CWMM 0.2.7 race condition on boot

CWMM 0.2.7 race condition on boot

Post by Isaa » Fri, 03 Oct 2003 14:07:36


I suspect that there is some race condition problem on startup with
0.2.7. It's similar to the experience with Warp 4 pre-FP1 and the
Warpcenter thing, if I recall.
Not sure, but I have NOT experienced this with previous versions of
CWMM. Here's the story:

- Absolutely CLEAN install of MCP2. I then followed all instructions on
installing CWMM 0.2.7 including all cleaning and re-boots and the
RUNME1ST.CMD and backup of old DLL's.
- First re-boot, I get the blue screen. WPS does not come up, I just
have the pointer.
- I fiddled with STARTUP.CMD to add a CMD.EXE to startup (after booting
to commandline) and 2nd re-boot. The desktop comes up just fine. I
restore STARTUP.CMD to the original and things still work.
- Works fine for several weeks. Last night, again, same thing. (No, NOT
due to installation of new software, there was none). Boots to blue
screen and hangs, so I still get PM, but not WPS, and no desktop.
- I repeated this 3 times with the same effect. Remove CWMM and it boots
to desktop fine. Install CWMM, and it hangs at blue screen. CWMM doesn't
even have to be fully de-installed, if I boot to command-line and
re-name the CWMM directory and re-boot, the WPS+Desktop will come up.

I have no other desktop enhancers, and the only thing starting up is the
Relish background reminder and the network stuff. AUTOSTART is set to
FOLDERS, TASKLIST and WARPCENTER only.

Has anyone else experienced this?
I won't say it's CWMM, but it's CWMM 0.2.7 interacting with _something_,
and I can't tell what's at fault.


Isaac
 
 
 

CWMM 0.2.7 race condition on boot

Post by Klaus Stae » Sat, 04 Oct 2003 05:15:37


I had a similar experience with CW-MM 0.26 before it was released to public!

It took me about 30 re-boots and installation attempts (including
several attempts by Chris Wohlgemuth to fix this) until I found out that
a Class installed by Worperfect 5.2a for OS/2 was the guilty. After
deregistering the Wordperfect Class everything went smooth.

Naybe it helps to find out whats going wrong.

Klaus Staedtler

 
 
 

CWMM 0.2.7 race condition on boot

Post by Christian » Sat, 04 Oct 2003 07:14:30

On Thu, 2 Oct 2003 20:15:37 UTC, Klaus Staedtler < XXXX@XXXXX.COM >




According to Chris W. the Quickmotion stuff is also a potential cause
for problems.
--
"In the beginning, the Universe was created. This has made a lot of
people very angry and been widely regarded as a bad move." - Douglas
Adams
 
 
 

CWMM 0.2.7 race condition on boot

Post by Isaa » Wed, 08 Oct 2003 13:44:38


Aah, yes, but how to find out what's wrong?
I don't have Wordperfect, nor Quickmotion. Nothing but the most common
programs like IBM Works, Lotus SS, StarOffice...

The biggest problem was that it wasn't consistent. As I mentioned, after
first install, it hung, but then it proceeded to work for weeks until
one day, it decided to give up again.

The only clue I have is that when it is booted up to PM, but NOT WPS
(i.e. it has "hung"), networking is busted. (I can play around by
starting a CMD.EXE session in STARTUP.CMD).
Now, I'm not sure if networking is supposed to be "not working" if the
WPS has not come up. I'm guessing no....
All the networking apps actually do come up, even the Java TCPIP config
is fine.

Isaac
 
 
 

CWMM 0.2.7 race condition on boot

Post by Klaus Stae » Thu, 09 Oct 2003 04:57:37


I my case it was intuition ;-). If you want to find it out really, have
a look at the registered classes (e.g. by using unimaint) and then
(after first saving the OS2.INI and OS2SYS.INI) start to deregister one
class. Test. If this doesn't solve, restore ini's and try with next ...

Slightly guessing that deregistering only classes which are added by
tools/applications additionally installed might be sufficient.

Klaus Staedtler
 
 
 

CWMM 0.2.7 race condition on boot

Post by Isaa » Thu, 09 Oct 2003 14:16:26


[snip]

Maybe yes, but CWMM is only 0.2.7 and it is replacing a very basic part
of the OS/2 system. Right away on startup, you may need to access audio
and image functionality, right?

Considering that there isn't likely to be any documentation on the
internal workings of OS/2 available to CWMM's author, it's not
surprising that things break now and then. Annoying, but not surprising.

Also, sometimes it still isn't CWMM's fault. I recall when some made the
switch from Classic Rexx to Object Rexx, a lot of things broke. Well, a
lot of people said that, obviously it's Object Rexx's fault....

But it wasn't. Eventually what people realized was that in most cases,
Object Rexx just had more stringent checks in place and essentially
rooted out bad coding that Classic Rexx let get away. I'm not saying
that it's the case with CWMM here, but it is possible.

