[postgis-devel] GEOS changes breaking PostGIS testcases

[Sandro Santilli]

On Thu, May 30, 2013 at 03:20:14PM -0700, dustymugs wrote:
> On 05/30/2013 01:36 PM, Sandro Santilli wrote:
> > FYI: I've committed improvements for the GEOS snapping algorithm, so that
> > ST_Snap is less likely to return invalid (collapsed) geometries.
> > 
> > Unfortunately some of the topology tests in PostGIS are expecting
> > outputs due to collapses and thus will now fail until tweaked.
> > 
> > I'm planning to change the expectances to match the new GEOS results
> > but of course doing so would mean the tests will fail with latest
> > GEOS release instead. Better ideas ?
> > 
> > Ticket: http://trac.osgeo.org/postgis/ticket/2346
> 
> I'd guess that this wouldn't be a problem for 2.1 since that version
> requires GEOS 3.4. As for 2.0, can we require a specific GEOS version
> for a minor version of PostGIS? Such as 2.0.4 requires GEOS 3.4?

Nope, but I guess I could release GEOS 3.3.9 with the change.
3.3.8 was released on Feb, 28, a new release would have these
changes:

- Bug fixes / improvements
    - Fix area boundary return from GEOSPointOnSurface (#623)
    - Speedup GEOSWKBReader_read (#621)
    - Fix RobustLineIntersector handling of invalid intersection points (#622)
    - Reduce likelyhood of invalid output from snap operation (#629, #501)

Only I'd like to wait a couple of weeks to allow for broader testing
of the new snapping algorithm.

PS: meawhile I've updated  the expectances for both postgis 2.0 and trunk

--strk;