PyPy3 tests hanging intermittently.

[jaraco]

Recently, I've observed the test suite hanging in Travis on test_develop.py on test 6. Travis ends up killing the process after 10 minutes.

Here's one example.

The first incident can be found in this build, which reports pypy 5.8.0-beta0. But going back through the dozens of prior builds that succeeded, there was no change in pypy.

That failing test is already skipped in AppVeyor as reported in #851. And that test makes two calls to subprocess.check_call which may be unreliable on PyPy3 as well.

[jaraco]

I've added a couple of commits in the bugfix/1202-pypy3 to attempt to delve into the failure a bit. Though I've twice run the tests and have yet to replicate the failure. I'll keep restarting the build (up to 20 times) until I replicate the failure.

[jaraco]

The third attempt of that build hung, and very early into the test:

setuptools/tests/test_develop.py::TestNamespaces::test_namespace_package_importable Processing /tmp/pytest-of-travis/pytest-0/test_namespace_package_importa0/myns.pkgA

So it seems the first line of output is making it through the subprocess, but then output stops. I suspect there's a bug in subprocess on pypy3, likely something that will get fixed elsewhere. I'm going to simply skip this test on PyPy 3 until it goes away.

[jaraco]

I tried to replicate the failure on my mac, which is running PyPy 5.9.0-beta, and the test passed over and over... so I'm just going to skip the test for now.

[jaraco]

I've moved those troubleshooting commits to this tag.

[pganssle]

I think the current status here is that we still don't know why this is happening. If anyone wants to figure out how it works, re-enable the test skipped in 9edbd7e, try to get it to consistently hang, and try to maybe make a MWE for it.

I do not recommend trying to fix this during a sprint (or if you do, please do it on a personal Appveyor instance), because anything that involves taxing the Appveyor build is going to block a lot of builds.