many fixes of little issues found by the cppcheck tool. In particular, a few memory leak were discovered.

Improved comments surrounding the RTCOPYD/REUSE_MOUNT configuration parameter.

am therefore removing this directory from svn.

The tapeNeedsToBeReleasedAtEndOfSession variable of the rtcpd_MainCntl()
function needs to be a static because its address is passed from the main
thread to the tape IO thread.  Making the variable static avoids the
possibility of memory corruption if the rtcpd_Main() function ends
prematurely before the main thread calling it has joined with the tape IO
thread.

I have modified the log level of the following vdqmd daemon message in the trunk:

2012-08-13T19:37:48.108835+02:00 lxc2dev4 vdqmd[11998]: LVL=Warn TID=12009 MSG="client has requested a forced unmount." REQID=6b015ba7-6d26-403d-aa3c-8fdd36ce7cc5

I have changed the log level from DLF_LVL_WARNING to DLF_LVL_SYSTEM.  This is because the default rtcpd daemon behaviour after this RFE is to not try an reuse mounts across stagers.

Implemented in the trunk.

Fixed bug #96453: stager_addprivilege gets Oracle error on 'long' lists of users : was a stupid mistake of using the wrong vector length

Fixed in trunk.

Fixed bug #96646: another "listtransfers -p -x" error case: dictionary update sequence element #0 has length 18; 2 is required

Fixed linking directive for pthread : -pthread is replaced by -lpthread. This is equivalent under linux and fixes the compilation for clang

The BridgeProtocolEngine::rtcpTapeErrReqRtcpdCallback() method currently
aborts the tape session if it detects an error by throwing an exception.
This is too abrupt and should be changed to trying to gracefully shutdown
the rtcpd session.

This commit introdcues the graceful shutdown of the rtcpd session in the
event of an error detected by the
BridgeProtocolEngine::rtcpTapeErrReqRtcpdCallback() method.

The rtcp_TransAckn() function needed to reset serrno to 0 before
calling netread_timeout() because if netread_timeout() returned
-1 then whatever was in serrno would be reported as the error
and errno would be ignored.

The castor::tape::tapebridge::BridgeProtocolEngine::rtcpTapeErrReqRtcpdCallback()
method currently hinders the shutdown of rtcpd in error conditions by
acknowledging an rtcpd RTCP_TAPEERR_REQ with the acknowledge of an
RTCP_TAPE_REQ message.  Rtcpd detects a protocol error which confuses the
shutdown sequence.

This commit correctly the acknowledge message send by the tapebridged daemon.

Added new tape errors to serrno manual page.

In order to improve logging for tape-operations and to facilitate the current
testing surrounding this bug, the serrno code of ENOMEM has been replaced by
the more specific error ETNOLBLINFOMEM meaning "No memory available for label
information" within the setlabelinfo function.

[sr #130991] 2.1.13-0 upgrade of NS db has effected enterfileclass, fixFileSize and checkreplicas.py
Relaxed schema version check for Nameserver and VMGR.

Fixed bug #96453: stager_addprivilege gets Oracle error on 'long' lists of users by addeing not null constraint on classid column of FileClass table

Fix which requires testing.

Added the unit-test class Ctape_reserveTest correctly this time (ismple forgot
to add it last time).

Ctape_reserve function now passes the unit-tests by return -1 and setting serrno
to EINVAL when more than one drive is requested within a drive reservation
request.

The ability to reserve drives from many DGNS was blocked by the previous
commit modifying the initlabelroutines() (SVN revision 24307).  This commit
adds the following two unit tests:

 Ctape_reserveTest::testCtape_reserveWithCountOf2
 Ctape_reserveTest::testCtape_reserveWith1DgnReservationWithDriveCountOf2

These two tests currently fail (test driven development). The first is a repeat
of trying to reserve drives from more than one DGN, though this time via a
different function Ctape_reservei().  The second and more important test is for
the case of using Ctape_reserve to try and reserve many drives from a single DGN.
In order to pass the unit-tests the Ctape_reserve() function needs to be
modified to return -1 and set serrno to EINVAL if either a DGN or a drive count
is anything other than 1.

The unit test now passes so any rtcpd installation built now will enforce single
dgn reservations.  In theory a program can still ask for many drives from the
same dgn.  To be continued..

rmlabelinfo() assumes we support multi-drive reservations and contains buggy
memory managment to deal with these cases.  Such reservations are no longer used
therefore the buggy code can be removed.

This commit adds a unit test which fails (test driven development) because the
future implementation of initlabelroutines() should only accept 1 as a parameter.
initlabelroutines() is the creator of the C structures that rmlabelinfo() tries
to remove.