I am trying to use VS2008 memory leak tool, but I have failed to build it at all.
The simplest scenarios works well, but when I try to use CObject - it does not compile
Here is the code (Its a newly create console application)
#include "stdafx.h"
#ifdef _DEBUG
#ifndef DBG_NEW
#define DBG_NEW new ( _NORMAL_BLOCK , __FILE__ , __LINE__ )
#define new DBG_NEW
#endif
#endif // _DEBUG
#define _AFXDLL
#include "afx.h"
class DRV : public CObject {};
int _tmain(int argc, _TCHAR* argv[])
{
DRV *d = new DRV;
}
This results : error C2059: syntax error : 'constant' in afx.h:
void* PASCAL operator new(size_t nSize);
If I try to move the #ifdef _DEBUG below the #include "afx.h", I get:
error C2661: 'CObject::operator new' : no overloaded function takes 4 arguments
on line:
DRV *d = new DRV;
So - what am I doing wrong?
Can I use the build in VS2008 memory leak detector?
Please help
Create file DebugNew.h and add this code to it:
#pragma once
#include "crtdbg.h"
#ifdef _DEBUG
#define DEBUG_NEW new( _NORMAL_BLOCK, __FILE__, __LINE__)
#else
#define DEBUG_NEW
#endif
In cpp file:
#include "stdafx.h"
#include "DebugNew.h"
#ifdef _DEBUG
#define new DEBUG_NEW
#endif
int _tmain(int argc, _TCHAR* argv[])
{
CrtSetDbgFlag( _CrtSetDbgFlag(_CRTDBG_REPORT_FLAG) | _CRTDBG_LEAK_CHECK_DF);
char *d = new char[100];
}
DebugNew.h file defines new operator, which allows to include source line information for every allocation. #define new DEBUG_NEW line redefines default new to DEBUG_NEW, only in Debug build. This line should be placed after all #include lines in all .cpp files. CrtSetDbgFlag enables memory leak allocation in debug build - when the program exits, all unreleased allocations are printed. Since new operator is redefined, they are printed with source line information.
For MFC projects, you only need to add lines
#ifdef _DEBUG
#define new DEBUG_NEW
#endif
to every .cpp file. All other things are already done by MFC. MFC project, created by MFC application Wizard, already contains all required stuff by default. For example, create Win32 Console Application with MFC support using Wizard - memory leaks tracking is working. You only need to add new DEBUG_NEW redefinition to every new file added to the project.
Related
I'm testing an antidebug solution with ptrace method; and i compile the program by using ndk21e cross-compile.
The problem is that it compiles successfully with gcc, but fails with ndk cross-compile.
ndk cross-compile compiles all other programs without problems
#include <stdlib.h>
#include <stdio.h>
#include<sys/ptrace.h>
#include <dlfcn.h>
#include <string.h>
int main(int argc, char **argv) {
void *handle;
long (*go)(enum __ptrace_request request, pid_t pid);
// get a handle to the library that contains 'ptrace'
handle = dlopen ("/lib/x86_64-linux-gnu/libc.so.6", RTLD_LAZY);
// reference to the dynamically-resolved function 'ptrace'
go = dlsym(handle, "ptrace");
if (go(PTRACE_TRACEME, 0) < 0) {
puts("being traced");
exit(1);
}
puts("not being traced");
// cleanup
dlclose(handle);
return 0;
}
And it shows the error like the picture as follow:
gcc compileresult and cross-compile error result
How can i solve this problem. Thanks.
I have found a strange behavior (probably an issue) while trying to compile a simple application with MS VC 15.3.1 (after applying VC 2017 Upgrade 3):
#include <iostream>
#include <algorithm>
#include <string>
// compiles if commenting out the line below
#include <vector>
#include <list>
class A
{
std::string s;
public:
A(const std::string& str) : s(str)
{
}
A(A&& other)
{
// compiles if changing std::swap<std::string>() to std::swap()
std::swap<std::string>(s, other.s);
}
};
int main(int argc, char *argv[])
{
return 0;
}
the compiler emits error :
1>d:\program files (x86)\microsoft visual studio\2017\enterprise\vc\tools\msvc\14.11.25503\include\vector(2131):
error C2039: '_Alloc': is not a member of 'std::basic_string<char,std::char_traits<char>,std::allocator<char>>'
This code compiles without any issues with VC 15.2, VS2015 and VS2013 toolsets.
I'm trying to create a linux kernel module that will disable the data cache. I'm trying to use the v7_exit_coherency_flush(all) function in arch/arm/include/asm/cacheflush.h, and this function calls v7_flush_dcache_all, which I found is in arch/arm/mm/arch-v7.S.
