Thanks. Uploading your fix now. Should have done this a while ago, just slipped between the cracks. (In general, do feel free to upload unobjectionable fixes to my packages, I really don't mind. Especially if you also push to the packaging repo.)
I'm thinking it might make sense to ditch the non-reentrant version of this library. I doubt there would be any measurable performance loss. Can you imagine any objections? --Barak.
