[PATCH 2/3] mgcp: Synchronize conn mode bits and output enabled flags

[Jacob Erlbeck]

Hi

On 20.01.2014 08:18, Holger Hans Peter Freyther wrote:
> On Thu, Jan 16, 2014 at 04:50:40PM +0100, Jacob Erlbeck wrote:
>>> +	/* Check that LOOPBACK enables all output */
>> +	OSMO_ASSERT((MGCP_CONN_LOOPBACK & MGCP_CONN_RECV_SEND) ==
>> +		    MGCP_CONN_RECV_SEND)
> 
> I think osmo_static_assert would work here too?

Yes, it would.

> 
>> -			if (endp->conn_mode != -1)
>> +			if ((endp->conn_mode & CONN_UNMODIFIED) == 0)
>>  				printf("Connection mode: %d, "
>>  				       "BTS output %sabled, NET output %sabled\n",
>>  				       endp->conn_mode,
> 
> In the next step we could change the test output to illustrate the
> tri-state from on/off and not modified to previous test?

Ok, I'd print the output_enabled values to stderr then.

> 
>> +			endp->conn_mode = MGCP_CONN_NONE | CONN_UNMODIFIED;
>> +			endp->net_end.output_enabled = 0;
>> +			endp->bts_end.output_enabled = 0;
> 
> do we still need these flags? Or only because of the above printf?

Yes, we do. Since conn_mode (the SEND_RECV bits) is modified, the
output_enabled must be adjusted, so that the invariant holds (this is
the central topic of this patch). I didn't want to have a modified
behaviour of the test, since this ought to be a semantic-neutral
modification (ideally with a unmodified ok file). See the next patch
"mgcp/test: Don't reset conn_mode between messages", that addresses this
issue.

Jacob

> 
>