bug#47244: 28.0.50; SIGSEGV in long-runnning Emacs

[martin rudalics]

> I had removed it, sorry.  I'll put it in for the next time.

Don't.  I completely forgot that we do replace_buffer_in_windows_safely
when that fails which even recreates *scratch* when no other buffer is
found.

> I think so.  Both the *Server* window and the *Group* window will, under
> normal circumstances, be the only window on the frame.
>
>> p XWINDOW (selected_window)->parent
>> p XWINDOW (selected_window)->prev
>> p XWINDOW (selected_window)->next
>> p XWINDOW (selected_window)->top_line
>> p XWINDOW (selected_window)->pixel_height
>>
>> print?
>
> (gdb) p XWINDOW (selected_window)->parent
> $15 = XIL(0)
> (gdb) p XWINDOW (selected_window)->prev
> $16 = XIL(0)
> (gdb) p XWINDOW (selected_window)->next
> $17 = XIL(0x555556e96635)
> (gdb) p XWINDOW (selected_window)->top_line
> $18 = 0
> (gdb) p XWINDOW (selected_window)->pixel_height
> $19 = 995

These confirm what you say above.  Maybe we get that dead buffer via
`set-window-configuration'?  Could

          else if (!NILP (w->start))
            /* Leaf window has no live buffer, get one.  */

have failed?  Can a "live window" have w->start equal nil?

martin