api: remove nullable from StatusOr value methods

[panchenko]

It someone wants it nullable, that should be defined via the type parameter, e.g. StatusOr<@Nullable Foo> foo

[ejona86]

I'm a little bit worried that this is getting into specifics of one Nullable annotation vs another; I know they aren't all equivalent and there were limitations lifted with later annotations. Given that everything is @Nullable, we know this annotation has no behavior right now, which means verifying behavior is out of the question.

[panchenko]

I was trying to use this type in our code, with the result of unexpected warnings from IDE

There are plenty of such warnings in this repo as well, e.g.

grpc-java/xds/src/main/java/io/grpc/xds/ClusterResolverLoadBalancer.java

Lines 118 to 119 in 1a7042a

	ClusterResolutionResult result = statusOrResult.getValue();
	List<EquivalentAddressGroup> addresses = result.addresses;

getValue() is nullable, so result can be null, and then there is a warning for result.addresses

[ejona86]

All arguments/return types in gRPC should be Nullable from a tool's perspective. We don't have NonNull enabled at the package level. To my current understanding, if these changes here are causing changes in the IDE, then the IDE is broken.

[panchenko]

From a tool's perspective - default nullability is unknown.
However ,getValue() is explicitly annotated as @Nullable, that makes the result local variable also nullable. There are no null checks in this code, so the tool issues a warning about possible NPE. That's the whole point of using @Nullable. The tool (IDE) works correctly.

Why getValue() is @Nullable?

[ejona86]

The suggested approach doesn't actually work with javax.annotation.Nullable:

   grpc-java/api/src/main/java/io/grpc/StatusOr.java:69: error: annotation @Nullable not applicable in this type context
    private static final StatusOr<@Nullable Object> TESTING = null;

But the @Nullable isn't actually providing any value (as it doesn't actually restrict whether the method can return null, as no method is annotated non-null), so if it is causing problem we can remove it for now.