place a single-step breakpoint over the bar breakpoint, and then
+# remove it.  But, a regular breakpoint it planted there already, and
+# with always-inserted on, should remain planted when the step
+# finishes.
+gdb_test "si" "Breakpoint .* bar break .*"
+
+# If the breakpoint is still correctly inserted, then this jump should
+# re-trigger it.  Otherwise, GDB will lose control and the program
+# will exit.
+gdb_test "jump *\$pc" "Breakpoint .* bar break .*"
-- 
1.9.0