[postgis-devel] PATCH: Test PostgreSQL 8.3 patch for PostGIS

[Michael Fuhr]

On Sun, Jun 03, 2007 at 03:38:50PM +0100, Mark Cave-Ayland wrote:
> On Sat, 2007-06-02 at 21:30 -0600, Michael Fuhr wrote:
> > Here's a regression test patch for the PIP bug, which appears to
> > be fixed now.
> 
> Cool. Looking over Mark's commit it appears that he's committed
> something along these lines with his fix. Do your test cases match the
> ones in the commit here?
> http://postgis.refractions.net/pipermail/postgis-commits/2007-June/000072.html.

I think the cases I submitted are different.  I ran the latest
regress_ogc.sql in a not-yet-fixed database and didn't get any
mismatches vs. the expected output, so Mark's new test cases must
not have covered the case I found.  I did get a mismatch with the
cases I added until I built the latest code.

Incidentally, disjoint100-106 and disjoint150-156 are identical,
as are intersects100-106 and intersects150-106.  Is there a reason
for that?

> > I've also attached a patch to lwpostgis.sql.in to
> > create the bool-to-text cast only for PostgreSQL versions earlier
> > than 8.3 because those casts were just added to 8.3, causing the
> > CREATE CAST to fail.
> 
> *blinks* now that IS a surprise. While PostGIS should be free to
> determine the casting rules for its own types, it doesn't seem right to
> be adding casts for internal PostgreSQL data types. Do the regression
> tests pass without the cast in place?

Not against 8.3devel -- CREATE CAST fails because the cast already
exists so the regression tests abort immediately.  I haven't checked
against 8.2 and earlier yet.

-- 
Michael Fuhr