[commit] Fix gdb.mi watchpoint tests for remote

[commit] Fix gdb.mi watchpoint tests for remote

[Daniel Jacobowitz]

The non-MI watchpoint tests do the same thing: if the board file
requests we not try hardware watchpoints, tell GDB not to.  So
I made the MI watchpoint tests do the same, which fixes two
failures when using gdbserver.

Of course this makes the tests run with software watchpoints.
They happen to fail, because we don't realize when we're in
the epilogue early enough (true on both amd64 and powerpc).
Progress, anyway...

Tested x86_64-linux and committed.

-- 
Daniel Jacobowitz
CodeSourcery

2007-09-06  Daniel Jacobowitz  <dan@codesourcery.com>

	* gdb.mi/mi2-watch.exp: Set can-use-hw-watchpoints to 0 if required.
	* gdb.mi/mi-watch.exp: Likewise.

Index: gdb/testsuite/gdb.mi/mi2-watch.exp
===================================================================
--- gdb/testsuite/gdb.mi/mi2-watch.exp	(revision 181154)
+++ gdb/testsuite/gdb.mi/mi2-watch.exp	(revision 181155)
@@ -168,6 +168,11 @@ proc test_watchpoint_triggering {} {
     }
 }
 
+# Disable hardware watchpoints if necessary.
+if [target_info exists gdb,no_hardware_watchpoints] {
+    mi_gdb_test "-gdb-set can-use-hw-watchpoints 0" "\\^done" ""
+}
+
 mi_runto callee4
 test_watchpoint_creation_and_listing
 #test_rwatch_creation_and_listing
Index: gdb/testsuite/gdb.mi/mi-watch.exp
===================================================================
--- gdb/testsuite/gdb.mi/mi-watch.exp	(revision 181154)
+++ gdb/testsuite/gdb.mi/mi-watch.exp	(revision 181155)
@@ -168,6 +168,11 @@ proc test_watchpoint_triggering {} {
     }
 }
 
+# Disable hardware watchpoints if necessary.
+if [target_info exists gdb,no_hardware_watchpoints] {
+    mi_gdb_test "-gdb-set can-use-hw-watchpoints 0" "\\^done" ""
+}
+
 mi_runto callee4
 test_watchpoint_creation_and_listing
 #test_rwatch_creation_and_listing

Re: [commit] Fix gdb.mi watchpoint tests for remote

[Maciej W. Rozycki]

On Thu, 6 Sep 2007, Daniel Jacobowitz wrote:

> The non-MI watchpoint tests do the same thing: if the board file
> requests we not try hardware watchpoints, tell GDB not to.  So
> I made the MI watchpoint tests do the same, which fixes two
> failures when using gdbserver.
> 
> Of course this makes the tests run with software watchpoints.
> They happen to fail, because we don't realize when we're in
> the epilogue early enough (true on both amd64 and powerpc).
> Progress, anyway...

 In case you or somebody else is interested and/or working in this area 
now -- I have a patch in my queue that makes these tests be run twice: 
once with hw watchpoints and once with sw watchpoints.  I could merge it 
with your changes and make it jump the queue.  I cannot guarantee it will 
not introduce regressions, but this was the very idea of this change 
actually -- to discover rotting support for sw watchpoints for platforms 
normally using hw ones (it succeeds both ways for my usual MIPS platforms 
though).

  Maciej

Re: [commit] Fix gdb.mi watchpoint tests for remote

[Daniel Jacobowitz]

On Fri, Sep 07, 2007 at 12:59:39PM +0100, Maciej W. Rozycki wrote:
>  In case you or somebody else is interested and/or working in this area 
> now --

Interested, but not working on it... it's hard to fix.

> I have a patch in my queue that makes these tests be run twice: 
> once with hw watchpoints and once with sw watchpoints.  I could merge it 
> with your changes and make it jump the queue.  I cannot guarantee it will 
> not introduce regressions, but this was the very idea of this change 
> actually -- to discover rotting support for sw watchpoints for platforms 
> normally using hw ones (it succeeds both ways for my usual MIPS platforms 
> though).

No rush, but I do think that this is a good idea.

-- 
Daniel Jacobowitz
CodeSourcery