Test failures on OSX due to incorrect unicode normalization

[Ryan J Ollos]

Issue seems to be explained in ​SO:9758014/121694. The test failures are:

======================================================================
FAIL: test_unicode_branches (tracopt.versioncontrol.git.tests.PyGIT.UnicodeNameTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/rjollos/Documents/Workspace/trac-dev/teo-rjollos.git/tracopt/versioncontrol/git/tests/PyGIT.py", line 347, in test_unicode_branches
    self.assertEqual(u'tickét10980', branches[1][0])
AssertionError: u'tick\xe9t10980' != u'ticke\u0301t10980'
- tick\xe9t10980
?     ^
+ ticke\u0301t10980
?     ^^


======================================================================
FAIL: test_unicode_tags (tracopt.versioncontrol.git.tests.PyGIT.UnicodeNameTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/rjollos/Documents/Workspace/trac-dev/teo-rjollos.git/tracopt/versioncontrol/git/tests/PyGIT.py", line 364, in test_unicode_tags
    self.assertEqual([u'täg-t10980', 'v0.42.1'], tags)
AssertionError: Lists differ: [u't\xe4g-t10980', 'v0.42.1'] != [u'ta\u0308g-t10980', u'v0.42....

First differing element 0:
t\xe4g-t10980
ta\u0308g-t10980

- [u't\xe4g-t10980', 'v0.42.1']
?      ^^^

+ [u'ta\u0308g-t10980', u'v0.42.1']
?     + ^^^^^           +

[Jun Omae]

The following patch might be a work around for that and requires git 1.7.10+. Otherwise, we could change to unicode characters which are not decomposed on Mac OS X.

tracopt/versioncontrol/git/tests/PyGIT.py

@@ -299 +299 @@
         self.repos_path = tempfile.mkdtemp(prefix='trac-gitrepos-')
         # create git repository and master branch
         self._git('init')
+        self._git('config', 'core.precomposeunicode', 'true')  # ticket:12020
         self._git('config', 'core.quotepath', 'true')  # ticket:11198
         self._git('config', 'user.name', "Joé")  # passing utf-8 bytes
         self._git('config', 'user.email', "joe@example.com")

[Ryan J Ollos]

Replying to jomae:

The following patch might be a work around for that and requires git 1.7.10+.

I'm using Git 2.3.5 and precomposeunicode defaults to true. However, it appears the setting applies only to file names, not to branch names.

Otherwise, we could change to unicode characters which are not decomposed on Mac OS X.

That's a good idea. Committed to 1.0-stable in [13961], merged to trunk in [13962].