Infinite recursion in test programs

[bl0ckeduser]

Hi, this is to report that the following test programs --
reverseif.ns
namespace.ns,
have an infinite recursion problem.

For example, in "reverseif.ns":

&if &_if =
{ rot _if } &if = #The 'reversed if'.

"_if" and "if" point to the same
object, thus, the following:
{ rot _if } &if =
is self-referential, so its expansion
leads to infinite recursion.

The true problem seems to be
that the "if" object's contents don't
get copied to a new location proper
to "_if".

[nikki93]

I noticed this too... It's because earlier &x pushed the function object onto stack, while now it just pushes a symbol and then 'follows' the symbol later, so no copying is involved.

One way to fix this is to also have another prefix, like 'x for example, that pushes a function object onto the stack... But I think this makes things messier.

What do you suggest?

[nikki93]

(closed issue by mistake - reopened it)

[bl0ckeduser]

I suggest that the interpreter handle this as a special case and make copies of function objects when function symbol assignment is encountered.

This is pretty easily coded, and as a proof-of-concept, the patch below, which implements the idea, seems to work pretty
well, as after applying it, all the test programs run correctly.

diff --git a/src/nsbuiltins.c b/src/nsbuiltins.c
index 46e0b13..80b0427 100644
--- a/src/nsbuiltins.c
+++ b/src/nsbuiltins.c
@@ -39,6 +39,23 @@ void ns_assign()

     struct ns_obj *p = ns_searchNamespace(ns_currNamespace, s.u.sym->arr);

+    if(NS_ISSYM(o))
+    {
+        struct ns_obj *q = ns_searchNamespace(ns_builtinsSpace, o.u.sym->arr);
+    
+        if(q->type == TY_FUNC)
+        {
+                if(p->type != TY_EMPTY)
+                {
+                        p->type = TY_FUNC;
+                        p->u.f = q->u.f;  
+                }
+                else
+                        ns_addToNamespace(ns_currNamespace, s.u.sym->arr, ns_makeFuncObj(q->u.f));
+                return;
+        }
+    }
+
     if (p->type != TY_EMPTY)
         *p = o;
     else

(Following this scheme, ns_parentAssign also has to be patched in a very similar way).

(EDIT: rephrased first paragraph for clarity)