[postgis-devel] PostGIS 2.0 regress failures compared to PostGIS 1.5: failing functions used to work return answers in PostGIS 1.5

[Paragon Corporation]

Paul,

I'm not sure its worthwhile to post these as tickets since I think some are
things that perhaps should not have returned answers in PostGIS 1.5.  There
are some true failures really.  One of which I have already posted.

I'm also not sure how many of these will vaporize once you are done with
your fitting.

Attached is my first pass against  PostGIS 1.5 where I ran the same battery
of tests between my 1.5.2 install and PostGIS 2.0 (of a day or so ago) and
then queried to see which tests used to complete in 1.5.2 and no longer
complete in PostGIS 2.0.

I've attached the failing tests as an sql log table.  You can run the
log_sql statement to see the error you get.


Cursory observation:

Total of 1065 Test failures

But they are limited to the following functions:

SELECT DISTINCT func,spatial_class
FROM postgis20_regress_failures_pass1
ORDER BY spatial_class, func;

-- output of above query:
         func         | spatial_class
----------------------+---------------
 ST_Buffer            | geography
 ST_Intersection      | geography
 ST_Distance_Sphere   | geometry
 ST_Distance_Spheroid | geometry
 ST_DumpPoints        | geometry
 ST_GeogFromText      | geometry
 ST_GeographyFromText | geometry
 ST_GeomFromEWKT      | geometry
 ST_Length_Spheroid   | geometry
 ST_Length3D_Spheroid | geometry
 ST_Transform         | geometry


There are some functions like ST_Lenght3D_Spheroid that used to work with
for example CURVEPOLYGON.  Those I assume we may be able to ignore.  I'm not
even sure if those past answers were right.

The ones that really concern me are the 

ST_Buffer geography and 

the Distance_Spher* functions which I already noted in a ticket.

For example -- this query
SELECT ST_Buffer(ST_GeogFromText('SRID=4326;MULTILINESTRINGM((-71.0821
50.3036 1,50 -71 1),(-56.0821 50.3036 1,50 -56 1),(-41.0821 50.3036 1,50 -41
1),(-26.0821 50.3036 1,50 -26 1),(-11.0821 50.3036 1,50 -11 1),(3.9179
50.3036 1,50 4 1),(18.9179 50.3036 1,50 19 1),(33.9179 50.3036 1,50 34
1),(48.9179 50.3036 1,50 49 1),(-71.0821 50.3036 2,50 -71 2),(-56.0821
50.3036 2,50 -56 2),(-41.0821 50.3036 2,50 -41 2),(-26.0821 50.3036 2,50 -26
2),(-11.0821 50.3036 2,50 -11 2),(3.9179 50.3036 2,50 4 2),(18.9179 50.3036
2,50 19 2),(33.9179 50.3036 2,50 34 2),(48.9179 50.3036 2,50 49 2))'),
0.25);

Returns error:  (so I assume all the buffer and intersection failures for
geography are redundant and are a result of the underlying transform
failure.
lwgeom_transform: Cannot handle type 'MultiLineString'

This one I'll put in as a ticket since that is a sure issue.

Thanks,
Regina


-------------- next part --------------
A non-text attachment was scrubbed...
Name: postgis20_regress_failures_pass1.zip
Type: application/x-zip-compressed
Size: 45161 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20101218/d6a88271/attachment.bin>