Re: set-window-start in Elisp manual

[Richard Stallman]

I think what's happening is that, if window-start isn't at the
beginning of a line, compute_window_start_on_continuation_line
assumes that means it's a continuation line.  In the normal
course of events, that's a valid heuristic.  It's only with code
like the example from the manual that it isn't valid.

So I think we should change the manual.

However, at the same time it might be ok to make the conditions
for calling compute_window_start_on_continuation_line narrower.
Maybe it doesn't need to be called so often.