Re: [ros-dev] [ros-diffs] [hbelusca] 71692: [SUBST]: CORE-10681 #comment Apply part of Peter Hater's patch proposed in CORE-10681, that is, fixing the usage of QueryDosDevice API, but without the new IsDriveUsed functionali...

[Thomas Faber]

On 2016-06-28 21:21, hbelusca(a)svn.reactos.org wrote:
>   ---
> trunk/reactos/base/system/subst/subst.c	[iso-8859-1] (original)
>  +++ trunk/reactos/base/system/subst/subst.c	[iso-8859-1] Tue Jun 28 19:21:08 2016
>  @@ -75,81 +75,71 @@
>       if (_tcslen(Drive) > 2)
>           return FALSE;
>   
>  -    dwSize = sizeof(TCHAR) * MAX_PATH;
>  -    lpTargetPath = (LPTSTR) malloc(sizeof(TCHAR) * MAX_PATH);
>  -    if ( lpTargetPath)
>  -    {
>  +    dwSize = MAX_PATH;
>  +    lpTargetPath = (LPTSTR)malloc(sizeof(TCHAR) * dwSize);
>  +    if (!lpTargetPath)
>  +        return FALSE;
>  +
>  +    CharCount = QueryDosDevice(Drive,
>  +                               lpTargetPath,
>  +                               dwSize);
>  +    while (!CharCount &&
>  +           GetLastError() == ERROR_INSUFFICIENT_BUFFER)
>  +    {
>  +        lpTargetPath = (LPTSTR)realloc(lpTargetPath, sizeof(TCHAR) * dwSize); 
This is not how you use realloc, this causes a memory leak in case of
failure.
Also, why is there a cast here (also, for malloc above)?

Thanks.
-Thomas

>   +        if (lpTargetPath)
>  +        {
>  +            CharCount = QueryDosDevice(Drive,
>  +                                       lpTargetPath,
>  +                                       dwSize);
>  +        }
>  +    }
>  +
>  +    if (CharCount)
>  +    {
>  +        Result = _tcsncmp(lpTargetPath, _T("\\??\\"), 4) == 0 &&
>  +            ( (lpTargetPath[4] >= _T('A') &&
>  +            lpTargetPath[4] <= _T('Z')) ||
>  +             (lpTargetPath[4] >= _T('a') &&
>  +            lpTargetPath[4] <= _T('z')) );
>  +    }
>  +
>  +    free(lpTargetPath);
>  +
>  +    return Result;
>  +}