Forces ring close on conversion, fixing a regression
from 2.8 (and 2.14). See #13635
Adds test for identifying invalid polygons, currently only
testing for the unclosed-ring invalidity.
The test was verified to fail without the fixes included
in this same commit, and to pass in 2.14.
- Make QgsCurvePolygonV2::fromWkb() accept CompoundCurveM sub-geometries
- Make QgsGeometryCollectionV2::fromWkb() validate the sub-geometry type,
so that QgsGeometryCollectionV2 subclasses do not import incompatible
sub-geometries
- Make QgsGeometryCollectionV2::fromWkt() accept curve sub-geometries
- Make QgsMultiPolygonV2::addGeometry() accept only Polygon and not
CurvePolygon
- Add tests
OSRImportFromProj4() may accept strings that are not valid proj.4 strings,
e.g if they lack a +ellps parameter, it will automatically add +ellps=WGS84, but as
we use the original mProj4 with QgsCoordinateTransform, it will fail to initialize
so better detect it now.
Fixes#14844
When a OGR layer is of geometry type "unknown", we use the
geometry of the first feature to guess the geometry type. But
if this feature has no geometry, then we assume that we have no
geometry for the whole layer. This is a bit extreme. Let us
allow to probe a few features before giving up.
Fixes#15065
Fix#15061
MapServer honours EPSG axis order in WFS 1.1, but returns srsName in GetFeature
response with EPSG:XXXX syntax instead of urn EPSG srs. This confuses the GML
parser that thinks that no axis inversion should then happen.
The heuristics here consist in checking the envelope of the response with the
capabilities extent.
This should be safe and should work for layers with non global extent, but will
not detect all issues.
Proper fix is either to force WFS 1.0, or upgrade to MapServer 7.0 with WFS 2.0
Fixes#11371 (crash on ungrouping after moving the group) and more
undo/redo related issues.
Enable pending test for the crash (now passing) and add many more
undo/redo related ones (including signals testing).
Includes a new QgsGroupUngroupItemsCommand class
and its SIP bindings.
This reverts commit 7cab60bc4328e6cb4c57f8320c2d85d9383a2424.
The conversion was unreliable, as it was not correctly determining
that the feature's geometry was required and consequently not
fetching it from the provider. This isn't trivial to fix, so
I'm reverting this feature before release.
According to the specification, the values of LatLongBoundingBox
are supposed to be in the SRS, not necessarily in WGS84.
But some servers do not follow the spec and return values in WGS84,
so let's try to accomodate for this too.
Fix#14876
