jcifs-0.7.15 from tgz

jcifs-0.7.15 from tgz

Thu Oct 23 01:18:29 EDT 2003

jcifs-0.7.15 released

The name service code has been modified to return a different response with
each  0x1C  domain controller query. The range of servers is limited to the
top  jcifs.netbios.lookupRespLimit  entries  in  the  NetBIOS response. The
effect  that this will have is that the NtlmHttpFilter and NtlmServlet will
rotate  through  domain  controllers when authenticating WWW clients. Also,
because  different  transports are used, fewer sessions will be multiplexed
over  each  resulting  in  an  increase  in scalability proportional to the
number  of  domain  controllers  used.  This  behavior can be turned off by
setting the jcifs.http.loadBalance property to false (the default is true).

The  jcifs.smb.client.ssnLimit  property  default value has been changed to
250 from 100.

Mon Oct  6 23:53:28 EDT 2003

jcifs-0.7.14 released

Eric's  LMv2  patch  has been merged. There is some uncertainty surrounding
the signing of SMBs if GUEST credentials are negotiated. On a related note;
the  "Concurrent  modification" issue has been resolved. The oversight with
this  one  was  that  it  was  believed  that only a thead calling into the
transport  should  ever  trigger  tryClose() which is not exactly true. The
transport  thread  can  call tryClose but prior to adding verify() calls it
was  rare  and  only  happend  when  the transport reached a state where no
further  operations  could  move  forward anyway. So rather than syncronize
tryClose,  I  have  changed  verify  to simply return a boolean value which
set's the new ServerMessageBlock.verifyFailed member. This is then examined
by  the  calling  thread  who will tryClose (with the transport locked) and
throw  the "Unverifyable signature" if verifyFailed. Finally there has been
a  straight  forward  but potentially dangerous change in SmbFile. All list
operations  now  build  a  list  of names or files using an ArrayList. This
assisted  in  fixing  another  problem  where NetServerEnum2 responses were
observed    to   return   server   names   of   0   length.   This   caused
ArrayIndexOutOfBounds exceptions. These entries are now ignored.