disturbing, hard-to-catch redisplay bug

disturbing, hard-to-catch redisplay bug

[John Wiegley]

Well, I have seen this twice now, but I can't catch it.  If I
restart my Emacs clean, the problem doesn't appear.  It only happens
if I've been using Emacs for a while.

I did, however, at least capture it visually.  These two images show
the bug:

In the first, I am editing a list of filenames:

  http://www.gci-net.com/~johnw/root1.jpg.bz2

Now I press M-} to advance to the next whitespace line, at which
point it "appends" to the end of my lines with content from the new
buffer location:

  http://www.gci-net.com/~johnw/root2.jpg.bz2

This is not real, however, as C-l (or any other redisplaying event)
will cure it.  I *have* moved, but the text of my new location does
not start at column 0, but rather is at the end of the visual lines
from my previous location -- at least until the next redisplay.

I hope the pictures will trigger someone's memory.  As I said,
changing my environment in any significant way causes the bug to
fade away.

I only see this when my screen is split.  A one-window display does
not have the problem.  A three-window display only has the problem
in the "half frame height weight" (i.e., the display you get after
pressing C-x 2 twice from a full frame).  If I balance the
three-window display, the problem does not appear.

Very odd,
  John

Re: disturbing, hard-to-catch redisplay bug

[Kim F. Storm]

John Wiegley <johnw@gnu.org> writes:

> Well, I have seen this twice now, but I can't catch it.  If I
> restart my Emacs clean, the problem doesn't appear.  It only happens
> if I've been using Emacs for a while.
> 
> I did, however, at least capture it visually.  These two images show
> the bug:
> 
> In the first, I am editing a list of filenames:
> 
>   http://www.gci-net.com/~johnw/root1.jpg.bz2

Why did you bzip the jpg files?  My browser doesn't grook that, so
I didn't look at the images.  You can use inline images too. 

Anyway, RMS fixed a bug yesterday which was related to positioning the
cursor at the end of the second last visible line on a window when it
should actually have been on the first char of the last visible line
of that window.  Could be related...

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

Re: disturbing, hard-to-catch redisplay bug

[Richard Stallman]

Can you give specifically precise directions that we could reproduce it?
It is ok if we have to load a file of Lisp code in order to do so.