文章转载自：使用 Valgrind 检测 C++ 内存泄漏
Valgrind 的介绍
Valgrind 可以用来检测程序是否有非法使用内存的问题，例如访问未初始化的内存、访问数组时越界、忘记释放动态内存等问题。在 Linux 可以使用下面的命令安装 Valgrind：
wget ftp://sourceware.org/pub/valgrind/valgrind-3.13.0.tar.bz2 bzip2 -d valgrind-3.13.0.tar.bz2 tar -xf valgrind-3.13.0.tar cd valgrind-3.13.0 ./configure && make sudo make install
检测内存泄漏
Valgrind 可以用来检测程序在哪个位置发生内存泄漏，例如下面的程序： int main() { int *array = malloc(sizeof(int)); return 0; }
编译程序时，需要加上
-g选项：gcc -g -o main_c main.c
使用 Valgrind 检测内存使用情况：
valgrind --tool=memcheck --leak-check=full ./main_c
==31416== Memcheck, a memory error detector
==31416== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==31416== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==31416== Command: ./main_c
==31416==
==31416==
==31416== HEAP SUMMARY:
==31416== in use at exit: 4 bytes in 1 blocks
==31416== total heap usage: 1 allocs, 0 frees, 4 bytes allocated
==31416==
==31416== 4 bytes in 1 blocks are definitely lost in loss record 1 of 1
==31416== at 0x4C2DBF6: malloc (vg_replace_malloc.c:299)
==31416== by 0x400537: main (main.c:5)
==31416==
==31416== LEAK SUMMARY:
==31416== definitely lost: 4 bytes in 1 blocks
==31416== indirectly lost: 0 bytes in 0 blocks
==31416== possibly lost: 0 bytes in 0 blocks
==31416== still reachable: 0 bytes in 0 blocks
==31416== suppressed: 0 bytes in 0 blocks
==31416==
==31416== For counts of detected and suppressed errors, rerun with: -v
==31416== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
先看看输出信息中的HEAP SUMMARY，它表示程序在堆上分配内存的情况，其中的1 allocs表示程序分配了 1 次内存，0 frees表示程序释放了 0 次内存，4 bytes allocated表示分配了 4 个字节的内存。
另外，Valgrind 也会报告程序是在哪个位置发生内存泄漏。例如，从下面的信息可以看到，程序发生了一次内存泄漏，位置是main.c文件的第 5 行：
==31416== 4 bytes in 1 blocks are definitely lost in loss record 1 of 1 ==31416== at 0x4C2DBF6: malloc (vg_replace_malloc.c:299) ==31416== by 0x400537: main (main.c:5)
Valgrind 也可以用来检测 C++ 程序的内存泄漏，下面是一个正常的 C++ 程序，没有发生内存泄漏：
include
int main()
{
auto ptr = new std::string("Hello, World!");
delete ptr;
return 0;
}
使用 Valgrind 分析这段程序：
valgrind --tool=memcheck --leak-check=full --show-leak-kinds=all ./main_cpp ==31438== Memcheck, a memory error detector ==31438== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al. ==31438== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info ==31438== Command: ./main_cpp ==31438== ==31438== ==31438== HEAP SUMMARY: ==31438== in use at exit: 72,704 bytes in 1 blocks ==31438== total heap usage: 2 allocs, 1 frees, 72,736 bytes allocated ==31438== ==31438== 72,704 bytes in 1 blocks are still reachable in loss record 1 of 1 ==31438== at 0x4C2DBF6: malloc (vg_replace_malloc.c:299) ==31438== by 0x4EC3EFF: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.21) ==31438== by 0x40104E9: call_init.part.0 (dl-init.c:72) ==31438== by 0x40105FA: call_init (dl-init.c:30) ==31438== by 0x40105FA: _dl_init (dl-init.c:120) ==31438== by 0x4000CF9: ??? (in /lib/x86_64-linux-gnu/ld-2.23.so) ==31438== ==31438== LEAK SUMMARY: ==31438== definitely lost: 0 bytes in 0 blocks ==31438== indirectly lost: 0 bytes in 0 blocks ==31438== possibly lost: 0 bytes in 0 blocks ==31438== still reachable: 72,704 bytes in 1 blocks ==31438== suppressed: 0 bytes in 0 blocks ==31438== ==31438== For counts of detected and suppressed errors, rerun with: -v ==31438== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)
