Re: DEBUG_NEW for ATL apps? [modified] - ATL / WTL / STL Discussion Boards

Re: Destructors in STL?

Zac Howland11-Aug-06 4:40

Zac Howland

11-Aug-06 4:40

sawerr wrote:
But i know that we don't need to delete stl componenets

I don't think you quite understand C++ fundamentals. STL classes (string, vector, list, map, etc) all allocate memory on the heap to store whatever data you are throwing at them. string, for example, allocates an array of characters. As you fill up that array, it evaluates if it needs to allocate a bigger buffer, etc. All that is handled for you. The destructor cleans up that heap memory.

When you write:

void f()
{
	string s;
}

s is allocated on the stack. At the end of the scope (in this case, when the function returns), s's destructor is called, which cleans up anything s had allocated.

If you were to write this:

void f()
{
	string* s = new string;
}

you would have a memory leak (not good!).

If you decide to become a software engineer, you are signing up to have a 1/2" piece of silicon tell you exactly how stupid you really are for 8 hours a day, 5 days a week

Zac

Calling m_spClientSite->GetContainer from another thread

TClarke8-Aug-06 3:34

TClarke

8-Aug-06 3:34

Hi all,

I'm writing an ActiveX ctrl in which a DDE server resides. The DDE server has it's own thread. From that thread function calls are made to the main thread on behalf of some messages. For the processing of one particular message a JavaScript function on the web page must be called. This involves acquiring an IHTMLWindow2 interface. My problem is that I get a return value of E_UNEXPECTED whenever I call the interface querying functions from the DDE server's thread. For example.
hr = m_spClientSite->GetContainer(&spContainer);

I presume this is a threading issue as the method works fine in the main thread. However, I have little idea of how to resolve the situation.

Thanks in advance

Tom

The greatest hinderence to the truth is not ignorance, it is the illusion of knowledge.
Anon.

DEBUG_NEW for ATL apps?

Rob Caldecott7-Aug-06 5:49

Rob Caldecott

7-Aug-06 5:49

I am debugging a large ATL-based application, and am suffering from some memory leaks that are reported when the app exits (thanks to _CrtSetDbgFlag(_CRTDBG_LEAK_CHECK_DF|_CRTDBG_ALLOC_MEM_DF) in the satrtup code). However, unlike MFC apps (which used the DEBUG_NEW macro) the file name/line number of the relevant allocation calls are not displayed. Frown | :(

How can I find out where the leaks are coming from without paying $$$ for BoundsChecker/Purify - am amazed that ATL doesn't support this.

Re: DEBUG_NEW for ATL apps?

Michael Dunn7-Aug-06 7:41

Michael Dunn

7-Aug-06 7:41

Here's what I use:

#ifdef _DEBUG
#define new new(_NORMAL_BLOCK, __FILE__, __LINE__)
#endif

--Mike--
Visual C++ MVP Cool | :cool:

LINKS~! Ericahist | PimpFish | CP SearchBar v3.0 | C++ Forum FAQ

Re: DEBUG_NEW for ATL apps?

Rob Caldecott7-Aug-06 7:50

Rob Caldecott

7-Aug-06 7:50

Thanks Michael. Would this be safe to use in stdafx.h do you know?

Re: DEBUG_NEW for ATL apps?

Michael Dunn7-Aug-06 8:14

Michael Dunn

7-Aug-06 8:14

Sure

--Mike--
Visual C++ MVP Cool | :cool:

LINKS~! Ericahist | PimpFish | CP SearchBar v3.0 | C++ Forum FAQ

Re: DEBUG_NEW for ATL apps?

Jörgen Sigvardsson8-Aug-06 20:47

Jörgen Sigvardsson

8-Aug-06 20:47

If unsure, always put stuff like that after your system includes, but before your own. Smile | :)

--
Behold, for I am THE CORRUPTOR!

Re: DEBUG_NEW for ATL apps? [modified]

Stuart Dootson7-Aug-06 21:24

Stuart Dootson

7-Aug-06 21:24

Robert - what I often find to work quite nicely (without redefining 'new', which for some reason, I find slightly distaasteful - don't ask me why, probably a symptom of incipient OCD) is to use the allocation ID - add _crtBreakAlloc to a watch window, then set it to the ID of a leaking allocation. You'll then break within the debug malloc (actually in _heap_alloc_dbg) when that allocation is made, so have access to the full call stack etc.

<code>
Detected memory leaks!
Dumping objects ->
{53} normal block at 0x002F3CF8, 4 bytes long.
 Data: <    > CD CD CD CD 
Object dump complete.</code>

In this object dump, I have one leaked allocation, with ID 53. So, step into main for your app, then set _crtBreakAlloc to (decimal) 53 and run until you get a User Breakpoint exception dialog appearing. Press 'Break' and there you are!

Last modified: 08 August 2006 02:21:44 --

Re: DEBUG_NEW for ATL apps?

