"IllegalState No gpio chip yet initial" does not seem to be a Pi4j exception

[stefanhaustein]

If the project has its own exception hierarchy, wouldn't it make sense to use it consistently, so it can actually be used to catch all pi4j-releated issues?

[eitch]

PR welcome =)

[DigitalSmile]

That's a very good suggestion. While writing the FFM provider code I also noticed the issue... I even tried to make it better in FFM code, but eventually I decided not to do this in the new provider PR. Let me also think how can I contribute

[stefanhaustein]

Maybe we should first take a step back here and create a short list of general code behaviour that's expected to be consistent for pi4j in general?

For instance, I am not sure if all exceptions are supposed to be pi4j exceptions, or if the pi4j exceptions mainly exist to wrap checked exceptions? And then there is the recent discussion on another bug whether failed SPI reads/writes should throw or not. It might be better to answer these questions generally than individually....?

I am not saying we should make everything consistent immediately, but when some issue comes up, it would be good to have some guidelines in which way it should be fixed / aligned...

And perhaps before the actual v4 release we could do a round of consistency checks for the providers we decide to keep around?

[eitch]

That is a good idea. I'm of the opinion, that exceptions should be thrown if something isn't as expected and carrying on will lead to inconsistent state. So for instance when a write didn't write the expected number of bytes, then that is an exception. I propose defining these exceptions:

• Pi4jConfigurationException for wrong parameters on initialization
• Pi4jStateException for when providers aren't ready, etc. or when e.g. when wrong number of bytes are written

Other opinions? Perhaps i missed an exception type? I was thinking of two different for state: one for provider not ready and another for validation when executing, but that might be unnecessarily complicated.

[stefanhaustein]

I think @IAmNickNack mentioned elsewhere that he'd prefer to get access to the error code if one is available. We could make this an optional ctor param of Pi4jStateException.

Generally, I think it would be useful if we keep such general "high level contract" things written up somewhere to ensure consistency. One option might be to factor the existing contribution instructions out of README.md into CONTRIBUTING.md and mention such things there?