Rob Landley wrote in a message to Kim Cheung:

 RL> Well, the standard technique (if it's your code) is to stick
 RL> a bunch of printf("Hi there!") calls in various places until
 RL> you've got it narrowed down to the actual line that's doing
 RL> the crashing, at which point you do some thinking.  
 RL> If you didn't compile it, and/or don't have the source code,
 RL> you're in trouble.  

Sorry to butt in, but I think there is a better way :-)

What I (under Windows) do is the following. I start the debugger and let
the debugger run the app. When an exception occurs, the debugger is invoked
in assembler mode, showing the faulty instruction. Since I don't do
assembler (much), I alter the invalid pointer (segment:offset values,
replacing them with valid values (usually ds:0 or a known and sacrificable
variable)), and then stepping through the asm until I hit source code. 
It works with Turbo debugger (Windows), but not with CodeView (also
windows) because it won't let you continue after an exception 13. So the
possibility is that it might work with OS/2 also. It surely is the fastest
method I know (at the moment) 

mvg/wr
  
--- timEd/2 1.01.g3+