scrolling in shell buffer

scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

I have (setq scroll-conservatively 6).
I'm running a job in a shell buffer that produces a line of output
a couple of times a minute, with point at the end.

As the output appears, the buffer scrolls so that the cursor is always
at the bottom.  That is good.

If I switch to another buffer, then switch back after a minute or two,
it scrolls up for the new output but the cursor stays at the bottom.
That is good.

But when I switch to another buffer, wait several minutes, then switch
back, so that many additional lines of output have been generated, it
puts point in the middle of the window.

That's good for ordinary buffers, but it isn't good for a shell
buffer.  For a shell buffer, as long as point is at the end, display
should normally scroll point to the bottom of the window.

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> Date: Sun, 18 Dec 2016 08:22:21 -0500
> 
> But when I switch to another buffer, wait several minutes, then switch
> back, so that many additional lines of output have been generated, it
> puts point in the middle of the window.
> 
> That's good for ordinary buffers, but it isn't good for a shell
> buffer.  For a shell buffer, as long as point is at the end, display
> should normally scroll point to the bottom of the window.

AFAIK, the only way redisplay guarantees that point will always be at
the bottom of the window is if you set scroll-conservatively to a
value larger than 100.  Otherwise, sooner or later you will see
recentering under some conditions.

If you want this behavior for shell buffers, I suggest to customize
comint-move-point-for-output and/or comint-scroll-to-bottom-on-input.

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > AFAIK, the only way redisplay guarantees that point will always be at
  > the bottom of the window is if you set scroll-conservatively to a
  > value larger than 100.  Otherwise, sooner or later you will see
  > recentering under some conditions.

  > If you want this behavior for shell buffers, I suggest to customize
  > comint-move-point-for-output and/or comint-scroll-to-bottom-on-input.

comint-move-point-for-output is not pertinent.
It controls moving point.  This is not about moving point.

comint-scroll-to-bottom-on-input is not pertinent either.
It controls what happens on input.  This is not about input.

Customizing is not pertinent, since this issue I'm raising
is about the proper default.

I think that we should make the default behavior for shell buffers,
when point is at the end, to scroll to keep point at the bottom of the
window.

Does anyone disagree that this behavior would be right?

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Mon, 19 Dec 2016 08:07:16 -0500
> 
> Customizing is not pertinent, since this issue I'm raising
> is about the proper default.
> 
> I think that we should make the default behavior for shell buffers,
> when point is at the end, to scroll to keep point at the bottom of the
> window.
> 
> Does anyone disagree that this behavior would be right?

By default?  I don't think I agree.  It's just another Emacs buffer.
Also, various terminal emulators on GUI systems let me scroll the
console window past its end, and don't scroll it back each time the
shell or some command run from the shell outputs something.

Re: scrolling in shell buffer

[Johan Bockgård]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Richard Stallman <rms@gnu.org>
>> CC: emacs-devel@gnu.org
>> Date: Mon, 19 Dec 2016 08:07:16 -0500
>> 
>> Customizing is not pertinent, since this issue I'm raising
>> is about the proper default.
>> 
>> I think that we should make the default behavior for shell buffers,
>> when point is at the end, to scroll to keep point at the bottom of the
>> window.
>> 
>> Does anyone disagree that this behavior would be right?
>
> By default?  I don't think I agree.  It's just another Emacs buffer.
> Also, various terminal emulators on GUI systems let me scroll the
> console window past its end, and don't scroll it back each time the
> shell or some command run from the shell outputs something.

? It is already the default, and it's called
`comint-scroll-show-maximum-output'.

However, we also find this comment in `comint-mode':

    ;; Following disabled because it seems to break the case when
    ;; comint-scroll-show-maximum-output is nil, and no-one can remember
    ;; what the original problem was.  If there are problems with point
    ;; not going to the end, consider re-enabling this.
    ;; http://lists.gnu.org/archive/html/emacs-devel/2007-08/msg00827.html
    ;;
    ;; This makes it really work to keep point at the bottom.
    ;; (make-local-variable 'scroll-conservatively)
    ;; (setq scroll-conservatively 10000)

The last two lines were added by rms in 2002, and commented out by Glenn
in 2007.

Re: scrolling in shell buffer

[John Yates]

[-- Attachment #1: Type: text/plain, Size: 356 bytes --]

On Mon, Dec 19, 2016 at 8:07 AM, Richard Stallman <rms@gnu.org> wrote:

> I think that we should make the default behavior for shell buffers,
> when point is at the end, to scroll to keep point at the bottom of the
> window.
>
> Does anyone disagree that this behavior would be right?
>

​That mimics every "glass tty" I ever ​used.

/john

[-- Attachment #2: Type: text/html, Size: 1135 bytes --]

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > > I think that we should make the default behavior for shell buffers,
  > > when point is at the end, to scroll to keep point at the bottom of the
  > > window.
  > > 
  > > Does anyone disagree that this behavior would be right?

  > By default?  I don't think I agree.  It's just another Emacs buffer.

It is meant to resemble a terminal.  I think it should treat this case
like the other terminals we use.  The Linux console scrolls the cursor
to the bottom line, and so does the GNOME terminal.

You can scroll back through the text, but there no use for scrolling forward
and pusting the text out of the screen.

  > Also, various terminal emulators on GUI systems let me scroll the
  > console window past its end, and don't scroll it back each time the
  > shell or some command run from the shell outputs something.

Can you tell me the names of some that do this, and how I can see an
example for myself?

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > ? It is already the default, and it's called
  > `comint-scroll-show-maximum-output'.

That operates when the subprocess generates output.

The point is that just switching to the buffer
doesn't obey `comint-scroll-show-maximum-output'.
That scrolling happens the next time output arrives.

