On Wed, Jun 22, 2005 at 02:46:19PM -0400, Christopher Browne wrote:
> Nothing leaps out to me from the script, but as you say, it's pretty
> lengthy, so there's "a lot of there" there (to say it in the Clintonian
> way :-)).
> 
> What I'd suggest first is that you seek to solve the "because the error
> message, ddl file, and schema are lengthy" problem.
> 
> Try doing it in pieces; I'd try doing everything up to (but not
> including) the alteration to segmentversions.
> 
> That means needing to lock the set multiple times, which is pretty evil,
> but hopefully it does two things:
> 
> 1.  It should get you to the point where *some* of the changes are
> getting through, and
> 
> 2.  It should allow narrowing the possible sources of the problem.

Actually, the error occurs as soon as the first view is created
("shows"), which I had assumed and have since verified (the same error
occurs if I delete everything after the alternations to the
showversions table and the shows view -- what may have led you to
think segmentversions was involved is that I was adding a column with
the same name to that table as well).  So there aren't actually any
hidden factors that need to be simplified out in the ddl.

If seeing the result of operating only on showversions/shows is
helpful, I've updated the tarball to include 3 additional files; but
in fact nothing is significantly different except that the ddl is
shorter.

Cheers,

js

-- 
Jacob Smullyan
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://gborg.postgresql.org/pipermail/slony1-general/attachments/20050622/f9dcd329/attachment.bin