django.contrib.gis.tests.test_measure.{AreaTest,DistanceTest} fail in py3

[idella4]

PYTHONPATH=.:../Django-1.5.1-python3_2/lib/:../Django-1.5.1-python3_2/ python tests/runtests.py --settings=test_sqlite -v1

.......ss.sE........E.........ssssssss..................
======================================================================
ERROR: testAccess (django.contrib.gis.tests.test_measure.AreaTest)
Testing access in different units
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/var/tmp/portage/portage/dev-python/django-1.5.1/work/Django-1.5.1/django/contrib/gis/tests/test_measure.py", line 175, in testAccess
    self.assertAlmostEqual(a.sq_ft, 1076.391, 3)
TypeError: assertAlmostEqual() takes exactly 3 positional arguments (4 given)

======================================================================
ERROR: testAccess (django.contrib.gis.tests.test_measure.DistanceTest)
Testing access in different units
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/var/tmp/portage/portage/dev-python/django-1.5.1/work/Django-1.5.1/django/contrib/gis/tests/test_measure.py", line 43, in testAccess
    self.assertAlmostEqual(d.ft, 328.084, 3)
TypeError: assertAlmostEqual() takes exactly 3 positional arguments (4 given)

----------------------------------------------------------------------
Ran 4963 tests in 855.328s

FAILED (errors=2, skipped=175, expected failures=4)

(a.sq_ft, 1076.391, 3)
(d.ft, 328.084, 3)

Well I count 3 positional arguments, py3 counts 4 given.
tilt.
py3 can't count??

[Baptiste Mispelon]

Hi,

I can't reproduce the issue (using python 3.3 here). I tried running the contrib.gis tests on master, on stable/1.5.x and on the 1.5.1 tag too and they all pass.

I think your setup may be at fault here (this would have been caught by our CI server too). Can you try running the tests in a clean environment? Do you have a strange version of unittest on your path maybe?

I'm marking this as needsinfo for now. Please reopen when you have more info on how to reproduce the issue.

Thanks

PS: The "wrong" argument count comes from the fact that python automatically passes the instance (self) as a first parameter when calling a method.

[Simon Charette]

self is the first positional argument here.

[Claude Paroz]

Exact Python version might help, too.

[Ian Delaney <della5@…>]

it was py3.2.3. Hopefully I just need 3.2.4

[Ian Delaney <della5@…>]

I only run tests in a clean environment. An ebuild deletes the source code on each run.

Is PYTHONPATH=. python3.2 tests/runtests.py --settings=test_sqlite -v1 the best or most apt way to run the test suite?

[Baptiste Mispelon]

Is PYTHONPATH=. python3.2 tests/runtests.py --settings=test_sqlite -v1 the best or most apt way to run the test suite?

That should work, yes. AFAIK, the recommended way is PYTHONPATH=..:$PYTHONPATH python ./runtests.py --settings=test_sqlite (run from the tests directory) [1]

I'll try to get my hands on a 3.2 build to see if I can trigger the issue too.

Have you tried running the full test suite as well?
I checked and assertAlmostEqual is also used in the modeltests tests. Do those fail too?

[1] ​https://docs.djangoproject.com/en/dev/internals/contributing/writing-code/unit-tests/#quickstart

[Ian Delaney <della5@…>]

bmispelon , eeer the full testsuite? I thought I had. What am I missing??

On there-run both py3.2.4 and 3.3.0 failed on them again. So this is from the ebuild simply unpacking the source, cd into it and running the testsuite without any building so if that's not clean I don't know what is. You say that py counts the self so why is the tally then not 4???
I'm running it again from cd into tests and PYTHONPATH=..:$PYTHONPATH python ./runtests.py --settings=test_sqlite which ought be the same.

[Ian Delaney <della5@…>]

I'm guessing I've run the whole suite, it takes about 15 minutes, those couple of fails occur right at the end, so I'm guessing any modeltests are run and pass. If there were anything else fail I'd have entered it

[Claude Paroz]

Are you really using CPython? Using positional arguments to pass parameters to a method with mixed positional/keyword arguments is explicitely admitted (see ​http://docs.python.org/3.2/tutorial/controlflow.html#keyword-arguments). And as you are the first to signal this error, something must be special on your side.

