In https://github.com/grpc/grpc/issues/19910, it was pointed out that
raising an exception from a Future callback would cause the channel spin
thread to terminate. If there are outstanding events on the channel,
this will cause calls to Channel.close() to block indefinitely.
This commit ensures that the channel spin thread does not die. Instead,
exceptions will be logged at ERROR level.
By popular demand, we'll now be offering separate py_grpc_library and
py_proto_library targets sharing the same interface as within google3.
This change necessitated some modifications to how we pull in our own
Python-level dependencies and how we make those available to those
pulling in our project via Bazel.
There is now a grpc_python_deps() Bazel workspace rule that pulls in the
appropriate dependencies, which should be called from the client
project's WORKSPACE file. A test has been added to the bazel/test/
directory to verify that this behavior works as intended.
It's worth noting that the protobuf repository's usage of Starlark
bind() caused a great deal of trouble in ensuring that we could also
pull in six.
This change also required a change in the way generated proto code is
imported in the channelz and health-check modules, as well as in their
associated tests. We were importing them two different ways, each
relative. This resulted in two different module objects being imported
into the process, which were incompatible. I am not sure exactly what
caused this behavior to begin, as this should have been possible before
this PR. As a workaround, I am simply trying two different absolute
imports and using the one that works. This should function both inside
and outside of Bazel environments.
Previously, signal handlers were only given a chance to run upon receipt of an
entry in the RPC stream. Since there is no time bound on how long that might
take, there can be an arbitrarily long time gap between receipt of the signal
and the execution of the application's signal handlers.
Signal handlers are only run on the main thread. The cpython implementation
takes great care to ensure that the main thread does not block for an
arbitrarily long period between signal checks.
Our indefinite blocking was due to wait() invocations on condition variables
without a timeout.
This changes all usages of wait() in the the channel implementation to use a
wrapper that is responsive to signals even while waiting on an RPC.
A test has been added to verify this.
Tests are currently disabled under gevent due to
https://github.com/grpc/grpc/issues/18980, but a fix for that has been
found and should be merged shortly.
Esun Kim