My issue is that when I try to make my module, I get a warning
WARNING: "v7_flush_dcache_all [/home/pi/Documents/ARMHammer/kern/my_kernel/cache_disable.ko] undefined!
and when I try to insert the module I get an error
insmod: ERROR: could not insert module cache_disable.ko: Unknown symbol in module
So it looks like that ach-v7.S file isn't being read. I tried simply including it in my main file, but that produced a lot of errors, probably because its an assembly file.
I'm pretty much stuck at this point, is there someway I can include the assembly file in the Makefile, or maybe I'm not including all of the necessary .h files?
For what its worth, here's my main file
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/sched.h> /* For current */
#include <linux/tty.h> /* For the tty declarations */
#include <linux/version.h> /* For LINUX_VERSION_CODE */
#include <linux/mm.h>
#include <asm/cp15.h>
#include <asm/cacheflush.h>
#include <asm/glue-cache.h>
#include <asm/shmparam.h>
#include <asm/cachetype.h>
#include <asm/outercache.h>
// #include "my_cache-v7.h"
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Peter Jay Salzman");
static void print_string(char *str)
{
struct tty_struct *my_tty;
#if ( LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,5) )
my_tty = current->tty;
#else
my_tty = current->signal->tty;
#endif
if (my_tty != NULL) {
((my_tty->ops)->write) (my_tty, /* The tty itself */
#if ( LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,9) )
0, /* Don't take the string
from user space */
#endif
str, /* String */
strlen(str)); /* Length */
#if ( LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,9) )
((my_tty->ops)->write) (my_tty, 0, "\015\012", 2);
#else
((my_tty->ops)->write) (my_tty, "\015\012", 2);
#endif
}
}
static int __init print_string_init(void)
{
v7_exit_coherency_flush(all);
print_string("The module has been inserted. Hello world!");
return 0;
}
static void __exit print_string_exit(void)
{
print_string("The module has been removed. Farewell world!");
}
module_init(print_string_init);
module_exit(print_string_exit);
and my Makefile
obj-m += cache_disable.o
KDIR = /home/pi/linux/
all:
make -C $(KDIR) M=$(PWD) modules
clean:
make -C $(KDIR) M=$(PWD) clean
Also, if anybody knows an easier way to disable the cache, I'm all ears!
v7_exit_coherency_flush() is for power management code to take a CPU out of the kernel cleanly in order to power it off - it's not callable from random modules for very good reason. If you really want to lose data and crash the machine in weird and subtle ways, you might as well just bypass kernel functions entirely and use a trivial inline asm to hit the SCTLR directly*.
I dread to imagine what you're trying to achieve, but if you really want to run Linux (painfully slowly) with the cache off, you'll need to rebuild the kernel, turning off CONFIG_SMP in order to turn on CONFIG_CPU_DCACHE_DISABLE. That's the only vaguely supported method which might work.
* I'm not even going to explain that, it's that terrible an idea.
so somewhere along the lines of putting this app together I've started to get a runtime check failure stack corruption when the destructor for a cstring class member is called.
I've gotten to the point of trying to debug this by throwing bricks at the issue but still havent root caused it. At the current moment the class that the cstring resides in does nothing but initialize its private string members and set a pointer to another class to NULL.
Interestingly if I do not set the class pointer to NULL and comment out that line the corruption goes away. I think this is somewhat of a red herring, and that something is changing in the way the compiler is putting the code together when it pulls in the .h file that contains theCLog definitions and that would be used since I'm declaring a pointer to that object.
