Re: Scoping of procedure-level variables

Adam Glauser wrote:

Joe Pluta wrote:

David Gibbs wrote:

Pardon my denseness here ... but what does 'lexical' mean in this context?

Another word for static. I pretty much always use static.

Joe, what you're referring to here is a type of memory allocation, automatic vs. static variables. As I understand it, this is different from lexical vs. dynamic scope.

No it's not. I'm referring to scope. It just so happens that scope is pretty much inextricably tied to allocation: a traditional static variable is typically in the program's variable space which is allocated the first time the program runs (or from the heap for more dynamic models), while automatic variables are typically on the stack.

By allocating in program storage or on the heap, the value stays intact between program calls, which is what makes the scope static. Variables allocated on the stack are by definition lost when the procedure returns, and thus fit the definition of automatic.

The question that arises with multi-threading is how to deal with thread-specific statics. Clearly they can't be allocated to a single memory location like traditional statics, and yet they must remain intact between calls. Thus thread-specific static variables must use heap allocation.

Joe