Hello,
I was investigating wc*tomb* and mb*towc* functions in CRT and comparing their
behavior to other implementations.
Take the following example:
```
mbrtowc (NULL, s, 1, ps)
mbrtowc (NULL, s + 1, 1, ps)
```
Here, `s` is a pointer to multibyte (DBCS) character, but since n==1 mbrtowc
returns (size_t)-2 and updates ps. Next call completes converting multibyte
character. What's the return value? CRT returns 2 while glibc returns 1.
It seems to me that ISO C and POSIX specify different behavior for this case.
cppreference[1] has the following:
```
the number of bytes [1...n] of the multibyte character successfully converted
from s.
```
In this case, CRT seems good. We converted a DBCS character with length of 2.
POSIX[2] has the following:
```
between 1 and n inclusive
If the next n or fewer bytes complete a valid character (which is the value
stored); the value returned shall be the number of bytes that complete the
character.
(size_t)-2
If the next n bytes contribute to an incomplete but potentially valid
character, and all n bytes have been processed (no value is stored).
```
glibc seems good here. The first call *consumed* n (1) bytes. Second call
consumed 1 byte to complete it, so the return value is 1.
Any ideas here? Or should I try my luck on gnulib list?
I guess consistency with CRT would be preferred on Windows. I also attached a
simple program which you can compile with CRT or glibc to observe this.
- Kirill Makurin
[1] https://en.cppreference.com/w/c/string/multibyte/mbrtowc
[2] https://pubs.opengroup.org/onlinepubs/9799919799/functions/mbrtowc.html
#define __USE_MINGW_ANSI_STDIO 0
#include <assert.h>
#ifdef _WIN32
#include <fcntl.h>
#include <io.h>
#endif
#include <locale.h>
#include <stdio.h>
#include <stdlib.h>
#include <wchar.h>
#ifdef _WIN32
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#endif
/**
* With msvcrt.dll, define EMU to use mingw-w64's replacement for mbrtowc
*/
typedef size_t (* mbrtowc_t) (wchar_t *, const char *, size_t, mbstate_t *);
int main (void) {
#ifdef _WIN32
_setmode (_fileno (stdout), _O_U8TEXT);
_setmode (_fileno (stderr), _O_U8TEXT);
_wsetlocale (LC_ALL, L"Japanese_Japan.20932");
#else
setlocale (LC_ALL, "ja_JP");
#endif
mbrtowc_t func = NULL;
#if defined (_UCRT) || defined (EMU) || !defined (_WIN32)
func = mbrtowc;
#else
HANDLE msvcrt = LoadLibraryW (L"msvcrt.dll");
if (msvcrt == NULL) {
abort ();
}
func = (mbrtowc_t) GetProcAddress (msvcrt, "mbrtowc");
if (func == NULL) {
abort ();
}
#endif
/* EUC-JP (code page 20932) */
char lead1[] = {(char) 161};
char lead2[] = {(char) 162};
char valid1[] = {(char) 161, (char) 162};
char valid2[] = {(char) 162, (char) 161};
mbstate_t state = {0};
wchar_t wc = WEOF;
assert (func (&wc, lead1, 1, &state) == (size_t) -2);
#ifdef _WIN32
assert (func (&wc, lead2, 1, &state) == 2);
#else
assert (func (&wc, lead2, 1, &state) == 1);
#endif
wprintf (L"%lc\n", wc);
assert (func (&wc, lead2, 1, &state) == (size_t) -2);
#ifdef _WIN32
assert (func (&wc, lead1, 1, &state) == 2);
#else
assert (func (&wc, lead1, 1, &state) == 1);
#endif
wprintf (L"%lc\n", wc);
assert (func (&wc, valid1, 2, &state) == 2);
wprintf (L"%lc\n", wc);
assert (func (&wc, valid2, 2, &state) == 2);
wprintf (L"%lc\n", wc);
return 0;
}
_______________________________________________
Mingw-w64-public mailing list
Mingw-w64-public@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
