Debugging memory leaks and memory corruption problems is something of a black art; there is probably no one right way and no one right tool to tackle these problems, most often you will have to try out multiple approaches and multiple tools.

The DGBMEM toolset is a drop-in memory allocator for Linux that adds some debugging functionality, it overrides the standard library functions like malloc and free that are returning new heap memory and adds some extra debugging functionality/

In order to use this tool you can use your process as is, and you don't have link your process with any libraries. This works well for ELF executables.

DBGMEM is a relatively primitive tool if compared with other tools like valgrind; though DBGMEM may sometimes be better suited for testing systems under live condition and under heavy load.

The tools also helped me to solve problem with real world systems, in other words it helped to save my but with a real job - which I think is very much for a tool like this ;-)

DBGMEM is a very feature-rich tool (some say that this is a euphemism for bloat, but well, Features Are a Good Thing).

Features

DBGMEM offers the following tools.

OPTIONS COMMON TO ALL TOOLS

A report containing all memory blocks + stack trace at allocation time is produced at program exit, or by application request; at the same occasion all memory blocks are checked for overwrites/underwrites.
Your application can call service functions that checks all memory blocks for overwrites / underwrites, or to print out all heap memory blocks; alternatively the tool enables you can install two signal handler for each of these task.
Whenever a memory error is encountered the event is logged into a log file and to standard error stream; you have the option to

Continue as normal
Dump core
Attach the gdb debugger to the currently running process.

The tool hooks the following GLIBC functions, and track dynamically allocated memory that is returned by them

malloc
calloc
realloc
memalign
posix_memalign
valign
free
strdup
strndup
getcwd
new/new[]/delete/delete[] (for C++ version only)
malloc_usable_size returns size of requested block.

There is an option to initialize all allocated heap memory with 0xDD and to set all freed memory content to 0xFF.This is a strong feature, references to freed memory or memory that is not initialized will cause program errors, which are otherwise left undetected. If freed/un-initialized memory contains a pointer, then dereferencing of this pointer results in a core dump.
There is an option to install SIGSEGV/SIGBUS signal handler that also checks all heap blocks for overwrites.
You can run an option that will invoke the top program in parallel with the debugged program so that it exclusively monitors the memory consumption of the debugged process.

TOOL: SIMPLE TOOL

pinpoint and track memory leaks in live systems with minimal overhead;
provides some checks for heap corruption; this tool keeps additional housekeeping information in a header before the allocated block (arena header) and after the allocated block, so heap corruption check is no always perfect; the checking tool supports these situations better.
Checks for double free, free of memory block that was not allocated.

TOOL: CHECKING TOOL

efficiently check for heap corruption, pinpoint and track memory leaks in live systems Housekeeping information is kept off the heap in an anonymous shared memory segment, so the heap checking is even better here, though a bit slower.
Checks for free of memory block that was not allocated.
For a set of standard library function we also check for following errors

heap overwrites in protected set of functions
Reading past allocated heap memory range
Stack smashing (limited support, the check triggers only if you smash the function return address)
The set of checked functions

memcpy
memmove
memset
strncpy
strcat
strncat
strcmp
strncmp

License

This software is licensed as LGPL v2 (GNU lesser general public license v 2.1)

You can use this product with proprietary applications as is; if you choose to modify this software then you are required to contribute your changes back.

Downloading

Download the latest version of this library from one of the following locations.

https://sourceforge.net/project/showfiles.php?group_id=235291

http://dbgmem.sourceforge.net/

Installation

# extract archive

tar xvfz DBGMEM.tar.gz .

# goto root directory

cd dbgmem

# build it and run sanity tests

# it's a shell script, really ;-)

./make

# alternatively, you can build the cpp version of this product and run sanity tests

# cpp version checks for matches of malloc/new/new[]

# with free/delete/delete[] if you think that this is important.

./make cpp

# install to /usr/local/dbgmem

# you can change the installation directory by setting

# environment variable DBG_ROOT prior to installation

# export DBG_ROOT=[you directory here]

./make install

How to prepare for debugging

First of all check that your program meets the following requirements.

