Phlip, thanks for taking the time to put your practices into GTK code.
You're right, it is different from what I am trying to do, however I
can see the advantages in your approach and I'm going to give it a
try.

On Mon, Dec 17, 2001 at 12:55:30AM +0900, Phlip wrote:
> You seemed from your initial post to be doing it a different way. That's
> okay, but it might have been because you assumed a window must be seen to
> be clicked on. The documentation for libraries written test-last often
> neglects issues like how to click on an unseen window, or how to destroy
> the window before showing it.

I guess all visibility issues can be solved, where necessary and where
one cares, with a bunch of calls to .visible?.

What I like in the signal approach is that it is very close to the
actual user interaction, probably just a step behind the UI testing
tools that work on the screen itself.  Calling widget methods, on the
other hand, might or might not map closely to user interaction (e.g. I
did not see a method to simulate a button click).

Thanks again.

Massimiliano