Fix golden tests to use the "target" Bazel configuration (#581)

In practice, this means that `executable_semantics` will be built and run using the configuration specified on the command line, rather than e.g. always using `-c opt`.

Also fix a bug exposed by this change.

executable_semantics/BUILD

@@ -92,17 +92,12 @@ EXAMPLES = [
     "experimental_continuation9",
 ]
 
-[genrule(
-    name = "%s_out" % e,
-    srcs = ["testdata/%s.carbon" % e],
-    outs = ["testdata/%s.out" % e],
-    # Suppress command errors.
-    cmd = "$(location executable_semantics) $< > $@ 2>&1 || echo EXIT CODE: $$? >> $@",
-    tools = [":executable_semantics"],
-) for e in EXAMPLES]
-
 [golden_test(
     name = "%s_test" % e,
+    cmd = "'$(location executable_semantics) $(location testdata/%s.carbon)'" % e,
+    data = [
+        ":executable_semantics",
+        "testdata/%s.carbon" % e,
+    ],
     golden = "testdata/%s.golden" % e,
-    subject = "testdata/%s.out" % e,
 ) for e in EXAMPLES]

executable_semantics/interpreter/typecheck.cpp

@@ -149,8 +149,8 @@ auto TypeCheckExp(const Expression* e, TypeEnv types, Env values,
       return TCResult(new_e, t, types);
     }
     case ExpressionKind::Index: {
-      auto res = TypeCheckExp(e->GetFieldAccess().aggregate, types, values,
-                              nullptr, TCContext::ValueContext);
+      auto res = TypeCheckExp(e->GetIndex().aggregate, types, values, nullptr,
+                              TCContext::ValueContext);
       auto t = res.type;
       switch (t->tag) {
         case ValKind::TupleV: {

executable_semantics/syntax/BUILD

@@ -47,7 +47,7 @@ cc_test(
     name = "paren_contents_test",
     srcs = ["paren_contents_test.cpp"],
     env = {
-        # FIXME: Remove this when leaks are fixed.
+        # TODO(#580): Remove this when leaks are fixed.
         "ASAN_OPTIONS": "detect_leaks=0",
     },
     deps = [