gdalwarp does not always create proper masks when using the -cblend option

[winkey]

chunks within the envelope of the cutline but not intersecting the cutline +/- blend distance appear to be improperly masked.

these areas outside of the poly results in the source data written to the dest dataset when it should not be.

and sometimes these areas on the inside of a poly the source data is not written to the dest dataset when it should be

[winkey]

screensot of data outside of the cutline poly copyed to the dst dataset

[winkey]

if you test if the poly intersects the chunk +/- the blend distance here... http://trac.osgeo.org/gdal/browser/trunk/gdal/alg/gdalcutline.cpp?rev=26326#L90

this will fix the issues outside the poly but not inside

[winkey]

if its is also does not intersect the poly (forced to a linestring), this will fix the issues inside

[winkey]

commited fix to svn in rev 26791

also want to note time to blend a 16k x 16k pixel image with another image went from 8 min to 2 min on my system