With r79207 (bpo-8180) applied to trunk, seeing this failure (10.6.2, HFS+ case-sensitive file system):

======================================================================
ERROR: test_normalize (test.test_pep277.UnicodeFileTests)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/private/tmp/pp/usr/local/lib/python2.7/test/test_pep277.py", line 119, in test_normalize
    os.stat(name)
OSError: [Errno 2] No such file or directory: '@test_24797_tmp/\xe2\x80\x82\xe2\x80\x82\xe2\x80\x82A'

Ran 30 tests in 0.237s

FAILED (errors=1)
test test_pep277 failed -- Traceback (most recent call last):
  File "/private/tmp/pp/usr/local/lib/python2.7/test/test_pep277.py", line 119, in test_normalize
    os.stat(name)
OSError: [Errno 2] No such file or directory: '@test_24797_tmp/\xe2\x80\x82\xe2\x80\x82\xe2\x80\x82A'

Also this failure on py3k:

======================================================================
ERROR: test_normalize (test.test_pep277.UnicodeFileTests)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/private/tmp/pp3/usr/local/lib/python3.2/test/test_pep277.py", line 119, in test_normalize
    os.stat(name)
OSError: [Errno 2] No such file or directory: '@test_42408_tmp/\u2002\u2002\u2002A'

This patch should fix it...
"HFS Plus uses a variant of Normal Form D in which U+2000 through U+2FFF, U+F900 through U+FAFF, and U+2F800 through U+2FAFF are not decomposed."

<rant>
I believed there was only one Unicode...
But obviously the Apple's Unicode is something different.
</rant>

Florent Xicluna wrote:

Florent Xicluna <florent.xicluna@gmail.com> added the comment:

This patch should fix it...
"HFS Plus uses a variant of Normal Form D in which U+2000 through U+2FFF, U+F900 through U+FAFF, and U+2F800 through U+2FAFF are not decomposed."

Could you provide a reference link for this quote ?

It's rather strange that those ranges are not decomposed,
since they do contain combining code points.

Actually, the file system in question is what Apple calls a HFSX case-sensitive (see http://developer.apple.com/mac/library/technotes/tn/tn1150.html#HFSX). On a typical OS X system, you could encounter any combination of HFS+ case-insensitive, HFSX case-insensitive, HFSX case-sensitive file systems, along with other usual suspects, like NFS or SMB. I'll note again the warning in r33595.

Could you provide a reference link for this quote ?

I put the link in the patch:
http://developer.apple.com/mac/library/qa/qa2001/qa1173.html

With the patch for trunk, the test no longer fails on the given file system.

Fixed on trunk with r79426 and on 3.x with r79474