Maybe we don't need to include the apparently borked header after all? Hate to make changes to libev source though.

Trying latest batch magic from https://github.com/ogrisel/python-appveyor-demo/tree/master, claims 3.5 support

Also, this comes on the heels of adding back __weakref__ to Semaphore.
This becomes a field in the struct and the field is cleared in
tp_dealloc. It seems like this would trigger the memory leak bug when
taking references to a subclass, but testing seems to show that
doesn't happen. I guess PyPy handles that specially:

>>>> objgraph.show_growth()
>>>> _ = [weakref.ref(gevent.thread.LockType()) for i in range(1000)]; del _
>>>> objgraph.show_growth()
>>>>

[skip ci]

Make sure gevent.lock.Semaphore can be weak referenced. This must be explicitly declared in cython. Fixes #666.

bwc for gunicorn in corner case (https://github.com/benoitc/gunicorn/blob/master/gunicorn/workers/ggevent.py#L219) [skip ci]

internal documentation for the numerous attributes used in of WSGIHandler; also eliminates pylint warnings about setting undeclared attributes

DummyThread objects are much lighter weight and don't allocate and then immediately delete a Semaphore.

This matters for servers that handle each request in a new greenlet (e.g., gunicorn) because logging caused these objects to be created. Avoiding the Semaphore is especially important on PyPy.

Brought up on IRC.

This involves having read_request raise ValueError subclasses instead of directly logging errors and returning False. This method could already raise ValueError when parsing headers, so callers are prepared for this. It shouldn't be a performance issue because (1) the try block was already being established and (2) these are exceptional situations, not the common case.

Disable test_ssl.test_algorithms test temporarily because the host it uses is not resolvable. Also whitespace cleanup in the list for consistency.

Work around a the worst tuple leak in Cython code under PyPy when releasing semaphores. A complete fix needs a newer cython release.

Mark test__issue6.py flaky everywhere, I'm getting tired of re-running test environemnts on Travis one at a time because of it.

Needs a test case.

Comment-deprecate the undocumented rfile kwarg to WSGIHandler (not using the warnings module because this is a hot code path and also because DeprecationWarning is ignored by default under Py27 anyway) [skip ci]