layers). The way that the raster layer sets it's srs is a bit complex
- hopefully these changes fix the problem without introducing other
problems (seems ok from the testing that I've done).
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5726 c8812cc2-4d05-0410-92ff-de0c093fc19c
are listed in the geometry_columns table. For well-maintained
databases (where all the spatial tables of interest have been
listed in the geometry_columns table) this can considerably speed
up the initial display of available spatial tables.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5718 c8812cc2-4d05-0410-92ff-de0c093fc19c
* Propagate proxy information from the Server Selection dialog to the WMS provider (used by the QgsRasterLayer).
Possible bonus feature:
* Proxy information is saved and restored from the project file.
I cannot test this locally, so would appreciate somebody else giving it a go.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5716 c8812cc2-4d05-0410-92ff-de0c093fc19c
--with-projlibdir option to explicitly set the directory that the
proj4 library is in.
Fix an error in the summary at the end of the configure run too.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5715 c8812cc2-4d05-0410-92ff-de0c093fc19c
in the geometry_columns table to just those in the public schema. For some
databases which have a lot of ancillary stuff out of the public schema this
can significantly reduce the time that qgis takes to produce the list of
tables to load.
Also tidy up the cause of some console warning messages about unknown
geometry types that happened when a spatial table not in the geometry_columns
table had no non-null geometries in it.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5711 c8812cc2-4d05-0410-92ff-de0c093fc19c
qgis directory will update a file (qgssvnversion.h) that contains the
svn version of the qgis code, which is visible with the --help command
line argument or in the Help:About dialog box.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5710 c8812cc2-4d05-0410-92ff-de0c093fc19c
arguments, the extent is -1,-1,1,1, but if there, the extent is determined by t
he layers loaded.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5703 c8812cc2-4d05-0410-92ff-de0c093fc19c
combination of QFile and QTextStream which worked under Qt3, but not
under Qt4
- fixed generation of wkb version of delimited text geometry. Again,
it appeared to be a problem with code that worked under qt3, but
not under qt4.
- corrected the calculation of the bounding rectangle for delimited
text layers (was always including the (0,0) point).
- Bonus features:
- a default layer name is constructed from the filename
- the user is told about lines that didn't yield x/y coords
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5698 c8812cc2-4d05-0410-92ff-de0c093fc19c
QgsHttpTransaction needed to wait for the done() signal from QHttp, not just requestFinished(). When QHttp::setProxy() was called, it actually triggered its own QHttp request (before QHttp::get()). Previously QgsHttpTransaction interpreted the requestFinished() from setProxy() as the requestFinished() from get(), therefore not actually doing anything when a proxy was used.
I cannot test until a Windows build is available based on this commit (or somebody else can confirm instead and update #8 appropriately).
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5697 c8812cc2-4d05-0410-92ff-de0c093fc19c
The Qt3 idiom QToolButton::setPaletteBackgroundColor() has been replaced with a Qt4 idiom of extracting the QPalette from the QToolButton, changing the color in the Window ColorRole, then adding the QPalette back in again.
Only the QGIS Options and Project Properties dialogs have been converted. If this change fixes the bug in Windows (which I don't have access to a compiler for), then this idiom should be propagated to all instances of setPaletteBackgroundColor (e.g. vector symbology dialog).
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5696 c8812cc2-4d05-0410-92ff-de0c093fc19c
* (#134) If a feature is edited using the Identify tool and QgsAttributeDialog, now only user-edited values are saved to the mChangedAttributes of the layer, previously it was all values. This now means that only those user-edited values are attempted to be committed when the user stops editing and saves (similar to if those values were edited using the Attribute Table).
* (#131) Modify QgsVectorLayer::stopEditing so that if a commit fails, the editing state is left alone (previously editing was still turned off anyway). This should be a better fix than r5591. I triggered the bug in #131 while testing for #134, therefore getting a 2-for-1 fix.
Some bonus features to assist people in the triggered situations described in #131 and #134:
* The Postgres Provider now reports an error to the user if a commit of changed attributes fails. This brings it up to the same behaviour of if a commit of changed geometries fails. Previously there was no response given on an changed attribute error, leaving the user to think the commit was successful.
* The Postgres Provider now reports the contents of the SQL used in a failed UPDATE statement. This may help the user to pinpoint which of his edits caused the error.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5694 c8812cc2-4d05-0410-92ff-de0c093fc19c
lib-refactoring branch, to head. These were signed/unsigned int mods
and boolean comparison mismatches. Many of the changes in these
commits had already been done in head
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5681 c8812cc2-4d05-0410-92ff-de0c093fc19c
changed it's extent, e.g., by changing the where clause on a postgres layer).
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5679 c8812cc2-4d05-0410-92ff-de0c093fc19c
The symbology tab still needs to be cleaned-up in operation but at least it's set to "Color Image" by default now.
git-svn-id: http://svn.osgeo.org/qgis/trunk/qgis@5678 c8812cc2-4d05-0410-92ff-de0c093fc19c
