This issue is to refactor the create-package code in test_runpy into a helper module, as suggested in issue 15358.

This is a prerequisite to moving the pkgutil.walk_package() tests from test_runpy into test_pkgutil.

One trap to watch out for: it's best to call os.path.realpath() on any temporary directories created, since the interpreter tends to do that internally.

I'd previously dealt with this in script_helper.temp_dir, but it came up again when I tightened up the runpy.run_module tests to actually check that __file__ ended up with the expected value.

I'm going to address this issue in parts to make it easier to review and see what is going on.

The first patch I'm uploading shortly here: issue 15403.  That issue I created to address just the "file creation" part of the code.

The approach I'm taking is to keep the file creation code decoupled from the temp directory, sys.path, and import code.  My plan is to address the latter in separate patches by creating or modifying existing test support context managers, as necessary.

I created issue 15415 and uploaded a patch there with this issue in mind.

test_runpy contains supporting _make_pkg() and _del_pkg() methods that,
among other things, create and delete a temp directory.  That logic
would be better handled by a temp directory context manager.

The patch uploaded to issue 15415 creates such a context manager and
also unit-tests it.  It also reduces the duplication of temp
directory logic between test/script_helper.py and test/support.py.

Hi Nick, I was wondering if you'd have time to take a look at the patches I've uploaded to issue 15403 and issue 15415 at some point.

Once those are addressed, I think we will be in a much better position to address most of this issue.

test.support is now a subpackage in both 3.3 and default, but test.support.package_helper still needs to be created (see comments on #15403)

The chain of dependencies goes from here to #15403, #18576 and back here.  How do we break this loop?

@BreamoreBoy, it looks like the next step would be to revisit issue15403 and go through the steps outlined by @ncoghlan.