bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

[Mark H Weaver]

Emacs 26.1 has introduced a regression compared with 25.3 in the
handling of 'comint-scroll-show-maximum-output', which I prefer to set
to nil.  Here's how to reproduce the problem:

* Run "emacs-26.1 -Q", with or without "-nw" as you prefer.
* M-x shell
* M-: (setq comint-scroll-show-maximum-output nil)
* Press RET repeatedly until the window starts scrolling.

Observe that after the point reaches the bottom of the window, it is
subsequently kept at the bottom of the window by aggressively scrolling
one line at a time.

If you perform the same steps above with emacs-25.3, you'll see that it
always recenters, so that scrolling occurs only once for every ~N/2
lines of output, where N is the window height.

Note that (setq comint-scroll-show-maximum-output nil) still has some
effect.  After performing the steps above, try this:

* C-l
* Press RET repeatedly until the window starts scrolling.
* M-: (setq comint-scroll-show-maximum-output t)
* C-l
* Press RET just once.

Observe that with 'comint-scroll-show-maximum-output' set to t, the
first RET after C-l immediately scrolls the window to show the maximum
output.

If you can suggest a workaround or a patch to fix this issue, I would be
grateful.

Thanks very much for your work on Emacs.

     Mark

bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

[Eli Zaretskii]

On May 30, 2018 8:56:13 AM GMT+03:00, Mark H Weaver <mhw@netris.org> wrote:
> Emacs 26.1 has introduced a regression compared with 25.3 in the
> handling of 'comint-scroll-show-maximum-output', which I prefer to set
> to nil.  Here's how to reproduce the problem:
> 
> * Run "emacs-26.1 -Q", with or without "-nw" as you prefer.
> * M-x shell
> * M-: (setq comint-scroll-show-maximum-output nil)
> * Press RET repeatedly until the window starts scrolling.
> 
> Observe that after the point reaches the bottom of the window, it is
> subsequently kept at the bottom of the window by aggressively
> scrolling
> one line at a time.
> 
> If you perform the same steps above with emacs-25.3, you'll see that
> it
> always recenters, so that scrolling occurs only once for every ~N/2
> lines of output, where N is the window height.
> 
> Note that (setq comint-scroll-show-maximum-output nil) still has some
> effect.  After performing the steps above, try this:
> 
> * C-l
> * Press RET repeatedly until the window starts scrolling.
> * M-: (setq comint-scroll-show-maximum-output t)
> * C-l
> * Press RET just once.
> 
> Observe that with 'comint-scroll-show-maximum-output' set to t, the
> first RET after C-l immediately scrolls the window to show the maximum
> output.
> 
> If you can suggest a workaround or a patch to fix this issue, I would
> be
> grateful.
> 
> Thanks very much for your work on Emacs.
> 
>      Mark

It's a deliberate change, by popular demand: shell-mode now sets
scroll-conservatively to a large value locally in shell buffers.  Reset
that in your mode hook, and Bob's your uncle.

If you are interested in the reasons for the change, see the log
message of commit d7ac7b15ca9, it points to the relevant discussion.

(This should have been in NEWS, sigh...)

bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

[Mark H Weaver]

Hi Eli,

Eli Zaretskii <eliz@gnu.org> writes:
> It's a deliberate change, by popular demand: shell-mode now sets
> scroll-conservatively to a large value locally in shell buffers.  Reset
> that in your mode hook, and Bob's your uncle.

That's very helpful, thank you!

> If you are interested in the reasons for the change, see the log
> message of commit d7ac7b15ca9, it points to the relevant discussion.

Thanks for the pointer.  I acknowledge that I'm in the minority here,
but I greatly prefer less frequent scrolling, so that I have more time
to visually process the text before it moves and disrupts my ability to
process it.

Ideally, both settings of 'comint-scroll-show-maximum-output' would
behave as expected, without mode hooks.

That said, I don't have time to work on this myself, and it's certainly
of very little importance.  The mode hook workaround is good enough for
me, and I guess there are very few people with my preferences, so I'm
willing to let this go.  Feel free to close this bug as "wontfix", or to
keep it open, as you prefer.

Thanks again,

     Mark

bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

[Andreas Schwab]

On Mai 30 2018, Mark H Weaver <mhw@netris.org> wrote:

> Observe that after the point reaches the bottom of the window, it is
> subsequently kept at the bottom of the window by aggressively scrolling
> one line at a time.

What is the value of scroll-conservatively?

Andreas.

-- 
Andreas Schwab, SUSE Labs, schwab@suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE  1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."

bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

[Eli Zaretskii]

tags 31653 notabug
thanks

> From: Mark H Weaver <mhw@netris.org>
> Cc: 31653@debbugs.gnu.org
> Date: Wed, 30 May 2018 03:56:47 -0400
> 
> That said, I don't have time to work on this myself, and it's certainly
> of very little importance.  The mode hook workaround is good enough for
> me, and I guess there are very few people with my preferences, so I'm
> willing to let this go.  Feel free to close this bug as "wontfix", or to
> keep it open, as you prefer.

Done.

bug#31653: 26.1; comint-scroll-show-maximum-output nil regression

[Eli Zaretskii]

> Date: Wed, 30 May 2018 09:56:49 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> 
> (This should have been in NEWS, sigh...)

Now done.