[Zope-dev] zope-tests - FAILED: 21, OK: 31

[Brian Sutherland]

On Thu, Oct 20, 2011 at 11:58:09AM -0400, Tres Seaver wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> > [1]    FAILED  ZTK 1.0dev / Python2.4.6 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051353.html
> >
> > 
> > 
> > [2]    FAILED  ZTK 1.0dev / Python2.5.5 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051351.html
> >
> > 
> > 
> > [3]    FAILED  ZTK 1.0dev / Python2.6.5 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051352.html
> 
> These
> > 
> are all failing due to a missing version pin for 'six', which
> should not happen:  the ZTK 1.0 package set should not include the new
> version of 'zope.schema' which introduced this dependency.
> 
> 
> > [4]    FAILED  ZTK 1.1 / Python2.5.5 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051398.html
> >
> > 
> > 
> > [5]    FAILED  ZTK 1.1 / Python2.6.5 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051397.html
> >
> > 
> > 
> > [6]    FAILED  ZTK 1.1 / Python2.7.2 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051396.html
> 
> This
> > 
> failure is due to the unexpected presence of a 'Data.fs' file in
> what is supposed to be a newly-created tempdir.  I don't understand
> how that is supposed to work, but the bug seems real.
> 
> 
> > [7]    FAILED  ZTK 1.1dev / Python2.5.5 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051354.html
> 
> Syntax
> > 
> errors due to Py3k compatibility changes in 'zope.schema'.  The
> "ZTK 1.1dev" checkouts pull in the trunk of 'zope.schema', which is
> now backward-incompatible (it drops support for Python 2.5).  I think
> we should be creating a '3.8' branch of 'zope.schema' (copied from
> 3.8.1), and make the "ZTK 1.1dev" checkout use that.

Thanks for fixing these. I didn't realize that the buildouts would break
because of my change. I should have though.

(I've also given up looking at the buildout perma-failures a long time ago)

> Meanwhile, I
> think we should update the 'zope.schema' trunk to indicate that it is
> targeting a new major version, 4.0, due to new dependencies and
> dropping Python 2.5 compatibility.

I've done this.

> > [8]    FAILED  ZTK 1.1dev / Python2.6.5 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051356.html
> 
> There
> > 
> is an odd 'zope.testrunner' failure here:
> 
>  Usage: test [options] [MODULE] [TEST]
> 
>  test: error: option --profile: invalid choice: 'cProfile' (choose from )
> 
> 
>  Failure in test
> /home/ccomb/ztk1.1dev-slave/Python2.6.5-Linux-64bit/build/src/zope.testrunner/src/zope/testrunner/testrunner-subunit.txt
>  Failed doctest test for testrunner-subunit.txt
> 
> And also the "unexpected Data.fs" issue which affected #4, 5, and 6.
> 
> 
> > [9]    FAILED  ZTK 1.1dev / Python2.7.2 Linux 64bit 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051355.html
> 
> This
> > 
> is the "unexpected Data.fs" issue which affected #4, 5, and 6.
> 
> 
> I'm ignoring the 'z3c.*' failures:  I still want these tests either
> fixed or else dropped from the daily report.
> 
> 
> > [21]   FAILED  winbot / ztk_dev py_254_win32 
> > https://mail.zope.org/pipermail/zope-tests/2011-October/051382.html
> 
> This
> > 
> is the same Python 2.5 incompatibility problem which afflicts #7.

This is a tricky issue, there are 2 parts:

    * Supporting python 2.5 in zope.schema is not hard, but will make
      the code quite ugly (no class decorator for implements)
    * I'm not sure if the next version of the ZTK will need to support
      Python 2.5

So there seem to be 3 options to fix this buildout issue:

    1. Point the ZTK trunk at the zope.schema 3.8 branch
    2. Make the zope.schema code uglier and support 2.5
    3. Declare that the next ZTK version drops support for 2.5 and stop
       the buildbot for Python 2.5 + ZTK trunk

I'm partial to option 1 to get the buildbot working and defer the Python
2.5/ugly code decision to later.

-- 
Brian Sutherland