MySQL 5.7 GIS test failures: InterfaceError: (-1, 'error totally whack')

[Tim Graham]

Not sure if it's a Django issue, but the following tests fail with InterfaceError: (-1, 'error totally whack'). The commonality seems to be that these are all ​new functions in MySQL 5.6.

gis_tests.distapp.tests.DistanceFunctionsTests.test_distance_simple
gis_tests.distapp.tests.DistanceFunctionsTests.test_distance_order_by
gis_tests.geoapp.test_functions.GISFunctionsTests.test_union
gis_tests.geoapp.test_functions.GISFunctionsTests.test_sym_difference
gis_tests.geoapp.test_functions.GISFunctionsTests.test_intersection
gis_tests.geoapp.test_functions.GISFunctionsTests.test_difference
gis_tests.geoapp.test_functions.GISFunctionsTests.test_diff_intersection_union

[Tim Graham]

From the ​MySQL 5.7 release notes:

The ST_Buffer(), ST_Difference(), ST_Distance(), ST_Intersection(), ST_IsSimple(), ST_SymDifference(), and ST_Union() functions have been reimplemented to use the functionality available in Boost.Geometry. The functions may raise an exception for invalid geometry argument values when the previous implementation may not have.

An example failing query from gis_tests.geoapp.test_functions.GISFunctionsTests.test_union:

SELECT geoapp_city.id, geoapp_city.name,
AsText(geoapp_city.point),
AsText(ST_Union(geoapp_city.point, GeomFromText('POINT (-95.3631510000000020 29.7633739999999989)', 4326))) AS union
FROM geoapp_city
WHERE geoapp_city.name = 'Dallas'

A minimal query to reproduce is:

SELECT ST_Union(geoapp_city.point, ST_GeomFromText('POINT (-95.3631510 29.763373999)', 4326)) FROM geoapp_city

Any advice, Claude?

[Claude Paroz]

Sorry, no clue :-(

[Sergey Fedoseev]

SRID of geometry from the table is 0, SRID of the given point is 4326, ​MySQL requires arguments to have the same SRID.

[Tim Graham]

Not sure I understand. For the union test, City.objects.get(name='Dallas').point.srid gives 4326 which is the same as the Point in the test. If you are able, feel free to send a patch, Sergey.

[Sergey Fedoseev]

​PR contains only the partial solution because existing geometries have SRID 0 and will cause a crash.

[Sergey Fedoseev]

It seems that I figure out a better solution to the problem.

[Sergey Fedoseev]

​PR

[Tim Graham <timograham@…>]

In 9031a4c1:

Fixed #26657 -- Made GeomValue omit SRID for MySQL.

This fixes some test failures on MySQL 5.7+.

[Tim Graham <timograham@…>]

In 9805726:

[1.10.x] Fixed #26657 -- Made GeomValue omit SRID for MySQL.

This fixes some test failures on MySQL 5.7+.

Backport of 9031a4c13bdb1ee4d0f1253fcfcad73d26c25740 from master

[Tim Graham <timograham@…>]

In 4178488:

Refs #26657 -- Fixed a crash induced by invalid WKT returned by MySQL 5.7.5+.

[Tim Graham <timograham@…>]

In 3771019d:

[1.9.x] Fixed #26657 -- Made GeomValue omit SRID for MySQL.

This fixes some test failures on MySQL 5.7+.

Backport of 9031a4c13bdb1ee4d0f1253fcfcad73d26c25740 from master

[Tim Graham <timograham@…>]

In 25581e6:

[1.10.x] Refs #26657 -- Fixed a crash induced by invalid WKT returned by MySQL 5.7.5+.

Backport of 4178488881f4e703c4e47ea78720dd61e0230ce5 from master