[RFC][PATCH] Fixes to gdb.mi testsuite

[Yao Qi]

On 05/29/2013 10:38 PM, Simon Marchi wrote:
>> MI notification =cmd-param-changed is not suppressed by mistake in
>> >GDB.  I've post a patch
>> ><http://sourceware.org/ml/gdb-patches/2013-05/msg00976.html>  to fix
>> >it.  Once it is committed, this part can simplified.
> Thanks.
>

The patch was committed.

>>> >>+        }
>>> >>+        default {
>>> >>+            fail "-exec-arguments output not matched"
>>> >>+            return -1
>>> >>+        }
>>> >>+        }
>>> >>+    }
>>> >>+    }
>>> >>+
>>> >>        send_gdb "${run_prefix}run $args\n"
>> >
>> >Looks the $args are passed to -exec-run command, but -exec-run command
>> >doesn't pass args to the inferior.  See mi/mi-main.c:mi_cmd_exec_run.
>> >If we can teach MI command -exec-run pass argument to inferior (not
>> >sure it is a good idea or not), we don't have to use -exec-arguments
>> >here.
> When $use_mi_command is false, then the command is simply "run"
> (${run_prefix} is empty), which passes the arguments to the inferior.

Yes, if CLI command 'run' can pass arguments to inferior, why doesn't MI 
command '-exec-run' do the same thing?  Unless there are some 
difficulties I don't realize to teach MI command '-exec-run' to pass 
arguments to inferior, I still prefer this way.

> What do you think? I left out the check for "llength $args > 0". If no
> args are passed to mi_run_cmd_full, we want to make sure to erase any
> arguments that could have been set previously. Also, should I use
> gdb_test_no_output instead of send_gdb for -exec-arguments and set args?

We don't use gdb_test_no_output in MI test cases.  I am not the people 
to approve or reject your patch, so probably you can update your patch 
on top of cvs trunk to discard the matching on "=cmd-param-changed", 
post a clean patch again, and see what other people think about it.

-- 
Yao (齐尧)