p_psPlaneTransformLinearInvert results in NaN

[eugene]

I had errors in the results from p_psPlaneTransformLinearInvert, with invalid
NaN coming out of the transformation. The existing code comments out the matrix
inversion form of the solution and uses a difficult-to-read solution. I suspect
there is an error in this math (D not A in the X1 = ... ?). the IfA-supplied
code had an error as well in the math, exchanging the x[1][0] and y[0][1] terms
and getting the signs wrong on the [0][0] term arithmetic. I re-wrote the math
using more explicit names than the a,b,c,d,e,f variables.

It is also important to note that this inversion is only valid if x->nX == 1 and
y->nY == 1, and if x->nY <= 1, and y->nX <= 1.

I'm supplying these changes in cvs tag eam_rel8_b0

[Paul Price]

[This has no impact on my stac code, since I don't use the linear inversion.]

[gusciora@…]

I fixed that bug a few days ago in our current CVS tree: there was a problem
with the math and I derived a determinant version of the code.

I'm not sure I entirely agree with the conditional on nX and nY. I had imagined
that there would be linear polynomials with nX and nX greater than 1, however
the coefficients being 0.0 for the higher order terms.

[eugene]

I agree: you can have a linear polynomial if the higher terms are masked. But
you also have to check and exclude polynomials which are of order 0.

[gusciora@…]

This is in our current CVS development tree.

[Paul Price]

Bug has been resolved.... closing.

[Paul Price]

Bugs have been fixed... closing.