Re: shell can't stand still

[Peter Dyballa]

Am 20.03.2005 um 01:21 schrieb Richard Stallman:

>     When in the *shell* buffer a running command produces output it's 
> quite
>     impossible to scroll back using the page-up key. It seems as if the
>     last output lines are being subtracted from the amount of lines to
>     scroll up ...
>
>     In GNU Emacs 22.0.50.1 (powerpc-apple-darwin7.8.0, GTK+ Version 
> 2.4.9)
>       of 2005-03-16 on Latsche.local
>
> That really surprises me.  Once point is no longer at eob, it should
> stop moving.
>
> Can you provide a precise test case, starting from emacs -Q?

Yes, here is an easy example: start CVS Emacs with -Q, create a shell, 
and make a new Emacs from CVS!

I did so. I waited till make's output hit the bottom and automatic 
scrolling started since the cursor was kind of driven by the text 
output and then I scrolled back with <prior> which is bound to 
scroll-down. Text was scrolled down and the text cursor was placed on 
the first (top-most) line in the first column. Then the text was by 
some automatic means scrolled up, which made me scroll it down again 
and so on. I could see that I came back to the place I left since it 
had short lines

        xrdb.c:92:1: warning: "malloc" redefined
        In file included from config.h:941,
                         from xrdb.c:24:
	s/darwin.h:334:1: warning: this is the location of the previous 
definition
        xrdb.c:93:1: warning: "realloc" redefined
	s/darwin.h:335:1: warning: this is the location of the previous 
definition
        xrdb.c:94:1: warning: "free" redefined
	s/darwin.h:336:1: warning: this is the location of the previous 
definition
	gcc -I/sw/include -L/sw/lib -c -fpascal-strings -fno-common -DMAC_OSX 
-I../mac/src  -I/sw/include -Demacs -DHAVE_CONFIG_H -DUSE

It was a ping-pong game between two positions. I too scrolled text up 
by pressing <next> and the same effect happened: the text was scrolled 
back, down. Repeatedly. Some of these scroll effects were a bit short 
so that the region of short lines changed its position in the buffer 
upwards.

And there is another anomaly (close to a bug): I wanted by means of the 
mouse (a trackpad actually, that is made by a software to work like a 
mouse wheel when I hold down a key) to mark that region of short lines. 
It did not work, because this marked up region scrolled up and down by 
some single lines and it was quite hard to catch these movements. I 
gave up when I only had one long line marked.

Something similiar to this happens when I mark a region and now come 
close to the upper or lower ends of the buffer. Then the text scrolls 
away, but much to fast so that I can't make a reasonable selection.

In usual Aqua applications of Mac OS X the scrolling effect is not as 
big as wish, while in X it often seems to exaggerate. xset tells me 
this:

Pointer Control:
  acceleration:  2/1    threshold:  4

which I can't classify since I don't remember what I else had ...


Since the compilation just ended I scrolled with my trackpad to the 
end. At about 60% from top was the shell prompt. When I hit RET the 
text 'jumped' to the bottom of this buffer. Is this normal? To me it 
makes no sense to position the new shell prompt at the bottom -- some 
text output from the launched application could happen and I wouldn't 
see it!

--
Greetings

  Pete

In a world without walls and fences, who needs gates and windows?