ogr_geo_utils.cpp: Check LatA == 0.0 and cos_Heading == 0.0? Or with epsilons?

[Kurt Schwehr]

Even,

Did you really want an exact == 0.0 for these?

if( cos_complement_LatA == 0.0 && cos_Heading == 0.0 )

​https://trac.osgeo.org/gdal/browser/trunk/gdal/ogr/ogr_geo_utils.cpp?rev=39193#L170

[Even Rouault]

I'm not sure of all conditions that can lead to dfDenomin being 0, and appropriate thresholds. And potentially when it is very close to 0 the result might be inaccurate/wrong too. This function is indeed a bit fragile on corner cases that oss-fuzz finds. But for real world cases, those situations are quite unlikely. So I'm basically on a mode of "do fixes as much as oss-fuzz complains and not more".

[Even Rouault]

In 39389:

OGR_GreatCircle_ExtendPosition: avoid division by zero. Fixes ​https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=2424. Fixes #6931. Credit to OSS Fuzz