bug#63967: 28.2; switch-to-buffer in normal window fails if minibuffer window is active

[Gregory Heytings]

> > I see.  Apparently we don't work the same way.  I didn't have time to 
> > investigate this issue further after bisecting, but the first thing I 
> > would have done is to determine which of the four calls to 
> > Fset_frame_selected_window in that changeset is responsible for that 
> > bug.
>
> You assume that the only way to change the selected-window on the C 
> level is by calling Fset_frame_selected_window?  That's false; just try 
> grepping the C sources for "selected_window =".  And that's even before 
> you consider the possibilities of indirect setting, when the actual 
> setting is in Lisp via some proxy.

I don't think it's useful to argue, I'm just observing that we don't work 
the same way.  But no, I don't assume anything, on the contrary.  Just 
commenting out each one of the Fset_frame_selected_window in that 
changeset in turn, and checking whether doing that fixes the bug, shows 
that the culprit is the call to Fset_frame_selected_window in 
minibuffer_unwind.  Doing that does not involve any guesswork, and the 
result is 100% accurate.  Of course, that's not the end of the story, and 
of course, it could have turned out that none of these four calls is the 
culprit (in which case I would probably have checked the five calls to 
set_window_buffer).