|
|
|
|
|
Hidden Value details
commands and procedures in MPE that can improve your productivity
with HP e3000 systems. Send your tips to john@burke-consulting.com,
or fax them to 512.331.3807.
Edited by John Burke
Within FTP, SITE
STREAM works for me from one MPE/iX box to another MPE/IX box. But
what about using it from an NT DOS FTP session? The remote help for
SITE makes it appear as though it should work.
John Burke replies:
From a
Windows 2000 machine (and presumably also an NT machine), after
connecting to a MPE/iX system, enter quote site stream FILENAME to
stream the job FILENAME.
I need an example
of how to verify that an INPUT variable conforms to a specific
format. Im looking for CI code to do the test. For instance,
this input should be 10 numeric characters:
INPUT
_CN;PROMPT=Paste or type 10-digit Customer Number here :
;READCNT=10
[Keven Miller and
Tracy Pierce both showed what to do, but the most aesthetically
pleasing answer (you know what I mean if you ever programmed in APL)
was from Jeff Vance:]
setvar num
while
len(setvar(num,input(Paste or type...,,10))) <> 10
or &
not numeric(num)
endwhile
# now we have a 10
digit string in the num variable ...
In the days of
old, one way to increase the speed of your logon was to block your
UDC files as high as possible to reduce the number of disk IOs that
were needed. Will this have any effect on MPE/iX 6.5 and later, or
can I leave it alone?
Jeff Woods and
Jeff Vance reply:
MPE/iX reads
disc in a multiple of 4K pages, attempts to cache frequently used
pages, and, when using a serial read via the file system API, tries
to read multiple pages at a time. I would expect that blocking factor
has no noticeable impact on UDC performance, but ensuring the file is
allocated in a single (or as few as feasible) extents would.
Contiguous disc pages can be read together (up to some maximum
limit), but parts of the file scattered around one or more discs will
obviously require separate reads.
Also, minimizing the
number of cataloged UDCs (both number of separate files and number of
UDCs within a file) does cause a noticeable increase in logon
performance.
After recently
updating our HP system we began to get the following error messages
displayed on the system console: /#J1/176/ Stat on
/etc/bootptab No such file or directory. What does this
mean?
Peter Osborne and
Mark Bixby reply:
If you are
not using bootp, comment out the bootps dgram.....
line in your INETDCNF.NET.SYS file. Then restart JINETD.
I would like to
run touch.hpbin.sys as part of a logon UDC to update the access
date/time of all the files in my group. However, I am struggling to
run this program with the correct parameters in a non-shell way. How
can I do this?
Steve Miller and
Ted Ashton reply:
Wildcarding
in Posix is actually controlled by the shell the shell expands
the wildcarding and passes the program the result. What we have done
in similar cases is actually run the shell with the -c parm to pass
it the command we want:
:sh.hpbin.sys
-c touch /ACCOUNT/GROUP/*
Is there a method
or utility available to change a files dates to a
user-specified date (other than changing the current date on the
system)?
Steve Miller replies:
Use a -t
option with touch and specify the date and time in YYYYMMDDHHMM.
Heres the format:
:sh.hpbin.sys
-c touch -t 200109151030 myfile
This should
set the mod and access date/time to 09/15/2001 10:30 am for myfile.
Add the -a option to just change the access date.
I just downloaded
and applied the patch to an MPE/iX 6.5 PowerPatch 2 system (along
with all the other General Released patches that Patchman found) that
delivers Apache 1.3.9 (We had been using 1.3.4 for a long time). Now
Apache does not run and I get the following error message in
error_log, repeated several times per second:
[alert]
(22)Invalid argument: setuid: unable to change to uid: -1
What has gone
wrong?
Mark Bixby replies:
Newer
versions of MPE Apache starting with 1.3.9 have become aware of the
User and Group directives in httpd.conf, and apparently the default
User and Group in the old 1.3.4 httpd.conf was bogus. See the new
/APACHE/PUB/conf/httpd.conf.sample that the Apache 1.3.9 patch should
have created on your system for the correct values of User and
Group.
Is there a command
file or utility where I can input a jobname, like JINETD, and get its
corresponding spool files?
Larry Simonsen,
Mark Bixby and David Darnell reply:
:listspf
o@;seleq=[jobname=jinetd]
I want to use the
first six characters in a file as a variable, but I cant figure
out how to do it.
Jon Backus replies:
If the file
is a simple ASCII file then you could do INPUT var;READCNT=6<file
and it would read the first six characters contained within the
file.
If NMMGR has been
set up to require a password, how do you get rid of it? I know you
need to run NMMGR in maintenance (or character) mode to set up a
password, but typing password <cr> does nothing.
John Burke replies:
Check out
NMSTART.PUB.SYS. Doug Werth of Beechglen first mentioned this some
time ago on 3000-L. The tip was included in the Hidden Value column
of the 3000 NewsWire and was made part of the Best Of ... compilation
from last year. This compilation, plus all columns since, are
available in searchable form at www.burke-consulting.com.
We have a user
whose UDC file contains an OPTION LOGON UDC. Sometimes this user
streams a batch job and, in this case, the OPTION LOGON interferes
with the rest of the job. Is there any parameter which can be used in
the job that will prevent execution of the OPTION LOGON UDC?
Jon Backus replies:
There
isnt a job parameter, but you could modify the UDC to check to
see if the process is a session or a job (if !HPJOBTYPE =
J then) and then skip the parts you dont want
executed for jobs.
Jeff Vance adds:
If a session
is created from within a job, HPJOBTYPE will be S in that session. To
detect this more unusual case, HPTYPEAHEAD can be set to true. This
will work in a normal session but will fail with CIWARN 8174 in a
session coming from a job.
We will soon be
replacing our disc drives and will need to do a store and restore.
When using TurboStore and three tape drives, which way is the
fastest, a serial or a parallel store/restore? We are also open to
any suggestions on how to ensure an easy move of the data (we will be
using buldacct to create the directory and catalogs).
Gilles Schipper replies:
Probably
parallel would be best for speed. It is not necessary to use buldacct
to create your directory and catalogs. Storing and restoring your
directories is much better. You really only need to use buldacct if
migrating groups/accounts from one volume set to another.
Using buldacct for
any other purpose will cause you more grief than you want,
particularly if you utilize Posix directories (and they are used
whether you know it or not).
John Burke adds:
I agree with
Gilles about the DIRECTORY option being easier and less prone to
error in your situation, but would add these cautions:
1. If you are using user
volumes, you must EXPLICITLY LIST all volume sets including the
system volume set. From the HELP subsystem:
DIRECTORY: Specifies that
the system accounting directory plus all HFS directories are to be
stored. This option requires System Manager (SM) or system supervisor
(OP) capability. If ONVS or SPLITVS is not specified, the DIRECTORY
defaults to writing ONLY the system directory. Otherwise, the
directories of the specified volume sets are written. This allows
operators and managers to store or copy private volume sets in their
entirety.
2. It doesnt hurt,
and is in fact a good idea, to run BULDACCT occasionally, just in
case.
Copyright The 3000 NewsWire. All
rights reserved.
|