Philipp Stephani
schrieb am Fr., 9. Sep. 2016 um
18:29 Uhr:
> Philipp Stephani schrieb am Fr., 9. Sep. 2016 um
> 18:20 Uhr:
>
> Eli Zaretskii schrieb am Fr., 9. Sep. 2016 um 18:07 Uhr:
>
> > From: Philipp Stephani
> > Date: Fri, 09 Sep 2016 15:59:02 +0000
> > Cc: 24372@debbugs.gnu.org
> >
> > Eli Zaretskii schrieb am Di., 6. Sep. 2016 um 18:03 Uhr:
> >
> > > From: Philipp Stephani
> > > Date: Mon, 05 Sep 2016 21:16:40 +0200
> > >
> > > emacs -Q -eval '(setq blink-cursor-delay 0.0)'
> > >
> > > Move point around in the scratch buffer (e.g. press C-b a couple of
> > > times): the cursor stays visible, as it should be. Then put the mouse
> > > focus on a different GTK window (not Emacs window), put the mouse
> focus
> > > back on Emacs, and move point again: the cursor is hidden, making it
> > > impossible to see until you stop moving.
> >
> > Does it happen even if you wait with cursor motion until after the
> > cursor blinks one time, i.e. if you start moving point with the cursor
> > already visible after Emacs gets focus?
> >
> > Yes. No matter what state the cursor is in and how often it has already
> blinked, it becomes invisible when
> > moving.
>
> So what is the importance of moving focus out of the Emacs frame and
> then back into it? Is the problem reproducible without that? Or are
> you saying that focus-out followed by focus-in event somehow changes
> the behavior wrt displaying the cursor?
>
>
> Yes. The cursor only become invisible after a focus-out/focus-in event.
> This isn't surprising given that blink-cursor-mode changes focus-in-hook
> and focus-out-hook. Probably the bug is hidden somewhere in the complex
> interaction between the various blink-cursor timers and hooks.
>
>
> A simpler recipe that doesn't need explicit focus events is
>
> emacs -Q -eval '(progn (setq blink-cursor-delay 0.0)
> (blink-cursor-suspend) (blink-cursor-check))'
>
> and then start moving point.
>
OK, I guess one issue is that setting blink-cursor-delay doesn't restart
blink-cursor-idle-timer. (Similarly, changing blink-cursor-interval doesn't
restart blink-cursor-timer.) While obviously we can't fix that when using
setq, I'd suggest adding custom setters to the variables nevertheless.
The direct cause of the issue seems to be that, when blink-cursor-delay is
idle, after every command blink-cursor-start is called immediately, which
hides the cursor. I guess that is so that in the default case where
blink-cursor-delay = blink-cursor-interval the cursor frequency is
constant. However, this causes the cursor to be hidden very quickly when
blink-cursor-delay < blink-cursor-interval. Maybe in that case
blink-cursor-start should show instead of hide the cursor, and
run-with-timer should be called with an argument of (blink-cursor-interval
- blink-cursor-delay), so that the cursor is at least shown for
blink-cursor-interval. WDYT?