General Book Discussion

Re: LeakDiag confusion

Anonymous — Wed, 18 Jun 2008 16:58:56 GMT

kirants:
I've been playing around with LeakDiag and am left all confused. I have been reading the book, looking at LeakDiag help and still not able to understand how one is supposed to interpret those log files.
Some of my questions:
1. What version of LeakDiag is the latest. The download I have is 1.25 and is dated 2002. However, somewhere here the one of the posts at the end allude to a 1.31.xx.yyy version. http://groups.google.com/group/microsoft.public.windbg/browse_thread/thread/f7be3ad0aef62120/657ccfeb114b7bd2?hl=en&lnk=st&q=LeakDiag#657ccfeb114b7bd2
2. I tried a simple application like below:
#include "stdafx.h"
#include <conio.h>
#include <memory>
int _tmain(int argc, _TCHAR* argv[])
{
getch();
char* p = (char*)malloc(30);
getch();
return 0;
}
using VS2005 IDE. I run this app, it stops at first getch at which point I run leakdiag, select this app, select C Runtime Allocator, hit Start.
I now, switch back to the target application console , punch in a key to let it go past the malloc , and switch back to LeakDiag and hit Log. I would've thought this should dump the malloced entry in the log file, but I see that the log file is not generated at all.
What is going on ?
P.S. I have tried an MFC application also, putting in a memory leak like above and then generating logs. I do see the logs, but there is no stack trace whatsoever which can be deciphered although I am pointing to the folder containing the exe.
BTW, I am running WinXP SP2

Re: LeakDiag confusion

Anonymous — Wed, 11 Jun 2008 17:37:34 GMT

kirants:
I've been playing around with LeakDiag and am left all confused. I have been reading the book, looking at LeakDiag help and still not able to understand how one is supposed to interpret those log files.
Some of my questions:
1. What version of LeakDiag is the latest. The download I have is 1.25 and is dated 2002. However, somewhere here the one of the posts at the end allude to a 1.31.xx.yyy version. http://groups.google.com/group/microsoft.public.windbg/browse_thread/thread/f7be3ad0aef62120/657ccfeb114b7bd2?hl=en&lnk=st&q=LeakDiag#657ccfeb114b7bd2
2. I tried a simple application like below:
#include "stdafx.h"
#include <conio.h>
#include <memory>
int _tmain(int argc, _TCHAR* argv[])
{
getch();
char* p = (char*)malloc(30);
getch();
return 0;
}
using VS2005 IDE. I run this app, it stops at first getch at which point I run leakdiag, select this app, select C Runtime Allocator, hit Start.
I now, switch back to the target application console , punch in a key to let it go past the malloc , and switch back to LeakDiag and hit Log. I would've thought this should dump the malloced entry in the log file, but I see that the log file is not generated at all.
What is going on ?
P.S. I have tried an MFC application also, putting in a memory leak like above and then generating logs. I do see the logs, but there is no stack trace whatsoever which can be deciphered although I am pointing to the folder containing the exe.
BTW, I am running WinXP SP2

Re: LeakDiag confusion

Anonymous — Fri, 06 Jun 2008 19:09:56 GMT

kirants:
I've been playing around with LeakDiag and am left all confused. I have been reading the book, looking at LeakDiag help and still not able to understand how one is supposed to interpret those log files.
Some of my questions:
1. What version of LeakDiag is the latest. The download I have is 1.25 and is dated 2002. However, somewhere here the one of the posts at the end allude to a 1.31.xx.yyy version. http://groups.google.com/group/microsoft.public.windbg/browse_thread/thread/f7be3ad0aef62120/657ccfeb114b7bd2?hl=en&lnk=st&q=LeakDiag#657ccfeb114b7bd2
2. I tried a simple application like below:
#include "stdafx.h"
#include <conio.h>
#include <memory>
int _tmain(int argc, _TCHAR* argv[])
{
getch();
char* p = (char*)malloc(30);
getch();
return 0;
}
using VS2005 IDE. I run this app, it stops at first getch at which point I run leakdiag, select this app, select C Runtime Allocator, hit Start.
I now, switch back to the target application console , punch in a key to let it go past the malloc , and switch back to LeakDiag and hit Log. I would've thought this should dump the malloced entry in the log file, but I see that the log file is not generated at all.
What is going on ?
P.S. I have tried an MFC application also, putting in a memory leak like above and then generating logs. I do see the logs, but there is no stack trace whatsoever which can be deciphered although I am pointing to the folder containing the exe.
BTW, I am running WinXP SP2

