Re: bug in 23.2.92 with anything

[martin rudalics]

> IIUC, you are saying that there will be still additional problems left
> even after this change.

The selected_window nil hack assures that point swapping is inhibited
until selected_window is set to some other value.  Our change assures
that point swapping is inhibited only for the first invocation of
Fselect_window.  If Fselect_window is called back via Fselect_frame,
point _is_ swapped.  This might even be TRT, but it does not preserve
the current behavior.

> If so, I might agree, but what we are trying
> to do here is to avoid the situation whereby selected_window is nil,
> because that bombs inside window_text_bottom_y.  Using the parameter
> allows us to communicate to select-window that it should not swap out
> point, without setting selected_window to nil.  Would you agree that
> this part of the problem is fixed by the change you posted?  If not,
> please explain why not.

If selected_window is no more set to nil, the crash in its present form
will disappear.  The question is whether the change will change the
current behavior in the non-crashing case.  Using a global variable, one
can emulate the current behavior without running into the crash.

martin