How do I find where my application is crashing

[Adam Hearn]

Sorry if this difficult to understand but I'm pulling my hair out and could
do with some good ideas please...

I've developed an application which in .NET as a Windows service which is
simply crashing but with no signs/details on where the crash is occuring.
Please not that the crash is happening on the server and only when the
server is running on a Win2K3 host. App is developed as an N-Tier app
(client-processing-processing-data) with (.NET Remoting & DCOM for
communications). VS.NET 2003 with .NET Framework V1.1 (will be testing under
V1.1 SP1 'soon').

It's a rather complicated application with a few layers so I'll try to
explain as best I can.... Here are all the bits:
A) .NET Service (C#)
B) Business Logic for application (C#)
C) .NET wrapper for 3rd party library (C#)
D) .NET to C++ Interface layer for 3rd party library (Managed C++)
E) 3rd party library - DLL (C++ API)
F) Interop to company's legacy component
G) Legacy components (ATL/C++ DCOM)
H) Data layer supporting 3rd party DLL to legacy component (Mostly unmanaged
C++ but a little bit of Managed C++ is present for event propogation)
I) New GUI (C#)

I initiates a connection to A via B (configured via Remoting Configuration
file)
I initiates an operation to process some data.
B allocates a thread from a ManagedThreadPool - class downloaded (from
GotDotNet, MS written)
In that thread B loads E via C and D (B has no direct access to D)
E dynamically loads H (GetProcAddress calls)
H calls G for data when needed.
H generates events which are received/processed by B which if needed are
evented back to I

The problem is pinpointed to when I run a Release copy of H, the server will
usually 'stop' without warning. A debug version never shows this problem.
However, the problem I have is that the process just *stops*. No error, log,
dialog, or popup is generated at all! I've attached a Remote Debugger to the
image (albeit Release mode) and there are *no* exceptions caught. It starts
work and then all threads just stop. I note that the exit codes of the
threads are 0x502.

I've got exception handling in all the layers that I have source code to
(i.e. except the 3rd party and a bit of the legacy component). I've even
added a C->C++ exception translator (which helped find some issues we were
having a while back), the .NET/C# assemblies all have full exception
handling using the MS Application Block. We continue receive and record all
other errors - it's just this one occasion where it looks like it was a
process exit rather than a failure but this isn't expected at all!

If anyone could point me in a direction I'd very much appreciate it! Sorry
for the information overload!

[Steve McLellan]

Hi,

Typically problems like this are caused by corruption somewhere that becomes
obvious when compiled as a Release build because of optimizations and
whatnot. Try reducing the compiler and linker optimization settings 1 by 1
if possible to find out whether a particular setting's breaking anything.
Failing that, are you able to by a third party analysis tool or something?
Debugging things like this is soul destroying, especially with the system
you're looking at. Can you isolate the offending subsystem? You may be able
to build a testbed to give it a work out, but you'll probably find that it
runs fine unless it's inconvenient for it to crash.

If you can isolate it down a bit, someone'll be able to guess why a certain
setting etc could be causing a problem.

Steve

"Adam Hearn" <ah****@nospam.com> wrote in message
news:eU**************@TK2MSFTNGP14.phx.gbl...

Sorry if this difficult to understand but I'm pulling my hair out and could do with some good ideas please...

I've developed an application which in .NET as a Windows service which is
simply crashing but with no signs/details on where the crash is occuring.
Please not that the crash is happening on the server and only when the
server is running on a Win2K3 host. App is developed as an N-Tier app
(client-processing-processing-data) with (.NET Remoting & DCOM for
communications). VS.NET 2003 with .NET Framework V1.1 (will be testing under V1.1 SP1 'soon').

It's a rather complicated application with a few layers so I'll try to
explain as best I can.... Here are all the bits:
A) .NET Service (C#)
B) Business Logic for application (C#)
C) .NET wrapper for 3rd party library (C#)
D) .NET to C++ Interface layer for 3rd party library (Managed C++)
E) 3rd party library - DLL (C++ API)
F) Interop to company's legacy component
G) Legacy components (ATL/C++ DCOM)
H) Data layer supporting 3rd party DLL to legacy component (Mostly unmanaged C++ but a little bit of Managed C++ is present for event propogation)
I) New GUI (C#)

I initiates a connection to A via B (configured via Remoting Configuration
file)
I initiates an operation to process some data.
B allocates a thread from a ManagedThreadPool - class downloaded (from
GotDotNet, MS written)
In that thread B loads E via C and D (B has no direct access to D)
E dynamically loads H (GetProcAddress calls)
H calls G for data when needed.
H generates events which are received/processed by B which if needed are
evented back to I

The problem is pinpointed to when I run a Release copy of H, the server will usually 'stop' without warning. A debug version never shows this problem.
However, the problem I have is that the process just *stops*. No error, log, dialog, or popup is generated at all! I've attached a Remote Debugger to the image (albeit Release mode) and there are *no* exceptions caught. It starts work and then all threads just stop. I note that the exit codes of the
threads are 0x502.

I've got exception handling in all the layers that I have source code to
(i.e. except the 3rd party and a bit of the legacy component). I've even
added a C->C++ exception translator (which helped find some issues we were
having a while back), the .NET/C# assemblies all have full exception
handling using the MS Application Block. We continue receive and record all other errors - it's just this one occasion where it looks like it was a
process exit rather than a failure but this isn't expected at all!

If anyone could point me in a direction I'd very much appreciate it! Sorry
for the information overload!