[kzeitler@lucent.com: Re: Crash with mouse-avoidance-mode]

[kzeitler@lucent.com: Re: Crash with mouse-avoidance-mode]

[Richard Stallman]

Would someone please investigate this and DTRT?

------- Start of forwarded message -------
To: rms@gnu.org
Subject: Re: Crash with mouse-avoidance-mode
From: Klaus Zeitler <kzeitler@lucent.com>
Date: Tue, 08 Aug 2006 11:49:29 +0200
In-Reply-To: <E1GAIpZ-0003dk-PE@fencepost.gnu.org> (Richard Stallman's message of "Tue, 08 Aug 2006 00:04:37 -0400")
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
X-Spam-Status: No, score=0.0 required=5.0 tests=none autolearn=failed 
	version=3.0.4

>>>>> "Richard" == Richard Stallman <rms@gnu.org> writes:
    Richard> 
    Richard> Can you tell us a precise recipe for reproducing this?

Nearly precise :-).

1. emacs -Q
2. (setq mouse-avoidance-animation-delay 0.1)
3. M-x mouse-avoidance-mode proteus
4. Move cursor carefully/slowly close to mouse pointer so that avoidance-mode
   will try to move the pointer away. I need only about 3 or 4 tries till the
   animation goes haywire and the mouse pointer starts jumping around


- -- 
 ------------------------------------------
|  Klaus Zeitler      Lucent Technologies  |
|  Email:             kzeitler@lucent.com  |
 ------------------------------------------
- ---
To invent, you need a good imagination and a pile of junk. -- Thomas Edison
------- End of forwarded message -------

Re: [kzeitler@lucent.com: Re: Crash with mouse-avoidance-mode]

[Nick Roberts]

Richard Stallman writes:
 > Would someone please investigate this and DTRT?
 > 
 > ------- Start of forwarded message -------
 > To: rms@gnu.org
 > Subject: Re: Crash with mouse-avoidance-mode
 > From: Klaus Zeitler <kzeitler@lucent.com>
 > Date: Tue, 08 Aug 2006 11:49:29 +0200
 > In-Reply-To: <E1GAIpZ-0003dk-PE@fencepost.gnu.org> (Richard Stallman's message of "Tue, 08 Aug 2006 00:04:37 -0400")
 > MIME-Version: 1.0
 > Content-Type: text/plain; charset=us-ascii
 > X-Spam-Status: No, score=0.0 required=5.0 tests=none autolearn=failed 
 > 	version=3.0.4
 > 
 > >>>>> "Richard" == Richard Stallman <rms@gnu.org> writes:
 >     Richard> 
 >     Richard> Can you tell us a precise recipe for reproducing this?
 > 
 > Nearly precise :-).
 > 
 > 1. emacs -Q
 > 2. (setq mouse-avoidance-animation-delay 0.1)
 > 3. M-x mouse-avoidance-mode proteus
 > 4. Move cursor carefully/slowly close to mouse pointer so that avoidance-mode
 >    will try to move the pointer away. I need only about 3 or 4 tries till the
 >    animation goes haywire and the mouse pointer starts jumping around

I don't know about the haywire bit, maybe it's meant to do that, but the
crash was related to sit-for which Chong has already fixed.


-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: [kzeitler@lucent.com: Re: Crash with mouse-avoidance-mode]

[Klaus Zeitler]

>>>>> "Nick" == Nick Roberts <nickrob@snap.net.nz> writes:
    Nick> 
    Nick> I don't know about the haywire bit, maybe it's meant to do that, but
    Nick> the crash was related to sit-for which Chong has already fixed.

No it's not meant to do that. If you try it first with the default
mouse-avoidance-animation-delay (0.01), you'll see how it is supposed
to act. And it does nearly the same with the value I've been using for
a long time (0.1). But with recent changes (I suspect sit-for) sometimes
the mouse pointer keeps jumping around or dithering and will not stop as
long as no new input is entered. Though I'm not sure if this is a Solaris
specific problem or not.

Klaus

-- 
 ------------------------------------------
|  Klaus Zeitler      Lucent Technologies  |
|  Email:             kzeitler@lucent.com  |
 ------------------------------------------
---
Pardon this quote. Database under reconstruction.

Re: [kzeitler@lucent.com: Re: Crash with mouse-avoidance-mode]

[Chong Yidong]

Klaus Zeitler <kzeitler@lucent.com> writes:

>>>>>> "Nick" == Nick Roberts <nickrob@snap.net.nz> writes:
>     Nick> 
>     Nick> I don't know about the haywire bit, maybe it's meant to do that, but
>     Nick> the crash was related to sit-for which Chong has already fixed.
>
> No it's not meant to do that. If you try it first with the default
> mouse-avoidance-animation-delay (0.01), you'll see how it is supposed
> to act. And it does nearly the same with the value I've been using for
> a long time (0.1). But with recent changes (I suspect sit-for) sometimes
> the mouse pointer keeps jumping around or dithering and will not stop as
> long as no new input is entered. Though I'm not sure if this is a Solaris
> specific problem or not.

I think this is an avoid.el bug that only surfaced with the new
sit-for.  mouse-avoidance-mode works by activating a timer to run
mouse-avoidance-fancy-hook every 0.1 seconds.  If the animation delay
is long, the timer can run again in the middle of the animation, which
nudges the mouse pointer in another random direction.

The way to solve this is to introduce a variable
mouse-avoidance-animating-pointer which is non-nil while animating the
mouse, and making mouse-avoidance-fancy-hook do nothing if
mouse-avoidance-animating-pointer is non-nil.

(BTW, mouse-avoidance-fancy-hook violates the naming conventions, and
should be renamed mouse-avoidance-timer-function.)

I'll check in changes to fix these issues shortly.

Re: [kzeitler@lucent.com: Re: Crash with mouse-avoidance-mode]

[Stefan Monnier]

> I think this is an avoid.el bug that only surfaced with the new
> sit-for.  mouse-avoidance-mode works by activating a timer to run
> mouse-avoidance-fancy-hook every 0.1 seconds.  If the animation delay
> is long, the timer can run again in the middle of the animation, which
> nudges the mouse pointer in another random direction.

> The way to solve this is to introduce a variable
> mouse-avoidance-animating-pointer which is non-nil while animating the
> mouse, and making mouse-avoidance-fancy-hook do nothing if
> mouse-avoidance-animating-pointer is non-nil.

Or to turn this repeated timer into a one-shot timer which is manually
re-created at the end of the animation step, so that the 0.1s refers not to
the interval between the beginning of each animation step but between the
end of one and the beginning of the next.


        Stefan