[Bug gdb/31331] Wenum-constexpr-conversion should be fixed, soon treated as a hard error

["carlosgalvezp at gmail dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=31331

--- Comment #29 from Carlos Galvez <carlosgalvezp at gmail dot com> ---
Hmm, applying Aaron's suggestion above, I found what I believe to be a bug.

With this change:

+template <typename T>
+struct enum_decays_to_signed
+{
+  static constexpr bool value = std::is_signed_v<decltype(true ?
std::declval<T>() : std::declval<std::underlying_type_t<T>>())>;
+};
+
 template<typename T>
 struct enum_underlying_type
 {
   DIAGNOSTIC_PUSH
   DIAGNOSTIC_IGNORE_ENUM_CONSTEXPR_CONVERSION
   typedef typename
-    integer_for_size<sizeof (T), static_cast<bool>(T (-1) < T (0))>::type
+    integer_for_size<sizeof (T), enum_decays_to_signed<T>::value>::type
     type;
   D

I get:

./dwarf2/cooked-index.h:212:22: error: use of deleted function ‘constexpr void
operator~(enum_type) [with enum_type = cooked_index_flag_enum;
<template-parameter-1-2> = void; <template-parameter-1-3> = void]’
  212 |     flags = flags & ~IS_PARENT_DEFERRED;
      |                      ^~~~~~~~~~~~~~~~~~


This is because that enum has an underlying type of "unsigned char":

enum cooked_index_flag_enum : unsigned char


Unsigned char does decay/promote to signed integer when performing arithmetic
on it:

https://godbolt.org/z/Tfv7e3f8Y

So current master is allowing operator~ on an enum of type unsigned char, which
will decay to signed int before applying operator~ (which is what we want to
avoid, since bitwise operations shall be performed on unsigned types only).

-- 
You are receiving this mail because:
You are on the CC list for the bug.