[postgis-users] postgis_restore.pl loads no data (pg-8.1, postgis-1.1)

[Michael Fuhr]

On Mon, Jan 16, 2006 at 01:32:38PM +0100, strk at refractions.net wrote:
> Still hung? What is your computed doing during the hung ?
> Can you show initial lines of top(1) output ?
> Can you show output/error of postgis_restore.pl run ?

I just ran some tests with PostgreSQL 8.1.2 and PostGIS 1.1.1CVS
on FreeBSD 6.0-STABLE and Solaris 9.  Here's what I did; both systems
behaved the same (skip to the bottom of this message to see what I
think is wrong):

  createdb test1
  createlang plpgsql test1
  psql -d test1 -f lwpostgis.sql
  psql -d test1 -f spatial_ref_sys.sql
  pg_dump -Fc test1 > test1.dump
  sh utils/postgis_restore.pl lwpostgis.sql test2 test1.dump > restore.log

If I now connect to test2 I see that it has no tables, not even
geometry_columns or spatial_ref_sys.  On line 747 of restore.log I
see this:

  Producing ascii dump test1.dump.ascii

and on lines 2562ff I see this:

  END;
  '
  Sourcing lwpostgis.sql
  Dropping geometry_columns and spatial_ref_sys
  Now source test1.dump.ascii manually
  LANGUAGE 'plpgsql' VOLATILE STRICT; -- WITH (isstrict);

Something doesn't look right there -- three lines have been
intermingled with a CREATE OR REPLACE FUNCTION command.  The above
output is from the FreeBSD box; the Solaris box did the same thing
but the lines are intermingled elsewhere (lines 3017ff).

At the end of restore.log I see this:

  ---------------------------------------------------------------
  -- END
  ---------------------------------------------------------------
  COMMIT;
  COMMIT
  DROP TABLE geometry_columns;DROP TABLE spatial_ref_sys;
  DROP TABLE
  DROP TABLE

In the current directory is test1.dump.ascii; loading that file
restores the tables:

  psql -d test2 -f test1.dump.ascii

I see the following lines near the end of the 1.1 postgis_restore.pl
script:

  print "Now source $dumpascii manually\n";
  exit(1);

In 1.0 those lines are commented out.  The code that follows loads
the .dump.ascii file, but in 1.1 that code isn't reached because
of the exit(1) statement.

-- 
Michael Fuhr