Re: PATCH: ewoc.el to permit node representation abutment

[ttn@gnu.org]

   From: Stefan Monnier <monnier@iro.umontreal.ca>
   Date: Fri, 05 May 2006 09:20:08 -0400

   Cool.  I assume you tested it with the most important case: when the
   pretty-printer returns the empty string (so you can temporarily hide
   elements, which is the main reason why I wanted this feature
   implemented).

actually, i didn't test that case (having in mind the munging of text
properties and `buffer-invisibility-spec' to implement hiding, instead).
i just started pcvs on a smallish project and typed `g' a few times.
anyway, looks like it will be handled in the process of solving...

   The problem case is when we have at the same spot many empty
   elements, and the one we refresh is one of the middle ones, so the
   preceding ones shouldn't move but the subsequent ones should.  It
   seems your code doesn't pay attention to this detail and will either
   move the preceding ones as well or fail to move the subsequent ones.

thanks for illustrating a concrete failure scenario.  i had vague doubts
but now they are firmer.  looks like slightly more hair is called for.

   [eta reduction]

it would merit "bugfix" if the header and footer were ever refreshed,
but i see that `ewoc-refresh' avoids them -- ugh.  (there seem to be
many design enhancement opportunities in ewoc.el...)

thi