See also assertAlmostEqual definition: ​http://hg.python.org/cpython/file/687295c6c8f2/Lib/unittest/case.py#l527

[Ian Delaney <della5@…>]

hmm well this

======================================================================
FAIL: testAccess (django.contrib.gis.tests.test_measure.AreaTest)
Testing access in different units
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/mnt/gen2/TmpDir/portage/dev-python/django-1.5-r1/work/Django-1.5/django/contrib/gis/tests/test_measure.py", line 175, in testAccess
    self.assertAlmostEqual(a.sq_ft, 1076.391)
AssertionError: 1076.391041670972 != 1076.391 within 7 places

======================================================================
FAIL: testAccess (django.contrib.gis.tests.test_measure.DistanceTest)
Testing access in different units
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/mnt/gen2/TmpDir/portage/dev-python/django-1.5-r1/work/Django-1.5/django/contrib/gis/tests/test_measure.py", line 43, in testAccess
    self.assertAlmostEqual(d.ft, 328.084)
AssertionError: 328.0839895013123 != 328.084 within 7 places

----------------------------------------------------------------------

resulted from taking out the last arg (3) out of the two instances of self.assertAlmostEqual.
The 7 default comes into play. The absence of the , 3 indeed took away the <takes exactly 3 positional arguments (4 given)> error.

But why only me? Am I using CPython?. Well I'm not using pypy and I think that only leaves cpython.
I'm re-running it again with what makes sense to me to get it to pass.

self.assertAlmostEqual(a.sq_ft, 1076.391, places=3)

The self I don't know offhand how to exclude without breaking.
The places= being an optional ought be honoured, making for 2 args and 1 optional, no?
My system is just a run o' the mill gentoo python developer system, with all 6 pythons and 3 pypys emerged.

[Ian Delaney <della5@…>]

---

Ran 4963 tests in 883.139s

OK (skipped=177, expected failures=4)
Destroying test database for alias 'default'...

that would be a yes

[Ian Delaney <della5@…>]

diff -ur Django-1.5.orig/django/contrib/gis/tests/test_measure.py Django-1.5/django/contrib/gis/tests/test_measure.py
--- django/contrib/gis/tests/test_measure.py 2013-02-27 03:04:14.000000000 +0800
+++ django/contrib/gis/tests/test_measure.py 2013-05-28 04:40:18.983011224 +0800
@@ -40,7 +40,7 @@

"Testing access in different units"
d = D(m=100)
self.assertEqual(d.km, 0.1)

• self.assertAlmostEqual(d.ft, 328.084, 3)

+ self.assertAlmostEqual(d.ft, 328.084, places=3)

def testAccessInvalid(self):

"Testing access in invalid units"

@@ -172,7 +172,7 @@

"Testing access in different units"
a = A(sq_m=100)
self.assertEqual(a.sq_km, 0.0001)

• self.assertAlmostEqual(a.sq_ft, 1076.391, 3)

+ self.assertAlmostEqual(a.sq_ft, 1076.391, places=3)

def testAccessInvaliA(self):

"Testing access in invalid units"

this works for me, hope you find it appliccable

[Ian Delaney <della5@…>]

see there

testuser@archtester ~/cvsPortage/gentoo-x86/dev-python/django $ grep AlmostEqual -r /mnt/gen2/TmpDir/portage/dev-python/django-1.5-r1/work/Django-1.5/tests/modeltests/
/mnt/gen2/TmpDir/portage/dev-python/django-1.5-r1/work/Django-1.5/tests/modeltests/aggregation/tests.py:        self.assertAlmostEqual(vals["friends__age__avg"], 34.07, places=2)
/mnt/gen2/TmpDir/portage/dev-python/django-1.5-r1/work/Django-1.5/tests/modeltests/aggregation/tests.py:        self.assertAlmostEqual(vals["authors__age__avg"], 38.2857, places=2)
/mnt/gen2/TmpDir/portage/dev-python/django-1.5-r1/work/Django-1.5/tests/modeltests/aggregation/tests.py:        self.assertAlmostEqual(vals["amazon_mean"], 4.08, places=2)

, places= is used in all 3 instances, and they all passed for me

[Ian Delaney <della5@…>]

I'm totally miffed by why yours passes under py3 and mine doesn't. There must be something 'special on my side' logically. Any guesses? I got nothing.
The adding of places= simply makes the code consistent between tests. I won't re-open this since that really needs pinning that special quality, however the patch won't break anything and will make for consistency. I've added the patch to the 1.5 versions in portage.

[Claude Paroz]

I think I found the bug: ​http://bugs.python.org/issue10353

Now the question is for you why you are still affected by this bug fixed some time ago. Either some older Python installation is left somewhere on your system, or a very recent Python commit reinstated this bug. In any case, this is not Django's fault, as self.assertAlmostEqual(a.sq_ft, 1076.391, places=3) or self.assertAlmostEqual(a.sq_ft, 1076.391, 3) are strictly equivalent.

[Ian Delaney <della5@…>]

right. well my python installs are recent and current, sp I'd go for something re-instating it. I'll see if some other python devs have some idea. thx