Fix tuple pattern matching issue where some elements would violate ca…

…ps. (#2658)

When pattern-matching tuples that have the same number of elements,
ponyc was treating the entire match as violating capabilities when
only one of the elements was violating. In truth we should be able
to recognize that tuples are distinct from the other elements.

This patch fixes `is_tuple_match_tuple` to return MATCHTYPE_REJECT
as soon as any pairwise element comparison returns MATCHTYPE_REJECT.

The fix was pinpointed by @Praetonus when I discussed this issue with
him.

src/libponyc/type/matchtype.c

@@ -273,7 +273,7 @@ static matchtype_t is_tuple_match_tuple(ast_t* operand, ast_t* pattern,
         break;
     }
 
-    if(ok == MATCHTYPE_DENY)
+    if(ok != MATCHTYPE_ACCEPT)
       break;
 
     operand_child = ast_sibling(operand_child);

test/libponyc/matchtype.cc

@@ -309,6 +309,7 @@ TEST_F(MatchTypeTest, Tuples)
     "interface Test\n"
     "  fun z(c1: C1, c2: C2, c3: C3, t1: T1, t2: T2,\n"
     "    c1c1: (C1, C1), c1c2: (C1, C2), c1c3: (C1, C3),\n"
+    "    c2i1: (C2, I1), c3i1tag: (C3, I1 tag),\n"
     "    t1t2: (T1, T2), i1: I1, i2: I2)";
 
   TEST_COMPILE(src);
@@ -328,6 +329,11 @@ TEST_F(MatchTypeTest, Tuples)
   ASSERT_EQ(
     MATCHTYPE_REJECT, is_matchtype(type_of("i2"), type_of("c1c1"), NULL, &opt));
 
+  ASSERT_EQ(
+    MATCHTYPE_REJECT, is_matchtype(type_of("c2i1"), type_of("c3i1tag"), NULL, &opt));
+  ASSERT_EQ(
+    MATCHTYPE_REJECT, is_matchtype(type_of("c3i1tag"), type_of("c2i1"), NULL, &opt));
+
   // We can't make types with don't cares in as the type of a parameter. Modify
   // t1t2 instead.
   ast_t* t1t2 = type_of("t1t2");