iterable[T] (#17196)

* fix failing test toSeq in manual which now works
* changelog
* reject proc fn(a: iterable)
* add iterable to spec
* remove MCS/UFCS limitation that now works

compiler/types.nim

@@ -460,8 +460,8 @@ const
     "lent ", "varargs[$1]", "UncheckedArray[$1]", "Error Type",
     "BuiltInTypeClass", "UserTypeClass",
     "UserTypeClassInst", "CompositeTypeClass", "inferred",
-    "and", "or", "not", "any", "static", "TypeFromExpr", "out ",
-    "void"]
+    "and", "or", "not", "any", "static", "TypeFromExpr", "concept", # xxx bugfix
+    "void", "iterable"]
 
 const preferToResolveSymbols = {preferName, preferTypeName, preferModuleInfo,
   preferGenericArg, preferResolved, preferMixed}
@@ -645,6 +645,11 @@ proc typeToString(typ: PType, prefer: TPreferedDesc = preferName): string =
     of tyDistinct:
       result = "distinct " & typeToString(t[0],
         if prefer == preferModuleInfo: preferModuleInfo else: preferTypeName)
+    of tyIterable:
+      # xxx factor this pattern
+      result = "iterable"
+      if t.len > 0:
+        result &= "[" & typeToString(t[0]) & ']'
     of tyTuple:
       # we iterate over t.sons here, because t.n may be nil
       if t.n != nil:
@@ -1191,7 +1196,7 @@ proc sameTypeAux(x, y: PType, c: var TSameTypeClosure): bool =
     result = sameTypeOrNilAux(a[0], b[0], c) and
         sameValue(a.n[0], b.n[0]) and
         sameValue(a.n[1], b.n[1])
-  of tyGenericInst, tyAlias, tyInferred:
+  of tyGenericInst, tyAlias, tyInferred, tyIterable:
     cycleCheck()
     result = sameTypeAux(a.lastSon, b.lastSon, c)
   of tyNone: result = false