Device-os: Particle.keepAlive() API does not work

[technobly]

Bug Report

Expected Behavior

You can call Particle.keepAlive(30); at any time and it should ping the server every 30 seconds. The system should keep track of the last TX/RX time and compare against the keepAlive interval (which if changed to a shorter setting may kick off a ping).

Observed Behavior

No activity in the log output every 30 seconds as expected when keepAlive is set before we are connected to the Cloud.

Steps to Reproduce

• Run the test app and observe the log output.
• Fails to work in THREADED or NON-THREADED mode
• Tested with system firmware 0.7.0-rc.4 and 0.8.0-rc.1

Test App

SerialLogHandler logHandler(LOG_LEVEL_ALL);

SYSTEM_THREAD(ENABLED);

void setup() {
    Particle.keepAlive(30); // send a ping every 30 seconds
}

Workaround

Call keepAlive after we are connected to the Cloud.

void loop() {
    static bool once = false;
    if (!once && Particle.connected()) {
        Particle.keepAlive(30); // send a ping every 30 seconds
        once = true;
    }
}

[ScruffR]

Maybe the "expected" behaviour should be that you can call Particle.keepAlive() anytime (prior or after connection) or at least like Particle.variable(), Particle.function() and Particle.subscribe() before up to a few seconds after the connection and still work as expected.

[technobly]

Expected behavior is just that, that you can call it any time and it just sets the ping interval. The system can keep track of the last TX/RX time and compare against the keepAlive interval (which if changed to a shorter setting may kick off a ping).

[kubark42]

Agreed with @technobly that the expected behavior is that there be no race condition for setting the keepAlive interval. Once set anywhere in the code, this should be propagated and preserved.

On a personal note, I spent a while independently rediscovering this bug, including an hour spent debugging with Particle's tech support. The bug also exists in 0.6.4 and 0.7.0. Think the fix will be backported?

P.S. As a follow-up, I found that the workaround can be simplified by keeping the code in setup(), as follows:

void setup() 
{
  // .
  // .
  // .

  Particle.connect();

  // Wait endlessly until the Particle is connected
  while (Particle.connected() == false) {
  }

  // Set the keep alive time. If this is too short, we won't be able to contact the Electron from the
  // Particle cloud.
  // NOTE: THIS MUST BE DONE AFTER THE PARTICLE IS CONNECTED. The upshot is
  // that in MANUAL or SEMI_AUTOMATIC modes this needs to be called after
  // Particle.connect() is successful.
  Particle.keepAlive(30);

  // .
  // .
  // .
}