since python 2.7.2 
maildir.items() doesn't return anything:

>>> from mailbox import Maildir
>>> x = Maildir('test')
>>> x.items()
[]

untils something was written on the maildir.

>>> x.add('bubu')
'1319447613.M259111P6097Q1.deeppurple'
>>> x.items()
[('1319447013.M913187P6096Q1.deeppurple', <rfc822.Message instance at 0x7f25f5b612d8>), ('1319447613.M259111P6097Q1.deeppurple', <rfc822.Message instance at 0x7f25f5b61680>)]

i verified now, and the 2.7.1 version is ok, the bug is present since the 2.7.2

3.1 is ok
3.2.2 is affected

I'm assigning this to myself so I don't lose it, but if someone wants to work on a patch please do.  It'll be a bit before I can look at it.

Since this is a regression I'm upping the priority further to release blocker, so that we don't put out another release with the regression still in it.

More specifically, this happens if the Maildir instance is created two seconds before items() is called:

>>> import time
>>> from mailbox import Maildir
>>> x = Maildir('test')  # has messages
>>> time.sleep(2.5)
>>> x.items()
[]

This happens because __init__() doesn't populate _toc, and mtimes haven't changed either, so _refresh() doesn't populate _toc either.

Attached a patch. It makes _refresh() re-read the _toc uncoditionally when called the first time.

The patch by Petri looks good.
I've uploaded a variant which set the initial mtimes to epoch instead of creating a _toc_read attribute. The tests are not changed.

Re-uploaded, because we can set initialize _last_read to 0 instead of time.time(). It is more consistent.

Finally, the patch can be only 5 lines...

The last patch looks good to me and is a clever approach indeed :)

New changeset 5f27a9f67a34 by Petri Lehtinen in branch '3.2':
Fix Maildir initialization so that maildir contents are read correctly.
http://hg.python.org/cpython/rev/5f27a9f67a34

New changeset 91a0f6879173 by Petri Lehtinen in branch 'default':
Fix Maildir initialization so that maildir contents are read correctly.
http://hg.python.org/cpython/rev/91a0f6879173

New changeset 0b754ee12dbd by Petri Lehtinen in branch '2.7':
Fix Maildir initialization so that maildir contents  are read correctly.
http://hg.python.org/cpython/rev/0b754ee12dbd

In the absence of any complaints, I went on and committed the fix. Thanks for reporting the issue!

test fails on x86 Windows7 2.7 buildbot
(ok on 3.2 and 3.3)


test_unix_mbox (test.test_mailbox.MaildirTestCase) ... D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\dbapi.py:649: DeprecationWarning: buffer() not supported in 3.x
  b = sqlite.Binary(chr(0) + "'")
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\types.py:70: DeprecationWarning: buffer() not supported in 3.x
  val = buffer("Guglhupf")
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\types.py:234: DeprecationWarning: buffer() not supported in 3.x
  val = buffer("Guglhupf")
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\types.py:350: DeprecationWarning: buffer() not supported in 3.x
  result = self.con.execute('select ? as "x [bin]"', (buffer(zlib.compress(testdata)),)).fetchone()[0]
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\userfunctions.py:39: DeprecationWarning: buffer() not supported in 3.x
  return buffer("blob")
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\userfunctions.py:200: DeprecationWarning: buffer() not supported in 3.x
  self.assertEqual(val, buffer("blob"))
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\userfunctions.py:237: DeprecationWarning: buffer() not supported in 3.x
  cur.execute("select isblob(?)", (buffer("blob"),))
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\userfunctions.py:255: DeprecationWarning: buffer() not supported in 3.x
  ("foo", 5, 3.14, None, buffer("blob"),))
D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\sqlite3\test\userfunctions.py:347: DeprecationWarning: buffer() not supported in 3.x
  cur.execute("select checkType('blob', ?)", (buffer("blob"),))
test test_mailbox failed -- Traceback (most recent call last):
  File "D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\test\test_mailbox.py", line 803, in test_reread
    self.assertFalse(refreshed())
AssertionError: True is not false

ok

======================================================================
FAIL: test_reread (test.test_mailbox.TestMaildir)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "D:\cygwin\home\db3l\buildarea\2.7.bolen-windows7\build\lib\test\test_mailbox.py", line 803, in test_reread
    self.assertFalse(refreshed())
AssertionError: True is not false

----------------------------------------------------------------------
Ran 280 tests in 5.250s

This is a transient failure, other builds are successful.

Maybe the line "time.sleep(2.01 + self._box._skewfactor)" could be changed to "time.sleep(2.5 + self._box._skewfactor)" in test_mailbox.TestMaildir.test_reread.

Closing because it's not related to this issue.

Even better would be to set skewfactor to something negative in this
test, too. I was told that long sleeps in tests are frowned upon.