The log shipping test (and only the log shipping test) fails on v7.4,
which seems to be attributable to the (obvious) syntax error seen
below.

In the file for the ".y22679" test, generated with PG 7.4, there is no
"\." in the dump file that is used to set up log shipping.

In the other one (from a test of v8.1), the "\." is present.


cbbrowne at dba2:/tmp> diff  -b slony-regress.w10354/logship_dump.sql slony-regress.y22679/logship_dump.sql 
101a102
> \.
cbbrowne at dba2:/tmp> tail slony-regress.w10354/logship_dump.sql slony-regress.y22679/logship_dump.sql 
==> slony-regress.w10354/logship_dump.sql <==
	raise notice ''Slony-I: Process set % sync % time %'', p_set_id, p_new_seq, p_sync_time;

	update "_slony_regress1".sl_setsync_offline set ssy_seqno = p_new_seq, ssy_synctime = p_sync_time
		where ssy_setid = p_set_id;
	return p_new_seq;
end;
' language plpgsql;

copy "_slony_regress1".sl_sequence_offline from stdin;
commit;

==> slony-regress.y22679/logship_dump.sql <==

	update "_slony_regress1".sl_setsync_offline set ssy_seqno = p_new_seq, ssy_synctime = p_sync_time
		where ssy_setid = p_set_id;
	return p_new_seq;
end;
' language plpgsql;

copy "_slony_regress1".sl_sequence_offline from stdin;
\.
commit;

Unfortunately, I think this is attributable to Stephane Schildkneckt's
recent change to the following that added the leading 'E':

    printf "select E'\\\\\\\\.';"

Can we verify that this usage breaks in PG 7.4?  I expect I need to
revert this change.  (Sorry, Stephane!)
-- 
let name="cbbrowne" and tld="ca.afilias.info" in String.concat "@" [name;tld];;
<http://dba2.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)