BF: remove a submodule with a remote without refs + misc fixes around

[yarikoptic]

now if there is a remote without any refs (this happens e.g. with git-annex repo with special remotes), it would refuse to remove the submodule
(with 2.0.8 it also didn't work but with a different msg "AssertionError: Remote datalad did not have any references")

[ankostis]

@yarikoptic would it be possible to add a test-case into the PR with the annex command, replicating the situation you just described?

[yarikoptic]

ok @ankostis -- but along the way had to fix 2 other bugs... apparently some tests (which used @with_rw_directory) were not run at all! so you can expect failures of those tests which were effected

[yarikoptic]

and I am not sure why that is the case -- I was expecting submodules repo to be the repo of the submodule ... but I guess it can make sense to refer to parent which contains the submodule, but then is there an attribute for submodules repo (if present)?

[ankostis]

@Byron any comment? I'm not familiar with git-modules.

[yarikoptic]

reincarnating @with_rw_directory lead to one failing test

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

FAIL: test_lock_reentry (git.test.test_config.TestBase)

----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/build/gitpython-developers/GitPython/git/test/lib/helper.py", line 95, in wrapper
    return func(self, path)
  File "/home/travis/build/gitpython-developers/GitPython/git/test/test_config.py", line 100, in test_lock_reentry
    self.failUnlessRaises(IOError, GitConfigParser, fpl, read_only=False)
AssertionError: IOError not raised
-------------------- >> begin captured logging << --------------------

git.util: INFO: Test TestBase.test_lock_reentry failed, output is at '/tmp/test_lock_reentryfSovNi'

I am not sure on state of locking/context manager expectations atm and see recent work touching locking... so I will need to leave this to you guys (@ankostis ?) to figure out

[ankostis]

@yarikoptic so you mean that git.test.test_config:TestBase.test_lock_reentry() that is failing is included in the CI for the first time?

[yarikoptic]

Yes, at least since some time in the past

[yarikoptic]

If your really want to know, I could check later since when

[ankostis]

Indeed @yarikoptic your 31fd955 fixed an important regression introduced by #519,
where I had practically disabled 19 TCs!
(so the ALL GREEN was an illusion :-/, check #519 for the actual list of TC errors on Windows)

So, I'm cherry picking your 2528d11 and 31fd955 but modified to include only your helper.py fixes and pushing them into my own repo, to try to fix the failing TC; for the test_submodles.py changes, they have to wait, and I will collect them and re-synthesize a new branch to merge.

[ankostis]

On Travis, PY3.3 failed due to file-handles; increased ulimit 96-->100 solved the problem.
On Appveyor, this is a drama...

[ankostis]

Forgot to thank you.
@yarikoptic If you may, add yourself on the Authors, on this branch, and I will cherry-pick it, just add a comment here to notify me.

[yarikoptic]

That is OK, if you like to see me there - add, otherwise I am just fine. Cheers. Thanks for attending to my PRs quickly