This is the mail archive of the mailing list for the libstdc++ project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: V3 PATCH: numeric_limits<> support, fix PR/3865

Neil Booth <> writes:

| Gabriel Dos Reis wrote:-
| > | > wchar_t is a keyword in C++, how do you easily test its signness with
| > | > if(0 or #if at compile-time and having that information in accordance
| > | > with the compiler's own idea?
| > | 
| > | if (L'\0' - 1 < 0)
| > |    wchar_t_is_signed;
| > | else
| > |    wchar_t_is_unsigned;
| > 
| > That doesn't work with -fshort-wchar.
| It should do.

But it doesn't.  

  { "-fshort-wchar",
    N_("Override the underlying type for wchar_t to `unsigned short'") },

Now, C++ rules require

     L'\0' - 1

be evaluated after "the usual arithmetic conversions are performed".
The usual arithmetic conversions here say that (C++, 4.5/2):

  An rvalue of type wchar_t (3.9.1) or an enumeration type (7.2) can
  be converted to an rvalue of the first of the following types that
  can represent all the values of its underlying type: int, unsigned
  int, long, or unsigned long. 

If "short" is 16-bit wide and int is 32-bit wide  -- common assumptions,
especially on most (all?) GNU/Linux -- then the value 
L'\0' - 1 is negative and wchar_t_is_signed is defined which is
clearly bogus since the underlying type will be "unsigned short".

-- Gaby

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]