=> [[a=-%r1, b=%r1]] NO!
Remove the extra variable:
=> [[a = - c, b = %r2]] OK
Also works fine if you call algsys directly:
At first I thought this had something to do with the
variable *ordering* in the first case, but in fact it
happens with all orderings. The problem is the number
Note that this is *not* an artificial situation. It is
common to have CREs with more ratvars than they use
after arithmetic operations, e.g. rat(a+b)-rat(a).