[DUG] Long running apps & weirdness

[Paul Heinz]

Trevor Jones wrote: 
>   Unfortuantely the weirdness can only be explained as I did:
> 
>   The kind of things that are being reported as going wrong 
> make no sense whatsoever.
> 
>   A list index out of bounds exception when doing nothing 
> that involves lists.
> 
>   Strange options being invoked when they are definitely turned off.
> 
>   App print drivers mysteriously vanishing.
> 
>   Sort of like BDS2006 (or Windows itself for that matter).

Are you capturing full stack traces for all exceptions? 

We used to use our own code that did that but we now use madExcept which
we've modified to catch full stack traces at first point of exception
raising. There is also EurekaLog which is another commercial components
to do this for you - I've not used it personally. Actually, I think even
the JEDI library might have some free stack trace handling support.
Adding some MAPI or SMTP support to your application makes it easy for
the affected sites to email traces to you.

Also, have you considered adding checkpoints for heap consistency? If
you're using D5 upwards - I'd suggest using FastMM4 in full debug mode
as a starting point. I've recently modified our version locally to allow
for easy heap consistency checks so I can place probe points through the
code to catch any memory corruption due to buffer overruns, double-free
or use after free style errors, all of which can show themselves as the
kind of symptoms you're describing.

And it might not be your code. 

Printer drivers load and run in your applications address space so buggy
printer drivers can definitely cause memory corruptions as we have found
from bitter experience. We found one Kyocera driver which could
bluescreen NT 4.0 servers which is pretty amazing since they run in
userspace. To do that, it must have been corrupting the Win32 subsystem
memory in a way that NT couldn't recover from. Are there any
commonalities in printers and drivers used?

And the cheaper 'unified memory architecture' video cards drivers
essentially scribble on normal system memory rather than card video
memory to do their work so bugs in those can cause strange symptoms. Are
there any commonalities in video cards and drivers used?

TTFN,
  Paul.