Rob Caldecott7-Aug-06 21:57

Rob Caldecott

7-Aug-06 21:57

Excellent - I assume this means the allocation IDs are reused between executions? If so, this is a great tip Stuart. Thanks.

Re: DEBUG_NEW for ATL apps?

Stuart Dootson8-Aug-06 0:05

Stuart Dootson

8-Aug-06 0:05

Yes, and they stay the same, so long as the allocation pattern doesn't change (which is likely to be the case if you don't alter your code too much).

FileMaker Pro

khwo6-Aug-06 20:59

khwo

6-Aug-06 20:59

Hi All,
Anyone of you know how to use the FileMaker Pro Software to develop an ERP system?

Thanks.

Re: FileMaker Pro

User 2155976-Aug-06 21:34

User 215597

6-Aug-06 21:34

What does ATL/WTL/STL have to do with it?

S o h a i l K a d i w a l a

modified 21-Apr-21 21:01pm.

stl containers allocation in heap or stack?

sawerr4-Aug-06 12:44

sawerr

4-Aug-06 12:44

Hi I am new to stl.
I see that we don't need to delete our containers
For example:
vector <char> vec;
But why?
Here, Does vec allocate in heap or stack by default.I heard that Although we don't create vec without new , it is allocated in heap and automatically delete it.Is this true?
If true i tried this:
vector <char>* vec = new vector<char>;
delete vec;
it doesn't give me error.Here this mustn't work.Am i wrong?

I looked to google, there are so many things about stl but i can't find explanation about relationship between stl containers and heap,stack allocation(or new ,delete).In what circumstances are stl containers allocated in heap or stack.And In what circumstances do we need to delete our containers?

Can you please explain me.

Thanks...

Re: stl containers allocation in heap or stack?

Jörgen Sigvardsson4-Aug-06 13:14

Jörgen Sigvardsson

4-Aug-06 13:14

It's basic C++.

class X {
   int* p;
public:
   X() : p(new int[100]) { }
   ~X() { delete [] p; }
};
  
void function() {
   X x;
}

x is an object allocated in the stackframe of function. X::p however is allocated on the heap, as it's done so by new. std::vector is no different.

--
Simulcast on Crazy People's Fillings

Re: stl containers allocation in heap or stack?

sawerr4-Aug-06 20:07

sawerr

4-Aug-06 20:07

thank you

Re: stl containers allocation in heap or stack?

Kevin McFarlane7-Aug-06 0:41

Kevin McFarlane

7-Aug-06 0:41

Have a look at the tutorials on this site.

Typically you just declare the STL containers on the stack.

vector<int> vec;
vector<MyClass> vec;

You can add simple types and objects to the vector and you don't need to worry about memory management <i>unless </i>those individual simple types or objects were themselves dynamically allocated.

Kevin

Re: stl containers allocation in heap or stack?

Jörgen Sigvardsson17-Aug-06 12:20

Jörgen Sigvardsson

17-Aug-06 12:20

template <typename T> 
struct Deleter { void operator(T* pObj) { delete pObj; } };
  
template <typename Collection>
void DeleteAllObjects(Collection& coll)
{
   std::for_each(coll.begin(), coll.end(), Deleter<Collection::value_type>());
   coll.clear();
};

--
Based on a True Story

Re: stl containers allocation in heap or stack?

Kevin McFarlane18-Aug-06 8:44

Kevin McFarlane

18-Aug-06 8:44

Yes, I have a similar utility functor myself, but I haven't done any C++ for 18 months. I'm in the .NET universe now. I guess I may do some C++/CLI some time. Smile | :)

Kevin

[ ! ] Crashing Connection points in the .exe.. Help!!

Eytukan4-Aug-06 3:18

Eytukan

4-Aug-06 3:18

Problem with firing events
I created an ATL/COm dll. and I'm testing it with a VB client. The client behaves fine in the IDE (VB6.0)
But once we create the .exe and run it, It crashes the application when an Event is fired from the DLL
I narrowed down to the point where it actually crashes. It's failing when we say,

this->Fire_testEvent()

inside the implementation,
it fails here.

<br />
HRESULT Fire_testEvent()<br />
	{<br />
		CComVariant varResult;<br />
		T* pT = static_cast<T*>(this);<br />
		int nConnectionIndex;<br />
		int nConnections = m_vec.GetSize();<br />
		<br />
		for (nConnectionIndex = 0; nConnectionIndex < nConnections; nConnectionIndex++)<br />
		{<br />
			pT->Lock();<br />
			CComPtr<IUnknown> sp = m_vec.GetAt(nConnectionIndex);<br />
			pT->Unlock();<br />
			IDispatch* pDispatch = reinterpret_cast<IDispatch*>(sp.p);<br />
			if (pDispatch != NULL)<br />
			{<br />
				VariantClear(&varResult);<br />
				DISPPARAMS disp = { NULL, NULL, 0, 0 };<br />
				</pre>pDispatch->Invoke(0x6, IID_NULL, LOCALE_USER_DEFAULT, DISPATCH_METHOD, &disp, &varResult, NULL, NULL);//<----ERROR<br />
</pre> 			}<br />
		}<br />
		return varResult.scode;<br />
	<br />
	}<br />

Please help me .. what could be wrong.. It's killing me.

