I have backport the automatic listen path generator from CVS HEAD...

The generating of listen paths has something of a sordid history:

- In v1.0, it had to be handled by hand by users, which was fine for
  simple clusters, but nasty for bigger clusters with many nodes.

- For v1.1, I wrote an automagic listen path generator which Jan
  rewrote, which seemed nice, but which had some problems.

- For v1.2, Florian Pflug rewrote it, fixing the problems then seen.

Unfortunately, we then saw some further problems.  So...

- For v2.0, Jan rewrote it again.

Our folk were, internally, running into problems with v1.2 for some of
our test cases, so I tried a backport of the new v2.0 code.

  Tests done thus far:
  - I added a "testlistenpath" test to the set of tests, which is working
    fine

   - JP Fletcher did some testing of cases where he saw things break,
     and this change addressed those cases, and examination of sl_listen
     showed that the paths did indeed look apropos...

With this having worked well, I have committed the backport to the 1.2
branch, so it will be in v1.2.12.

<http://lists.slony.info/pipermail/slony1-commit/2007-September/002016.html>

If we encounter further issues with listen path generation, it seems
pretty likely that it's easier to fix the new code than to worry about
trying to roll back to elder code...

If people have any test cases that tend to exercise this, it's worth
testing this.
-- 
(format nil "~S@~S" "cbbrowne" "acm.org")
http://www3.sympatico.ca/cbbrowne/finances.html
Rules of  the Evil  Overlord #6.  "I will not  gloat over  my enemies'
predicament before killing them." <http://www.eviloverlord.com/>