Compile program without frame pointer omission (either no -O option or both -O and -fnoframepointer); this is required if you want it to produce meaningful stack trace.
Compile program with debug symbols (gcc option -g). The same applies to any libraries that you want to be checked.
Please check that you are not using the gcc options -static-libgcc or -static linker options. In this case the debugger will not see your memory allocations.
Also beware of the -nodefaultlibs linker option.
If you plan to stop your process with SIGINT or SIGTERM signals, please check that you do not have a signal handler for these signals; the tools signal handlers will not take effect in this situation.

Now I am out of reasons why this stuff would not work for you, please contact me if you run into problems.

Running the debugger

The debugger is invoked via the /users/local/dbgmem/scripts/run script (you can change the root installation dir though)

/users/local/dbgmem/scripts/run -d check -a gdb -s 7 test2

Runs checking tool that does leak detection and adds checks for heap overwrites (option -d check)
Whenever there is a memory error, invoke the debugger right now and attach it onto the running process (-a gdb)
For each allocation we will record seven stack frames (options -s 7)

/users/local/dbgmem/scripts/run -d simple -b -s 7 test2

runs leak detecting debugger (option -d simple)
allocated and freed memory will be initialized (option -b ; this is a very strong option)
for each allocation we will record seven stack frames (options -s 7)

Lets see what happens with simple debugger

/user/local/dbgmem/scripts/run -d simple -b -s 7 test2

debug engine: simple

output report: dbgmemlog.log

frames per entry: 7

command: test2

DBGMEM: running command with memory debugger...

DBGMEM: Please stop debuged process by signal other with SIGKILL (with Ctl+C SIGINT or SIGTERM)

DBGMEM: debugee started with process id 6701

Entry: 0x8427300 size: 10 generation: 1

status [memory low mark overwritten ptr=0x8427300 size=10]

frame 0 : 0x804841b

frame 1 : 0x8048468

frame 2 : 0x8048449

frame 3 : 0x8048468

frame 4 : 0x8048449

frame 5 : 0x8048468

frame 6 : 0x8048449

Entry: 0x84272b8 size: 10 generation: 1

status [memory upper mark overwritten ptr=0x84272b8 size=10]

frame 0 : 0x80483fa

frame 1 : 0x8048468

frame 2 : 0x8048449

frame 3 : 0x8048468

frame 4 : 0x8048449

frame 5 : 0x8048468

frame 6 : 0x8048449

DBGMEM: debugee exit code is 0

DBGMEM: Running analyser script, please wait...

DBGMEM: created report files: dbgmemlog.log

When a memory error occurs then the event is traced to standard error file;

A report is generated after running the debugged process (see file dbglog.log)

The report consists of the following sections

information returned by mallinfo (man page 3) information at the time that the report has been generated
memory errors; memory overwrites, twice free, etc. etc.
leak reports
A leak report can be divided into 'generations', i.e. distinct phases of the program execution can be tagged (say the initialization/setup phase is generation 0 and everything after setup is phase 1 - see ‘Memory Leaks section’ for more information)

======================

MALLOC INFO (MALLINFO)

======================

non-mmaped space allocated from system: 135168

number of free chunks: 1

number of fastbin blocks: 0

number of mmapped regions: 0

space in mmapped regions: 0

maximum total allocated space: 0

space available in freed fastbin blocks: 0

total allocated space: 328

total free space: 134840

top-most, releasable (via malloc_trim) space: 134840

DBGMEM arena size: 12

MALLINFO EOF

Entry: 0x9228120 size: 10 generation: 1 status [memory low mark overwritten ptr=0x9228120 size=10] frame: 0 : (0x804841b) ~"test_malloc2 + 107 in section .text\n" ~"Line 44 of \"test2.c

frame: 1 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 2 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 3 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 4 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 5 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 6 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

Entry: 0x9228100 size: 10 generation: 1 status [memory upper mark overwritten ptr=0x9228100 size=10] frame: 0 : (0x80483fa) ~"test_malloc2 + 74 in section .text\n" ~"Line 36 of \"test2.c

frame: 1 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 2 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 3 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 4 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 5 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 6 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

MEMALLOC: EOF

============

LEAK SUMMARY

============

===============

GENERATION [0]

===============

All-size: 80 | Entry: 0x92280e8 size: 8 Entry: 0x92280d0 size: 8 Entry: 0x92280b8 size: 8 Entry: 0x92280a0 size: 8 Entry: 0x9228088 size: 8 Entry: 0x9228070 size: 8 Entry: 0x9228058 size: 8 Entry: 0x9228040 size: 8 Entry: 0x9228028 size: 8 Entry: 0x9228010 size: 8

frame: 0 : (0x80483e8) ~"test_malloc2 + 56 in section .text\n" ~"Line 32 of \"test2.c

frame: 1 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 2 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 3 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 4 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 5 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 6 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

