How to effectively handle infinite loops?

[Sebight]

Hello, I have successfully implemented most of Berry support into our firmware and now I am wondering how can I effectively detect infinite loops and stop them.

I saw this macro:

But even after decreasing it to as down as value 2, I still get stuck in an infinite loop. The good thing is that it triggers the ESP watchdog, so one possible solution is clearing the saved Berry code if the reset reason was WDT task. However, this could bring up unwanted behaviour.

Any other ideas or experience, maybe the way you handle it in Tasmota? @s-hadinger

Thank you!

[s-hadinger]

Sure, the heartbeat is managed here: https://github.com/arendst/Tasmota/blob/4fbe3590bea16b3d43fc870bda13807e5d223534/tasmota/tasmota_xdrv_driver/xdrv_52_9_berry.ino#L292-L301

And the observability callback is registered here: be_set_obs_hook(berry.vm, &BerryObservability); /* attach observability hook */

So every 1 million instructions, it has the opportunity to raise an exception and kill the infinite loop.

[Sebight]

Perfect, thanks for the resource. I'll fully implement it tomorrow. Just from playing around with be_raise, it seems to call ESP32's abort function. Is that the expected behaviour, or should it call a catchable C++ exception?

Thank you!

[s-hadinger]

Uncaught exceptions result in abort indeed. I always call Berry code with be_pcall() that reports any exception inside the code, for ex: https://github.com/arendst/Tasmota/blob/4fbe3590bea16b3d43fc870bda13807e5d223534/tasmota/tasmota_xdrv_driver/xdrv_52_9_berry.ino#L154-L160

be_pcall() is "protected call", i.e. it will catch any error or exception and not raise to higher levels

[Sebight]

Got it, I was trying out a quick example, and I called the be_raise before even calling the function using be_pcall, but I understand it now.

Thanks a lot for the swift response and plenty of examples.