Hi Denis,

On 17/02/2011 17:14, Denis Kenzior wrote:
> Hi Guillaume,
>
> On 02/17/2011 07:48 AM, Guillaume Zajac wrote:
>> ---
>>   gatchat/test-server.c |    6 ++++++
>>   1 files changed, 6 insertions(+), 0 deletions(-)
>>
>> diff --git a/gatchat/test-server.c b/gatchat/test-server.c
>> index e574d64..c0a5a4d 100644
>> --- a/gatchat/test-server.c
>> +++ b/gatchat/test-server.c
>> @@ -137,6 +137,12 @@ static void ppp_disconnect(GAtPPPDisconnectReason reason, gpointer user)
>>   	g_at_ppp_unref(ppp);
>>   	ppp = NULL;
>>
>> +	if (reason == G_AT_PPP_REASON_NET_FAIL) {
> Don't you mean G_AT_PPP_REASON_LINK_DEAD?
>

In case of test-server, I didn't reproduce the crash with/without the 
fix, I applied it to test server first as we don't have dial_cb for 
emulator yet:

I did ./gsmdial -i localhost -p 12346

At the end of the exchange:
ppp_lcp_down_notify() is called
disconnect reason is G_AT_PPP_REASON_PEER_CLOSED
then the GAtServer is well resumed.

In case of emulator (based on Gustavo/ZhenZhua dial_cb), 
ppp_lcp_down_notify() is called but just after this one, 
ppp_ipcp_up_notify() fails and sets disconnect reason to 
G_AT_PPP_REASON_NET_FAIL.
Then we get a crash after ppp_disconnect because we call the hdlc write 
handler e.g. can_write_data() into gathldc.c although we have unref the 
PPP server.

So I called ofono_emulator_remove() that will call the unref GAtServer 
etc... when we get G_AT_PPP_REASON_NET_FAIL into ppp_disconnect

I don't get G_AT_PPP_REASON_LINK_DEAD disconnect reason when emulator 
crashes.

So should we call ofono_emulator_remove()/GAtServer_unref() in both 
disconnect reason cases?
Or is there a specific action to do when G_AT_PPP_REASON_NET_FAIL?

>> +		g_at_server_unref(server);
>> +		server = NULL;
>> +		return;
>> +	}
>> +
>>   	g_at_server_resume(server);
>>   	g_at_server_set_debug(server, server_debug, "Server");
>>
> Regards,
> -Denis
>

Kind regards,
Guillaume