Tim Spriggs

I've long since cleared out my cache of go 1.6 cached objects so I can't reproduce anymore. Sorry. It seemed to be related to having both 1.6 and 1.7 objects...

NullAway works specifically by looking for an `@Nullable` annotation. AFAICT, you could roll your own or if you were willing to include the JSR305 library you could use `javax.annotations.Nullable` From...

Maybe we should optionally retain them since proto3 is "in the wild" with this behavior now? It's possible some people are relying on not having extra fields once the local...

I think a schema (maybe checked in/versioned) should be a valid way to ensure all of the details (eg: nullable) line up correctly through a codebase. Doing it any other...

FWIW, I still see this in errorprone 2.11.0 (and 2.13.1)

I can reproduce in my codebase with 2.13.1. Here is a minimal reproducer for me as `com/example/Foo.java`: ```java package com.example; import javax.inject.Inject; import java.util.Map; public class Foo { private record...

If you use a hash instead of a list then the performance shouldn't suffer nearly as much: ``` + def end_mapping + mapping = pop + keys = {} +...

Could you subclass `Supplier` and annotate the subclass' return value with `@Nonnull`? eg: ``` interface NonNullSupplier extends Supplier { @Nonnull T get(); } ```

If we were to correctly assume that `get()` from `java.util.function.Supplier` is nullable then the `NonNullSupplier` interface makes sense. If we incorrectly assume that `get()` is Nonnull then we can't create...

> This I don't understand. If you have a NonNullSupplier why would you want to allow an implementation to return null? You want NullAway to mark it as an error....