geometry regress failure

[robe]

I thought the issue in #3759 was a PostgreSQL 10 one which mysteriously disappeared.

But now our seeing it on winnie in her 9.5 run, but a different record different.

Checking for shp2pgsql ... found
Checking for pgsql2shp ... found
Checking for raster2pgsql ... found
TMPDIR is /projects/postgis/tmp/2.4.0dev_pg9.5_geos3.7.0dev_gdal2.1.3w64
Creating database 'postgis_reg' 
Preparing db 'postgis_reg' using: CREATE EXTENSION postgis
PostgreSQL 9.5.2, compiled by Visual C++ build 1800, 64-bit
  Postgis 2.4.0dev - r15406 - 2017-05-26 20:07:15
  scripts 2.4.0dev r15406
  raster scripts 2.4.0dev r15406
  GEOS: 3.7.0dev-CAPI-1.11.0 8fe2ce6
  PROJ: Rel. 4.9.1, 04 March 2015
  SFCGAL: 1.3.0
  GDAL: GDAL 2.1.3, released 2017/20/01

tickets .. failed (diff expected obtained: /projects/postgis/tmp/2.4.0dev_pg9.5_geos3.7.0dev_gdal2.1.3w64/test_89_diff)
-----------------------------------------------------------------------------
--- tickets_expected	2016-09-26 01:59:38 -0400
+++ /projects/postgis/tmp/2.4.0dev_pg9.5_geos3.7.0dev_gdal2.1.3w64/test_89_out	2017-05-26 16:13:04 -0400
@@ -284,7 +284,7 @@
 #3368|\x660001011fb98788d35ed6fbcdc831c580012b959f01469d8d0e9305ff8618ed08b1b607e302a614fe70bc4682b303b4379ab503a228eeb603e2138eb802900cc0b802ba04dab802b801a8840cb0229cca06f401c216f403a016ea05d0c301a8b301cc189226ac15ee27f811a029b40eac2ae00a902b8407c82b9e03d0e305801dfad3239504ce7e8d01e2f701f6019b04d1e2279bf901ff01
 #3375|GEOMETRYCOLLECTION(POINT(0 -7))
 #3399|1|1
-#3399|10|10
+#3399|10|9
 #3399|100|100
 #3399|1000|1000
 ERROR:  invalid KML representation
-----------------------------------------------------------------------------

Problem query is:

-- #3399

WITH g as (

select 'POLYGON((1 0, 0 1, 1 2, 2 1, 1 0))'::geometry as geom

),

n as (

select n from unnest(ARRAY[-1,0,1,10,100,1000]) n

),

pts as (

  select n,(st_dump(st_generatepoints(geom, n))).geom from g,n

)

select '#3399' as t, n, count(*) from
g, pts
where st_contains(g.geom, pts.geom)
group by n
ORDER BY n;

Might be bot specific I guess but still concerning.

Problem is intermittent. I did a manual run multiple times and it passed. So perhaps some dirty memory somewhere or only happens if system is taxed.

[robe]

should change to use ST_Intersects in the tests.

[robe]

In 15596:

ST_GeneratePoints check to use ST_Intersects instead of ST_Contains
Minimize failures when point falls on the boundary.
Closes #3765 for PostGIS 2.4.0