On Sun, Sep 26, 2010 at 8:42 PM, Tres Seaver <tseaver@palladion.com> wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

On 09/26/2010 11:52 AM, Marius Gedminas wrote:

...

On Sun, Sep 26, 2010 at 11:05:03AM -0400, Tres Seaver wrote:

...

On 09/26/2010 07:58 AM, Zope Tests Summarizer wrote:

...

Subject: FAILED : Zope Buildbot / ztk slave-ubuntu64 From: jdriessen at thehealthagency.com Date: Sat Sep 25 09:57:03 EDT 2010 URL: http://mail.zope.org/pipermail/zope-tests/2010-September/020550.html

This failure is due to 'test-ztk' timing out.  I'm willing to bet that the 'zope.testrunner' doctests which attempt to test use of pdb are to blame:  they cause the tests to hang on my machine, anyway.

I can reproduce that in r116884, but it appears to be fixed in the latest revision (116919), which bumped zope.testing to 3.10.0.

...

I would favor commenting out those tests, or even removing them altogether.  I don't think the "feature" is worth any price at all in terms of the overally ecosystem.

Being able to use pdb to debug doctests is occasionally valuable.

I don't mind that such debugging is possible: I do mind very much having a meta-doctest which tries to demonstrate that feature cause runs of the test suite to hang.

Post-mortem debugging is a wildly useful feature to me. If it became unstable in the test runner, I'd be tempted to fork it. When those tests were written, they didn't cause hangs or intermittent test failures. If they are now, then I suggest that someone broke something. Jim -- Jim Fulton