The patch prevents some tests from running multiple times in a test session.

Hm, I was hoping more for patches to the C code that would make these
initializations reentrant (e.g. by testing a flag in C that says "I'm
already initialized").

Here is a version of test_freeze that doesn't depend on stdout output on
load.

Please check in the frozen-related fixes.

Your changes to test_tcl break that test even when run without -R!

For test_pkg, I would rather see a fix that undoes all the changes to
sys.modules (note that it already restores sys.path and deletes the
files it's created, but that's not enough).

How do you like;

class TestPkg(unittest.TestCase):

    def setUp(self):
        self.root = None
        self.syspath = list(sys.path)
        self.sysmodules = sys.modules.copy()

    def tearDown(self):
        sys.path[:] = self.syspath
        sys.modules.clear()
        sys.modules.update(self.sysmodules)
        del self.sysmodules
        cleanout(self.root)

Or I could use the paranoid approach: create set from sys.modules.keys()
in setUp + tearDown and remove the items that are in the second set but
not in the first set.

Fix for test_frozen comitted in r58953

If that makes the test pass with and without -R::, go for it!

On Nov 12, 2007 11:14 AM, Christian Heimes <report@bugs.python.org> wrote:
>
> Christian Heimes added the comment:
>
> How do you like;
>
> class TestPkg(unittest.TestCase):
>
>     def setUp(self):
>         self.root = None
>         self.syspath = list(sys.path)
>         self.sysmodules = sys.modules.copy()
>
>     def tearDown(self):
>         sys.path[:] = self.syspath
>         sys.modules.clear()
>         sys.modules.update(self.sysmodules)
>         del self.sysmodules
>         cleanout(self.root)
>
> Or I could use the paranoid approach: create set from sys.modules.keys()
> in setUp + tearDown and remove the items that are in the second set but
> not in the first set.
>
>
> __________________________________
> Tracker <report@bugs.python.org>
> <http://bugs.python.org/issue1414>
> __________________________________
>

ping?

I don't know how to fix the problem. You have to assign the bug to
somebody who has experience with Tcl/Tk.

Can you open a separate item for the Tkinter leak so you can close this one?

The bug report for test_tcl is at http://bugs.python.org/issue1532