nrroots from rootsur.lib gives wrong results

[steenpass]

According to the online manual, nrroots from rootsur.lib gives "the number of real roots" of a polynomial, but it gives strange results in some cases:

> option(noloadLib);
> LIB "rootsur.lib";
> ring r = 0,x,ds;
> nrroots((x-1)*x*(x+1));
1
> 

One could argue that this result is correct in some sense because ds is a local monomial order and locally at 0, there is only one root. But then what about the following cases?

> nrroots((x-1)*(x+1));
2
> nrroots((x-1)*(x+1)*(x+2));
1
> 

[hannes]

a global ordering is required for proc sturmha (etc.) fixed with rev. 14388