Before, I mentioned using the z command as a way to figure out where an error HRESULT was occuring.  (Really, the best way is to have some decent tracing so you can see the failures in a debugger). 

 

wt is better at this - it's much much faster and adds more flexibility.  wt does a few different things, including a sort of profiler view of the calls made - but the tracing is the most useful to me.  Here's the command I end up using a lot:

 

wt -l 5 -m shell32 -oR

 

What does it do?  Well, we execute the function and trace up to 5 levels deep (within shell32 - one level beyond that) and we output the return value.  So, in this way you can trade off between more tracing and less iterations tracking down the problem or less spew and more repeat debugging.  I find 5 levels deep usually helps me spot the problem in one pass - and it's fast.

 

wt - the next best thing to (but nowhere near as good as) decent tracing in your debug code.