VOICE Home Page: http://www.os2voice.org
[Previous Page] [Next Page]
Editor's note: these tips are from OS/2 users and in some cases can not be verified
by myself. Please heed this as a warning that if you are not sure about something,
don't do it.
Yes! The "Examples & Demos" in Word 6.0a-c *can* be made to work
Many theories arrived. None helped, thanks anyway.
Trial and Error reveals:
1. It doens't work in NT, either.
2. It doesn't work in 98, either. (It crashes in '98; merely refuses to work in NT).
3. All it takes is to add to the [Microsoft Word] section of winword6.ini in the WINOS2 folder this line:
Then you get the nice, on screen movie-like demos and examples.
And, somehow, learning that the install is crippled for NT, too, not just OS/2,
cheers me. Guess it shows there's a little bit of OS/2 in NT, after all.
The problem are not the animations. You can get this behaviour on complex pages
without animated GIFs.
Here is a description of the problem:
When downloading a page with NS 4.04 or 4.61 all of a sudden CPU utilization goes to 100%. The CPU meter and the watch in WarpCenter stop. Netscape itself responds to user interaction absolutely normal (e.g. opening a menu or the preferences dialog) and the process indicator of NS works smoothly. In most cases, opening the preferences dialog and keeping it open until the pages is completely loaded brings down the CPU utilization to normal values. If you switch to another applications while this problem occurs there is a high chance that the system stalls completely with the need to reboot. This problem is very similar to the '100% CPU load while calculating complex tables' problem. The big difference is, that the latter is not a real bug - there is no system stability problem. While the table is rendered NS's interface is not very responsive (e.g. the status information stops and jumps a lot), and after completing the table the system reverts to a normal state.
Only under special circumstances the problem occurs. This is the reason why many people (and IBM) don't seem to be able to reproduce this. First you need a complex page. Then your system has to be configured like this:
+ You use a proxy on the same machine (like WBI).
+ You have a DNS running on the same computer (like BIND).
+ You use a 32bit TCP/IP stack (e.g. TCP/IP 4.1 or MPTS WR8610 or WR8620).
I am not quite sure if all of the above is required to run into the problem. Nor do I know if this happens if you use another proxy (like SQUID) or another nameserver.
If you use a local proxy there is a high chance that you have turned off the file cache within NS since this would be a second time caching beside what the proxy does. Most likely you have turned on the memory cache of NS so that pages visited in a session will be redisplayed faster (e.g. when using the back button). To cure the problem you now should do this:
TURN OFF NETSCAPE'S MEMORY CACHE!!! (set it to 0)
The drawback is, that animations will only run once. Configuring the file cache, so that it holds some amount of data (e.g. 1 to 2 MB) will bring back the animations.
I hope this clarifies the situation and solves a problem some of you encountered.
Michael Warmuth Austria - The place in the
http://www.os2forum.or.at/ heart of Europe where no
http://www.osiconsult.co.at/ kangaroos are hopping around
If you experience boot problems with SciTech's drivers, there is a quick way
to "get back up" The only requirement is that the GENGRADD drivers work
with your video card.
After a boot or PM failure...
1. boot to commandline by invoking ALT+F1 and the "OS/2..." blurb when the system start booting OS/2 (BEFORE the OS/2 Warp graphic logo)
2. Edit the CONFIG.SYS and change to GENGRADD.
b) Press escape (to put the cursor in the "editing" position)
c) find the line
and change it to
d) delete or rename the \OS2\SVGADATA.PMI file
You should now be up in Generic SVGA support.
Hope this helps. (this would be great for the README on future betas)
To make gif animations stop after one cycle, read this page:
http://simmons.starkville.ms.us/tips/081097/In short, search for NETSCAPE2.0 and for ANIMEXTS1.0 and change them to be something else - I just changed the . to a , (in netscape.exe).
I used the command line:
java -classpath d:\java11\lib\classes.zip;install.zip;. install
"java -classpath" is the JDK version of "jre -cp (I think).
Note that I included the classes.zip file in the classpath along with install.zip as well as the current directory, the trailing ";." .
There is a bug I documented with IBM concerning the fact that chkdsk.sys does
not seem to be able to give the memory back when the system is done running Chkdsk.
They were looking into triggering a call back to the device driver, possibly when
PMShell starts, letting it know that it is time to release the memory.
