#1152 inverse of matrix with taylor-exprs wrong

[Stavros Macrakis]

matrix([ 1, taylor(x,x,0,1)],
[taylor(y,y,0,1), 1 ]) ^^ -1
=>
matrix([ -1/(y^2-1) , y/(y^2-1) ],
[ y/(y^2-1) , -1/(y^2-1) ])

which is very wrong (it has apparently conflated variables x and y -- depending on the previous session history, it may return a result in x or in y). This happens both with ratmx=false and ratmx=true.

To calculate the correct answer:

subst([a=taylor(x,x,0,1),b=taylor(y,y,0,1)],
matrix([1,a],[b,1])^^-1 )
=>
matrix([1+x*y+...,-x+...],[-y+...,1+x*y+...])

Not sure whether this is a problem in the matrix code or the taylor code.

[Barton Willis]

Logged In: YES
user_id=895922
Originator: NO

A related bug is 939022.