-----------------------------------

Sum of all user memory blocks for generation [0]: 80

Number of blocks this generation: 1

DBGMEM overhead: 12

===============

GENERATION [1]

===============

All-size: 8 | Entry: 0x9228140 size: 8

frame: 0 : (0x80483e8) ~"test_malloc2 + 56 in section .text\n" ~"Line 32 of \"test2.c

frame: 1 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 2 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 3 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 4 : (0x8048449) ~"test_malloc2 + 153 in section .text\n" ~"Line 56 of \"test2.c

frame: 5 : (0x8048468) ~"test_malloc + 20 in section .text\n" ~"Line 63 of \"test2.c

frame: 6 : (0x80484e6) ~"main + 121 in section .text\n" ~"Line 82 of \"test2.c

-----------------------------------

Sum of all user memory blocks for generation [1]: 88

Number of blocks this generation: 1

DBGMEM overhead: 12

See section “Memory leaks in more detail” for more information on this subject.

DBGMEM help printout , summary of options

run [-d simple|check] [-f log_file] [-a] [-s NUMBER] [-b] [-e] [-snext <signum>] [-sdump <signum>] <cmd-to-debug>

DBGMEM Memory debugger;

Starts a process with the DBGMEM memory debugger. Please run your test sequence and then terminate

the process with any signal other than SIGKILL, use SIGTERM or SIGINT instead.

Be friendly to the environment ;-)

Options:

-d simple|check choice of debug engine. possible values for this options:

simple - mainly good for detecting memory leaks

check - memory leaks detection + check for pointers in common library functions

The default value is simple.

-f outputfile set output logfile. if not set then the name is dbgmemlog.log

-s NUMBER number of stack frames that are kept per memory allocation.

we record the stack where allocation did happen.

The default value is five.

-a core|gdb|skip what to do when memory error is detected. value

core generate core

gdb attach the gnu debugger

skip continue as if nothing happened

default value is skip.

-b fill byte is set - uninitialised memory is set to 0xDD,

freed memory set to 0xEE

slows thing down but helps to find errors due to accessing

uninitialised or freed memory.

default is off.

-p <n> whenever an allocated block is checked for overwrites, then

adjacent <n> memory

blocks are checked too.

(this is called healthy paranoia mode)

default value is none, only one block is checked for validity

-e Install signal handler for SIGSEGV and SIGBUS.

the signal handler checks if heap is corrupted so that all bad

memory blocks are printed out.

and then dumps core.

this option is off by default

-leak <n1> <n2> install signal handlers that will help with leak detection

<n1> if number is not null then installs signal handler to

signal number <n1> which

checks if heap is corrupted

advances generation tag

<n2> if number is not null then installs signal handler to

signal number <n2> which

dumps all memory blocks to a raw log file.

this option is off by default

-t <seconds> run top command to print memory statistics of debugee; the statistics

are gathered every <seconds> seconds. The output file is <outputfile>.top

this option is off by default

-y symfile use argument symbol file in order to resolve addresses of functions in stack trace.

-v verbose mode; the debugee process prints options received to stdandard error.

-version print version of this software

-h print this help message

[cmd-to-debug] command line of process that will be started and debugged.

Note: this debugged process will set its core limit to unlimited.

Memory leaks in more detail

Memory leaks cause the process to consume more and more memory, leading to general performance degradation; swapping, and eventually to the process termination due to the fact that the process has run out of free memory. DBGMEM can help you to find these issues effectively.

Terminology

Let first look at the different stages in the life of a process.

The lifetime of a process can be divided into following logical stages:

[--- INITIALIZATION ---]

[--- ACTIVE STAGE ---]

[--- SHUTDOWN ---]

[--- EXIT ---]

Initialization

The process is initializing. Usually here the configuration is read, caches are initialized and objects and resources used throughout further stages are created and initialized.

Active state

The process is servicing request and does something useful;

During this stage the process completes one or more logical units of processing,

Each such unit may be the processing of a request or series of requests from a network, the processing of a batch data job, or the processing of an interactive user requests. Typically during each logical unit of processing some resources and memory is allocated and then either released or leaked.

