`unnecessary_null_comparison` warning doesn't show consistently (if it should at all; and it shows an incorrect message...?) about sdk

[scheglov]

FYI, this was implemented intentionally in #51093.

[scheglov]

I suspect that consistency between "no initializer" and "initialized with null" cannot be easily implemented. We have a long term experiment that might help with this, but I don't think we have immediate plans when it will be ready. Theoretically we could have "is the variable ever assigned not null value" flag, but it seems that the value of this feature is quite low. Prove me wrong, of course, I might undervalue it.

I agree that the error message is confusing.

[skylon07]

That's good insight. I agree with you, I think seeing that inconsistency (my first point) would lead to at worst a "huh, that's weird" kind of scenario -- I mostly brought that up in case it was some kind of simple oversight (which it appears it is not). I do hope to see changes to the error message though (my second point), assuming that's a quick-and-easy fix. That message is what led me to digging in and making sure I understood variable initialization correctly, since the error seemed to imply it could have "not been null" by being uninitialized (which isn't true).

At the end of the day though, I'd say all of this put together is still just a minor inconvenience at worst.