W/ WarpServer (Warp3 base) and FP41, a server w/ three 18GB volumes, and 256MB
RAM, our server code (Rexx based) which eats up all of that RAM was abending with
the strangest errors you ever saw. Huge swap file was the clue that we didn't have
all of our RAM. It was so bad that we were unable to start additional CMD sessions!
I have solved the problem as follows:
C: Pri CHKDSK FAT
C: Pri MAINT FAT
D: Log SERVER HPFS
Both of the C drives were created w/ BootOS2 - the one called ChkDsk does nothing
more than ChkDsk every HPFS volume while booting, using the + for each drive letter
and ChkDsk.sys to all ChkDsk to get enough memory. Once inside WPS of that partition,
we SetBoot /IBD:D to get into the server.
MAINT is our normal Maint partition created again with BootOS2, but I have the
undocumented /FORCE option on that HPFS.IFS and no autocheck so that it is sure
to boot w/o error on a dirty drive.
Our servers use CronRGF, and I have a job to reboot the server at specific times,
SetBoot /IBD:D, and once a week we ChkDsk all the drives, SetBoot /IBA:MAINT
One other tip on the /FORCE paramater. I thought I would be able to use my standard
MAINT partition, having startup.cmd start a Rexx process, putting up a message saying
that ChkDsk would start checking all specified HPFS drives in 20 seconds if the
window was not closed. Unfortunately since the dirty bit is still set on the drive,
as soon as ChkDsk /F is run on the drive OS/2 pops a full screen error message asking
you to run ChkDsk /F - so much for no user intervention! Thus I had to add a ChkDsk
partition. This code still comes in handy because we have all of our servers monitored
by HP OpenView, so our weekly ChkDsk against the drives is done w/ the MAINT partition
where I have a thin MPTS installed, enough TCP/IP to run FTPD, TELNETD, FTP, and
TELNET - thus OpenView thinks the box is on-line and we don't get paged. This only
works with TCP/IP 4.0 and lower - TCP/IP 4.1 and 4.21 get very mad when I do this
If the server hard lock's up then the night watch people on-site call us, we
walk them through hitting the power button since it has come to that, and since
CHKDSK is the default partition in BootManager, everything cleans up nicely.
I suggest using FAT for the MAINT type partitions BECAUSE of a BUG I found in
HPFS where it is Trapping the system during boot time w/ /CRECL:64 paramater set
because it currupts its own spare block on our active servers. Since one drive is
dirty, you are unable to install HPFS.IFS, thus FAT MAINT partitions come in handy.
So, DO NOT USE /CRECL:64, because IBM can't recreate it on a test system, and we
can't afford the down time on a production server while IBM tries to get enough
debug material. This bug has been documented since around Warp 3 FP5 time frame.
I only found the bug about two months ago!!!
Product Suggestion - GammaTech saved the day! Since the MAINT partition was FAT,
I was able to remove the HPFS.IFS line from it's config.sys with boot disks, then
add NetBIOS to the partition, and install ThinIFS/SrvIFS to get to another backup
server with lots of disk space. GammaTech was able to RAW READ the drive and recover
almost ever file on all three 18GB volumes.
Lueck Data Systems
When you run your batch file through the "start" command, you can set
the title line as the 1st parameter of the "start" command, enclosed in
doublequotes; for instance
start "My Title" /C /F myprog.cmd
When you run your batch file as a program object, the title of the program object,
as specified in its settings notebook, is passed to the OS/2 window.
Have you tried LOADDSKF itself?
LoadDskF 1.14p 05-26-92
(c) Copyright IBM Corporation 1989, 1992
LoadDskF [D:][path]filename.dsk d: [options]
The file must contain an image of the diskette that you wish to create
in the specified drive. The destination diskette is checked to ensure
the same format as the diskette from which the file was originally
The file must not be on the diskette in the destination drive.
/F The target diskette will be formatted to match the input image.
/Y Bypasses the user input (y/n) prior to overwriting the target diskette.
/Y/Q Omits the warning message before overwriting the target diskette.
/S Silent Mode - Don't Beep after writing diskette.
/C Used to write 3.5" images to 5.25" diskettes or vice versa.
See the documentation for usage and limitations.
This program will work on DOS and OS/2 Operating Systems.
