Nested functions in Locals analysis about pythran

[coyotte508]

Related to nested functions, the assignment of the locals of a module node.

I don't understand why it's not stored using .copy() as it is everywhere else.

Because as it is now, it would store the locals encoutered until and including the locals within the first nested function seen. But after that, this line would prevent any more locals to be stored in that self.results[module_node].

[serge-sans-paille]

It is assumed that locals() in Pythran always yields set([]) for a module. And this particular set of locals is never updated, it's just here for consistency with other nodes, as far as I can tell.

[coyotte508]

but since it's not copied, it shares an address with Locals.locals, and so when Locals.locals is updated, the locals for the module are updated too. (current behavior)

If a .copy() was added, it would stay as set([]) and not be updated.