bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Drew Adams]

At least on MS Windows, when you scroll down, you keep scrolling past
the end of the buffer, without (noticeable) limit.
 
I cannot believe that anyone would consider this a "feature".
It is in fact a regression wrt Emacs 22 (21, 20,...).
 

 

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Date: Mon, 15 Jun 2009 09:34:55 -0700
> Cc: 
> 
> At least on MS Windows, when you scroll down, you keep scrolling past
> the end of the buffer, without (noticeable) limit.

Scroll how?  Please give a complete recipe, starting with "emacs -Q".
FWIW, I cannot reproduce this on MS-Windows using the latest pretest.

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Drew Adams]

> > At least on MS Windows, when you scroll down, you keep 
> > scrolling past the end of the buffer, without (noticeable) limit.
> 
> Scroll how?  Please give a complete recipe, starting with "emacs -Q".
> FWIW, I cannot reproduce this on MS-Windows using the latest pretest.

Sorry, I meant scroll by dragging (downward) the scroll-bar handle with the
mouse (e.g. mouse-1).

But I see now that scrolling does *not* continue forever. It does continue past
eob, sometimes as much as a whole blank screen, but it stops at some point.

Point does not move past eob, of course; it is just the window that scrolls.

emacs -Q
grab vertical scroll bar with mouse-1 and drag down

Buffers with different contents have different amounts of blank overshoot scroll
(i.e. past eob). It seems that the max overshoot is a full blank screen. It also
seems that larger buffers have more overshoot. So to see this with a full blank
window of overshoot, try scrolling a large buffer.

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Date: Mon, 15 Jun 2009 10:44:40 -0700
> 
> emacs -Q
> grab vertical scroll bar with mouse-1 and drag down
> 
> Buffers with different contents have different amounts of blank overshoot scroll
> (i.e. past eob). It seems that the max overshoot is a full blank screen. It also
> seems that larger buffers have more overshoot. So to see this with a full blank
> window of overshoot, try scrolling a large buffer.

Still cannot reproduce this.  The worst case I see is that the very
last line of the buffer ends up on the top visible line of the window.

If that's what you see, it's normal, and Emacs 22.x and 21.x behave
exactly the same.

If that's not what you see, can you pick up some of the files in the
Emacs distribution, and describe what you see scrolling each one of
them?  Please define and describe the ``overshoot'' more precisely.

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Drew Adams]

> Still cannot reproduce this.  The worst case I see is that the very
> last line of the buffer ends up on the top visible line of the window.
> 
> If that's what you see, it's normal, and Emacs 22.x and 21.x behave
> exactly the same.

No, that's not what I see. The last line can be scrolled completely off of the
screen, so all that can be seen is blank space.

And yes, I do see the same thing in Emacs 22.3. But *not* in Emacs 21.3 or Emacs
20.7 (-q for all three).

Even if what you say were the case, no, that's wouldn't be normal.

> If that's not what you see, can you pick up some of the files in the
> Emacs distribution, and describe what you see scrolling each one of
> them?  Please define and describe the ``overshoot'' more precisely.

emacs -Q
C-x C-f /path/to/allout.el
C-x 1
Scroll down using mouse on scroll-bar slider. You can scroll so that no text is
visible, all text is above the (single) window. Same thing even if there are
other windows above or below.

Same with any other large file.

Try it with a small file, autoarg.el. The overshoot possible is only about 6
lines in that case. The larger the file, the larger the overshoot, with a full
blank window maximum (last line just off the window).

Clear enough? Do you see something different from this?

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Cc: <3576@emacsbugs.donarmstrong.com>
> Date: Mon, 15 Jun 2009 13:04:59 -0700
> 
> emacs -Q
> C-x C-f /path/to/allout.el
> C-x 1
> Scroll down using mouse on scroll-bar slider. You can scroll so that no text is
> visible, all text is above the (single) window. Same thing even if there are
> other windows above or below.

Yep.  Emacs 21.4 stops on the last text line, Emacs 22.3 and 23.0.94
stop on EOB.

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Drew Adams]

> > emacs -Q
> > C-x C-f /path/to/allout.el
> > C-x 1
> > Scroll down using mouse on scroll-bar slider. You can 
> > scroll so that no text is visible, all text is above the
> > (single) window. Same thing even if there are
> > other windows above or below.
> 
> Yep.  Emacs 21.4 stops on the last text line, Emacs 22.3 and 23.0.94
> stop on EOB.

What do you mean by "stop on EOB"? If end of the buffer is the newline
immediately following the last line of non-newline chars, then how can showing
up to a screenful of blank vertical space be considered "stopping at EOB"?

Do you mean that it decides that the final newline has as a right to be scrolled
to the top of the screen? Is that what "stopping at eob" means?

If this is just an unfortunate result of the way things happen to be currently
implemented, that's one thing. But I find it hard to believe that this would
have been a design goal - that someone would start out intentionally to produce
this behavior. What's the advantage or use case?

