I've tested this new mode some time ago, but still had some problems with a "frozen" player (which then was released after some clicking and pressing various keys

) -- maybe I got the wrong mode? But this was still the 3.1.2 based version...
About the "mousemode=49" - It is intended that the game is in "pause mode" after you click with the mouse.
You can "virtually" click the way on the screen, snap and drop, go on - and if you toggle pause off the player run the given way.
The auto pause can be toggle by pressing the key "b".
Unfortunately the "frozen" problem always exists - a click in the "door area" should relase this.
In your last mail you suggested me that I should take the return values from DigField, MovePlayer, SnapField.
Upps. I always forgot that I am indirectly do it, as suggested.
The problem is that this routines return MV_ACTION if the "next" step is possible, but don't return "next step is never possible".
Arrgh. difficult to explain.
What do you mean by "removing reported bugs"?
e.g one bug I reported in my last german mail.
giving more than one directory argument dont work in 3.1.2 neither in 3.2.0.
> In this "test candidate" there is a little resource problem, if someone
> switch between the "bigger" levelsets like "BD2K3" and "SnakeBite" too
> often. At the moment I dont exactly know where this comes from.
Hmmm... Are you sure that this only happens with your modified version of R'n'D?
I found this bug/leak! After debugging this I found a bug in my code AND the nasty leak in the SDL lib (loading sound files) - it seems that sometime s the LOAD_WAV Routine dont free the file handle and the RWOPS memory. After writing a workaround the bug/leak is gone.
I uploaded a new version of RndTest.
And the answer is ... 42 !