Possible bug in Geography ST_Covers

[robe]

I would expect the answer to my UTM to agree with Geography answer. Also the platte-care answer is woo really screwed. That I think is a bug in the casting or somethng I don't understand about the casting which I'll put in as a separate ticket.

SELECT ST_AsText(pt.the_geog) As wkt_pt, ST_Covers(poly.the_geog, pt.the_geog) As geog,
	ST_Covers(
		ST_Transform(CAST(poly.the_geog As geometry),32611), 
		ST_Transform(CAST(pt.the_geog As geometry),32611)) As utm,
	ST_Covers(
		CAST(poly.the_geog As geometry), 
		CAST(pt.the_geog As geometry)
	) As pca
FROM (SELECT ST_GeographyFromText('SRID=4326;POLYGON((-119.5434 34.9438,-119.5437 34.9445,-119.5452 34.9442,-119.5434 34.9438))') As the_geog) 
	As poly
    CROSS JOIN 
	(VALUES
		( ST_GeographyFromText('SRID=4326;POINT(-119.5434 34.9438)') ) ,
		( ST_GeographyFromText('SRID=4326;POINT(-119.5452 34.9442)') ) ,
		( ST_GeographyFromText('SRID=4326;POINT(-119.5434 34.9438)') ),
		( ST_GeographyFromText('SRID=4326;POINT(-119.5438 34.9443)')  )
	)
	As pt(the_geog);

I'm expecting all these to yield an answer of true. Instead I get

          wkt_pt          | geog | utm | pca
--------------------------+------+-----+-----
 POINT(-119.5434 34.9438) | t    | t   | f
 POINT(-119.5452 34.9442) | f    | t   | f
 POINT(-119.5434 34.9438) | t    | t   | f
 POINT(-119.5438 34.9443) | t    | t   | f

The pca answer is particularly troubling which I think is a separate issue.

[pramsey]

I get the right (?) answer

          wkt_pt          | geog | utm | pca 
--------------------------+------+-----+-----
 POINT(-119.5434 34.9438) | t    | t   | t
 POINT(-119.5452 34.9442) | t    | t   | t
 POINT(-119.5434 34.9438) | t    | t   | t
 POINT(-119.5438 34.9443) | t    | t   | t
(4 rows)

as of r4640. Could be the precision issues that manifest in your cunit failures. Those remain on my list (though a ticket wouldn't hurt).

[robe]

That is the answer I would expect. Mine is still wrong even after r4640. I thought it was maybe a bbox issue again but && returns true on all accounts.

ST_Distance also returns 0 on all accounts. Anyrate I'll put in another ticket for Cunit issue once I've cataloged the various ways its failing on different boxes.

[pramsey]

I'm still getting the right answer. Has anything changed on your end?

[robe]

Yap works now.