使用 Valgrind 分析 C++ 程序时，有一些问题需要留意。例如，这个程序并没有发生内存泄漏，但是从HEAP SUMMARY可以看到，程序分配了 2 次内存，但却只释放了 1 次内存，为什么会这样呢？
实际上这是由于 C++ 在分配内存时，为了提高效率，使用了它自己的内存池。当程序终止时，内存池的内存才会被操作系统回收，所以 Valgrind 会将这部分内存报告为 reachable 的，需要注意，reachable 的内存不代表内存泄漏，例如，从上面的输出中可以看到，有 72704 个字节是 reachable 的，但没有报告内存泄漏。
检测越界访问
C++ 程序经常出现的 Bug 就是数组越界访问，例如下面的程序出现了越界访问：
#include
#include
int main()
{
std::vector v(10, 0);
std::cout << v[10] << std::endl;
return 0;
}
使用 Valgrind 分析这段程序，Valgrind 会提示越界访问：
g++ -std=c++11 -g -o main_cpp main.cpp valgrind --tool=memcheck --leak-check=full ./main_cpp ==31523== Memcheck, a memory error detector ==31523== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al. ==31523== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info ==31523== Command: ./main_cpp ==31523== ==31523== Invalid read of size 4 ==31523== at 0x400AD7: main (main.cpp:7) ==31523== Address 0x5ab5ca8 is 0 bytes after a block of size 40 alloc'd ==31523== at 0x4C2E216: operator new(unsigned long) (vg_replace_malloc.c:334) ==31523== by 0x4010D3: __gnu_cxx::new_allocator::allocate(unsigned long, void const*) (new_allocator.h:104) ==31523== by 0x401040: std::allocator_traits<std::allocator >::allocate(std::allocator&, unsigned long) (alloc_traits.h:491) ==31523== by 0x400F91: std::_Vector_base<int, std::allocator >::_M_allocate(unsigned long) (stl_vector.h:170) ==31523== by 0x400E7E: std::_Vector_base<int, std::allocator >::_M_create_storage(unsigned long) (stl_vector.h:185) ==31523== by 0x400D1E: std::_Vector_base<int, std::allocator >::_Vector_base(unsigned long, std::allocator const&) (stl_vector.h:136) ==31523== by 0x400C11: std::vector<int, std::allocator >::vector(unsigned long, int const&, std::allocator const&) (stl_vector.h:291) ==31523== by 0x400AB9: main (main.cpp:6)
Invalid read of size 4表示越界读取 4 个字节，这个操作出现在main.cpp文件的第 7 行。另外可以看到，vector分配了一块 40 字节的内存，程序越界访问紧急着这块内存之后的 4 个字节。
检测未初始化的内存
另一种经常出现的 Bug，就是程序访问了未初始化的内存。例如：
#include
int main()
{
int x;
if (x == 0)
{
std::cout << "X is zero" << std::endl;
}
return 0;
}
使用 Valgrind 检测这个程序：
g++ -std=c++11 -g -o main_cpp main.cpp valgrind --tool=memcheck --leak-check=full ./main_cpp ==31554== Memcheck, a memory error detector ==31554== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al. ==31554== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info ==31554== Command: ./main_cpp ==31554== ==31554== Conditional jump or move depends on uninitialised value(s) ==31554== at 0x400852: main (main.cpp:6)
输出中提示了main.cpp文件的第 6 行访问了未初始化的内存。