int _tmain(int argc, _TCHAR* argv[])
{
DWORD a = 0xBABA; //just to help catch the corrupter
DWORD b = 0xDFDF;
CStringW startat = L"\\\\anetworkshare\\fre";
CStringW lookfor = L".inf";
DirEnum myEnum(startat,lookfor);
ULONG en = a + b;
en = a - b;
return 0;
}
DirEnum.cpp
DirEnum::DirEnum(CString startingdir,CString fileFilter)
{
m_plogfile = NULL; //If you comment out this line corruption goes away
m_startingdir = L"";
m_extfilter = L"";
if(startingdir.GetLength() > 0)
m_startingdir = startingdir;
if(fileFilter.GetLength() > 0)
m_extfilter = fileFilter;
//following commented out to tshoot
//CLogBase& ref = ref.GetInstance();
//logBase = &ref;
//m_plogfile = new CLog(L"DirEnumerator",L"logfile.txt",logINFO);
}
Now I suspect that something in the log.h file is causing a change to occuur in the ATL or CString libraries but I dont know what. Heres the log.h file
#pragma once
//#define _ATL_CSTRING_EXPLICIT_CONSTRUCTORS // some CString constructors will be explicit
#ifndef UNICODE
#define UNICODE
#endif
#ifndef _UNICODE
#define _UNICODE
#endif
using namespace std;
#ifndef TYPEDEF_H
#define TYPEDEF_H
#include <string>
#include <sstream>
#include <iostream>
#include <fstream>
#include <tchar.h>
#include <time.h>
//simple defines to allow the TCHAR library to be used
typedef std::basic_string<TCHAR> tstring;
typedef std::basic_ostream<TCHAR> tostream;
typedef std::basic_istream<TCHAR> tistream;
typedef std::basic_ostringstream<TCHAR> tostringstream;
typedef std::basic_istringstream<TCHAR> tistringstream;
typedef std::basic_ofstream<TCHAR> tofstream;
#if defined(UNICODE) || defined(_UNICODE)
#define tcout std::wcout
#define tcin std::wcin
#else
#define tcout std::cout
#define tcin std::cin;
#endif
#endif
#if defined DEBUG || defined (_DEBUG)
#define TOCONSOLE
#endif
typedef enum LOGLVL{logERROR =0,logWARN,logINFO,logDEBUG};
//CLogBase os a singleton log class. Intent is that you can establish a reference from anywhere in the project and write to the same file
// without having locking issues or threading issues
class CLogBase
{
public:
static CLogBase& GetInstance(CString logname = L"log.txt",LOGLVL lvl = logWARN);
~CLogBase(void);
tostringstream& GetLog(LOGLVL level);
tostringstream& GetStream(LOGLVL);
void Forceflush();
private:
CLogBase(CString file,LOGLVL lvl);
//our outstream
tostringstream m_os;
tostringstream m_dummy;
tofstream m_filestream;
CString m_filename;
LOGLVL m_reportlvl;
//Private declarations to prevent copy constructors from being invoked; these are do nothig implimentations
CLogBase(CLogBase const&);
void operator=(CLogBase const&);
};
class CLog
{
public:
CLog(CString component);
CLog(CString component,CString logname,LOGLVL lvl);
~CLog();
void Log(LOGLVL,CString message);
void CLog::Flush();
tostringstream& CLog::GetStream(LOGLVL lvl);
private:
CString m_componentname;
CLogBase* m_logBase;
};
I thought I would answer this as I found the issue. this was not a coding problem per se but a visual studio issue.
What happened was that I was storing the direnum.h file and .cpp file in a different directory than the one used for the main project. referencing the header with #include "..\somedir\direnum.h"
at one point in time visual studio reported the file as locked and did I want to overwrite \ cancel etc. I choose overwrite but what seemed to happen was that this caused VS to COPY the files to the current project. All my troubleshooting attempts were being edited in the local somename.h file whtne opened in the editor but the compiler was doing the correct thing and pulling down the .h file from the location above.
removing switching to the now local copy of direnum.h and recompiling fixed this as it was compiling part of the code as ANSI and the other part as WCHAR
Maybe this sounds stupid, but I want to create a vector/array of SRWLocks dynamically during runtime. The following code compiles both ways, but the array thing doesn’t work. Probably the question should be, can I create an SRWLock object using the new operator?
If this is not how it’s done, what’s the right way to create an array of SRWLock objects?
#include "stdafx.h"
#include <ppl.h>
#include <Windows.h>
#include <iostream>
using namespace std;
#define NOT_WORKING
int _tmain(int argc, _TCHAR* argv[])
{
#ifdef NOT_WORKING
// The following doesn't work
SRWLOCK *lock = new SRWLOCK[2];
int a = 10;
AcquireSRWLockExclusive(&lock[0]);
cout<<"Exclusive Lock Acquired";
a++;
ReleaseSRWLockExclusive(&lock[0]);
// The following doesn't work either
/*
SRWLOCK *lock_n = new SRWLOCK;
AcquireSRWLockExclusive(lock_n);
cout<<"Exclusive Lock Acquired";
a++;
ReleaseSRWLockExclusive(lock_n);
*/
#else
// This works
SRWLOCK lock;
int a = 10;
AcquireSRWLockExclusive(&lock);
cout<<"Exclusive Lock Acquired";
a++;
ReleaseSRWLockExclusive(&lock);
#endif
cout<<"Value of a:"<<a;
cin.get();
return 0;
}
User1577008's own answer:
Got it! One needs to use the InitializeSRWLock() function when lock objects are created dynamically. Somebody paste this answer, and help me close this question.