#56 close() should be handled better inside Connection.py

[Christian Reis]

If storage.close() or db.close() have been issued and a
commit() or connection.close() is called, evil errors
pop up.

See
http://mail.zope.org/pipermail/zodb-dev/2003-May/004983.html
for details.

Attached patch is a first cut at making this clearer.

[Christian Reis]

Bulletproof access to structures in Connection when *.close() has been called.

[Christian Reis]

testClose.py: test bulletproofing (run against unpatched ZODB for extra fun)

[Jeremy Hylton]

Logged In: YES
user_id=31392

Is this still an issue?

[Christian Reis]

Logged In: YES
user_id=222305

Let me check again, but I think it still is (with 3.2-CVS HEAD).

[Christian Reis]

Logged In: YES
user_id=222305

Yeah, it still is:

Traceback (most recent call last):
File "testClose.py", line 51, in ?
test()
File "testClose.py", line 29, in test
x.root()["foo"] = 100
File
"/mondo/local/lib/python2.3/site-packages/ZODB/Connection.py",
line 542, in root
return self['\0\0\0\0\0\0\0\0']
File
"/mondo/local/lib/python2.3/site-packages/ZODB/Connection.py",
line 150, in __getitem__
p, serial = self._storage.load(oid, self._version)
File
"/mondo/local/lib/python2.3/site-packages/ZODB/FileStorage.py",
line 687, in load
return self._load(oid, version, self._index, self._file)
File
"/mondo/local/lib/python2.3/site-packages/ZODB/FileStorage.py",
line 658, in _load
file.seek(pos)
ValueError: I/O operation on closed file

This happens when trying to write something to
conn.root()['xxx']. If you look at the testcase (just
comment out the ClosedError bits, which require the patch)
you'll see what I mean.

I can help out here, but I'd like guidance as to what to do.