如何将wstring缓冲区发送到子进程的stdin？

Question

I'm having trouble writing a WString to the STDIN of a child process. If I have only acii character string (eg: @WSX3edc), the code works fine, but if it contains a non ascii character (eg: @WSX3edcß) it fails.

The child process is 7zr.exe (7Zip cmd line version). The input I'm writing to the STDIN is the password to extract the file.

// inject password
wPassword.append(password);
wPassword.append(L"\n"); \\For carriage return
...
DWORD dwBytesToWrite = wPassword.length()*sizeof(wchar_t);
DWORD dwBytesWritten = 0;
char szBuffer[1024] = "\0";
wcstombs(szBuffer, wPassword.c_str(),wcslen(wPassword.c_str())+1);
dwBytesToWrite = strlen(szBuffer);
if (!WriteFile(hInput, szBuffer, dwBytesToWrite, &dwBytesWritten, NULL)) {
    std::cout<<"write file failed"<<GetLastError()<<std::endl;
    goto Cleanup;
}

The writefile always succeed but some how the file extraction is not successful due to faulty password injecting mechanism.

Createprocess for this looks like : (si object has the STDIN and STDOUT streams set using a CreatePipe earlier)

if(!CreateProcess((LPWSTR)cmd, (LPWSTR)cmdArgs, NULL, NULL, TRUE, NORMAL_PRIORITY_CLASS,
        NULL, NULL, &si, &pi)) {
         std::cout<<"7zr.exe process creation failed "<<GetLastError()<<std::endl;
         goto Cleanup;
}

Note : 7zr.exe works just fine with this particular password , if we run it on command-line and paste this password. The extraction works fine.

Answer 1

If the narrow character set doesn't have the relevant password character, you can't use this approach. Instead find what option 7zr has for specifying the password. I don't have an executable called 7zr but I do have 7z , and the command 7z | find /i "pass" 7z | find /i "pass" worked nicely.

---

In other news:

• The variable dwBytesToWrite is initialized with one value, only to be reassigned a few lines later, without having been used.

• goto Cleanup is generally ungood in C++. If you want guaranteed cleanup use a destructor (the technique called RAII, read up on it).

• Microsoft's Hungarian notation, with prefixes such as sz and dw , is generally an abomination. It once, in the 1980's, supported the help system in Microsoft's Programmer's Workbench. AFAIK that product hasn't existed the last 30 years or so.

• The C cast in (LPWSTR)cmd can easily introduce a bug. Use const_cast where you want to cast constness. Then it would be more clear that this cast is incorrect: you need a mutable buffer.

• Instead of reporting a failure to the standard output stream, via std::cout , consider using the standard error stream, via either std::cerr or std::clog . Better, don't do i/o at the place where a failure is detected, but throw an exception to let the calling code deal with it. The calling code can't remove output that's already, uh, output.

Answer 2

wcslen(wide) returns the number of wide characters in its argument wide ( see ).

wcstombs(narrow,wide,len) writes no more than len bytes to narrow ( see ).

Now if we always had one wide character = one narrow character = one byte, it wouldn't have much sense to have two varieties of characters, would it?

As soon as you have a wide character that translates to more than one narrow character, there is undefined behaviour.

Since your szBuffer is of fixed size, you could just as well write

wcstombs(szBuffer, wPassword.c_str(), sizeof(szBuffer));