I don't feel strongly one way or another, either. I think I would
lean a little toward allowing it to take the NULL, rather than
forcing a small publisher to supply an empty callback.
But if you want to keep it the way it is, than I think the "
MQTTAsync_publish.c" example app needs to be fixed.
Thanks,
Frank
On 01/11/2015 05:57 PM, Ian Craggs
wrote:
Hi Frank,
when I wrote the library, it seemed to make sense to enforce the
setting of the message arrived callback, to make sure any
incoming messages were handled. There are some servers where
a message can be sent to the client unsolicited, without a
subscription. This isn't standard MQTT behaviour of course,
although I'm not sure that the spec actually forbids it.
I'm not strongly attached to my decision to return an error if
you don't set message arrived when calling setCallbacks, so I
would be happy to change it.
Ian
On 09/01/15 17:43, Frank Pagliughi wrote:
Hello Ian, and all,
In the latest Paho C Async library, the publish example,
MQTTAsync_publish.c, sets the callbacks as so (~line 126):
MQTTAsync_setCallbacks(client, NULL, connlost, NULL,
NULL);
As makes sense with a publish-only client, it sets the 4th
parameter for a "message arrived" callback to NULL.
The implementation of the MQTTAsync_setCallbacks function,
however, rejects the NULL value for message arrived (ma)
parameter, in MQTTAsync.c (~line 1719):
if (m == NULL || ma == NULL ||
m->c->connect_state != 0)
rc = MQTTASYNC_FAILURE;
else
...
Which of these should be the correct behavior? I'm guessing
that the library should accept a NULL value for the callback
(?).
Thanks,
Frank
|
