Regression with GEOS 3.12.0 in Fedora Rawhide #1855
Description
Expected behavior and actual behavior.
Since the GEOS package was updated from 3.11.2 to 3.12.0 in Fedora Rawhide (the development version of the distribution), we are encountering a single test failure:
=================================== FAILURES ===================================
______________ OperationsTestCase.test_parallel_offset_linestring ______________
self = <tests.test_parallel_offset.OperationsTestCase testMethod=test_parallel_offset_linestring>
def test_parallel_offset_linestring(self):
line1 = LineString([(0, 0), (10, 0)])
left = line1.parallel_offset(5, 'left')
assert_geometries_equal(left, LineString([(0, 5), (10, 5)]))
right = line1.parallel_offset(5, 'right')
assert_geometries_equal(right, LineString([(10, -5), (0, -5)]), normalize=True)
right = line1.parallel_offset(-5, 'left')
assert_geometries_equal(right, LineString([(10, -5), (0, -5)]), normalize=True)
left = line1.parallel_offset(-5, 'right')
assert_geometries_equal(left, LineString([(0, 5), (10, 5)]))
# by default, parallel_offset is right-handed
assert_geometries_equal(line1.parallel_offset(5), right)
line2 = LineString([(0, 0), (5, 0), (5, -5)])
> assert_geometries_equal(line2.parallel_offset(2, 'left', resolution=1),
LineString([(0, 2), (5, 2), (7, 0), (7, -5)]))
E AssertionError:
E Geometries are not equal:
E
E Not equal to tolerance 1e-07
E x: array(<LINESTRING (0 2, 5 2, 5.39 1.962, 5.765 1.848, 6.111 1.663, 6.414 1.414, 6....>,
E dtype=object)
E y: array(<LINESTRING (0 2, 5 2, 7 0, 7 -5)>, dtype=object)
tests/test_parallel_offset.py:32: AssertionError
=========================== short test summary info ============================
FAILED tests/test_parallel_offset.py::OperationsTestCase::test_parallel_offset_linestring
======== 1 failed, 257 passed, 2 skipped, 3 xfailed, 1 xpassed in 0.99s ========
Steps to reproduce the problem.
It’s hard to say how best to reproduce this outside of Fedora infrastructure. It is possible that this might always occur when compiling with GEOS 3.12.0.
Operating system
Fedora Linux Rawhide
Shapely version and provenance
I’m a co-maintainer of the python-shapely package in Fedora Linux. We build from the PyPI sdist (version 2.0.1) without any downstream patches.
Additional comments
I tried packaging a snapshot of https://github.com/shapely/shapely/tree/maint-2.0, but this didn’t seem to make a difference.
Is there anything else I can try to understand this better?