ESP32 gets no IP and returns no error if esp_netif_create_default_wifi_sta() called before esp_event_loop_create_default() (IDFGH-10321)

[mik13ST]

I am learning ESP32 and ESP-IDF. I was trying to make it connect to WiFi. I looked at the https://github.com/espressif/esp-idf/tree/master/examples/wifi/getting_started/station example and copied relevant method calls until things started working. In this code

void wifi_init_sta(void)
{
    s_wifi_event_group = xEventGroupCreate();

    ESP_ERROR_CHECK(esp_netif_init());

    ESP_ERROR_CHECK(esp_event_loop_create_default());
    esp_netif_create_default_wifi_sta();

    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
    ESP_ERROR_CHECK(esp_wifi_init(&cfg));

I skipped esp_event_loop_create_default() because I didn't know why I need it at first. At a later point I found out why and added this call somewhere before esp_event_handler_instance_register() calls. This worked, I got the WIFI_EVENT_STA_CONNECTED event in my event handler (again, derived from the example), but I never got the IP_EVENT_STA_GOT_IP nor the device ever asked for IP (checked on my router). No runtime errors or warning logs were generated that would hint this being the problem.

I think this should result in an error of sorts. It's not obvious there is a tight coupling (or dependency) between these two methods and the example doesn't mention this tight coupling either. This was very difficult to troubleshoot for a beginner like me, I spent a couple of hours on this.

The reason for this call order requirement is because esp_netif_create_default_wifi_sta() calls esp_event_handler_register() which as a result relies on already existing event loop created by the user.

Also, the esp_netif_create_default_wifi_sta documentation doesn't mention this requirement either.

The implementation looks like this:

/**
 * @brief User init default station (official API)
 */
esp_netif_t* esp_netif_create_default_wifi_sta(void)
{
    esp_netif_config_t cfg = ESP_NETIF_DEFAULT_WIFI_STA();
    esp_netif_t *netif = esp_netif_new(&cfg);
    assert(netif);
    esp_netif_attach_wifi_station(netif);
    esp_wifi_set_default_wifi_sta_handlers();
    return netif;
}

The esp_wifi_set_default_wifi_sta_handlers() call internally returns errors but this function doesn't handle them in any way. I think even assert() as seen in assert(netif) would be better than this. Please don't ignore errors.

An alternative could be the esp_wifi_set_default_wifi_sta_handlers function would create its own event loop so it wouldn't depend on the user's event loop.

[KaeLL]

Good point. Wi-Fi deinitialization has the same issue, where the order of function calls isn't clear.