There is some kind of bug in the WHDLoad/WinUAE combination that causes the ATR Xmas demo to set $dff1fc (FMODE) to $ffff. It doesn't occur on the disk version in WinUAE, and it doesn't occur on my Amiga using WHDLoad.
All this talk people have mentioned about must require 68030 etc is all crap. The chipset, CPU has nothing to do with it. When the title screen becomes corrupt (using any CPU, I use 68020 for example) you can do the following:
Shift-F12 to enter debugger
W dff1fc 0 2
g
Instantly the screen will fix itself, the keyboard will work correctly, WHDLoad will quit when you press the quit key etc. I have tried to put a memory watchpoint on $dff1fc for write but WinUAE seems to ignore it and doesn't not show me when the register changes. If you type "e" in the debugger at the start, you can see FMODE is 0, then later on it's suddenly $ffff.
Therefore it is extremely difficult to trace this "bug" when it doesn't occur on a real Amiga