Resend PUBACK and PUBREC irrespective of the DUP flag #83
Conversation
We already do this, since |
| @@ -1054,7 +1054,7 @@ static MQTTStatus_t handleIncomingPublish( MQTTContext_t * pContext, | |||
| * state engine. This will be handled by ignoring the | |||
| * #MQTTStateCollision status from the state engine. The publish | |||
| * data is not passed to the application. */ | |||
| else if( ( status == MQTTStateCollision ) && ( publishInfo.dup == true ) ) | |||
| else if( status == MQTTStateCollision ) | |||
There was a problem hiding this comment.
If we're making this change specifically for AWS IoT Core, we should at least log a warning, if not an error. If the DUP flag isn't set, it seems to be a protocol violation (3.3.1.1):
The DUP flag MUST be set to 1 by the Client or Server when it attempts to re-deliver a PUBLISH Packet
There was a problem hiding this comment.
Please see the updated logs.
source/core_mqtt.c
Outdated
|
|
||
| if( publishInfo.dup == false ) | ||
| { | ||
| LogError( ( "DUP flag is 0 for duplicate incoming publish packet." ) ); |
There was a problem hiding this comment.
Should we have a more ominous sounding error message like "Protocol violation: ..."? I think we should be clear that this isn't according to the spec.
There was a problem hiding this comment.
I would simply say "DUP flag is 0 for duplicate packet (MQTT-3.3.1.-1)" that gives all the breadcrumbs you need to figure it out. If you just google MQTT-3.3.1.-1 you will hit it directly.
| @@ -955,7 +955,7 @@ static MQTTStatus_t deserializeConnack( const MQTTPacketInfo_t * pConnack, | |||
| if( ( pRemainingData[ 0 ] & MQTT_PACKET_CONNACK_SESSION_PRESENT_MASK ) | |||
| == MQTT_PACKET_CONNACK_SESSION_PRESENT_MASK ) | |||
| { | |||
| LogWarn( ( "CONNACK session present bit set." ) ); | |||
| LogInfo( ( "CONNACK session present bit set." ) ); | |||
There was a problem hiding this comment.
Thanks. I'm pretty sure this was a leftover from v4_beta, where the session present flag wasn't given back to the application
yourslab
changed the title
In the OASIS MQTT Version 3.1.1 OASIS Standard, the following is stated:
Also, when AWS IoT Core resends a PUBLISH with the same packet ID as the first that wasn't PUBACK'd, the DUP flag will always be set to 0. Therefore, our library will return
MQTTStateCollisionand will NOT send a PUBACK in response.For QoS 2, the OASIS spec does not seem to explicitly state anything about whether or not to resend a PUBREC in response to a resent PUBLISH with the DUP flag set to 0. However, I think the same logic still applies, so the PUBREC must still be resent in that case. Feel free to disagree.