minibuffer and scroll-conservatively

minibuffer and scroll-conservatively

[Lars Hansen]

Start todays emacs with -Q and set scroll-conservatively to some
non-zero number, say 1.
Then type eg. M-x  ABC. Now the minibuffer shows "M-x ABC" as expected.
But if you type the Danish character A-ring (a capital A with a ring
above) everything seems to disappear from the minibuffer.
I am sorry to say that I don't know how to input A-ring without having a
Danish keyboard.
The reason that the minibuffer contents disappear must be some kind of
scrolling, because hitting left-arrow will make the minibuffer show the
expected contents, including the A-ring.
If you make the same experiment with scroll-conservatively set to 0, the
minibuffer contents does not disappear, but it scrolls down slightly
when A-ring is typed; that character must somehow be a bit to high.

The behaviour described above does not occur with emacs 21; here you
only get the small scroll even with a non-zero value of
scroll-conservatively.

Re: minibuffer and scroll-conservatively

[Gaëtan LEURENT]

Lars Hansen wrote on 21 Jun 2005 17:47:08 +0200:

> Start todays emacs with -Q and set scroll-conservatively to some
> non-zero number, say 1.
> Then type eg. M-x  ABC. Now the minibuffer shows "M-x ABC" as expected.
> But if you type the Danish character A-ring (a capital A with a ring
> above) everything seems to disappear from the minibuffer.
> I am sorry to say that I don't know how to input A-ring without having a
> Danish keyboard.

You can usually enter it with Compose-A-A in X11, and this indeed
triggers the bug.

-- 
Gaëtan LEURENT

Re: minibuffer and scroll-conservatively

[Richard M. Stallman]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 829 bytes --]

    Start todays emacs with -Q and set scroll-conservatively to some
    non-zero number, say 1.
    Then type eg. M-x  ABC. Now the minibuffer shows "M-x ABC" as expected.
    But if you type the Danish character A-ring (a capital A with a ring
    above) everything seems to disappear from the minibuffer.

I tried this under X, using the Emacs input method Latin-1-Postfix to
insert the Å, and the failure happened.  What occurred was that the
minibuffer contents all scrolled off the top of the window.

I think the reason for this is that the Å character is taller than
ordinary letters; taller, in fact, than the height of the window.
I think this confuses scrolling calculations so that they believe
that character doesn't fit on the screen while allowing point
to be visible.

Kim, could you investigate this and fix it?

Re: minibuffer and scroll-conservatively

[Kim F. Storm]

"Richard M. Stallman" <rms@gnu.org> writes:

>     Start todays emacs with -Q and set scroll-conservatively to some
>     non-zero number, say 1.
>     Then type eg. M-x  ABC. Now the minibuffer shows "M-x ABC" as expected.
>     But if you type the Danish character A-ring (a capital A with a ring
>     above) everything seems to disappear from the minibuffer.
>
> I tried this under X, using the Emacs input method Latin-1-Postfix to
> insert the Å, and the failure happened.  What occurred was that the
> minibuffer contents all scrolled off the top of the window.
>
> I think the reason for this is that the Å character is taller than
> ordinary letters; taller, in fact, than the height of the window.
> I think this confuses scrolling calculations so that they believe
> that character doesn't fit on the screen while allowing point
> to be visible.
>
> Kim, could you investigate this and fix it?

Can you try this patch, to see if the result is acceptable...

There are still problems: If I enter M-x ABÅjg, the result is not
perfect -- but what's the alternative?


