Daniel McBrearty wrote:
-------------------------------
(LOL ... Well, I was trying ...)



So do all your systems rely on a watchdog then?


Whenever possible.


When software is used to protect against nasty physical problems, like intermittent spikes on a pin, that protection can be switched in for the final product, but NOT during development and testing. Otherwise, the problem might get WORSE in the field, and you'll kick yourself that you didn't know about it in test.



Suppose there was a PCB routing problem - the switch line was too near to a mains signal. There are spikes on the switch pin, but small enough to be suppressed by the debounce. Nobody knows it's there.




Now I would have spotted them with my scope anyway, and got rid of them, but I would still use anti-bugging anyway..

And noone has yet explained to me why a switch pin on a SBC should be more likely to get noise than an edge triggered int pin - which is more sensitive and maybe more critical.


because its coming from a mechanical system . Hell it could VIBRATE and give an output ! - and it could be perfectly electrically noise free. Have you been watching John Carmack's Armadillo Aerospace rocket project ? John Carmack (Wolfenstein 3D, Doom, Quake), now doing real time rocket control systems, with a poor success rate, often caused by inattention to noise sources like I tend to worry about....
By all means, play it safe in the field ... but ... you either trust your hardware or you don't.
This is one game where paranoia is healthy and employment enhancing