animate control regression

[Mike McCormack]

Dimitrie O. Paun wrote:

> We should just get rid of the thread and the critical section altogether,
> comctl32 6.0 is documented not to support it anymore. But before we do
> that, let's try to fix this, so we have good code in CVS :)

If you get rid of the thread, programs that don't run a message loop or 
that block for a time after receiving certain messages won't animate 
smoothly.

What we had there worked, and it took quite a while to get rid of the 
races and lockups.  Why can't we just revert to what we had?

Mike