ZorbaImpl::init might call GlobalEnvironment::init with a NULL store pointer

I removed the store == NULL test from the assertion.

Actually, would it make sense to completely remove the store parameter, make init a 0-ary function, and just use the store manager's singleton for initializing the global environment, since the variable store may only have a given value?

Ghislain, yes, what you say makes sense, given the current implementation. But the current API reflects our long-term goal of separating zorba from the store (into separate libraries). The application may link with several stores and start zorba with a different store each time. In this scenario, the store must be passed as an input to zorba.