*** xdisp.c	26 Jun 2005 23:25:32 +0200	1.1028
--- xdisp.c	27 Jun 2005 12:24:09 +0200	
***************
*** 11294,11300 ****
    window_height = window_box_height (w);
    if (row->height >= window_height)
      {
!       if (!force_p || w->vscroll)
  	return 1;
      }
    return 0;
--- 11294,11300 ----
    window_height = window_box_height (w);
    if (row->height >= window_height)
      {
!       if (!force_p || MINI_WINDOW_P (w) || w->vscroll)
  	return 1;
      }
    return 0;
***************
*** 12729,12735 ****
      }
  
    /* Don't let the cursor end in the scroll margins.  */
!   if (check_margins)
      {
        int this_scroll_margin, cursor_height;
  
--- 12729,12736 ----
      }
  
    /* Don't let the cursor end in the scroll margins.  */
!   if (check_margins
!       && !MINI_WINDOW_P (w))
      {
        int this_scroll_margin, cursor_height;

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: minibuffer and scroll-conservatively

[Richard M. Stallman]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 512 bytes --]

    There are still problems: If I enter M-x ABÅjg, the result is not
    perfect -- but what's the alternative?

Would it be better to lose some height at the top, rather than only at
the bottom?  Currently, I think, g and q cannot be distinguished in
the default font.

In the long run, perhaps the best solution is to allow some extra
space under the minibuffer for the sake of such situations.  But that
can't be done until the Emacs frame height no longer has to be an
integral multiple of the line height.

Re: minibuffer and scroll-conservatively

[Kim F. Storm]

"Richard M. Stallman" <rms@gnu.org> writes:

>     There are still problems: If I enter M-x ABÅjg, the result is not
>     perfect -- but what's the alternative?
>
> Would it be better to lose some height at the top, rather than only at
> the bottom?  Currently, I think, g and q cannot be distinguished in
> the default font.

Neither is A and Å if you remove at the top...

But does it matter that much?  The minibuffer is used for entering
some text, and in most cases, the user will know what he typed...

>
> In the long run, perhaps the best solution is to allow some extra
> space under the minibuffer for the sake of such situations.  But that
> can't be done until the Emacs frame height no longer has to be an
> integral multiple of the line height.

That is a good approach.

Alternatively, we do have code to automatically grow the minibuffer if
the text wraps, so perhaps it could be tweaked to also grow to show
the extra height too.  But I don't have time to work on that.

I suggtest that we install my (imperfect) patch, and add a note to
TODO, that we need to address this issue after the release.

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: minibuffer and scroll-conservatively

[Lars Hansen]

Kim F. Storm wrote:

>I suggtest that we install my (imperfect) patch, and add a note to
>TODO, that we need to address this issue after the release.
>  
>
I agree. In my test your patch makes CVS Emacs work like Emacs 21, so at
least there is no regression.

Re: minibuffer and scroll-conservatively

[Lars Hansen]

Kim F. Storm wrote:

> I suggtest that we install my (imperfect) patch, and add a note to
> TODO, that we need to address this issue after the release.

Has this been forgotten?

Re: minibuffer and scroll-conservatively

[Kim F. Storm]

Lars Hansen <larsh@soem.dk> writes:

> Kim F. Storm wrote:
>
>> I suggtest that we install my (imperfect) patch, and add a note to
>> TODO, that we need to address this issue after the release.
>
> Has this been forgotten?

Thanks for the reminder.  I have installed the change.
-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: minibuffer and scroll-conservatively

[Richard M. Stallman]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 621 bytes --]

    > Would it be better to lose some height at the top, rather than only at
    > the bottom?  Currently, I think, g and q cannot be distinguished in
    > the default font.

    Neither is A and Å if you remove at the top...

Maybe that depends HOW MUCH is removed at the top.  That's why I am
suggesting to try taking off a little at the top and a little at the bottom.
Maybe there's a combination that avoids both kinds of lossage, or maybe
there isn't.

    I suggtest that we install my (imperfect) patch, and add a note to
    TODO, that we need to address this issue after the release.

Would you please do that?

Re: minibuffer and scroll-conservatively

[Miles Bader]

On 6/28/05, Kim F. Storm <storm@cua.dk> wrote:
> I suggtest that we install my (imperfect) patch, and add a note to
> TODO, that we need to address this issue after the release.

A comment next to the affected code (with a pointer to TODO?) would
also be nice.

-Miles
-- 
Do not taunt Happy Fun Ball.