[Zope-dev] Post-mortem debugging support in the test runner (Re: Zope Tests: 55 OK, 16 Failed, 3 Unknown)

[Jim Fulton]

On Sun, Sep 26, 2010 at 8:42 PM, Tres Seaver <tseaver at 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