[gdal-dev] OGRLayer::SetFeature fails with out of date OGRFeature

Ari Jolma ari.jolma at gmail.com
Wed Apr 28 02:34:57 EDT 2010

Even Rouault wrote:
> Ramiro,
> It is illegal to change the layer definition with CreateField() if you have 
> pre-existing features of that layer. This is documented here : 
> http://www.gdal.org/ogr/classOGRFeatureDefn.html#40e681d8464b42f1a1fac655f16ac3dd

Illegal yes, but IMHO segfault should never be an option in a case like 
this. Just consider an interactive program with real-time scripting with 
one of the bindings languages.

The segfault happens in this case in OGRFeature::IsFieldSet 
(ogr_feature.h:193), which does not check that iField is within legal 
limits. The method could check that iField is less or equal to 
poDefn->nFieldCount. That should not be a problem. Probably similar 
checks could/should be done in other methods as well.

At least in shape driver the layer does not assume that the feature 
schema is exactly the same as in layer: it asks the data from the 
feature using type casting methods (GetFieldAsInteger and friends).

I'm going to make a bug report from this.

Best regards,


> There's no way to check if the feature is "valid". So, just don't do it ;-)
> Le Tuesday 27 April 2010 21:17:17 Ramiro Gonzalez, vous avez écrit :
>> My application uses OGR to edit ESRI Shapefiles.
>> If I follow this steps:
>> 1_ Create a feature(OGRFeature::CreateFeature) using the feature definition
>> from an existing layer(layerA)
>> 2_ Add a field (OGRLayer::CreateField) to layerA
>> 3_ Try to save the feature (OGRLayer::SetFeature) in layerA
>> I receive a SIGSEGV from function OGRFeature::SetField. I believe this
>> happens because I modify the layer definition in step 2 leaving the feature
>> definition out of date.
>> I'd like to know if there is any way to check if the feature is valid
>> before calling SetFeature.
>> Thanks
>> Ramiro
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/gdal-dev

More information about the gdal-dev mailing list