> But you reported that it didn't work, somehow.  So it doesn't have the
 > same effect as setting the variable manually, evidently.  (And that
 > change isn't installed, so only you have it in your sources.)

I lied.  I don't have them any more.

 >>  > I suspect different
 >>  > cursor motion commands.  If you want to analyze these more precisely,
 >>  > produce 3 termscripts where the keyboard commands were exactly the
 >>  > same, then using Diff might show the differences.
 >>
 >> Too difficult :-(
 >
 > Why is that?  You just need to count down- and up-arrow keystrokes.

I have to count _and_ hit a key at the same time.

 > Or am I missing something?

Obviously not.

 > Anyway, I'm out of ideas, so please try that, and maybe we will have a
 > lead.

I attach two termscript files.  The first one (ts-base) is standard, the
second (ts-rmwnil) has `resize-mini-windows' set to nil (you can see it
in the final window).  I did F10, hit the down key 10 times and then F10
again.  In the ts-base case I had the "--" artefact and the sticky
highlighting as in one of the caps.

Please tell me if the differences you see have any significance.  I
can't interpret them and diffing without newlines is a pain.  I can then
continue doing anything you suggest (as long as we keep the number of
keystrokes reasonably low).

 > Which reminds me: if you increase the height of that terminal window,
 > such that the longest menu fits in its entirety without reaching the
 > mode line, does the problem still happen?

Looks like your crystal ball knows no weekend ;-) In this case there's
no problem.

One more strange thing I noted: If I invoke

script some FILE

in the terminal just before calling emacs (to avoid calling
`open-termscript' there) the bug does not happen either.

martin