Re: LeakDiag confusion

Anonymous — Tue, 03 Jun 2008 07:27:31 GMT

kirants:
I've been playing around with LeakDiag and am left all confused. I have been reading the book, looking at LeakDiag help and still not able to understand how one is supposed to interpret those log files.
Some of my questions:
1. What version of LeakDiag is the latest. The download I have is 1.25 and is dated 2002. However, somewhere here the one of the posts at the end allude to a 1.31.xx.yyy version. http://groups.google.com/group/microsoft.public.windbg/browse_thread/thread/f7be3ad0aef62120/657ccfeb114b7bd2?hl=en&lnk=st&q=LeakDiag#657ccfeb114b7bd2
2. I tried a simple application like below:
#include "stdafx.h"
#include <conio.h>
#include <memory>
int _tmain(int argc, _TCHAR* argv[])
{
getch();
char* p = (char*)malloc(30);
getch();
return 0;
}
using VS2005 IDE. I run this app, it stops at first getch at which point I run leakdiag, select this app, select C Runtime Allocator, hit Start.
I now, switch back to the target application console , punch in a key to let it go past the malloc , and switch back to LeakDiag and hit Log. I would've thought this should dump the malloced entry in the log file, but I see that the log file is not generated at all.
What is going on ?
P.S. I have tried an MFC application also, putting in a memory leak like above and then generating logs. I do see the logs, but there is no stack trace whatsoever which can be deciphered although I am pointing to the folder containing the exe.
BTW, I am running WinXP SP2

Re: LeakDiag confusion

Hans De Smaele — Tue, 29 Jan 2008 10:16:52 GMT

Hi,

together with the LeadDiag tool, there come two packages LDGrapher and LDParser, giving a nice view of the leak. The latest version of LeakDiag I got is 1.31 (obtained it via Microsoft Premier Support). I explained the usage of it in my article about production debugging (http://www.microsoft.com/belux/msdn/nl/community/articles/feb05_applicationdebugging.mspx). Hope this helps.

Re: LeakDiag confusion

kirants — Thu, 27 Dec 2007 20:32:29 GMT

Hi Mario,

I think I might have nailed the reason at least for Visual Studio 6.0 compiled exe. All along, my Visual Studio 6.0 project settings were using static linking with CRT. When I did switch to multi-threaded dll option to use the crt dlls, I am able to see the C runtime allocator working fine.

However,

I still see that with VS2005 compiled executable, this doesn't happen. Using dependencywalker I see that malloc is infact being imported from MSVCR80.dll.

Does LeakDiag take care of CRT dlls other than MSVCRT.dll ?

Thanks

Kiran

Re: LeakDiag confusion

Anonymous — Thu, 27 Dec 2007 16:36:22 GMT

Hi Mario,

Thanks for the reply. First off, kudos on the book. It is well written and I think you have succeeded in revealing a lot of hidden gems for debugging. LeakDiag is one such discovery for me :)

As for the issue, this is what I found today. All the while I was choosing the wrong allocator. I would've though using C runtime allocator would trap calls to malloc, but alas, it does not. With C Runtime allocator, this is the log I got:

- <SUMMARY_INFO>

<FILENAME>C:\tmp\AALeak\Debug\AALeak.exe</FILENAME>

- <LEAKDIAG_SETTINGS>

</LEAKDIAG_SETTINGS>

- <MEMORY_INFO>

</MEMORY_INFO>

</SUMMARY_INFO>

Note that the Leaks section is empty.

Subsequently, I did the same test using the Windows Heap allocator, and I do see the stack trace now. This is what I see:

- <LEAKS ver="1.25.28.2201">

- <STACK numallocs="01" size="066" totalsize="066">

- <STACKSTATS>

</STACKSTATS>

</STACK>

</LEAKS>

Question. Why this behavior? Per documentation on LeakDiag C runtime allocator should trap malloc, new, callocs etc.

Re: LeakDiag confusion

marioh — Thu, 27 Dec 2007 04:24:56 GMT

Hi,

Did you set the correct symbol path in LeakDiag? Also, can you share parts of the log file that was generated?

Thanks,

Mario

LeakDiag confusion

kirants — Thu, 27 Dec 2007 01:34:10 GMT

I've been playing around with LeakDiag and am left all confused. I have been reading the book, looking at LeakDiag help and still not able to understand how one is supposed to interpret those log files.

Some of my questions:

1. What version of LeakDiag is the latest. The download I have is 1.25 and is dated 2002. However, somewhere here the one of the posts at the end allude to a 1.31.xx.yyy version. http://groups.google.com/group/microsoft.public.windbg/browse_thread/thread/f7be3ad0aef62120/657ccfeb114b7bd2?hl=en&lnk=st&q=LeakDiag#657ccfeb114b7bd2

2. I tried a simple application like below:

#include "stdafx.h"
#include <conio.h>
#include <memory>

int _tmain(int argc, _TCHAR* argv[])
{
getch();
char* p = (char*)malloc(30);
getch();
return 0;
}

using VS2005 IDE. I run this app, it stops at first getch at which point I run leakdiag, select this app, select C Runtime Allocator, hit Start.

I now, switch back to the target application console , punch in a key to let it go past the malloc , and switch back to LeakDiag and hit Log. I would've thought this should dump the malloced entry in the log file, but I see that the log file is not generated at all.

What is going on ?

P.S. I have tried an MFC application also, putting in a memory leak like above and then generating logs. I do see the logs, but there is no stack trace whatsoever which can be deciphered although I am pointing to the folder containing the exe.

BTW, I am running WinXP SP2