<marquee scrollamount="1" scrolldelay="1" direction="up" height="10" step="1">--[ Jig | [Dance]

]--

[My Current Status]
I dont know why the hell the script for voting 5 is disabled only for me?? Mad | :mad:

Re: [ ! ] Crashing Connection points in the .exe.. Help!!

Roger Stoltz4-Aug-06 4:12

Roger Stoltz

4-Aug-06 4:12

Are you by any chance firing the event from a secondary thread?

If you want to fire events from other threads than the one that created the server, you have to marshal the sink interface pointer.
Have a look at Michael Lindig's solution here[^], I've used it quite a lot.

If you're not using multithreading, post again and describe your app further.

--
Rog

It's supposed to be hard, otherwise anybody could do it!

Regarding CodeProject: "resistance is pointless; you will be assimilated"

Re: [ ! ] Crashing Connection points in the .exe.. Help!!

Eytukan4-Aug-06 4:19

Eytukan

4-Aug-06 4:19

Yup exactly, I fire it from a worker thread. So that's the culprit? hmm.. thanks for your valuable help. Thanks you so much. I'll go through the link. and comeback here again. But Roger I have a question. How does it work, when I test the application in the IDE ? and what makes it fail after I create the .Exe file? Unsure | :~

<marquee scrollamount="1" scrolldelay="1" direction="up" height="10" step="1">--[ Jig | [Dance]

]--

[My Current Status]
I dont know why the hell the script for voting 5 is disabled only for me?? Mad | :mad:

Re: [ ! ] Crashing Connection points in the .exe.. Help!!

Roger Stoltz4-Aug-06 4:34

Roger Stoltz

4-Aug-06 4:34

I'm glad I could help you, even if I guessed about your problem.

Regarding VB and why it works running from the IDE:
I don't know the technical reason for it, but VB is quite forgiving when running from the IDE. I guess (again... Roll eyes | :rolleyes:

) that VB makes some kind of qualified guess and if you're "lucky" it works.

Aahh, I forgot a few really important things: you have to set up the apartment as well from your secondary thread with a call to ::CoInitialize(). If you haven't done so yet, this could be the reason that it works when running from the IDE. If it's an in-process server it would be the same thing as using a function pointer even if the call is made from another thread. If you properly set up the secondary apartment you could get a HRESULT = RPC_E_WRONG_THREAD, especially in you're running the server out-of-process.

It's supposed to be hard, otherwise anybody could do it!

Regarding CodeProject: "resistance is pointless; you will be assimilated"

Re: [ ! ] Crashing Connection points in the .exe.. Help!!

Eytukan4-Aug-06 5:28

Eytukan

4-Aug-06 5:28

Had to solve other problems in my project. Frown | :(

I should spend a while to fully understand it and implement. anyway I'm leaving office with a complete peace of mind only because of Mr. Roger Stoltz. Smile | :)

. thanks!! Rose | [Rose]

.. see you on monday. Happy weekend Beer | [beer]

<marquee scrollamount="1" scrolldelay="1" direction="up" height="10" step="1">--[ Jig | [Dance]

]--

[My Current Status]
I dont know why the hell the script for voting 5 is disabled only for me?? Mad | :mad:

Re: [ ! ] Crashing Connection points in the .exe.. Help!!

Roger Stoltz4-Aug-06 5:35

Roger Stoltz

4-Aug-06 5:35

Have a nice weekend yourself.
You saved my weekend with your kind replies! Big Grin | :-D

Careful with the beer, mate! Your belly measures are at risk!

It's supposed to be hard, otherwise anybody could do it!

Regarding CodeProject: "resistance is pointless; you will be assimilated"

Re: [ ! ] Crashing Connection points in the .exe.. Help!!

Eytukan8-Aug-06 3:36

Eytukan

8-Aug-06 3:36

Hello buddy Smile | :)

Just fixed my thread-event problem. Jig | [Dance]

It's working fine now Smile | :)

. Thank you so much Roger. I was really shocked to find the pDispatch->Invoke(0x1,line of code crashing the application. I was clueless what on earth makes it fail with the .exe. Hmmm.. now everything's fine Smile | :)

chilling out a bit.. Beer | [beer]

.. thanks again Rose | [Rose]

[Roger].

<marquee scrollamount="1" scrolldelay="1" direction="up" height="10" step="1">--[ Jig | [Dance]

]--

[My Current Status]
I dont know why the hell the script for voting 5 is disabled only for me?? Mad | :mad:

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.