bugfix: implicit items/pairs work better

compiler/semexprs.nim

@@ -589,9 +589,10 @@ proc semStaticExpr(c: PContext, n: PNode): PNode =
 
 proc semOverloadedCallAnalyseEffects(c: PContext, n: PNode, nOrig: PNode,
                                      flags: TExprFlags): PNode =
-  if efWantIterator in flags:
+  if flags*{efInTypeOf, efWantIterator} != {}:
-    result = semOverloadedCall(c, n, nOrig, {skIterator})
+    # consider: 'for x in pReturningArray()' --> we don't want the restriction
-  elif efInTypeOf in flags:
+    # to 'skIterator' anymore; skIterator is preferred in sigmatch already for
+    # typeof support.
     # for ``type(countup(1,3))``, see ``tests/ttoseq``.
     result = semOverloadedCall(c, n, nOrig, 
       {skProc, skMethod, skConverter, skMacro, skTemplate, skIterator})

tests/compile/tcolonisproc.nim

@@ -0,0 +1,12 @@
+
+proc p(a, b: int, c: proc ()) =
+  c()
+
+ 
+p(1, 3): 
+  echo 1
+  echo 3
+    
+p(1, 1, proc() =
+  echo 1
+  echo 2)

todo.txt

@@ -82,7 +82,6 @@ version 0.9.XX
 Not essential for 1.0.0
 =======================
 
-- investigate the implicit items/pairs issue
 - optimize method dispatchers
 - ``with proc `+`(x, y: T): T`` for generic code
 - new feature: ``distinct T with operations``