Andy Piper <andy(a)xemacs.org&gt; writes: > (this is 0 for normal scrolling). I have then modified window_scroll() so > that if you are trying to scroll an increment < (height of the top line) / > height of the default face then this offset will get incremented by an > amount == height of the default face and point is left alone. I really should have a look at your patch...Sigh.. What does this do for the line_start_cache? My current opinion about the some of the assert(cur_elt>=0) crashes that particular assert makes unwarranted assumptions and that your change will increase the number of cases of this. That isn't a problem with your patch but rather with the line_start_cache, it contains loads of undocumented assumptions. However, Hrvoje forgot to ask the obvious question: What does (window-start) give in the case of a clipped first line?