CString Problem

The same code executes properly with me , in release mode tooo..

Whats the problem with you ??
Can you attach the code ??

Hi...

moodfy the code like this
pUserInfo = new UserInfo; //(UserInfo *)malloc(sizeof(UserInfo));

it should work.

regards,
vikram patwardhan

Hi guys,
Thanks a lot for checking. Here is my code.

/* This structure will be useful as a recordset for all getting a list of the Users */
struct UserInformation
{
CString cstrUserName;
BOOL blnUsrCanChangePswd;
UserInformation *Next;
};

/* Function for Getting a list of all Users */
UserRecordset* GetAllUsers()
{
UserInformation *ptrRsUser = NULL;

ptrRsUser = (UserInformation *)malloc(sizeof(UserInformation));
if(ptrRsUser != NULL)
{
ptrRsUser->cstrUserName = "asdf";
ptrRsUser->blnUsrCanChangePswd = FALSE;
ptrRsUser->Next = NULL;
}
}

Its strange if this doesnt give error.

Thanking You,
Sonali....

HAVE A GUD DAY!!!!!

Originally posted by sss0379
Hi guys,
Thanks a lot for checking. Here is my code.

/* This structure will be useful as a recordset for all getting a list of the Users */
struct UserInformation
{
CString cstrUserName;
BOOL blnUsrCanChangePswd;
UserInformation *Next;
};

/* Function for Getting a list of all Users */
UserRecordset* GetAllUsers()
{
UserInformation *ptrRsUser = NULL;

//ptrRsUser = (UserInformation *)malloc(sizeof(UserInformation));
do as 0xC0000005 suggested!:
ptrRsUser = new UserInformation;
if(ptrRsUser != NULL)
{
ptrRsUser->cstrUserName = "asdf";
ptrRsUser->blnUsrCanChangePswd = FALSE;
ptrRsUser->Next = NULL;
}
}

Its strange if this doesnt give error.

Thanking You,
Sonali....

HAVE A GUD DAY!!!!!

Originally posted by myth7676

i didn't replace that and it still works ?! This also may work, but it isn't right:

int main()
{
char *p = "ABCDEF";
p[7] = 'D';
}

What am I trying to say? You were given the correct advice -- get rid of the malloc(), and please replace it with operator new.

Using malloc for such a class invokes undefined behavior, just like the code above that writes beyond the boundaries of an array. It may work, it may crash, you don't know. Right now, your program is ill-formed.

Classes and structures that either have constructors, or whose members contain constructors must be constructed using operator new / new[] if they are created dynamically. The bottom line is never use malloc(), calloc(), or other 'C' allocation functions on classes and structures that contain non-trivial constructors.

Your class has a CString member, and CString has a non-trivial default constructor, therefore your usage of malloc() is illegal and leads to undefined behavior. The malloc() does not call the proper constructors for the class, and your object is not constructed correctly, therefore you are leaving yourself wide open for all kinds of errors. Pretend you've never seen malloc() in your life and the only thing you know of is new / new[] / delete / delete[] :)

Regards,

Paul McKenzie

But CString's = operator is overloeaded operator and handles memory allocation tooo. So there shouldnt be any prblem if malloc is used for object allocation.
On My machine win2k , there is no problem with the code given by sonali , without new operator.
See MFC code.

void CString :: AssignCopy(int nSrcLen, LPCTSTR lpszSrcData)
{
AllocBeforeWrite(nSrcLen);
memcpy(m_pchData, lpszSrcData, nSrcLen*sizeof(TCHAR));
GetData()->nDataLength = nSrcLen;
m_pchData[nSrcLen] = '\0';
}

const CString& CString :: operator=(LPCTSTR lpsz)
{
ASSERT(lpsz == NULL || AfxIsValidString(lpsz));
AssignCopy(SafeStrlen(lpsz), lpsz);
return *this;
}

#ifdef _UNICODE
const CString& CString :: operator = (LPCSTR lpsz)
{
int nSrcLen = lpsz != NULL ? lstrlenA(lpsz) : 0;
AllocBeforeWrite(nSrcLen);
_mbstowcsz(m_pchData, lpsz, nSrcLen+1);
ReleaseBuffer();
return *this;
}

Here AllocBeforeWrite takes care if CString dont have allocated area to keep the assigned string.

void CString :: AllocBeforeWrite(int nLen)
{
if (GetData()->nRefs > 1 || nLen > GetData()->nAllocLength)
{
Release();
AllocBuffer(nLen);
}
ASSERT(GetData()->nRefs <= 1);
}

void CString :: AllocBuffer(int nLen)
// always allocate one extra character for '\0' termination
// assumes [optimistically] that data length will equal allocation length
{
ASSERT(nLen >= 0);
ASSERT(nLen <= INT_MAX-1); // max size (enough room for 1 extra)

if (nLen == 0)
Init();
else
{
CStringData* pData;
#if !defined(_DEBUG) && !defined(_WIN64)
if (nLen <= 64)
{
pData = (CStringData*)_afxAlloc64.Alloc();
pData->nAllocLength = 64;
}
else if (nLen <= 128)
{
pData = (CStringData*)_afxAlloc128.Alloc();
pData->nAllocLength = 128;
}
else if (nLen <= 256)
{
pData = (CStringData*)_afxAlloc256.Alloc();
pData->nAllocLength = 256;
}
else if (nLen <= 512)
{
pData = (CStringData*)_afxAlloc512.Alloc();
pData->nAllocLength = 512;
}
else
#endif
{
pData = (CStringData*)
new BYTE[sizeof(CStringData) + (nLen+1)*sizeof(TCHAR)];
pData->nAllocLength = nLen;
}
pData->nRefs = 1;
pData->data()[nLen] = '\0';
pData->nDataLength = nLen;
m_pchData = pData->data();
}
}

Originally posted by Krishnaa
But CString's = operator is overloeaded operator and handles memory allocation tooo. So there shouldnt be any prblem if malloc is used for object allocation.You are invoking undefined behavior. Again, you never allocate a structure witn malloc that contains members with non-trivial constructors. How is the constructor for CString ever called if you call malloc? Classes and structures such as the one posted by the OP must be created with operator new, or else the constructors for the objects are not called. That is fundamental ANSI C++ rules. To argue against it is like "spitting in the wind" :)
On My machine win2k , there is no problem with the code given by sonali , without new operator.Again, undefined behavior. I can post code that "works" -- it doesn't mean it's correct. Did you see the tiny example I posted? Is it correct? No. Does it crash? It doesn't on my machine. Does it mean it's right? No. Is it ill-formed? Yes.

Also, why are you looking at the internals of CString? What if it changes in the next version? The bottom line is that there is no need to know what is going on inside of CString or any non-trivial class during construction to see whether malloc is "OK" -- malloc is never OK. Just call "new" and the proper constructor for the object is invoked. This is why a constructor was created -- so that the object can be constructed correctly. If not, why write a constructor?

Regards,

Paul McKenzie

Yes Paul , This is a valid answer that Constructors should be called for C++ types...I was confused for a some time even though I was well known to this funda ...it happens with me ..:( :(