It doesn't make sense to warn about a switch case not being an
enumerator if the enum type has no enumerators; such a type is being
used as an opaque typedef rather than a true enumerated type.  This
usage makes more sense with C++11 enums with explicitly specified
underlying type, but I don't think it's necessary to limit the patch
to that case.

Tested x86_64-pc-linux-gnu, applying to trunk.
commit 673265c3d4cd77c7b0e927cc1547444ef869ad3a
Author: Jason Merrill <ja...@redhat.com>
Date:   Thu May 18 22:45:24 2017 -0400

            * c-warn.c (match_case_to_enum_1): Don't warn about enums with no
            enumerators.

diff --git a/gcc/c-family/c-warn.c b/gcc/c-family/c-warn.c
index e67ffb7..8971156 100644
--- a/gcc/c-family/c-warn.c
+++ b/gcc/c-family/c-warn.c
@@ -1069,6 +1069,10 @@ warnings_for_convert_and_check (location_t loc, tree 
type, tree expr,
 static void
 match_case_to_enum_1 (tree key, tree type, tree label)
 {
+  /* Avoid warning about enums that have no enumerators.  */
+  if (TYPE_VALUES (type) == NULL_TREE)
+    return;
+
   char buf[WIDE_INT_PRINT_BUFFER_SIZE];
 
   if (tree_fits_uhwi_p (key))
diff --git a/gcc/testsuite/g++.dg/cpp1z/byte2.C 
b/gcc/testsuite/g++.dg/cpp1z/byte2.C
new file mode 100644
index 0000000..6a395c1
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp1z/byte2.C
@@ -0,0 +1,13 @@
+// { dg-options "-std=c++17 -Wall" }
+
+#include <cstddef>
+
+bool white_space(std::byte x) {
+  switch (x) {
+  case std::byte{' '}: case std::byte{'\t'}: case std::byte{'\v'}:
+  case std::byte{'\f'}: case std::byte{'\n'}:
+  return true;
+  default:
+    return false;
+  }
+}

Reply via email to