esp_mqtt_client_publish Semaphore lock if the wifi/network is not connected yet (IDFGH-4169)

[umer-ilyas]

steps to reproduce issue

• wifi initialized but not connected to AP yet
• mqtt client started esp_mqtt_client_start
• if i call esp_mqtt_client_publish then MQTT_API_LOCK(client); does not return

I have not figured out where is sempore is acquired but it is something to do with wifi connection

• after few seconds wifi connects and MQTT_API_LOCK(client); gets the lock and everything return to normal

if for some reason wifi AP is not available MQTT_API_LOCK(client); in function esp_mqtt_client_publish will stuck forever

[david-cermak]

Hi @umer-ilyas

Thanks for reporting. The reason why the publishing gets stuck is that the esp_mqtt_client_start() tries to connect to the broker in the initial phase. This should however fail in the condition you've described and unlock the mutex for the publish to continue. And as far as I can see it works like that when tried to publish just after starting the client and before connecting to the AP:

I (1015) MQTT_EXAMPLE: [APP] Before publish
D (1015) MQTT_EXAMPLE: Event dispatched from event loop base=MQTT_EVENTS, event_id=7
I (1025) MQTT_EXAMPLE: Other event id:7
E (1035) TRANS_TCP: [sock=54] connect() error: Host is unreachable
E (1045) MQTT_CLIENT: Error transport connect
D (1045) MQTT_EXAMPLE: Event dispatched from event loop base=MQTT_EVENTS, event_id=0
I (1055) MQTT_EXAMPLE: MQTT_EVENT_ERROR
D (1055) MQTT_CLIENT: Reconnect after 10000 ms
D (1065) MQTT_EXAMPLE: Event dispatched from event loop base=MQTT_EVENTS, event_id=2
I (1075) MQTT_EXAMPLE: MQTT_EVENT_DISCONNECTED
D (1075) MQTT_CLIENT: mqtt_enqueue id: 969, type=3 successful
D (1085) OUTBOX: ENQUEUE msgid=969, msg_type=3, len=23, size=23
D (1085) MQTT_CLIENT: Publish: client is not connected
I (1095) MQTT_EXAMPLE: [APP] After publish

This is what I tried to do:

    esp_mqtt_client_start(client);
    ESP_LOGI(TAG, "[APP] Before publish");
    esp_mqtt_client_publish(client, "/topic/qos1", "data_3", 0, 1, 0);
    ESP_LOGI(TAG, "[APP] After publish");

and it seemed to work as expected:

• client started (mutex locked when trying to reach the broker)
• publish stuck on the lock waiting for the mutex
• client connection failed unlocking the mutex
• publish done (msg moved to the outbox)

Are you seeing a different behaviour?

[umer-ilyas]

@david-cermak i was expected the same behaviour as you mentioned
but i am getting a different behaviour
let me digg little bit deeper and get back

[david-cermak]

@umer-ilyas Any update? Closing for now. Feel free to reopen once applicable.