[C++,/,RFC] PR 51242

Message ID 50B38DEC.8090004@oracle.com
State New
Headers show

Commit Message

Paolo Carlini Nov. 26, 2012, 3:42 p.m.

this PR is about the rejection (in C++11 mode of course) of:

enum class MyEnum { A = 1 };

struct MyClass
   MyEnum Field1 : 3;

whereas the corresponding unscoped enum case already works. As noticed 
by Jon, it seems that the below straightforward patchlet is enough to 
solve the problem but then we have a warning issue:

51242.C:5:19: warning: ‘MyClass::Field1’ is too small to hold all values 
of ‘enum class MyEnum’ [enabled by default]

which is easily explained: the underlying type is fixed (to an implicit 
int type), thus in finish_enum_value_list, fixed_underlying_type_p is 
true and the code "restricting" the ENUM_UNDERLYING_TYPE for diagnostic 
purposes doesn't run. In other terms, we warn for the same reason we for 
example -Wnarrowing warn (before erroring out) for:

enum class Code {
   SUCCESS = 0

Code a;

short r[] = {a};

That seems intended, but especially annoying in the bitfield case. An 
idea could be giving a name to the "too small to hold all values" 
warning and making possible disabling it. What do you suggest?



Jason Merrill Feb. 15, 2013, 5:40 p.m. | #1
The patch is OK.  To deal with the warning, I would suggest putting the 
constrained type somewhere other than ENUM_UNDERLYING_TYPE that's shared 
between enums with fixed and non-fixed underlying types; having it in 
ENUM_UNDERLYING_TYPE isn't really correct anyway.



Index: decl2.c
--- decl2.c	(revision 193814)
+++ decl2.c	(working copy)
@@ -1030,7 +1030,7 @@  grokbitfield (const cp_declarator *declarator,
   if (TREE_CODE (value) == VOID_TYPE)
     return void_type_node;
       && (POINTER_TYPE_P (value)
           || !dependent_type_p (TREE_TYPE (value))))