C / C++ / MFC

3-Aug-07 0:22

Hi,

CString* strData;
When I am calling "delete [] strData" , I am getting a crash . What are the possible reasons for getting the crashes.

Taruni

Sam_c3-Aug-07 0:26

Sam_c

3-Aug-07 0:26

you didn't initialise strdata as new?

CString *strData = new CString();

then use

delete strData;

Code Project Lounge 101 by John Cardinal
Beer | [beer]

Taruni3-Aug-07 0:38

3-Aug-07 0:38

I have initialised. I forgot to mention.

Before deleting , I am checking whether strData exists or not as follows.

if(strData)
{
delete [] strData;
}

The if condition is satisfied , but a crash when deleting strData.

Taruni

KarstenK3-Aug-07 1:35

KarstenK

3-Aug-07 1:35

check *strData: that a pointer in your code. D'Oh! | :doh:

Greetings from Germany

Paresh Chitte3-Aug-07 0:45

Paresh Chitte

3-Aug-07 0:45

Why you need [] in delete as it is a single pointer variable ?

It is applicable when you are trying to delete an array of pointer.

CString* strData = new CString;<br />
<br />
if(NULL != strData)<br />
{<br />
 delete strData;<br />
}

Regards,

Paresh.

Taruni3-Aug-07 0:52

3-Aug-07 0:52

Thank You Paresh.

But I am trying to delete an array of pointers of CString type.

Taruni

Paresh Chitte3-Aug-07 0:56

Paresh Chitte

3-Aug-07 0:56

How did you allocate memory to strData pointer ? Please post the code snippet.

Regards,

Paresh.

Taruni3-Aug-07 1:09

3-Aug-07 1:09

CString* strData;
int nCount;
if(nCount > 0)
{
strData = new CString[nCount];
}

.
.
.
delete[] strData;

I have understood the problem , Paresh . The nCount is 0. So never the memory is allocated . But what should be the condition to test whether strData is actually initialised (strData is an array of pointers of CString type) before it is actually deleted.

Taruni

CPallini3-Aug-07 2:37

3-Aug-07 2:37

Taruni wrote:
But what should be the condition to test whether strData is actually initialised

There is such a condition only if you're disciplinate programmer:

CString* strData = NULL; // always initialize pointers
int nCount;
if(nCount > 0)
{
strData = new CString[nCount];
}
...
if (strData)
{
  delate [] strData;
  strData = NULL; // always mark deleted pointers
}

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

Cedric Moonen3-Aug-07 1:01

Cedric Moonen

3-Aug-07 1:01

Taruni wrote:
But I am trying to delete an array of pointers of CString type.

So, you did something like this ?
CString pArray = new CString[10];

If yes, then you need to use delete[] instead:
delete[] pArray;

Cédric Moonen
Software developer

Charting control [v1.2]

Taruni3-Aug-07 1:11

3-Aug-07 1:11

Thank you Cedric Moonen.

CString* strData;
int nCount;
if(nCount > 0)
{
strData = new CString[nCount];
}

.
.
.
delete[] strData;

I have understood the problem . The nCount is 0. So never the memory is allocated . But what should be the condition to test whether strData is actually initialised (strData is an array of pointers of CString type) before it is actually deleted.

Taruni

Paresh Chitte3-Aug-07 1:13

Paresh Chitte

3-Aug-07 1:13

Initialise strData with NULL and check strData against NULL before deleting.

CString* strData = NULL;<br />
int nCount;<br />
if(nCount > 0)<br />
{<br />
strData = new CString[nCount];<br />
}<br />
<br />
.<br />
.<br />
.<br />
if(NULL != strData)<br />
{<br />
delete[] strData;<br />
}

Regards,

Paresh.

Nemanja Trifunovic3-Aug-07 2:06

3-Aug-07 2:06

Paresh Chitte wrote:
Initialise strData with NULL

Yes.

Paresh Chitte wrote:
and check strData against NULL before deleting.

That's not necessary. Deleting NULL is guaranteed to be a no-op.

CPallini3-Aug-07 2:43

3-Aug-07 2:43

Nemanja Trifunovic wrote:
That's not necessary. Deleting NULL is guaranteed to be a no-op.

The test maybe useful to mark again with NULL the deleted pointer:

if (strData)
{
  delete [] strData;
  strData = NULL;
}

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

CPallini3-Aug-07 2:41

3-Aug-07 2:41

Paresh Chitte wrote:
if(NULL != strData)
{
delete[] strData;
}

I suggest to replace the above with:

if(NULL != strData)
{
  delete[] strData;
  strData = NULL;
}

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

Cedric Moonen3-Aug-07 1:14

Cedric Moonen

3-Aug-07 1:14

Initialize your pointer to NULL:
CString* strData = NULL;
You should always always initialize your pointers to NULL.

Cédric Moonen
Software developer

Charting control [v1.2]

Nemanja Trifunovic3-Aug-07 2:36

3-Aug-07 2:36

Cedric Moonen wrote:
You should always always initialize your pointers to NULL.

Unless you initialize them with something else Wink | ;)

I.e., this is acceptable:

<br />
CString* strData = new CString[count];<br />

In any case this new/delete construct is overused and leads to code that is not exception-safe and is hard to maintain. Wouldn't it be easier and safer to simply write something like:

<br />
vector<CString> strData;<br />

or even:

<br />
CStringArray strData;<br />

I agree

CPallini3-Aug-07 2:52

3-Aug-07 2:52

Nemanja Trifunovic wrote:
Unless you initialize them with something else

I.e., this is acceptable:

CString* strData = new CString[count];

I agree on the above and suggest the following (weaker than the Cedric Moonen one)
gold rule of the disciplined coder:
You should always initialize your pointers
Smile | :)

Nemanja Trifunovic wrote:
In any case this new/delete construct is overused and leads to code that is not exception-safe and is hard to maintain. Wouldn't it be easier and safer to simply write something like:

I don't blame for new/delete usage. You've to be disciplined though.

Smile | :)

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

Nemanja Trifunovic3-Aug-07 3:19

3-Aug-07 3:19

CPallini wrote:
I don't blame for new/delete usage. You've to be disciplined though.

A LOT of discipline and hard work is needed to make this code exception-safe:

A* a = new A;
... // what if this trows?
B* b = new B;
... // what if this trows?
C* c = new C;
... // what if this trows?
delete c;
delete b;
delete a;

CPallini3-Aug-07 3:37

3-Aug-07 3:37

Of course, one can choose the proper data depending on the coding scenario. BTW at the very end of your reasonment there's managed code!! Big Grin | :-D

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

Nemanja Trifunovic3-Aug-07 3:54

3-Aug-07 3:54

CPallini wrote:
BTW at the very end of your reasonment there's managed code!!

Nope!

Managed code would help in this particular scenario, but if we were allocating resources other than memory (files, handles, db connections...) it wouldn't help at all.

What I am advocating here is the use of a simple and effective idiom known as RAII[^]. It enables automatic release of the allocated resources in a deterministic way without use of a GC.

CPallini3-Aug-07 4:09