boto test failures

[Michael Schwartz]

Hi,

I have some changes I want to commit to boto, but I'm seeing the following test failure (which is unrelated to my changes; I see the failure when I pull from the head of the develop branch). Does anyone know about this failure?

Thanks,

Mike

% python test.py -t s3

nose command: test.py -a !notdefault,s3

....E........................

======================================================================

ERROR: test_error_code_populated (tests.integration.s3.test_connection.S3ConnectionTest)

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

Traceback (most recent call last):

  File "/usr/local/google/mfschwartz/boto/tests/integration/s3/test_connection.py", line 245, in test_error_code_populated

    c.create_bucket('bad$bucket$name')

  File "/usr/local/google/mfschwartz/boto/boto/s3/connection.py", line 436, in create_bucket

    data=data)

  File "/usr/local/google/mfschwartz/boto/boto/s3/connection.py", line 472, in make_request

    override_num_retries=override_num_retries)

  File "/usr/local/google/mfschwartz/boto/boto/connection.py", line 913, in make_request

    return self._mexe(http_request, sender, override_num_retries)

  File "/usr/local/google/mfschwartz/boto/boto/connection.py", line 859, in _mexe

    raise e

gaierror: [Errno -5] No address associated with hostname

[Mitchell Garnaat]

Hi Mike -

No, I have not seen that error in my local tests.  I'm pulling the latest code now and will try to reproduce.  It looks like DNS error.

Mitch

[Michael Schwartz]

Hi Mitch,

The test that's failing purposefully picks an invalid DNS name, so I think this problem was caused by a change in the failure handling. Here's an excerpt of the log trace for the failure I saw:

...

Sun, 12 Aug 2012 16:25:07 GMT

/bad$bucket$name/

boto: DEBUG: encountered gaierror exception, reconnecting

boto: DEBUG: establishing HTTPS connection: host=bad$bucket$name.s3.amazonaws.com, kwargs={}

...

I looked at the change log for boto/connection.py and then did a sequence of git resets to various points, to try to isolate when the test failure was introduced. Unfortunately it was difficult to isolate to a specific checkin because the test infrastructure was changing around the time this failure was introduced, and I was unable to get the tests to run at some of the points where I tested. Here's what I found (most current log point first) -- it looks like the test failure was introduced sometime around June 14: 

d1fa49ff6c006b03a8b4a390f9cd9c2b3fb79a24: fails

6a501a42910d75f3aaa52e1e898ee7b495ac3318: tests won't run

9fabfca45294665cc3a3310acf53308f24add4fb: tests won't run

173478682e779c678f3b4edd69f9953fb59944bd succeeds

The change I'd like to commit has nothing to do with any of these changes (my change involves mods to the StorageUri factory in boto/__init__.py). Would you mind if I checked in this change now (i.e., before the test failures above are fixed)?

Thanks,

Mike

[Mitch Garnaat]

Sure, Mike.  Go ahead and commit your changes whenever you are ready.

Mitch

[Michael Schwartz]

Thanks Mitch.

Mike

[Mitchell Garnaat]

I have been unable to reproduce this error.  With the latest code from github:

(boto-dev)3c07545ae5b6:tests garnaat$ python test.py -t s3

nose command: test.py -a !notdefault,s3

.............................

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

Ran 29 tests in 157.360s

OK

I'm not sure how to proceed.

[Michael Schwartz]

Hi Mitch,

Hmm. I ran the tests on another machine outside of Google (at Rackspace), and I'm seeing the same errors. Strange.

Mike