The delayMicrosecond(100) in PubSubClient::wait_for causes wdt and/or exception failure

[av1024]

The delayMicoroseconds does not update WDT so I have lock-ups, wdt or exception failures on stress-tests (stop MQTT broker, turn off WiFi etc).

Look like replacing in by yield() solves an issue.

[Testato]

the Arduino esp8266 core document that delayMicroseconds() stop the wifi stack, but the delay() not. So maybe we can try whith delay(1), 1millisecondo instead of 0.1. delay pauses the sketch for a given number of milliseconds and allows WiFi and TCP/IP tasks to run

whit the yeld() you removed completely the delay because it is like delay(0)

Probably the best is use a millis timing whithout use of delay, not only here but in all the library ?

[Imroy]

Yes, I had forgotten that delayMicroseconds() doesn't call yield. What if yield is called right after?

[Testato]

Probably the best is use a millis timing whithout use of delay, not only here but in all the library ?

[av1024]

Why do you need 100us delay here? Is it 1ms be enough? Or simply use yield() (equals to delay(0)) and millis()

[av1024]

@Testato yield() is enough because it used for switch scheduler, but time interval measured by millis()

[Imroy]

I'm not sure why I chose 100us. It doesn't seem to be inherited from knolleary's original code, so I can't blame him :P I think I just didn't want it to spin its wheels checking a simple condition and to have a small delay in there. But I forgot about yielding. And it worked for me(tm).

[av1024]

And it worked for me(tm).

For me too, But not for latest git Arduino's 8) I have got a mysterious resets on not clean wifi area but I hope they are gone...

[igrr]

Guys, this code was there since original Nick's version. I remember it caused problems so I had a workaround in my code which called yield() internally if it detected that you are polling WiFiClient::available() in a loop. It seems to be broken now — I will check that.

My suggestions is to replace delayMicroseconds(100) with yield();