This is what happens when you "reproject" a vector overlay (gshhs_land)
using ogr2ogr:
gdal_ogr2ogr_tm.png.
The rat's nest has been overlaid onto a normal (unprojected) gshhs_land.
The GeoTIFF and its
WKT.
gdal_rasterize_smear_v0_line.png (v0_landmass)
gdal_rasterize_smear_v0.png (polygon rather than linestring)
GSHHS shapefiles: v1.3 LINESTRINGs and v1.6 POLYGONs
My fixed geoloc code: gdalgeoloc.cpp
The patch which the maintainer won't apply: gdalgeoloc.diff
Without the patch: gdalwarp.png or gdalwarp.jpg
MODIS reprojection (250m band but not 250m geo), notice how various parts of the island are seen twice: gdalwarp_modis250m_geo1k.png
This shows how badly the MODIS image has been geolocated: gdal_rasterize_offset.png
