verified alignment of our custom utf8_codecvt_facet against the base class from which it is derived - std::codecvt.

This commit is contained in:
Robert Ramey
2014-02-27 08:02:01 -08:00
parent 3dd2674052
commit 094070e5e4

View File

@ -7,7 +7,7 @@
#define BOOST_UTF8_CODECVT_FACET_HPP
// MS compatible compilers support #pragma once
#if defined(_MSC_VER)
#if defined(_MSC_VER) && (_MSC_VER >= 1020)
# pragma once
#endif
@ -85,10 +85,6 @@
#include <boost/config.hpp>
#include <boost/detail/workaround.hpp>
#ifndef BOOST_NO_CXX11_HDR_CODECVT
#include <codecvt>
#endif
#if defined(BOOST_NO_STDC_NAMESPACE)
namespace std {
using ::mbstate_t;
@ -96,6 +92,19 @@ namespace std {
}
#endif
#if !defined(__MSL_CPP__) && !defined(__LIBCOMO__)
#define BOOST_CODECVT_DO_LENGTH_CONST const
#else
#define BOOST_CODECVT_DO_LENGTH_CONST
#endif
#if ! defined(BOOST_NO_CXX11_NOEXCEPT)
#define BOOST_CODECVT_NOEXCEPT noexcept
#else
#define BOOST_CODECVT_NOEXCEPT throw()
#endif
// maximum lenght of a multibyte string
#define MB_LENGTH_MAX 8
BOOST_UTF8_BEGIN_NAMESPACE
@ -119,9 +128,13 @@ protected:
) const;
virtual std::codecvt_base::result do_out(
std::mbstate_t & state, const wchar_t * from,
const wchar_t * from_end, const wchar_t* & from_next,
char * to, char * to_end, char * & to_next
std::mbstate_t & state,
const wchar_t * from,
const wchar_t * from_end,
const wchar_t* & from_next,
char * to,
char * to_end,
char * & to_next
) const;
bool invalid_continuing_octet(unsigned char octet_1) const {
@ -144,7 +157,9 @@ protected:
// == total octets - 1.
int get_cont_octet_out_count(wchar_t word) const ;
virtual bool do_always_noconv() const throw() { return false; }
virtual bool do_always_noconv() const BOOST_CODECVT_NOEXCEPT {
return false;
}
// UTF-8 isn't really stateful since we rewind on partial conversions
virtual std::codecvt_base::result do_unshift(
@ -152,50 +167,31 @@ protected:
char * from,
char * /*to*/,
char * & next
) const
{
) const {
next = from;
return ok;
}
virtual int do_encoding() const throw() {
virtual int do_encoding() const BOOST_CODECVT_NOEXCEPT {
const int variable_byte_external_encoding=0;
return variable_byte_external_encoding;
}
// How many char objects can I process to get <= max_limit
// wchar_t objects?
// This is the "official version which implements the standar
// library interface
// wchar_t objects?
virtual int do_length(
std::mbstate_t & state,
const char * from,
const char * from_end,
std::size_t max_limit
) const
#if BOOST_WORKAROUND(__IBMCPP__, BOOST_TESTED_AT(600))
throw()
#endif
{
return do_length(const_cast<const std::mbstate_t &>(state), from, from_end, max_limit);
}
// this version handles versions of the standard library which pass
// the state with a "const". This means early dinkumware standard library
virtual int do_length(
const std::mbstate_t & state,
BOOST_CODECVT_DO_LENGTH_CONST std::mbstate_t &,
const char * from,
const char * from_end,
std::size_t max_limit
) const;
// Largest possible value do_length(state,from,from_end,1) could return.
virtual int do_max_length() const throw () {
virtual int do_max_length() const BOOST_CODECVT_NOEXCEPT {
return 6; // largest UTF-8 encoding of a UCS-4 character
}
};
BOOST_UTF8_END_NAMESPACE
#endif // BOOST_UTF8_CODECVT_FACET_HPP