mapserver-4.8.0-rc2 wfs filter bug

[rchamber@…]

wfs filter doesn't work with layers that have a tileindex.

vtable is accessed in FLTApplyFilterToLayer() in mapogcfilter.c without calling
msLayerOpen()...as a result layer.connectiontype may not be correctly
initialized e.g. MS_TILED_SHAPEFILE.

[rchamber@…]

testcase

[rchamber@…]

(From update of attachment 415)
unzip

[rchamber@…]

patch

[assefa]

looking into this.

[assefa]

e-mail conversation 

from Assefa

>I was looking into bug 1615 and would like to have your comments on
>> it.  What we end up in this case is that a tile layer is using
>> functions of a flat shape file.  From what I can see, a possible
>> source of problem in this case would be that function
>> msInitializeVirtualTable does not include some logic that is
>> included for example in msLayerOpen :
>> 
>>  if(layer->features && layer->connectiontype != MS_GRATICULE )
>>     layer->connectiontype = MS_INLINE;
>> 
>>   if(layer->tileindex && layer->connectiontype == MS_SHAPEFILE)
>>     layer->connectiontype = MS_TILED_SHAPEFILE;
>> 
>>   if(layer->type == MS_LAYER_RASTER )
>>     layer->connectiontype = MS_RASTER;
>> 
>>  Do it make sense to have these in msInitializeVirtualTable  ?


From Jani

Hello,

Thanks for looking this. I think it would be a good thing to move this
logic to msInitializeVirtualTable.  In the perfect world msLayerOpen
should be the first action with layer, but we are any way assuring
that vtable is initialized by initializing it if it's null, so it
should be ok and correct thing to move this logic out of msLayerOpen.

At the same time, could you also commit this fix:

Index: maplayer.c
===================================================================
RCS file: /data2/cvsroot/mapserver/maplayer.c,v
retrieving revision 1.108
diff -u -r1.108 maplayer.c
--- maplayer.c  29 Oct 2005 02:03:43 -0000      1.108
+++ maplayer.c  31 Jan 2006 16:39:44 -0000
@@ -1138,7 +1138,7 @@
 createVirtualTable(layerVTableObj **vtable)
 {
     *vtable = malloc(sizeof(**vtable));
-    if ( ! vtable) {
+    if ( ! *vtable) {
         return MS_FAILURE;
     }
     return populateVirtualTable(*vtable);

I think both of these fixes should be back ported to 4.8, so it would be nice
if they are within one commit.

[assefa]

This should be fixed in trunk. Please reopen if there are an y issues.