To the esteemed Debugging Experts out there 
-
I sent a question about this once before in a broader context, and i’m trying again with a narrower question.
I have an NT Service that handles client transactions and essentially relays them to a communications server. The service will now and then get into a confused state, and it would be great if we could simply nuke the process and force a Dr.Watson and get a USER.DMP.
The way we’ve done this in the far past under NT 4 was simply to do a zero divide.
More recently, perhaps under late versions of NT 4 (probably post-SP4 but it may have nothing to do with that) and under W2K systems, we have been using this technique
but it has not been working. We get messages that say that the USER.DMP file cannot be read by WinDbg (we’re using the latest GA version on the web).
Is there something we need to do before or after the zero divide, to insure that we get a valid crash dump? delays or something? is there something to configure on the machine other than the options set by DRWTSN32.EXE? What are the correct options?
Has anyone else had trouble getting valid USER.DMP files?
Sigh
Richard J. Pennenga
Avaya, Inc.
http://www.avaya.com
x/0 gives me a corrupted USER.DMP fileIs there any reason why you can’t use
a DbgBreakPoint, or the user-mode equivalent? Divide by zero seems like
using a tactical nuke on a single tank.
“Pennenga, Richard J (Rich)” wrote in message
news:xxxxx@windbg…
To the esteemed Debugging Experts out there -
I sent a question about this once before in a broader context, and i’m
trying again with a narrower question.
I have an NT Service that handles client transactions and essentially relays
them to a communications server. The service will now and then get into a
confused state, and it would be great if we could simply nuke the process
and force a Dr.Watson and get a USER.DMP.
The way we’ve done this in the far past under NT 4 was simply to do a zero
divide.
More recently, perhaps under late versions of NT 4 (probably post-SP4 but it
may have nothing to do with that) and under W2K systems, we have been using
this technique
but it has not been working. We get messages that say that the USER.DMP
file cannot be read by WinDbg (we’re using the latest GA version on the
web).
Is there something we need to do before or after the zero divide, to insure
that we get a valid crash dump? delays or something? is there something to
configure on the machine other than the options set by DRWTSN32.EXE? What
are the correct options?
Has anyone else had trouble getting valid USER.DMP files?
Sigh
Richard J. Pennenga
Avaya, Inc.
http://www.avaya.com
Excellent idea!
Let me see if i got that right: that will go to WinDbg if it is “attached”, and throw an exception to go to DrWatson otherwise - so i get either control, or my USER.DMP - is that right? If so, that works well for me!
Please confirm.
-----Original Message-----
From: Phil Barila [mailto:xxxxx@Seagate.com]
Sent: Wednesday, April 24, 2002 2:46 PM
To: Kernel Debugging Interest List
Subject: [windbg] Re: x/0 gives me a corrupted USER.DMP file
x/0 gives me a corrupted USER.DMP fileIs there any reason why you can’t use
a DbgBreakPoint, or the user-mode equivalent? Divide by zero seems like
using a tactical nuke on a single tank.
“Pennenga, Richard J (Rich)” wrote in message
news:xxxxx@windbg…
To the esteemed Debugging Experts out there -
I sent a question about this once before in a broader context, and i’m
trying again with a narrower question.
I have an NT Service that handles client transactions and essentially relays
them to a communications server. The service will now and then get into a
confused state, and it would be great if we could simply nuke the process
and force a Dr.Watson and get a USER.DMP.
The way we’ve done this in the far past under NT 4 was simply to do a zero
divide.
More recently, perhaps under late versions of NT 4 (probably post-SP4 but it
may have nothing to do with that) and under W2K systems, we have been using
this technique
but it has not been working. We get messages that say that the USER.DMP
file cannot be read by WinDbg (we’re using the latest GA version on the
web).
Is there something we need to do before or after the zero divide, to insure
that we get a valid crash dump? delays or something? is there something to
configure on the machine other than the options set by DRWTSN32.EXE? What
are the correct options?
