On 12/23/2010 12:16 PM, Joel Brobecker wrote:
>> +# Get things started.
>> +
>> +gdb_exit
>> +gdb_start
>> +gdb_reinitialize_dir $srcdir/$subdir
>> +gdb_load ${binfile}
> 
> Use "clean_restart"....

Done.


>> +# which is in library might be compiled in Thumb.
>> +gdb_test "set displaced-stepping off" ""
> 
> Use gdb_test_no_output.

Done.

>> +gdb_test "set displaced-stepping on" ""
> 
> Same here.
> 

Done.

>> +gdb_test_multiple "continue" "continue to test_call" {
>> +	-re ".*bl test_call_subr.*" {
>> +	    pass "continue to test_call"
>> +	}
>> +	-re "Displaced stepping is only supported in" {
>> +	    xfail $testfile
>> +	    return
> 
> I'm not sure about the xfail. Why did you decide to use it?

Because I misunderstood xfail.

> I thought we were supposed to use kfail in this case, since this is
> a debugger issue? xfail should be used when the failure is caused
> by the debugging environment (compiler bug, missing feature in kernel,
> etc).

Yes, we should use kfail here.

Now, run arm-disp-step.exp with with
\{-mthumb,-marm\}\{-march=armv5t,-march=armv7-a\}, get result like this,
		=== gdb Summary ===

# of expected passes		60
# of known failures		2

-- 
Yao (齐尧)