[skip ci]

[skip ci]

Fixes #1236

[skip ci]

* Switch off prospector, go back to plain pylint.

prospector wasn't working on CI for some reason.

Fix the outstanding pylint errors.

* Explicitly request pylint 2 on Python 3

* Also need to update astroid.

* pylint warning generated on Travis (case-sensitive filesystem.)

* Sleeping updates the loop's current time always.

That had been disabled for libuv, but now that we run timers in our own code that's not a concern anymore.

Fixes #1227.

* Try Python 3.5.5 because of the weird Illegal Instruction crash we suddenly see with 3.5.4

* Increase the value; test_sendall_timeout was often failing to actually timeout.

* Really python3.5

[skip ci]

[skip ci]

[skip ci]

[skip ci]

Allow weakreferences to Queue objects. Fixes #1217.

[skip ci]

[skip ci]

Fix gevent.os.nb_write/read not always closing the IO event the opened, which is a problem on libuv.

Also fix the formatting in changelog_1_0.rst so it renders legibly.

I checked with restview --long --pypi and the long description still
renders nicely, as do the pages created by sphinx

I thought the use of .. include:: would make README.rst still render
with all its contents, but sadly github doesn't enable
that (https://github.com/github/markup/issues/172)

 Let Event have weak references.

Catch the ValueError that sys._getframe() can raise when there are no Python frames.

This can happen in the embedded case, when the CPython C api is used to directly call Greenlet.spawn, which itself is now implemented in C.

Fixes #1212.

Move the __weakref__ to the base class for Cython because we did see a failure in 3.6 too: https://travis-ci.org/gevent/gevent/jobs/379718754#L753

Fixes #1211

Note that this can crash on CPython 3.7b4; see https://github.com/cython/cython/issues/2270

Refs #1207 and #1209.

[skip ci]

[skip ci]

Run leakchecks on Python 3.7.

Fix ABCs in the MRO of gevent.local. Fixes #1201

Try to make it take less than the 200s timeout.

Update config.guess/sub to latest versions.

Fixes #1197

Switch to objgraph to handle the measurement for us. That cleared up a
few of the obscure issues with references to
functions/getset_descriptors and the like. (Possibly because it keys
by string names and we were keeping type objects alive.)

Many of the real failures were due to re-using exception instances,
which is bad because of chaining.

Most of the @ignore_leakcheck are for performance, only one is for a
real issue---and that test was skipped already on CI anyway for being
too flaky.

[skip ci]

[skip ci]