I've never seen this behavior in any other app (though I'm sure you'll come up
with some other app that has the same behavior). ;-)

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Cc: <3576@emacsbugs.donarmstrong.com>
> Date: Mon, 15 Jun 2009 20:10:27 -0700
> 
> > Yep.  Emacs 21.4 stops on the last text line, Emacs 22.3 and 23.0.94
> > stop on EOB.
> 
> What do you mean by "stop on EOB"? If end of the buffer is the newline
> immediately following the last line of non-newline chars, then how can showing
> up to a screenful of blank vertical space be considered "stopping at EOB"?

EOB == (point-max)

Emacs 21.4 stops when the last line of the buffer is at the topmost
line of the window.  Emacs 22 and 23 put EOB on the topmost line of
the window.

> Do you mean that it decides that the final newline has as a right to be scrolled
> to the top of the screen? Is that what "stopping at eob" means?
> 
> If this is just an unfortunate result of the way things happen to be currently
> implemented, that's one thing. But I find it hard to believe that this would
> have been a design goal - that someone would start out intentionally to produce
> this behavior. What's the advantage or use case?
> 
> I've never seen this behavior in any other app (though I'm sure you'll come up
> with some other app that has the same behavior). ;-)

I was just stating the fact, not embracing it.

Now, can you please stop shouting at me?  My only sin is that I
confirmed your report.

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Drew Adams]

> > > Yep.  Emacs 21.4 stops on the last text line, Emacs 22.3 
> > > and 23.0.94 stop on EOB.
> > 
> > What do you mean by "stop on EOB"? If end of the buffer is 
> > the newline immediately following the last line of non-newline
> > chars, then how can showing up to a screenful of blank vertical
> > space be considered "stopping at EOB"?
> 
> EOB == (point-max)
> 
> Emacs 21.4 stops when the last line of the buffer is at the topmost
> line of the window.  Emacs 22 and 23 put EOB on the topmost line of
> the window.

Emacs 22/23 does not put the last non-whitespace line at the top of the window.
It apparently puts (point-max) at the top, which in most cases (e.g. Lisp
libraries) is a line with only a newline character.

Why should they put the last line (whitespace or not) at the _top_ of the
window, instead of the bottom of the window? Why would someone want to see an
extra screen of blank space, with the final newline at the top of the window?
That just doesn't seem very user friendly.

Do you know of a use case for that, which might justify that as the standard
behavior?

If not, I propose that this be fixed by returning to the Emacs 20/21 behavior.

> > Do you mean that it decides that the final newline has as a 
> > right to be scrolled to the top of the screen? Is that what
> > "stopping at eob" means?
> > 
> > If this is just an unfortunate result of the way things 
> > happen to be currently implemented, that's one thing.
> > But I find it hard to believe that this would have been a
> > design goal - that someone would start out intentionally to
> > produce this behavior. What's the advantage or use case?
> > 
> > I've never seen this behavior in any other app (though I'm 
> > sure you'll come up with some other app that has the same
> > behavior). ;-)
> 
> I was just stating the fact, not embracing it.
> 
> Now, can you please stop shouting at me?  My only sin is that I
> confirmed your report.

No one has shouted at you. Don't be so defensive.

This is not about you, at all. It's merely a bug report. Thank you for
confirming it.

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Stefan Monnier]

> Why should they put the last line (whitespace or not) at the _top_ of the
> window, instead of the bottom of the window?

That's the wrong question.  The right question could be something like:

- Why should Emacs (or any other application ofr that matter) prevent
  you from displaying the last line anywhere else than the bottom of the
  window, what would be gained from it?
- Why does Emacs only let you display the first line of the buffer at
  the top of the window? [well, actually, if you play enough with
  overlays and before-strings you could move the first line further
  down, but that's cheating]
- Why hasn't anybody written a patch (after all these years of acrymony)
  so that Emacs can mimick the brain-dead dogmatic behavior of most
  other GUI apps's drag scroll?

> Do you know of a use case for that,

Of course: move the last line so it sits right next to some other line
in some other window.  I often do such things to compare two windows
line-by-line (it doesn't always involve the last line of one of the two
buffers, but sometimes does).

> which might justify that as the standard behavior?

The real justification is that it's difficult to implement and the
benefits are dubious.


        Stefan

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Lars Magne Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> That's the wrong question.  The right question could be something like:
>
> - Why should Emacs (or any other application ofr that matter) prevent
>   you from displaying the last line anywhere else than the bottom of the
>   window, what would be gained from it?

So this doesn't seem like a bug, so I'm closing the report.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

bug#3576: 23.0.94; scroll bar scrolls past eob - keeps scrolling

[Drew Adams]

> So this doesn't seem like a bug, so I'm closing the report.

Stefan's "real justification" was that the right behavior is "difficult to
implement".  That does not mean that this is not a bug.

Besides, it cannot be that difficult, since this is a _regression_ from the
fixed behavior in Emacs 18, 19, 20, and 21.

Please close it as "won't fix", if that's the decision, but not as "not a bug".