In QgsDateTimeFieldFormatter::representValue(), QDate and QDateTime fields don't need a triple conversion (QVariant->toString->fromString->toString) where a double conversion (QVariant->QDateTime->toString) is enough.
The lower limit was set to Jan 1st 100 CE only because QDateTimeEdit::setMinimumDateTime doesn't support dates < '0100-01-01'.
This limitation doesn't affect QDateTimeEdit::setDateTimeRange so it is possible to allow entering dates also from 1 CE to 99 CE which previously where not allowed.
Set to 10000 by default. Can be overriden through settings.
This is only for the preview in the source select dialog. (Attribute
table performs badly for much less columns)
Related to #36392
Fixes#36392Fixes#21976Fixes#17190
We are obliged to do 'at hand' parsing due to QT not handling CR-only
end of lines.
As we are at it, also limit each line to 1 MB to avoid potential denial
of service (which was what close to what happened here before the CR-only
parsing fix)
Add tests for parsing CR-only end of lines, and exercising the at-hand
buffering logic
Even though SAGA explicitly does't follow semver conventions, there are
advances, and not all changes on their part break the API. Since the
user is already warned in big letters that versions different from 2.3
are unsupported, in beta status, we've changed the requirement for SAGA
"beta" from 7.3 to everything in version 7.
manually in the plugin manager, then force a reload of the repositories
and don't rely on cached copies of the repo manifest
Otherwise it's not possible to force a refresh on certain types of
plugin repositories where it's not possible to modify the Cache-Control
attribute of the repo
Note that to avoid unnecessary server load, we still use cached
copies during startup and by default. It's only when a user explicitly
hits the "Reload all repositories" button that we force a non-cached
fetch.
Fixes#34351
Otherwise this provider cannot be initialized from standalone Python
scripts
Fixes#36661
Also cleanup 3d sip generation by correctly defining SIP_NO_FILE
in headers we don't want to expose, so that the sip_include script
can do its magic
This avoids the force conversion to a raster based pattern which currently occurs
when exporting maps/layouts to a vector format (e.g. PDF). The raster pattern
results in considerable quality loss, and the tiling edges of the raster brush
can sometimes be seen in outputs.
Additionally, fixes render corrupt when marker subsymbols have data defined properties
which affect the marker shape, such as data defined rotation or sizes
Refs #16100 (still needs fixing for line fill symbols)
The current approach was based on a srsid, that that might be unset for custom
CRS. So store a full QgsCoordinateReferenceSystem object internally and use
that for exchanges with QgsProjectionSelectionWidget
Consequence: we need to deprecate the "long crs()" method, and replace it
by a "QgsCoordinateReferenceSystem crsObject()" one. I'm not particularly
happy with the name but the existing one was not very well named...
Mark also the QgsVectorLayerSaveAsDialog( long srsid, ... ) constructor as
deprecated
Fixes the use case of https://github.com/qgis/QGIS/issues/34471#issuecomment-586181977
There were several non-compliance in the CoverageDescription output to a
DescribeCoverage request. The corrected result has been validated against the
OGC WCS 1.0.0 schema.
Fixes#36504
- left clicking in the combo box opens the dropdown, just like a standard
combo box
- right clicking on an unexpanded combo shows the correct context menu
for the widget, instead of the default line edit context menu
Fixes#31630
I couldn't reproduce a crash on my platform (Linux, Ubuntu 16.04, gnome classic),
but this smells pretty much like the issue of #30210 fixed per PR #30210
So using QgsFocusKeeper hoping that it would solve the issue for people affected
