[mod.computers.vax] Another possibility in "The case of the wondering DCLTABLES"

WCE8760@TAMVENUS.BITNET.UUCP (01/18/87)

Fellow VAXlings,

In the recent dialog of late, there has been a number of questions about
DCLTABLES.EXE getting transplanted to the sys$specific root of a multi-root
cluster. This strange occurrence has befallen me as well. These are the
conclusions that I came up with, but may not neccessarily reflect the same
situation that transpired at the original site that brought out the question.

I was installing some product with VMSinstal, I think it was FMS. In any case
the other processor, a 785, in my cluster started receiving compliants
concerning the inability to utilize the new product after reboot. It was
originally installed from the 8600.

Upon further investigation, it was noticed that the DCLTABLES.EXE had a copy
sitting up in the sys$specific:[syslib] directory root. We were supposed to be
running a common system disk with a sys$common root that held most of our
running images, including DCLTABLES.EXE. It had somehow during the vmsinstal,
managed to get placed in the wrong directory root.

After digging around in the kit's saveset and fooling around with the
vmsinstal.com file I realized what was happening. This could be considered to
be a foolish mistake on my part in the first place. I have about 40 layered
products on my cluster and unfortunately have gained a false sense of security
when it comes to software installations.

In any case, I was running the installation from my personal account which is
UIC [1,5]. Most system files are [1,4] or [1,1]. It turns out the installation
procedure wants to place files in vmi$root, which should be sys$common for the
DCLTABLES.EXE file on my system. However, the VMS logical search strings, in
collusion with VMS aliases use the UIC pattern that currently exists on the
file of interest to figure out where the file should be written root-wise.

Some utilities can bypass the alias problem with an account that has privs. EDT
for instance will always place the higher version in the same root, even if you
are using a differnet UIC than the original version.

SET COMMAND is not one of these such utilities. If you have a DCLTABLES.EXE
file in the sys$common:[syslib] directory with UIC [1,4], and you subsequently
issue the following command from an account with full privs and UIC [100,77].
You will find that there is now a new version of your tables file, but it now
resides in sys$specific:[syslib].

SET COMM/TABL=SYS$SYSROOT:[SYSLIB]DCLTABLES/OUT=SYS$SYSROOT:[SYSLIB]DCLTABLES -
 somefile.CLD

The plain fact is that this will happen with any file under similar
circumstances. I don't know all the utilities that this behavior occurs with,
but I can assure you that I now install everything from the system UIC. I
believe CREATE also exhibits this behavior

Chip Eagle
Academic VAX Systems Manager, Texas A&M University
College Station, Texas 77843
TEXnet: VENUS::WCE8760
BITnet: WCE8760@TAMVENUS
(409)845-7594