ifstream 's mistake in vc8, correct in vc7.1

When I use ifstream 's constructor to open a text file,
the text file's filename contain chinese char,
and the filename is in Multi-Byte Character Set style (const char *),
I success in vc7.1 but fail in vc8.

Trace the library's source code, I found in vc8, the ifstream will call
mbstowcs_s to convert the input filename form MBCS to wide-char,
mbstowcs_s will call _mbstowcs_l_helper to do that, but it get
the wrong string in my english ver. OS.

I am so curious why to do that, it is total excrescent.

PS:
I use API MultiByteToWideChar can get the correct string.

    // Convert ANSI string to Unicode
    int result = MultiByteToWideChar( CP_ACP, 0, src,
        src_len, wchar_buf,
        BUF_LEN);

----------------- test code ------------------------------

#include "stdafx.h"
#include <iostream>
#include <fstream>

using namespace std;

int _tmain(int argc, _TCHAR* argv[])
{
    ifstream ifs("c:/事.txt");

    if (!ifs)
    {
        return -1;
    }

    return 0;
}

Answer this question

ifstream 's mistake in vc8, correct in vc7.1

Dhirendra
Interesting, thanks for the info, I was trying on a Chinese machine (i.e. system locale, or language for non-Unicode programs set to Chinese in regional settings) and it worked.
Roundtwoit
As you see in the code, it uses the CRT's current locale setting (a global state setting set to "C" by default) As I said before you just need to call

setlocale(LC_CTYPE, "")

once at the beginning of your program.

Then all will work properly.
chipscap
You first need to call setlocale(LC_CTYPE, "") otherwise it will use the "C" locale (look at C:\Program Files\Microsoft Visual Studio 8\VC\crt\src\mbstowcs.c and look for _CLOCALEHANDLE )
Tatjana
Thanks, Ted. I'm not very familiar with Unicode stuff.
The problem occurs when I switch my app from mbcs to Unicode. So in mbcs mode(which i use ifstream), it works well (except that I have to call _tsetlocale(LC_CTYPE,_T(".ACP")) as you suggested to make 2005 correctly handle mbcs filename). The issue appears when i turn to Unicode, then the problem is related to the following funciton (by the way, when using unicode, I need not use setlocale, the overloadded w_char ver of "open" can handle the unicode filename correctly):
codecvt_base::do_in
it use following code to test if the current char in buffer is partial
switch (_Bytes = _Mbrtowc(_Mid2, _Mid1, _Last1 - _Mid1,
&_State, &_Cvt))
and it always return ok by default (in c locale, which _Cvt==0).
And just call _tsetlocale(LC_CTYPE,_T(".ACP")) will not influence the value of _Cvt, so I have call locale::global(locale(".ACP",locale::ctype)) explicitly. It seems that the c & c++ locale setting function are not related
Satpal
Ted. wrote:
You first need to call setlocale(LC_CTYPE, "") otherwise it will use the "C" locale (look at C:\Program Files\Microsoft Visual Studio 8\VC\crt\src\mbstowcs.c and look for _CLOCALEHANDLE )

Thx, I call setlocale( LC_ALL, ".ACP" ); first to solve this problem, But...... , I did not like the way MS change ifstream that much.
Gregor Bussmann
I found that the default locale of fstream is "C"
when i use the following code to read ansi file
wifstream f(_T("E:\\Table1.txt"));
wstring str;
f>>str;
then f will not correctly handle mbcs chars (for example, it will always treat one chinese character as two seperate chars. like read 0xb7 0x9a as L0xb7 L0x9a)
so i have to add following code after every ifstream object
f.imbue(locale(".ACP")); //my windows system is chinese
Is there anyway to globally set the "default" locale used by fstream to ACP
(i find that the default locale is set in the following code:
basic_streambuf()
: _Plocale(_NEW_CRT(locale))
which in turn call _Getgloballocale
it seems that just call _tsetlocale(LC_ALL,_T(".ACP")); doesn't take effect
)
Pamish
Thanks for your answer.
but as i tried, just call setlocale will not take effect
now i use "locale::global(locale(".ACP",locale::ctype))", that works well.
hoosierboy
Please feel free to log a suggestion issue showing your opinion at http://lab.msdn.microsoft.com/productfeedback/default.aspx

Thanks in advance for taking the time to share your views.

Thanks,
Ayman Shoukry
VC++ Team
Andy Knight
I encountered the same problem.
ifstream cannot correctly handle MBCS filename in ANSI mode
So I have to switch to use CFile instead
ifstream 's mistake in vc8, correct in vc7.1

Answer this question