Cyclomatic complexity for Smalltalk

I need to compute the cyclomatic complexity of Pharo methods.

There is a tool to do that in Moose but its implementation is quite unreliable since it explicitly lists the branching selectors and loop selectors.

Even the comment in the method says the implementation is wrong.

The problem is that in Smalltalk, potentially any selector can branch or perform loop. And you can not know it in advance.

Anyone here already tried to address this problem?

Thanks in advance,

Julien