I think that is a bug, and should be fixed.

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[raman]

Richard Stallman <rms@gnu.org> writes:


I agree with RMS here. And even if there are GUI terminals that do the
wrong thing, it doesn't mean that Emacs needs to race them to the bottom:-)> [[[ To any NSA and FBI agents reading my email: please consider    ]]]
> [[[ whether defending the US Constitution against all enemies,     ]]]
> [[[ foreign or domestic, requires you to follow Snowden's example. ]]]
>
>   > > I think that we should make the default behavior for shell buffers,
>   > > when point is at the end, to scroll to keep point at the bottom of the
>   > > window.
>   > > 
>   > > Does anyone disagree that this behavior would be right?
>
>   > By default?  I don't think I agree.  It's just another Emacs buffer.
>
> It is meant to resemble a terminal.  I think it should treat this case
> like the other terminals we use.  The Linux console scrolls the cursor
> to the bottom line, and so does the GNOME terminal.
>
> You can scroll back through the text, but there no use for scrolling forward
> and pusting the text out of the screen.
>
>   > Also, various terminal emulators on GUI systems let me scroll the
>   > console window past its end, and don't scroll it back each time the
>   > shell or some command run from the shell outputs something.
>
> Can you tell me the names of some that do this, and how I can see an
> example for myself?

-- 

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Tue, 20 Dec 2016 13:15:13 -0500
> 
>   > Also, various terminal emulators on GUI systems let me scroll the
>   > console window past its end, and don't scroll it back each time the
>   > shell or some command run from the shell outputs something.
> 
> Can you tell me the names of some that do this, and how I can see an
> example for myself?

I stand corrected.  I wrote what I wrote because I had a vivid mental
picture of that in my mind, but after looking around, I see that no
terminal emulator behaves like that.  I guess I was thinking about
Emacs buffers, where we indeed can scroll forward past the end of the
buffer using the scroll bar.

Sorry.

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > I stand corrected.  I wrote what I wrote because I had a vivid mental
  > picture of that in my mind, but after looking around, I see that no
  > terminal emulator behaves like that.

So I think we are all agreed that the change I proposed would be good.
Would someone like to implement it?

It would be sufficient to add a feature, enabled by some global
variable, to scroll the buffer (to put point on the last line if it's
at the end of the buffer) when the buffer is redisplayed for the first
time after being put in any given window with set-window-buffer.

set-window-buffer could set a flag in the window so that the next
redisplay of that window will do this scrolling if the buffer requests
it.

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Fri, 23 Dec 2016 11:11:49 -0500
> 
> So I think we are all agreed that the change I proposed would be good.
> Would someone like to implement it?
> 
> It would be sufficient to add a feature, enabled by some global
> variable, to scroll the buffer (to put point on the last line if it's
> at the end of the buffer) when the buffer is redisplayed for the first
> time after being put in any given window with set-window-buffer.
> 
> set-window-buffer could set a flag in the window so that the next
> redisplay of that window will do this scrolling if the buffer requests
> it.

If you set scroll-conservatively to a number larger than 100, locally
in the shell buffer, don't you get the behavior you want?  If you do,
then all we need is to provide such a local binding in shell buffers.

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > If you set scroll-conservatively to a number larger than 100, locally
  > in the shell buffer, don't you get the behavior you want?

No.  scroll-conservatively has no effect on the first redisplay after
I switch to the shell buffer.

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Sat, 24 Dec 2016 13:53:09 -0500
> 
>   > If you set scroll-conservatively to a number larger than 100, locally
>   > in the shell buffer, don't you get the behavior you want?
> 
> No.  scroll-conservatively has no effect on the first redisplay after
> I switch to the shell buffer.

Strange.  It did for me when I tried that.  I wrote a shell script
that produced a line of output every several seconds, then ran it in a
shell buffer, and switched away.  When I switched back to the shell
buffer, some seconds later, the buffer was full of lines and point was
shown on the last line.

Is it possible that you tried that when the buffer had less than
windowful of text?

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > Strange.  It did for me when I tried that.  I wrote a shell script
  > that produced a line of output every several seconds, then ran it in a
  > shell buffer, and switched away.  When I switched back to the shell
  > buffer, some seconds later, the buffer was full of lines and point was
  > shown on the last line.

It didn't work for me.  I ran a simple shell program

  while true; do sleep 15; echo foo; done

and when I switched back to the *shell* buffer it displayed
point (which was at eob) at mid-screen.

Then the next line came out and it scrolled to put point at the bottom.

I verified that scroll-conservatively was 100.

  > Is it possible that you tried that when the buffer had less than
  > windowful of text?

Just now, the line number at the end was over 29000.



-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Sun, 25 Dec 2016 15:44:28 -0500
> 
> I verified that scroll-conservatively was 100.

Please repeat the experiment with scroll-conservatively set to 101 or
greater.  It should be greater than 100 for this feature to kick in.

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > > I verified that scroll-conservatively was 100.

  > Please repeat the experiment with scroll-conservatively set to 101 or
  > greater.  It should be greater than 100 for this feature to kick in.

I did this with 101, and the problem is mostly gone.  Most of the time,
when I switch to that buffer, there is no empty space at the end.
However, once there was empty space.  So far I have not been able to
make it do that a second time.

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

If setting scroll-conservatively to 101 is the fix to make shell buffers
keep the last line at the end of the window, would someone please
install a fix to make comint locally do this?

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: scrolling in shell buffer

[Eli Zaretskii]

> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> Date: Mon, 23 Jan 2017 16:04:14 -0500
> 
> If setting scroll-conservatively to 101 is the fix to make shell buffers
> keep the last line at the end of the window, would someone please
> install a fix to make comint locally do this?

Done.