Re: input-pending-p after make-frame-visible

[Aaron Jensen]

On Wed, Oct 20, 2021 at 2:24 PM Eli Zaretskii <eliz@gnu.org> wrote:
>
> > From: Aaron Jensen <aaronjensen@gmail.com>
> > Date: Wed, 20 Oct 2021 13:47:57 -0400
> > Cc: martin rudalics <rudalics@gmx.at>, Alan Third <alan@idiocy.org>,
> >       Gregory Heytings <gregory@heytings.org>, emacs-devel@gnu.org,
> >       YAMAMOTO Mitsuharu <mituharu@math.s.chiba-u.ac.jp>
> >
> > > Yes.  But the change you propose is in readable_events, a function
> > > that has many more callers than just input-pending-p.  In particular,
> > > one of its callers is kbd_buffer_get_event, which is the API through
> > > which Emacs reads all of its input.  And now, under some
> > > circumstances, that API will ignore some events.  That's scary, at
> > > least for me.
> >
> > AFAICT kbd_buffer_get_event calls `readable_events (0)' only, which
> > will disregard any of the code I changed or added. The
> > READABLE_EVENTS_FILTER_EVENTS must be set for it to come into play.
> > Only input-pending-p sets that flag AFAICT.
>
> I don't see how this helps: who will remember that no caller of
> readable_events can ever use that flag without invoking this behavior
> whose justification we don't understand?

Which behavior are you saying the justification is not understood? If
it's the behavior I introduced, it should be understood as it is
understandable. It is there so that the flag
READABLE_EVENTS_FILTER_EVENTS can be actually respected instead of
partially respected.

Previously, that flag was only respected if USE_TOOLKIT_SCROLL_BARS
was set and the event was focus in or focus out. That was not the sum
total of events that we want to ignore, I believe. That is determined
by while-no-input-ignore-events.

> > > No, that must be a variable exposed to Lisp, so users could fiddle
> > > with it without rebuilding.
> >
> > Ah, I believe that variable exists already and can affect everything I
> > changed other than the reliance on USE_TOOLKIT_SCROLL_BARS:
> > while-no-input-ignore-events
>
> No, it must be a separate variable, because it's for people who do
> have non-nil while-no-input-ignore-events.

Ok, let's see if we can get to where we are comfortable without this first.