Tests cannot find timezone file which is present, then fail (should skip)

[koobs]

The following tests fail when testing with source obtained from GitHub. The timezone file is present, but the tests do not find/see them.

Beyond identifying the root cause of this issue, th tests should skip, not fail if the tarball cannot be found.

1. Include the timezone file as a data file for the tests to use (Edit: timezone file is present)

======================================================================
FAIL: testZoneInfoFileEnd1 (dateutil.test.test_tz.TZTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/home/koobs/repos/freebsd/ports/devel/py-dateutil/work/dateutil-2.5.0/dateutil/test/test_tz.py", line 310, in testZoneInfoFileEnd1
    "EDT", MISSING_TARBALL)
AssertionError: None != u'EDT' : This test fails if you don't have the dateutil timezone file installed. Please read the README

======================================================================
FAIL: testZoneInfoFileStart1 (dateutil.test.test_tz.TZTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/home/koobs/repos/freebsd/ports/devel/py-dateutil/work/dateutil-2.5.0/dateutil/test/test_tz.py", line 304, in testZoneInfoFileStart1
    MISSING_TARBALL)
AssertionError: None != u'EST' : This test fails if you don't have the dateutil timezone file installed. Please read the README

======================================================================
FAIL: testZoneInfoOffsetSignal (dateutil.test.test_tz.TZTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/home/koobs/repos/freebsd/ports/devel/py-dateutil/work/dateutil-2.5.0/dateutil/test/test_tz.py", line 317, in testZoneInfoOffsetSignal
    self.assertNotEqual(utc, None, MISSING_TARBALL)
AssertionError: None == None : This test fails if you don't have the dateutil timezone file installed. Please read the README

[pganssle]

@koobs How exactly did you download the source here? Are you using the latest version from master? Looks to me like the zoneinfo tarball is included in the github source.

[koobs]

@pganssle Latest release tag (I was updating the dateutil FreeBSD Port) Was the tarball only recently added, ie; after the last release was tagged?

[koobs]

I can confirm that the dateutil-zoneinfo.tar.gz is present in my repo checkout of 2.5.0 (via tag) and the PyPI 2.5.0 sdist.

Somehow the tests aren't finding them

[pganssle]

@koobs I think the tests are not included in the PyPI sdist anyway (see #193).

I'm unable to duplicate the issue on Ubuntu 15.04. I ran:

git clone git@github.com:dateutil/dateutil.git dateutil_clean
cd dateutil_clean
git checkout 2.5.0
python setup.py test

All tests passed, including the ones requiring the tarball, so I'm not quite sure what's going on in your system.

As for the suggestion that we skip the tests if the tarball is missing, I don't think I want to do that now that the tarball is included in the repository in the same way that I wouldn't want to skip all the "easter" related tests if for some reason easter.py wasn't in the source - the tarball is now an actual part of the library and something is wrong if it's missing.

[koobs]

@pganssle I had in my original description mentioned that the tests aren't 'yet' in the sdist, being aware of #193 :) I shouldn't have edited it out.

Skipping the test was suggested because it relies on a separate resource in the first place to execute. A skip if a resource is not available is correct, as the failure is elsewhere. I believe most test runners also allow developers to state 'consider skipped tests failures'. That would be the correct way to ensure 'nothing is ever missed', whether by future accidental omission of the timezone file, or otherwise.

If you don't want to skip the tests given the tarball is shipped with the library, then I suppose we need to isolate the root cause of the failure :)

Where do we start?

[koobs]

Tests are being run via setup.py test

[pganssle]

Yeah, I'm just saying that the resource not being available should now itself be considered a failure condition. There are other skips in the tests that are not failures, because they are platform-specific.

Can you check if it works if you run python updatezinfo.py before running the tests?

[koobs]

It works after running that command:

py-dateutil/work/dateutil-2.5.0] python2.7 updatezinfo.py
Downloading tz file from mirror 0
Updating timezone information...
"/tmp/tmpQUnGnq/factory", line 9: warning: time zone abbreviation differs from POSIX standard (Local time zone must be set--see zic manual page)
Done.

Ran 807 tests in 23.417s OK (skipped=16)

[pganssle]

@koobs Hm. And a fresh clone of dateutil 2.5.0 doesn't work? Maybe this is a python2.7 issue? I can try it with python 2.7 a bit later.

[koobs]

@pganssle Using the plain /tarball/ GitHub API for fetching tagged (2.5.0) sources, so they're clean. I'll test the tests with 3.3, 3.4 and 3.5 and get back to you

[pganssle]

@koobs Fresh clone works fine for me with python2.7. Can you clarify what exactly you mean by the "/tarball/ Github API? Exactly what URL are you pulling this tarball from?

[pganssle]

@koobs Did you ever figure out what was causing this issue? Ideally I'd like to resolve it before 2.5.1 is released, but I'm still unable to replicate.

[pganssle]

@koobs Let me know if the issue is resolved in 2.5.1.

[koobs]

@pganssle Apologies for the delay. I'll check/confirm 2.5.1 then update here

[koobs]

I can no longer reproduce these test failures this with 2.8.0, closing.