Re: [pypy-dev] stmgc: rethinking the GC?