[postgis-devel] gist broken on PG 8.0

[strk at refractions.net]

On Fri, Sep 17, 2004 at 10:50:50AM +0100, Mark Cave-Ayland wrote:
> 
> > -----Original Message-----
> > From: postgis-devel-bounces at postgis.refractions.net 
> > [mailto:postgis-devel-bounces at postgis.refractions.net] On 
> > Behalf Of strk at refractions.net
> > Sent: 16 September 2004 16:46
> > To: postgis-devel at postgis.refractions.net
> > Subject: [postgis-devel] gist broken on PG 8.0
> > 
> > 
> > I've been spending a good amount of time trying to debug
> > gist indexing for LWGEOM. I now discovered the bug
> > also show with HWGEOM (current CVS code - should be 0.9).
> > Does anyone experience segfaults with PG8.0.0beta2 and
> > postgis 0.9 / head CVS branches ?
> > 
> > Adding hooks in postgresql gist code I found out that
> > gistchose() get a Page as argument which has a 
> > maxoffsetnumber == 0, which makes code around gist.c:1477 
> > loop forever (up to memory fault).
> > 
> > Any help ?
> > 
> > --strk;
> 
> 
> Hi strk,
> 
> I've just done a fresh install with PG 8.0b2 and PostGIS 0.9. I loaded
> in an 8,000 geometry shapefile, build an GiST index, and tried a few
> queries without any problems. I also ran through the test_estimation.pl
> several times, and again didn't experience any crashes.

mmm.. dragons here...
I got segfaults just creating the index on an empty table
and adding a couple of POINT geoms (sometimes more then 2 but
no more then 10).
The faults under LWGEOM are more tested (HWGEOM issue 
was new to me). The maxoffsetnumber thing is verified under
LWGEOM, second point added is enough to exploit.

--strk;