//: So far we have local variables, and we can nest local variables of short //: lifetimes inside longer ones. Now let's support 'global' variables that //: last for the life of a routine. If we create multiple routines they won't //: have access to each other's globals. //: //: This feature is still experimental and half-baked. You can't name global //: variables, and so like in most tests they don't get checked for types (the //: only known hole in the type system, can cause memory corruption). We might //: fix these issues if we ever use globals. Or we might just drop the feature //: entirely. :(scenario global_space) def main [ # pretend address:array:location; in practice we'll use new 10:num <- copy 0 # refcount 11:num <- copy 5 # length # pretend address:array:location; in practice we'll use new 20:num <- copy 0 # refcount 21:num <- copy 5 # length # actual start of this recipe global-space:space <- copy 20/unsafe default-space:space <- copy 10/unsafe 1:num <- copy 23 1:num/space:global <- copy 24 ] # store to default space: 10 + (skip refcount and length) 2 + (index) 1 +mem: storing 23 in location 13 # store to chained space: (contents of location 12) 20 + (refcount and length) 2 + (index) 1 +mem: storing 24 in location 23 //: to support it, create another special variable called global space :(before "End is_disqualified Cases") if (x.name == "global-space") x.initialized = true; :(before "End is_special_name Cases") if (s == "global-space") return true; //: writes to this variable go to a field in the current routine :(before "End routine Fields") int global_space; :(before "End routine Constructor") global_space = 0; :(after "Begin Preprocess write_memory(x, data)") if (x.name == "global-space") { if (!scalar(data) || !is_space(x)) raise << maybe(current_recipe_name()) << "'global-space' should be of type address:array:location, but tried to write '" << to_string(x.type) << <