Tileindex memory leaks in raster code

[warmerdam]

Running a simple raster map with a tile index reports the following
memory leak (and others not shown here) using valgrind:

==17403== 4 bytes in 1 blocks are definitely lost in loss record 1 of 20
==17403==    at 0x400268D5: malloc (vg_replace_malloc.c:160)
==17403==    by 0x80A3D35: msDBFGetValueList (mapxbase.c:831)
==17403==    by 0x806D449: msLayerNextShape (maplayer.c:289)
==17403==    by 0x80B9841: msDrawRasterLayerLow (mapraster.c:1486)
==17403==    by 0x807201A: msDrawRasterLayer (mapdraw.c:930)
==17403==    by 0x80708B3: msDrawLayer (mapdraw.c:581)
==17403==    by 0x806FB2B: msDrawMap (mapdraw.c:295)
==17403==    by 0x804EA81: main (shp2img.c:159)

The problem seems to be that msDrawRasterLayerLow() reused the shapeObj, 
but the code in msLayerNextShape() does not try to clear old "values" lists
off the shapeObj passed in before assigning a new one in this bit of code:

    shape->values = values;
    shape->numvalues = layer->numitems;

Should the msDrawRasterLayerLow() be calling msFreeShape() each time through
the loop, instead of just at the end?

I believe this problem exists in the 4.2 branch as well as the development
tree.

[sdlime]

Is fixing this in 4.2 worth it?

[fwarmerdam]

I guess I had forgotten it was an issue in 4.2 as well. I don't think it would
be wise to try and change this in 4.2 at this point as it has a high risk
of introducing bugs.  It might be wise to fix in 4.2 release for a 4.2.1 
if there will be one. 

[warmerdam]

I have run valgrind on shp2img on the msautotest/gdal/tileindex.map map, and it does not show this memory leak. I believe it was resolved some time ago.

Closing.