Bug 1102926

Created attachment 900510 [details]
Proposed patch

When testing OpenStack Swift, usually the following happens right away
after a  git clone and  python setup.py build:

[zaitcev@guren swift-tip]$ PYTHONPATH=$(pwd) ./.unittests
............................
ERROR: test_run_server_conf_dir (test.unit.common.test_wsgi.TestWSGI)
Traceback (most recent call last):
  File "/q/zaitcev/hail/swift-tip/test/unit/common/test_wsgi.py", line 399, in test_run_server_conf_dir
    conf = wsgi.appconfig(conf_dir)
[......traceback.......]
  File "/usr/lib/python2.7/site-packages/pkg_resources.py", line 576, in resolve
    raise DistributionNotFound(req)
DistributionNotFound: xattr>=0.4

This used to happen when building in Koji, too, but we worked around it
by switching the whole dependency checking off by "rm requirements.txt".
So, it's just the development trees, where it's easy to work around
by editing requirements.txt.

However, it's somewhat unobvious and it requires to carry a modded
requirements.txt in your work tree. Because of that I submitted patches
with "-xattr>=0.4" to Gerrit a few times.

The root cause is the application wants a package called "xattr" and
we provide one called "pyxattr". Thus, "xattr" is not found (any version).
Note that our xattr code itself is fine (or almost, see bug 732692),
the problem is in Python dependencies only.

An easy workaround is this:

[root@guren pyxattr.rawhide]# ln -s ./pyxattr-0.5.1-py2.7.egg-info  /usr/lib64/python2.7/site-packages/xattr-0.5.1-py2.7.egg-info

But, doing that in an RPM spec is a little challenging and causes
some weirdest effects. For example, a bogus complaint about file
conflicts from  rpm -U.

So, I propose we carry a downstream patch, which adds a fake egg
with the name "xattr". Please see attached.

In case anyone ever wants to package python-xattr, it's not going
to conflict either as Python package or Fedora package, because
file paths are different and version is different.