[Qgis-user] QGIS big fail in deleting Shapes!!

[Even Rouault]

Le mercredi 20 janvier 2016 11:14:49, Jürgen E. Fischer a écrit :
> Hi Andreas,
> 
> On Wed, 20. Jan 2016 at 09:04:52 +0100, Neumann, Andreas wrote:
> > I also remember that Marco Hugentobler (Sourcepole) was contracted to
> > fix these Shapefile issues for the province of Vorarlberg - though I may
> > be wrong. So there is already someone investing to fix this bug. Marco -
> > correct me if I am wrong.
> 
> Yes, please.  I thought this was a fact - and didn't look at all into
> the problem.
> 
> Not sure if QGIS always does a REPACK (it used to after each delete - but
> that changes feature ids and interferes with internal stuff).

I've had a look at current master : it sets a "mShapefileMayBeCorrupted" flag 
when SetFeature() or DeleteFeature() are called, and if that flag is set at 
QgsOgrProvider::syncToDisc() time, it calls repack(). So it looks OK .

Just tested with QGIS master that deleting a feature leads to repacking when 
saving the edition. That said, that should work with GDAL 1.11 as well. The 
GDAL 2.0 fix is about also make REPACK repacking shapefiles whose geometries 
have been edited.

> I'm also
> not sure that GDAL 2 fixes the problem - the comments to the ticket are
> not conclusive in that regard 

Agreed...

> (GRASS is probably not a good test as it
> also uses GDAL and hence also should follow it's interpretation of what's
> deleted; see also
> https://lists.osgeo.org/pipermail/gdal-dev/2016-January/043483.html).
> 
> 
> Jürgen
> 
> 
> PS: @Saber may your GDAL 2.0.0dev come from experiments with this?  OSGeo4W
> has 1.11.3 (or 2.0.0 as experimental - which QGIS doesn't use - but no
> *dev).

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com