jcifs-1.2.4 from tgz
Wed Sep 21 01:53:28 EDT 2005
jcifs-1.2.4 released / Timeout Transport Exception, Bogus Signature Error,
and More
A NetBIOS keep-alive message (received after ~10 minutes) would break
message processesing with a timedout waiting for response Exception. This
has been fixed.
JCIFS would fail to validate responses with a status that is not zero.
Assuming we are calculating the verfication signature correctly I can only
assume the affected servers choose not to generate the correct signature
for error responses (perhaps for DOS reasons). Because JCIFS checked the
signature before the message status, an error response would fail with
"signature verification failure". This behavior has been changed so that
signatures are not verified if the status is non zero.
It was discovered that the new transport (as of 1.2.x) could not cleanly
recover from temporary server failure (e.g. a restart). This has been
fixed. Methods will still throw Exceptions but moment the server comes back
online the client gracefully recover.
Wed Aug 24 13:29:44 EDT 2005
jcifs-1.2.3 released / Port 445 Fixed
A mistake in the 1.2.2 release broke port 445 communication entirely. It
has been fixed. The exact error (with a sufficiently high loglevel) was
"Invalid payload size: 1".
Sat Aug 20 00:26:11 EDT 2005
jcifs-1.2.2 released / Exception "cannot assign requested address" Fixed,
Clusters, NetApp Filer, and More
There have been a number of small fixes. These are:
o The "cannot assign requested address" exception caused by trying to bind
the local address 127.0.0.1 has been fixed.
o In a cluster environment the NTLM HTTP Filter could fail with "account
currently disabled" or "Access denied" errors due to a deserialization
issue of "preauthentication" credentials stored in the HttpSession. The
initialization of default credentials has been changed however it is not
clear that the change will have any effect as I do not have a clustered
environment in which to test.
o The combination of plain text passwords and Unicode (largely specific to
Samba 3) has been fixed.
o A bogus debugging statement has been discovered and removed. Who left that
in there?!
o A Socket.shutdownOutput() call has been added to doDisconnect as it is
believed to reduce spurrious RST frames observed when abruptly shutting
down transports. These are believed to be harmless but they have been
associated with unsightly messages in Samba log files.
o The copyTo() method will now check to see if the source path is a child,
parent or equal to the destination path and if so throw a Source and
destination paths overlap exception.
o An additional debugging statement has been added to the NTLM HTTP Filter
domain controller interrogation code.
o The getDiskFreeSpace call could fail with NetApp Filer. It has been
repaired.
15 files changed: