Ticket #1135 (closed enhancement: fixed)
Small regression test updates
|Reported by:||ads||Owned by:||strk|
|Keywords:||Cc:||ads@…, andreas.scherbaum@…, ashwin.kumar3@…|
As posted earlier on the devel mailinglist:
http://postgis.refractions.net/pipermail/postgis-devel/2010-October/014441.html (the date is invalid because of server issues)
here are two patches for the regression tests. One patch is for 1.5.3, one for 2.0.0SVN.
From my mail to the mailinglist:
we recently implemented PostGIS in Greenplum (a PostgreSQL fork). The regression tests throwed some errors on us, but mainly only a different output order in the affected .expected files.
Greenplum is a distributed shared-nothing architecture. Unlike PostgreSQL, we will always get the rows in a different order - unless we order them explicitely, like it is required in the SQL standard.
Thinking about it, there is a (although very small) chance, that synchronized sequentiell scans in 8.3 and above will also break these regression tests - as example, if a backup is running at the same time.
Attached you can find a small patch which adds "order by 1" in some of the regression tests, and also changes the output order in some of the .expected files. The data itself is not changed. The patch works with PostgreSQL 9.04 and PostGIS 1.5.3.