The clients now block until the channel is in the READY state. This
fixes some test flakiness issues we have had.

It's been confusing that these tests have been called "interop" but
are not actually tests of interoperation.

Added tests for detecting invocation time defects arising out of the
runtime.

Cast iterables into iterators for stream based compression, empty
message, and metadata tests.

Restructure the consume_request_iterator method to handle exceptions
where the object being passed is not an iterator. Fixes #8231 and #8454.

Documentation for the new tests can be found https://github.com/grpc/grpc/blob/master/doc/negative-http2-interop-test-descriptions.md

Add a comma delimiter for "_invalid_metadata_test.InvalidMetadataTest".

Apparently Exception.message was introduced in Python 2.5 and
deprecated almost immediately afterward in Python 2.6.

Uses dynamic loading to paper-over the negative effects of losing
namespace packages in the previous commit.

Uses dynamic loading to paper-over the negative effects of losing
namespace packages in the previous commit.

Setuptools was updated and our hacky namespace-package-chickens came
back to roost. This removes the unsupported namespace package hacks.

Setuptools was updated and our hacky namespace-package-chickens came
back to roost. This removes the unsupported namespace package hacks.

It was a mistake that requests might be sent; the test specification
calls for no requests to be sent. It was a mistake that the response
future's cancelled() method was called; the cancelled() method returns
something more like "was this object's cancel() method called earlier?"
than "did the RPC terminate with status code CANCELLED?". Since it's
something that we'd well enough like to work I've retained the
cancelled() call with a different failure message.

ServicerContext.set_code takes a (grpc.)StatusCode, not an integer.

This reverts commit 3045a379.

Backports per-object grpc_init/deinit and separated-file grpc protoc
codegen (#7538, #8246, #8920).

Almost fixes #8753 (which still needs a test).

The server cleanup method is untested.

The join() function that exposes it is only called by the internals of threading.py, and we don't hold a reference to the server thread to explicitly join() it, and I'm not sure we should add a reference just for this purpose.

Moreover, the threading.py only calls join(None), the code path in question isn't even exercised by the internals of threading.py. Its just there to make sure we properly follow the join(timeout) semantics.

The requirement that any created managed call must have operations
performed on it is obstructing proper handling of the case of
applications providing invalid invocation metadata. In such cases the
RPC is "over before it starts" when the very first call to
start_client_batch returns an error.

Add use_tls, use_test_ca, and server_host_override flags to Python
stress client. These are needed to run the stress client against a local
server that is using tls.