submodule_update() throws error and does not mimick git behaviour

[bilderbuchi]

Hi!
Say you add a submodule to a repo by doing git submodule update --init, and then do git submodule update --init again to update it, git does not complain and just gives the same output as before.
In gitpython, this gives an error:

repo.submodule_update(init=True)
---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
<ipython-input-6-86ea9b78455b> in <module>()
----> 1 repo.submodule_update(init=True)

/usr/lib/python2.7/dist-packages/git/repo/base.pyc in submodule_update(self, *args, **kwargs)
    265                 take the previous state into consideration. For more information, please
    266         see the documentation of RootModule.update"""
--> 267                 return RootModule(self).update(*args, **kwargs)
    268 
    269         #}END submodules

/usr/lib/python2.7/dist-packages/git/objects/submodule/root.pyc in update(self, previous_commit, recursive, force_remove, init, to_latest_revision, progress, dry_run)
    293                         # update the submodule using the default method
    294             sm.update(recursive=False, init=init, to_latest_revision=to_latest_revision, 
--> 295                         progress=progress, dry_run=dry_run)
    296 
    297                         # update recursively depth first - question is which inconsitent

/usr/lib/python2.7/dist-packages/git/objects/submodule/base.pyc in update(self, recursive, init, to_latest_revision, progress, dry_run)
    376                                         os.rmdir(module_path)
    377                                 except OSError:
--> 378                                         raise OSError("Module directory at %r does already exist and is non-empty" % module_path)
    379                                 # END handle OSError
    380                         # END handle directory removal

OSError: Module directory at '/home/me/gitpython-test/somesubmodule' does already exist and is non-empty

Of course it exists, but imo gitpython should not error out about it (just like git does not). same happens when omitting the init=True option.
git.__version__: '0.3.2 RC1'

repo.git.submodule('update') works, though. although it does not display the output that git does, at least it doesn't seem to fail.

[Byron]

git-pythons submodule implementation will need a proper review, as it has many flaws (by now), not properly implementing gits own submodule handling.

[Byron]

The test-suite has been adjusted to test for your case specifically.
Fortunately I was unable to reproduce the issue in the upcoming release, which strengthens my impression that the issue was caused by submodules which used .git files (instead of the actual repository).

[Byron]

The archived live-stream can be found on youtube

[bilderbuchi]

thanks for fixing the bug! (although I have to admit I haven't used gitpython since the time when I reported the bug)

[Byron]

There is a time for everything, after all I didn't look at git-python for several years myself. And look who is back now ;).