(But sorry, no, I'm not going to spend the time
registering/deregistering every class on my system to find the broken
bit. I have to _use_ my system! :-)


Isaac
 
 
 

CWMM 0.2.7 race condition on boot

Post by Paul Ratcl » Thu, 09 Oct 2003 20:23:47

On Tue, 07 Oct 2003 19:57:37 GMT, Klaus Staedtler < XXXX@XXXXX.COM >




I would suggest starting with the ones which begin with CW.


What possible reason could there be for unrelated classes to influence
these replacement MM classes?
CWMM seems to cause so many problems, and the only common factor seems to
be CWMM.....
No other classes seem to cause these startup lockups, so you have to ask
yourself what is going on with CWMM. Just what is it about these things
that demands they be installed first, demands you clean you INI files
300 times etc. without locking up? I suspect there is a fundamental design
flaw in them somewhere.
 
 
 

CWMM 0.2.7 race condition on boot

Post by Klaus Stae » Fri, 10 Oct 2003 00:11:49


Paul you're right and I have my rest.

Klaus Staedtler
 
 
 

CWMM 0.2.7 race condition on boot

Post by Ilya Zakha » Fri, 10 Oct 2003 04:12:26

[A complimentary Cc of this posting was NOT [per weedlist] sent to
Paul Ratcliffe
< XXXX@XXXXX.COM >], who wrote in article < XXXX@XXXXX.COM >:


Paul, did you try to insult the *original design* of MMOS2 subsystem,
or a *particular CW implementation*? [I do not use CW (yet?), but
inquiring minds want to know ;-]

If the former, then (from experience with IBM's implementation) I
wholeheartedly agree! ;-) ;-) ;-)

3/4 serious,
Ilya
 
 
 

CWMM 0.2.7 race condition on boot

Post by Mark Kleba » Fri, 10 Oct 2003 07:01:20

No, please don't quit. Those classes work fantastically for me and
I'd like to see you continue ;-)

On Wed, 8 Oct 2003 15:11:49 UTC, Klaus Staedtler





--
 
 
 

CWMM 0.2.7 race condition on boot

Post by Klaus Stae » Fri, 10 Oct 2003 16:10:03


To make things clear: I'm not the developer. I only support the CW-MM
classes with some painted pixels, testing and answering questions.

Klaus Staedtler
 
 
 

CWMM 0.2.7 race condition on boot

Post by Christian » Sat, 11 Oct 2003 05:50:50

On Wed, 8 Oct 2003 11:23:47 UTC, Paul Ratcliffe



The "flaw" seems to be that the original MMPM/2 stuff leaves traces all
around the system, and they are hard to get rid of. But without getting
rid of them there will be problems due to the WPS' design.
--
"In the beginning, the Universe was created. This has made a lot of
people very angry and been widely regarded as a bad move." - Douglas
Adams
 
 
 

CWMM 0.2.7 race condition on boot

Post by Rainer Doe » Sat, 11 Oct 2003 07:31:16

On Thu, 02 Oct 2003 05:07:36 +0000, Isaac



1st::
The old tip, make a reference of Warpcenter in Startup Folder and kill
WARPCENTER from line AUTOSTART

2nd:
Make a test w/o System Sound; if success, move the sound card lines in
CONFIG.SYS at the end (sound card RUN statement last RUN in
CONFIG.SYS)

3rd:
In '96/97 a have many BSOD, it was the combination ATI Grahic card/ISA
sound card; no BSOD after change to Matrox
 
 
 

CWMM 0.2.7 race condition on boot

Post by Isaa » Tue, 14 Oct 2003 12:36:47


[snip]

> 1st::

Yes, already tried this. I should have mentioned it. FP1 from old W4 was
supposed to fix this, but this is something i always try anyways. But no
luck here.


Eh? Doesn't make sense to me. Doesn't the sound card drivers get loaded
from the CONFIG.SYS before the WPS comes up? Why should the order matter?
Yes, it might mean that disabling sound might cause it to work, but that
defeats a lot of the purpose of CWMM then.


Nope, should be no such problem here. It is a Thinkpad 770X which is
supposed to be fully supported under OS/2 and has had no problem for
years. Went from native Trident drivers to SNAP.


Isaac
 
 
 

CWMM 0.2.7 race condition on boot

Post by Paul Ratcl » Wed, 15 Oct 2003 20:54:40

On Wed, 8 Oct 2003 19:12:26 +0000 (UTC), Ilya Zakharevich



The design of MMOS2 is not too bad, maybe a bit over-engineered, but the
implementation leaves a lot to be desired and the documentation is worse.
The Multimedia Programming Ref. has to be *the most* frustrating INF file
on the system. I have spent a fair bit of time in the last week trying to
get something working and have more or less ended up not far from where
I started. That I got anywhere is more by luck than anything else.

As to CWMM, I have not dared to install it on anything because of all
the changes it does. Maybe I will try it in a VPC sandbox.
As far as the implementation of that goes, I haven't had chance to study
the source code so I can't comment, but I don't like the implication that
comes with installing this thing - "it locks up after you installed it, so
it is somebody else's fault, but not mine".
I don't know of any other WPS classes that take this attitude.