On 08/23/2012 08:03 PM, Denis Kenzior wrote:
> Hi Philippe,
>
> On 08/23/2012 11:27 AM, Philippe Nunes wrote:
>> GCF test case 31.8.1.2.3 is rejecting the user response.
>> Any subsequent USSD notification are not handled because USSD is
>> always in
>> state user-action.
>> ---
>> src/ussd.c | 12 ++++++++++++
>> 1 file changed, 12 insertions(+)
>>
>> diff --git a/src/ussd.c b/src/ussd.c
>> index 74888b2..21e0164 100644
>> --- a/src/ussd.c
>> +++ b/src/ussd.c
>> @@ -414,6 +414,18 @@ void ofono_ussd_notify(struct ofono_ussd *ussd,
>> int status, int dcs,
>> return;
>> }
>>
>> + if (status == OFONO_USSD_STATUS_TERMINATED&&
>> + (ussd->state == USSD_STATE_IDLE ||
>> + ussd->state == USSD_STATE_USER_ACTION)) {
>
> This check makes no sense, why bother checking for ussd->state ==
> USSD_STATE_IDLE?

The issue has been unveiled after a network initiated USSD request:

\r\n+CUSD: 1,"Type *70*635*562# and send",15\r\n
src/ussd.c:ofono_ussd_notify() status: 1 ACTION_REQUIRED, state: 0 IDLE

then we try to  give an answer:

AT+CUSD=1,"*70*635*562#",15\r
\r\n+CUSD: 2\r\n
\r\n+CME ERROR: 100\r\n


I presume that the same error could occur after a user USSD request. In 
this case, the state is still 'idle' and the "Terminated" notification 
results in a "NotificationReceived" signal with a NULL string. To avoid 
this, I need to check also for ussd->state == USSD_STATE_IDLE. That's 
why the first version was for me more appropriate.


>
>> + ussd_change_state(ussd, USSD_STATE_IDLE);
>> +
>> + if (ussd->pending == NULL)
>> + return;
>> +
>> + reply = __ofono_error_not_supported(ussd->pending);
>> + goto out;
>> + }
>> +
>
> I would handle it just like the 'Not Supported' case but with a new
> __ofono_error_network_terminated error reply. Or perhaps re-use
> __ofono_error_canceled.

In practice, I don't ever expect to send a reply since the 
ussd_response_callback/ ussd_callback is invoked before the TERMINATED 
notification. The error returned in the callback is __ofono_error_failed.
So, what about changing the state to 'idle' if not already set and 
simply return?

Regards,

Philippe.