Joshua Haberman
Joshua Haberman
I wasn't able to reproduce the output above. I get: ``` $ USE_BAZEL_VERSION=8.x bazel build rust/test:parent_proto --norun_validations Starting local Bazel server (8.4.2) and connecting to it... WARNING: For repository 'rules_cc',...
@Logofile where would be the best place to document this performance edge case?
I can handle manually importing this and resolving the compiler errors.
> Classes are not marked as final using the @typing.final decorator Marking classes as final in `.pyi` sounds like a great idea. > The stubs indicate classes have __slots__, when...
Assigning to @anandolee for further investigation.
> I could see a type-checker have special handling in non-strict mode: like allowing unknown attributes get/set in non-strict mode, but still disallowing it if __slots__ are set. I discovered...
Is there any way to make both type checkers happy? It sounds like the Python type checking semantics are not always consistent across tools.
If anyone can propose a solution that makes both mypy and PyCharm happy, we'd be happy to accept a contribution.
It's been a while since I looked at this, so my memory doesn't help. But looking at https://github.com/protocolbuffers/protobuf/issues/21691#issuecomment-3185957672, it seems that my testing showed that a statement like `msg.nonexistent_field= 1`...
@jasonkuster thanks for the summary! If reverting https://github.com/protocolbuffers/protobuf/pull/23156 would be the most helpful for the time being, I'm happy to do that.