Has anyone else had trouble getting valid USER.DMP files?
Sigh
Richard J. Pennenga
Avaya, Inc.
http://www.avaya.com
—
You are currently subscribed to windbg as: xxxxx@avaya.com
To unsubscribe send a blank email to %%email.unsub%%
Sorry, I can’t confirm it, you’ll just have to try it out. I imagine that
won’t be too hard for you to do, though. Worst case is it doesn’t work,
hoses your test machine, and you have to rebuild your test system. But that
is unlikely. 
Phil
“Pennenga, Richard J (Rich)” wrote in message
news:xxxxx@windbg…
Excellent idea!
Let me see if i got that right: that will go to WinDbg if it is “attached”,
and throw an exception to go to DrWatson otherwise - so i get either
control, or my USER.DMP - is that right? If so, that works well for me!
Please confirm.
-----Original Message-----
From: Phil Barila [mailto:xxxxx@Seagate.com]
Sent: Wednesday, April 24, 2002 2:46 PM
To: Kernel Debugging Interest List
Subject: [windbg] Re: x/0 gives me a corrupted USER.DMP file
x/0 gives me a corrupted USER.DMP fileIs there any reason why you can’t use
a DbgBreakPoint, or the user-mode equivalent? Divide by zero seems like
using a tactical nuke on a single tank.
[snip]
Yes, execpt in the case where the machine was booted for kernel
debugging. In which case the machine is halted, pending a continuation
signal from the kernel debugger.
You might want to check out the windbg SDK. There is a sample in there
about how an app can use dbgeng.dll to create a dump of itself. After
the dump is made you could then end the process by calling quit() or
whatever.
-----Original Message-----
From: Pennenga, Richard J (Rich) [mailto:xxxxx@avaya.com]
Sent: Wednesday, April 24, 2002 3:20 PM
To: Kernel Debugging Interest List
Subject: [windbg] Re: x/0 gives me a corrupted USER.DMP file
Excellent idea!
Let me see if i got that right: that will go to WinDbg if it is
“attached”, and throw an exception to go to DrWatson otherwise - so i
get either control, or my USER.DMP - is that right? If so, that works
well for me!
Please confirm.
-----Original Message-----
From: Phil Barila [mailto:xxxxx@Seagate.com]
Sent: Wednesday, April 24, 2002 2:46 PM
To: Kernel Debugging Interest List
Subject: [windbg] Re: x/0 gives me a corrupted USER.DMP file
x/0 gives me a corrupted USER.DMP fileIs there any reason why you can’t
use a DbgBreakPoint, or the user-mode equivalent? Divide by zero seems
like using a tactical nuke on a single tank.
“Pennenga, Richard J (Rich)” wrote in message
news:xxxxx@windbg… To the esteemed Debugging Experts out there - I
sent a question about this once before in a broader context, and i’m
trying again with a narrower question. I have an NT Service that handles
client transactions and essentially relays them to a communications
server. The service will now and then get into a confused state, and it
would be great if we could simply nuke the process and force a Dr.Watson
and get a USER.DMP. The way we’ve done this in the far past under NT 4
was simply to do a zero divide. More recently, perhaps under late
versions of NT 4 (probably post-SP4 but it may have nothing to do with
that) and under W2K systems, we have been using this technique but it
has not been working. We get messages that say that the USER.DMP file
cannot be read by WinDbg (we’re using the latest GA version on the web).
Is there something we need to do before or after the zero divide, to
insure that we get a valid crash dump? delays or something? is there
something to configure on the machine other than the options set by
DRWTSN32.EXE? What are the correct options? Has anyone else had trouble
getting valid USER.DMP files? Sigh Richard J. Pennenga Avaya, Inc.
http://www.avaya.com
—
You are currently subscribed to windbg as: xxxxx@avaya.com
To unsubscribe send a blank email to %%email.unsub%%
—
You are currently subscribed to windbg as: xxxxx@microsoft.com To
unsubscribe send a blank email to %%email.unsub%%