Manu Sridharan

Yeah I think our Rx support is a bit brittle and needs one-off support for cases like this. We can think about how to make it more robust.

Thanks for the report! We do have support for inferring nullness information from built-in language equality tests like`==` and `!=`. I imagine the test assertion handler could be enhanced to...

> We're currently using NullAway v0.9.1. BTW, I'd recommend updating to version 0.9.5 if you're able. In particular, 0.9.3 fixed a bug that could lead to missed nullness issues.

> In v0.9.5 we have a similar issue with the `isA` matcher (probably just exposed by some other improvement where something is detected as nullable which previously was not): >...

Couple notes based on some initial tinkering: * We'll need to handle assignments, e.g., you can't assign a `Pair` to `Pair` (with both `@NonNull`). Probably just want to support invariance....

FWIW Error Prone has some support for generics and null checking now, implemented by @bennostein. See the [inference](https://github.com/google/error-prone/tree/master/check_api/src/main/java/com/google/errorprone/dataflow/nullnesspropagation/inference) code and [this commit](https://github.com/google/error-prone/commit/75970cc19c1347579d356055583b5683c5be8f96#diff-a0dd292c7e34084af66261b0e00a1f2e). Perhaps we could do something similar in NullAway?...

Is there a brief explanation anywhere of the advantages?

> Is there a brief explanation anywhere of the advantages? For my own future reference, I think resource limitations on Travis CI are a good enough reason to move. Also,...

Thanks for the offer, @sherrif10! But, I think it will be easier for a maintainer still at Uber like @lazaroclapp to take care of this one, as it requires dealing...

Thanks for the report! I assume all these tests are with Guava 31.x? Regarding examples 3 and 4, these work as expected since NullAway does not support `@ElementTypesAreNonnullByDefault`. I feel...