home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Network Supervisor's Toolkit
/
Network Supervisor's Toolkit.iso
/
patches
/
_312ptch
/
libup4
/
details.doc
next >
Wrap
Text File
|
1996-07-10
|
5KB
|
101 lines
-----------
3.12g CLIB
-----------
349. Added "LONG *fileHandle;" to end of open/create call back
structures. This applies to file open, create file, and open/create
function calls.
350. If a pre-hook call back func. cancels a call (errors out) the following call
back funcs. in the chain will not be given their shot at it-now compatible
with the fshooks in 4.x. Also made it so the
post-hook call back functions cannot change the OS return code-now compat-
ible with fshooks in 4.x.In hook312.386 changed the check to see if the'
dynamic import of the "GenNS" call succeeded, we were comparing incorrectly.
351. Changed the way we protect the OS lock manager calls from
being called with the same connection number more than once at the same
time. The old way was broken for NCP Extension handlers and legitimate
users of Disable/EnableConnection(). We still protect, but in a
different way. Also the OS's callback had no context so return values
from the locking calls were sometimes lost. Fairly extensive changes
were made to correct that problem. If NCP extensions are used and any
of the locking calls in conjunction, the locking calls will fail, or
disable and enable connection calls in conjunction with lock calls, the
locking calls will fail. Both failures are EFAILURE.
352. Added a check to ExitCritSec to make sure the thread
was part of the thread group list. This might be a problem in 3.x
because the 3.x SetThreadGroupID does not link the thread into the
thread list. The SPAGENT.NLM exhibited GPIs during loading/whenever it
ended a critical section.
-----------
4.01e CLIB
-----------
877. Added code to disallow unloading until a NLM is running through it's
main routine and changed the SIGTERM context establishment to not
happen on a synchronized NLM because two threads changed the console
command thread's mappings.
879. Changed the way we protect the OS lock manager calls from being called
with the same connection number more than once at the same time. The
old way was broken for NCP Extension handlers and legitimate users of
Disable/EnableConnection(). We still protect, but in a different way.
Also the OS's callback had no context so return values from the locking
calls were sometimes lost. Fairly extensive changes were made to correct
that problem. If NCP extensions are used and any of the locking calls
in conjunction, the locking calls will fail, or disable and enable
connection calls in conjunction with lock calls, the locking calls will
fail. Both failures are EFAILURE.
880. Fixed a case where an GPI abend would occur when calling the
SPXGetConnectionStatus() function.
881. GetBinderyObjectDiskSpace() was checking the connectionID
passed in as if it was CURRENT_CONNECTIONID.
882. Made SetCurrentNameSpace() return -1 and set NWErrno to
ccode where it was not doing so.
885. In chmod(), I changed the targetNameSpace parm. from CURRENT_NAMESPACE
to DOSNameSpace. The call uses modify info. that is NS specific. We
had a case where the current NS was MAC and the MAC.NAM doesn't support
changing the file attributes.
888. SetCurrentConnection and SetCurrentFileServerID were setting the current
directory to SYS: whether they had rights or not. This would fail if
they were only attached. We now go to SYS:LOGIN if it fails.
890. Added "LONG *fileHandle;" to end of open/create call back
structures. This applies to file open, create file, and open/create
function calls.
892. SetCurrentConnection was failing because it was blocking,
and another thread was changing it during this time. We now use a local variable
to hold the connection.
893. If DeleteUser was called to delete a user and its thread
group user context was null it would abend.
895. In SetCurrentConnectionID the DirHandle was never
freed successfully because it was nulled before it was freed. The return
code was never checked.
897. Added CYieldWithDelay to FreeConnection in case it was hung up
on semaphores trying to be freed.
898. NCPConnectByAddress was not checking the transLen coming in.
On occasions (bugs) it comes in as a zero.
----------
4.02 DSAPI
----------
131. Made changes in each of these files to synchronize them with the
client NWNET code.
132. Put in a check for translate strings before getting nwstrlen
on schema names. We cannot assume that the string is in the local code
page (it may be in unicode).