Shutdown

The process has received a signal that will cause it to exit.

The process is cleaning up and is caches and is freeing resources.

This is a common but optional stage.

Exit

The process exits.

Causes for memory leaks

A leak is a resource that is allocated and not freed during ACTIVE STAGE of a process. If you allocated a block during initialization and forgot to clean it up then this has a fixed cost and will not have any affect (unless you have a shared memory space between processes).

There are the following types of leaks

Simply leaked; it is allocated and then we simply forgot to free it.
Added to caches for further use; this pattern may lead to a leak if an object is always added to a cache, but is never reused from that cache (bad cache leaks).

Leaks related to reference counting; Reference counting is a special form of garbage collection that is often implemented in C/C++ programs; there are potential problems such as

Leaking of reference counts; a reference counted object is only freed if all outstanding references are released. A leak can occur when at least one reference to an object remains, because somehow we forgot to clear an outstanding object reference. This scenario often occurs if a root object refers to all other objects but forgets to clear its references. Note that this is similar to bad cache leaks. Note that this scenario often happens in other programming languages such as Java.
Circular references; object A is referencing object B, which is also back-referencing object A; No matter what, the object A will always have at least one back reference from object A, so it will never be deleted.

Strategy for solving leak issues

The DBGMEM tool creates a detailed memory report that lists all memory blocks present at a particular point in time, this report includes the stack where each block was allocated; all heap objects that were allocated from the same stack location are summed up in one report entry.

Two reports taken at different points in time contain interesting information;

· On completing the SHUTDOWN sequence; right before the process exits normally This report is always generated by DBGMEM unless the process is killed by SIGKILL signal; you don't need to modify you program in order to get this report. This report lists the origin of most leak issues; the stack trace where the leaking memory blocks were allocated will give you a strong hint as to how to fix these issues, alas for some types of leaks related to reference counting you need more information.

· On completing the ACTIVE STAGE right before the SHUTDOWN sequence

We want to tag each memory allocation with the stage in the program lifetime when it was made. For instance we want to know if an allocation was done during INITIALIZATION stage, so we will not have to deal with it at all.

This stage is useful if you want to track reference counted leaks; usually these leaked references are removed during SHUTDOWN stage, when the root objects that are holding these references are deleted. When a lot of blocks occur in the report taken before SHUTDOWN stage, but are not present in the report taken before EXIT then this can indicate a leak. This analysis is not done by the tool, it is done by the avid reader, when he compares the two reports; the avid reader is also armed with knowledge of the workings of his program and will be able to conclude what is happening.

There are two ways to indicate the two state transitions

1. INITIALIZATION and ACTIVE STAGE

2. ACTIVE STAGE and SHUTDOWN stage

You can either

1. Enable DBGMEM command line option that sets up two signal handler, where each signal handler will do the requested state transition.

2. Modify the debugged program so that it calls the debug library to indicate the two events; this is the preferred solution if you want to run DBGMEM from within unit tests/system tests.

Example session for finding leaks

Lets run an example session – you will find test3.cpp in the src directory.

Option –v verbose option prints out additional output

This time –leak 31 33 option is given; the memory debugger installs two additional signal handler;

One signal handler is installed on signal 31 – this one increases the generation tag value and checks all memory blocks for overwrites.

The other on signal handler is installed on signal 33 – this one will dump all memory blocks.

~/dbgmem/src> ../scripts/run -v -d simple -leak 31 33 -s 10 ./test3

debug engine: simple

output report: dbgmemlog.log

frames per entry: 10

command: ./test3

DBGMEM: running command with memory debugger...

DBGMEM: Please stop debuged process by signal other with SIGKILL (with Ctl+C SIGINT or SIGTERM)

DBGMEM: debugee started with process id 1914

DBGMEM Options:

process id: 1914

ignore this process: off

allocator: simple

stack frames: 10

action on memory errors: continue

check adjacent nodes: off

init new & freed memory: off

SIGSEGV/SIGBUS error handler: off

Install signal for leak detection: on

signal to increment generation tag: 31

signal to dump heap contents: 33

log directory: /disks/uilstore6/disk602/mmoser/dbgmem/src

-\|/...

