shell32/tests: Fix a test on Win2k platform and above.

[Vincent Povirk]

On Wed, Mar 18, 2009 at 1:34 PM, Nicolas Le Cam <niko.lecam at gmail.com> wrote:
> Ok, did more testing ...
> It seems you're right, adding a do_events() call after each
> DestroyWindow() one seems to fix the test on Win2k.
> Do you think this is a better fix ?

No, that probably creates a race condition. I've seen the windows
implementation post messages after SHAppBarMessage returns, and it may
sometimes post the notification between do_events() and the next
SHAppBarMessage call.

In general, do_events() cannot reliably catch messages that are posted
as a result of a previous SHABM call.

I still think that removing the code resetting GWLP_USERDATA is the best fix.