[mapguide][MG129][Modified] Great Circle azimuth computation is numerically unstable

[Traian Stanev]

You can view the artifact detail at the following URL:

    https://mapguide.osgeo.org/servlets/Scarab/id/MG129

Type:
Defect

Artifact ID:
MG129 (Great Circle azimuth computation is numerically unstable)

Modified by:
Traian Stanev
traianstanev (traian.stanev at autodesk.com)

The following modifications were made to this artifact:
---------------------------------------------------------------------

-- Artifact Status changed:
Old value:
New
New value:
Resolved
-- Resolution Description set to new value:
Added a special code path for displacement angles less than 1e-7. The new code uses small angle approximations to avoid doing inverse cosine, which was the cuase of the failure.

I was debating whether to use 1e-6 or 1e-7 for the small angle cutoff -- either is small enough -- but LatLon buffers seemed to work ok with 1e-7, so that's what I left it at. If we see more crashes, we should make the cutoff angle bigger (1e-6)

Note that the great circle distance and great circle offset code also has some problematic code, which would break down for small values, but we can fix this once we see it fail.
-- Resolution set to new value:
Submitted



---------------------------------------------------------------------
This message was automatically generated by Project Tracker.