I'm not complaining about anything, but as you can see in the debugger, all XRAM locations are initialized to 0, which is hardly ever the case with the real thing. I think there should be atlest an option for this in simulator whether to fill uninitialized location with random data in order to get a closer approximation of the real world.

I remember once having defined an uninitialized image_buffer[] in XDATA which filled the LCD with all sort of garbage.

I think we must try to understand why a piece of code works as equally as why it does not (not forget to mention not relying too much on the simulator!)

The reason is clear now: strcat() or strcpy() in real hardware did't stop copying after buffer[1].

I think the falsifying debugger was at fault too! (okay, I'm complaining :-) )