Now lets assume that the application is now up and running; we can now signal that the system is in ACTIVE stage; In another console from the same user let’s issue a signal.

kill -31 1914

The debugger will write the following

DBGMEM: check heap memory blocks

DBGMEM: increasing generation value

From now on all allocations will be tagged as generation 1 and for our purpose the debugged process is now in ACTIVE stage.

Lets assume that you have run a test scenario, and now its time to enter SHUTDOWN stage. Again in the second terminal send the signal.

kill -33 1914

The debugger will write

DBGMEM: logging all heap blocks...

Now exit the process by running Ctrl+C in the debugger.

DBGMEM: debugee exit code is 0

DBGMEM: Running analyser script, please wait...

DBGMEM: created report files: dbgmemlog.log dbgmemlog.log_snapshot1

Now we have the report files;

dbgmem.log_snapshot1 – all memory blocks taken when process enters SHUTDOWN stage

dbgmem.log - when process is about to terminate.

Lets look at the following dbgmem.log_snapshot1 log file; all leaks are printed, loud and clear; sorted by the number of bytes leaked etc. etc. etc.

======================

MALLOC INFO (MALLINFO)

======================

non-mmaped space allocated from system: 135168

number of free chunks: 1

number of fastbin blocks: 0

number of mmapped regions: 0

space in mmapped regions: 0

maximum total allocated space: 0

space available in freed fastbin blocks: 0

total allocated space: 6024

total free space: 129144

top-most, releasable (via malloc_trim) space: 129144

DBGMEM arena size: 64

MALLINFO EOF

MEMALLOC: EOF

============

LEAK SUMMARY

============

===============

GENERATION [0]

===============

All-size: 1624 | Entry: 0x903d540 size: 1304 Entry: 0x903d0f8 size: 320

frame: 0 : (0x19c86e) 0x19c86e - (rva: 571502) /usr/lib/libstdc++.so.5.0.3

frame: 1 : (0x1890f1) 0x1890f1 - (rva: 491761) /usr/lib/libstdc++.so.5.0.3

frame: 2 : (0x188ffd) 0x188ffd - (rva: 491517) /usr/lib/libstdc++.so.5.0.3

frame: 3 : (0x188b6c) 0x188b6c - (rva: 490348) /usr/lib/libstdc++.so.5.0.3

frame: 4 : (0x80493fa) ~"std::__simple_alloc<char*, std::__default_alloc_template<true, 0> >::allocate(unsigned int) + 34 in section .text\n" ~"Line 238 of \"/usr/include/c++/3.2.3/bits/stl_alloc.h

frame: 5 : (0x80492e1) ~"std::_Vector_alloc_base<char*, std::allocator<char*>, true>::_M_allocate(unsigned int) + 17 in section .text\n" ~"Line 121 of \"/usr/include/c++/3.2.3/bits/stl_vector.h

frame: 6 : (0x8048f9e) ~"std::vector<char*, std::allocator<char*> >::_M_insert_aux(__gnu_cxx::__normal_iterator<char**, std::vector<char*, std::allocator<char*> > >, char* const&) + 238 in section .text\n" ~"Line 900 of \"/usr/include/c++/3.2.3/bits/stl_vector.h

frame: 7 : (0x8048d45) ~"std::vector<char*, std::allocator<char*> >::push_back(char* const&) + 81 in section .text\n" ~"Line 498 of \"/usr/include/c++/3.2.3/bits/stl_vector.h

frame: 8 : (0x8048c04) ~"foo_bar::foo_bar() + 138 in section .text\n" ~"Line 20 of \"test3.cpp

frame: 9 : (0x8048b2c) ~"__static_initialization_and_destruction_0(int, int) + 34 in section .text\n" ~"Line 34 of \"test3.cpp

All-size: 495 | Entry: 0x903d490 size: 99 Entry: 0x903d3e0 size: 99 Entry: 0x903d330 size: 99 Entry: 0x903d280 size: 99 Entry: 0x903d048 size: 99

frame: 0 : (0x19c86e) 0x19c86e - (rva: 571502) /usr/lib/libstdc++.so.5.0.3

frame: 1 : (0x19c9bf) 0x19c9bf - (rva: 571839) /usr/lib/libstdc++.so.5.0.3

frame: 2 : (0x8048bef) ~"foo_bar::foo_bar() + 117 in section .text\n" ~"Line 19 of \"test3.cpp

