Fix parallel testing - #6415

Hi there @astrofrog 👋 - thanks for the pull request! I'm just a friendly 🤖 that checks for issues related to the changelog and making sure that this pull request is milestoned and labelled correctly. This is mainly intended for the maintainers, so if you are not a maintainer you can ignore this, and a maintainer will let you know if any action is required on your part 😃.

Everything looks good from my point of view! 👍

@pllim - why does this have a windows label?

@astrofrog says "CircleCI" so that was why, but feel free to remove tag if that was wrong.

circleCI runs a 32bit linux box :)

Important note: enabling parallel containers in CircleCI is not useful, as it runs 4 separate containers, making it hard to test this. So for now I've disabled it (but we can still test the parallel option, there just won't be any speedup)

Oh... I thought it's 32-bit Windows, not sure why! My bad. Thanks for fixing the labels, @bsipocz !

Important note: enabling parallel containers in CircleCI is not useful, as it runs 4 separate containers, making it hard to test this. So for now I've disabled it (but we can still test the parallel option, there just won't be any speedup)

What about having then only 2 parallel threads, also maybe just for one module? If there is no speed up, then running 4 doesn't make any sense.

What about having then only 2 parallel threads, also maybe just for one module? If there is no speed up, then running 4 doesn't make any sense

There is no harm in running on 4 though, the runtime is not any worse. Since we can't do a build matrix on CircleCI, I think just running the full test suite with 4 threads is fine?

@eteq - just for info, this fixes a minor 'bug' with Python 2.7 in coordinates. It turns out that:

python setup.py test -t docs/time/index.rst

actually raises an exception because if run on its own, _site_registry from astropy.coordinates is not yet defined, and it tries to get downloaded. This doesn't happen in Python 3.6 because the remote_data exception is a URLError there (but it's an IOError in Python 2). Anyway, details, but bottom line is that the above test command now works with Python 2 with this PR.

This came out here because one of the threads encountered this as its first test, behaving as if i'd run the tests for that file alone.

Strictly speaking we should probably remove _site_registry after each test.

@bsipocz @eteq - do you fancy reviewing this?

Looks good to me!

Since this still has some issues on travis I went ahead and merged #6420 that contained only the version limitation for circleCI. Now all the other PRs can be rebased.