Is it possible to retrieve the call stack programmatically in VB6?

Question

When an error occurs in a function, I'd like to know the sequence of events that lead up to it, especially when that function is called from a dozen different places. Is there any way to retrieve the call stack in VB6, or do I have to do it the hard way (e.g., log entries in every function and error handler, etc.)?

Answer 1

You do have to do it the hard way, but it's not really all that hard... Seriously, once you've written the template once, it's a quick copy/paste/modify to match the function name in the Err.Raise statement to the actual function name.

Private Function DoSomething(ByVal Arg as String)

    On Error GoTo Handler

    Dim ThisVar as String
    Dim ThatVar as Long

    ' Code here to implement DoSomething...

    Exit Function

Handler:
    Err.Raise Err.Number, , "MiscFunctions.DoSomething: " & Err.Description

End Function

When you have nested calls, this unwinds as each routine hits its Handler and adds its name to the error description. At the top level function, you get a "call stack" showing the list of routines that were called, and the error number and description of the error that actually occurred. It's not perfect, in that you don't get line numbers, but I've found that you don't usually need them to find your way to the problem. (And if you really want line numbers, you can put them in the function and reference them in the Err.Raise statement using the Erl variable. Without line numbers, that just returns 0.)

Also, note that within the function itself, you can raise your own errors with the values of interesting variables in the message like so:

Err.Raise PCLOADLETTER_ERRNUM, , "PC Load Letter error on Printer """ & PrinterName & """"

(The syntax highlighting looks wonky in the preview... I wonder how will it look when posted?)

Answer 2

I'm pretty sure you have to do it the hard way. At a previous job of mine, we had a very elegant error handling process for VB6 with DCOM components. However, it was a lot redundant code that had to be added to every method, so much that we had home-grown tools to insert it all for you.

I can't provide too much insight on its implementation (both because I've forgotten most of it and there's a chance they may consider it a trade secret). One thing that does stand out was that the method name couldn't be derived at run-time so it was added as a string variable (some developers would copy-paste instead of using the tool and it would lead to error stacks that lied...).

HTH