Comments/questions to TCOOPER@VNET.IBM.COM on the Internet.
We've put the entire package out on ftp://service.boulder.ibm.com/ps/tools/loaddskf/ as well, see loaddskf.doc.
Depending on who is rendering the font and in what way will determine which fonts
work. So bitmap fonts usually don't work, truetype works, ATM depends. We definitely
know that S3 cards have had problems displaying 1004. The browser is actually the
best test of this though. Set your font to Tms Rmn. Go to http://members.aol.com/pspmikek/charsets
Select Latin-1 on the left. If the characters in the table do not match the GIF
below, it is definitely a video card problem.
Your FI.INI is probably not okay.
And place it in \OS2\INSTALL.
Run it. It will create a file called FI.NEW.
Rename the original FI.INI and name FI.NEW as FI.INI.
Then retry installing.
The following was posted to staroffice.com.support.announce
Dear StarOffice User,
with this posting we provide a small patch, to disable the need for register each time, when you make changes in the user settings of older StarOffice 5.1 versions. Please do not use this patch for the latest StarOffice 5.1 version from Sun. Download the below listed patch into the StarOffice directory, where all the DLL's (or shared Lib's) are located. After downloading execute this patchprogram.
StarOffice 5.1 patch for disabling the registration mechanism.
Solaris Intel ftp://ftp.stardivision.de/pub/support/so51/unxsoli/regpatch
Solaris Sparc ftp://ftp.stardivision.de/pub/support/so51/unxsols/regpatch
The patchprogram makes a backup copy of the DLL (shared Lib) and changed this
file afterwards. Please make sure, that StarOffice is not running during this change.
If you have questions or problems with the patch, please describe the problem in
our installation newsgroup for your operating system.
Manager StarOffice Competence Center
The program was probably written with a Borland compiler. Their DPMI support
sometimes confuses OS/2. You need to go in the Memory settings for the object and
change the DPMI setting from Auto to Enabled.
To get there, you might need to rename the executable first, or set up a Program
Object from a Template by setting all the memory settings, Full Screen, etc., first,
then (as the last step) put in the path, etc. for the executable. I don't recall
the details. I hope this gives you enough information to get you started.
It's been a long time .... I'll just have to check what I did back then:
OK, copy the entire ibmworks directory from a Warp4 CDROM \os2image\fi\bonuspak\
to \IBMWORKS on your WSEB root. Then call IBMWDESK.cmd from within that directory
and reboot. That's it.
I have several of them working. It takes a bit of doing, though, since the drivers
for the DFE530TX won't work with it (different chipset on the card). On the disk
that came with the card, there is a NDIS directory, with an OS/2 Subdirectory. In
there, you'll find the Lanserver drivers for this card. You need the driver, but
not the NIF file (the NIF file will not work).
Daniella Engert (DANIS506.ADD fame) has a patch utility.Also here is a post from Frank Field of an email from Phoenix Technologies that explains the RAM detection problem:
From the HARDWARE mailing list:
"I've written a utility to patch the OS2LDR for all recent versions of OS/2 so that this memory detection defect shouldn't be an issue any longer. I had only one machine to test it, a GigaByte GA7IX Athlon mobo, and it works as intended.
The utility is available on request from now on.
Not sure Dani would want her e-mail address posted in usenet .....but if you have her IDE drivers, you can probably find it in the readme files.
OS/2 and Windows NT/9x use different BIOS calls to determine
how much memory is available. OS/2 uses INT 15h,AX=E801h
while Windows uses INT 15h,AX=E820h.
Look here, http://www.ctyme.com/intr/int.htm
<http://www.ctyme.com/intr/int.htm> to see the details.
Award has had support for the E801 call since late in 1995.
Recently, though, this support was made conditional on ACPI
in order to pass an Intel "Component Validation Test". In other words a BIOS
with ACPI support could not have E801 support.
Very recently (October), this code was refined to allow the
E801 support to be included even if ACPI is supported.
In summary, your BIOS doesn't have support for the runtime
call OS/2 needs to correctly determine how much memory
above 64MB is available. It is possible that a BIOS upgrade
from Gigabyte might fix this.
I hope this clarifies what is happening for you, and regret that we can't make
the BIOS support OS/2.
[Previous Page ] [ Index] [Next Page ]
VOICE Home Page: http://www.os2voice.org