Still getting a test error in 0.9.7.x branch

[Kyle VanderBeek]

I'm trying to get the 0.9.7 final version rolled into an RPM for
Fedora, and keep running into a single test failure on
"test_project_do_two_engines" both from the 0.9.7 tarball and the tip
of the 0.9.7.x hg branch. It's a complaint from the caching layer,
but I'm not convinced this is the real culprit:

[...]
File "/usr/lib/python2.6/site-packages/beaker/container.py", line
383, in __init__
raise MissingCacheParameter("data_dir or dbm_dir is required")
MissingCacheParameter: data_dir or dbm_dir is required

I think it may be an artifact of the way many of us have to run the
test suite by setting PYTHONPATH instead of running "python setup.py
develop". This seems like it should work, and would be ideal for
anyone wanting to help with Pylons development running on a box they
don't have root access on or have to share with other people (making
modifying the python installation's path via "setup.py develop"
untenable):

PYTHONPATH=`pwd` python setup.py test

If anyone has any clues as to why this test fails while running tests
in this manner, I'd really appreciate a hint. I'm digging, but
haven't found it yet, and I'd really like to get the final version of
0.9.7 into Fedora so people can use it easily. Output of an attempted
run is attached. Thanks!

--
Kyle.
www.kylev.com

[Kyle VanderBeek]

On Jul 18, 11:44 am, Kyle VanderBeek <vanderb...@gmail.com> wrote:
> I'm trying to get the 0.9.7 final version rolled into an RPM for
> Fedora, and keep running into a single test failure on
> "test_project_do_two_engines" both from the 0.9.7 tarball and the tip
> of the 0.9.7.x hg branch.  It's a complaint from the caching layer,
> but I'm not convinced this is the real culprit:
>
> [...]
>   File "/usr/lib/python2.6/site-packages/beaker/container.py", line
> 383, in __init__
>     raise MissingCacheParameter("data_dir or dbm_dir is required")
> MissingCacheParameter: data_dir or dbm_dir is required
>
>

Actually, it was the Beaker that was the problem. The current stable
release of Beaker 1.3.1 has a bug that is fixed in the trunk. It
causes elements of the config to get overwritten. Fix here:

http://bitbucket.org/bbangert/beaker/changeset/24ef019fc6d6/

So, caution: as far as I can tell, you may end up with broken things
if you run Pylons with Beaker 1.3.1.

--
Kyle.
www.kylev.com