Currently, if two threads call grpc_completion_queue_pluck on the same
completion queue for different tags, there is a 50% chance that we
deliver the completion wakeup to the wrong poller - forcing the correct
poller to wait until its polling times out before it can return an event
up to the application.
This change tweaks our polling interfaces so that we can indeed wake a
specific poller.
Nothing has been performance tuned yet. It's definitely sub-optimal in a
number of places. Wakeup file-descriptors should be recycled. We should
have a path that avoids calling poll() followed by epoll(). We can
probably live without it right at the second though.
This code will fail on Windows at least (I'll do that port when I'm in the office and have a Windows
machine).
- Positive tests pass even if we will have to change the interface to
add the processor to the server credentials (will be done in a
separate pull request).
- ASAN leaks for the error case.
- The client should get a GRPC_STATUS_UNAUTHENTICATED as opposed to
GPRC_STATUS_INTERNAL.
- Right now it is a global function: would be better to have this per
(secure) port.
- Changed the interface of the auth_context slightly to make it more
friendly.
- Positive tests pass. Still need some work on error case (have a
negative case as well).
- Fixing cpp auth context tests so that they use the shiny new C API.