frame: 3 : (0x8048b2c) ~"__static_initialization_and_destruction_0(int, int) + 34 in section .text\n" ~"Line 34 of \"test3.cpp

frame: 4 : (0x8048b75) ~"global constructors keyed to global + 21 in section .text\n" ~"Line 32 of \"test3.cpp

frame: 5 : (0x804c8b1) ~"__do_global_ctors_aux + 25 in section .text\n"

frame: 6 : (0x80487b5) ~"_init + 21 in section .init\n"

frame: 7 : (0x804c826) ~"__libc_csu_init + 26 in section .text\n"

frame: 8 : (0x21f74b) 0x21f74b - (rva: 87883) /lib/tls/libc-2.3.2.so

frame: 9 : (0x8048969) ~"_start + 33 in section .text\n"

All-size: 420 | Entry: 0x903e3d8 size: 140 Entry: 0x903e0b8 size: 140 Entry: 0x903dd98 size: 140

frame: 0 : (0x8048a31) ~"leak_it(int, unsigned int) + 61 in section .text\n" ~"Line 46 of \"test3.cpp

frame: 1 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 2 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 3 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 4 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 5 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 6 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 7 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 8 : (0x8048aa8) ~"main + 70 in section .text\n" ~"Line 68 of \"test3.cpp

frame: 9 : (0x21f79a) 0x21f79a - (rva: 87962) /lib/tls/libc-2.3.2.so

All-size: 390 | Entry: 0x903e308 size: 130 Entry: 0x903dfe8 size: 130 Entry: 0x903dcc8 size: 130

frame: 0 : (0x8048a31) ~"leak_it(int, unsigned int) + 61 in section .text\n" ~"Line 46 of \"test3.cpp

frame: 1 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 2 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 3 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 4 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 5 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 6 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 7 : (0x8048aa8) ~"main + 70 in section .text\n" ~"Line 68 of \"test3.cpp

frame: 8 : (0x21f79a) 0x21f79a - (rva: 87962) /lib/tls/libc-2.3.2.so

frame: 9 : (0x8048969) ~"_start + 33 in section .text\n"

All-size: 360 | Entry: 0x903e248 size: 120 Entry: 0x903df28 size: 120 Entry: 0x903dc08 size: 120

frame: 0 : (0x8048a31) ~"leak_it(int, unsigned int) + 61 in section .text\n" ~"Line 46 of \"test3.cpp

frame: 1 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 2 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 3 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 4 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 5 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 6 : (0x8048aa8) ~"main + 70 in section .text\n" ~"Line 68 of \"test3.cpp

frame: 7 : (0x21f79a) 0x21f79a - (rva: 87962) /lib/tls/libc-2.3.2.so

frame: 8 : (0x8048969) ~"_start + 33 in section .text\n"

All-size: 330 | Entry: 0x903e190 size: 110 Entry: 0x903de70 size: 110 Entry: 0x903db50 size: 110

frame: 0 : (0x8048a31) ~"leak_it(int, unsigned int) + 61 in section .text\n" ~"Line 46 of \"test3.cpp

frame: 1 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 2 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 3 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 4 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 5 : (0x8048aa8) ~"main + 70 in section .text\n" ~"Line 68 of \"test3.cpp

frame: 6 : (0x21f79a) 0x21f79a - (rva: 87962) /lib/tls/libc-2.3.2.so

frame: 7 : (0x8048969) ~"_start + 33 in section .text\n"

All-size: 100 | Entry: 0x903daa0 size: 100

frame: 0 : (0x8048a31) ~"leak_it(int, unsigned int) + 61 in section .text\n" ~"Line 46 of \"test3.cpp

frame: 1 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 2 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 3 : (0x8048a11) ~"leak_it(int, unsigned int) + 29 in section .text\n" ~"Line 44 of \"test3.cpp

frame: 4 : (0x8048aa8) ~"main + 70 in section .text\n" ~"Line 68 of \"test3.cpp

frame: 5 : (0x21f79a) 0x21f79a - (rva: 87962) /lib/tls/libc-2.3.2.so

frame: 6 : (0x8048969) ~"_start + 33 in section .text\n"

-----------------------------------

Sum of all user memory blocks for generation [0]: 3719

Number of blocks this generation: 7

DBGMEM overhead: 448

===============

GENERATION [1]

===============

All-size: 140 | Entry: 0x903e6f8 size: 140