Leaktrace 使用方法

分类于 阅读次数： Waline： 本文字数： 1.7k 阅读时长 ≈ 6 分钟

概述

LeakTracer 是在检查 C++ 程序内存泄漏时编写的一个小工具。我无法让 dmalloc 显示我想要的内容，我只看到了提到的 __builtin_return_address gcc-extension。

要使用 LeakTracer，请使用提供的 LeakCheck 脚本运行您的程序。它使用 LD_PRELOAD 特性在你的函数之上“覆盖”一些函数（不需要重新编译）。如果您的平台不支持 LD_PRELOAD，您可以将 LeakTracer.o 对象文件添加到 Makefile 中的对象并运行您的应用程序。

LeakTracer 利用 gdb 去输出发生内存泄露所发生的位置，它是通过 override operator new, operator delete, operator malloc, operator free 来实现检测。

用法

加载 leaktracer 库的 3 种方法：

  • 将您的程序链接到 libleaktracer.a
  • 将您的程序链接到 libleaktracer.so。您需要将 -lleaktracer 选项作为链接命令的第一个选项
  • 使用 LD_PRELOAD 环境变量以确保它在任何其他库之前加载

将 leaktracer 添加到程序中：

  • 添加头文件 MemoryTrace.hpp
  • 添加 leaktracer::MemoryTrace::GetInstance().startMonitoringAllThreads() 函数，作为起始检测位置
  • 添加 leaktracer::MemoryTrace::GetInstance().writeLeaksToFile("/mnt/extsd/leaks.out") 函数，作为结束位置，生成检测报告
  • 编译选项中添加-funwind-tables 生成 backtrace 信息表，添加-rdynamic
  • 链接选项中添加-Wl,-Bstatic -lleaktracer 静态链接 libleaktracer，最好加上-g3，处理报告时可以显示对应的代码
  • 编译后保存生成的 bin 文件，备用

生成报告及处理：

生成的报告中一行信息如下：

1
2
3
4
5
6
# LeakTracer report diff_utc_mono=1588164717.365890
leak, time=1434.190531, stack=0x7f05d63801d5 0x7f05d638e594 0x7f05d638d8f8 0x7f05d638c28d 0x7f05d6382738, size=1024, data=***********************************************...
leak, time=1444.194638, stack=0x7f05d70f9c58 0x4027f5 0x7f05d6333830 0x402369, size=100, data=."%...............................................
leak, time=1434.190625, stack=0x4024d7 0x7f05d6333830 0x402369, size=100, data=..................................................
leak, time=1444.194222, stack=0x7f05d70f9a5c 0x4027f0 0x7f05d6333830 0x402369, size=100, data=. %...............................................
leak, time=1434.190651, stack=0x402500 0x7f05d6333830 0x402369, size=100,
  • leak：代表内存泄露
  • time：代表调用分配内存函数的时间（开机到当前的时间）
  • stack：调用栈
  • size：泄露的内存大小
  • data：申请时内存中的数据

明显可以看到调用栈，callstack 全是地址，我们以使用 helpers 文件夹中的 leak-analyze-addr2line，leak-analyze-gdb 二个工具进行解析。或者借助 gdb、objdump 或 map 文件等手段得到该泄露源的真正文件/行号或函数范围。

报告处理

  • 使用 addr2line 处理 leaks.out，生成可读的调用栈：

    1
    leak-analyze-addr2line {bin} leaks.out > leaks.addr2line.txt

    生成的报告示例如下：

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    $ ./helpers/leak-analyze-addr2line memleak_test_so leaks.out 
    Processing "leaks.out" log for "memleak_test_so"
    Matching addresses to "memleak_test_so"
    found 49 leak(s)
    100 bytes lost in 1 blocks (one of them allocated at 1444.194222), from following call stack:
    	??:0
    	/home/cll/99_temp/memory_leak/leaktracer/memleak_way1/memleak_test.cpp:110
    	??:0
    	??:?
    400 bytes lost in 1 blocks (one of them allocated at 1444.194550), from following call stack:
    	??:0
    	/home/cll/99_temp/memory_leak/leaktracer/memleak_way1/memleak_test.cpp:110
    	??:0
    	??:?
    328 bytes lost in 1 blocks (one of them allocated at 1434.190960), from following call stack:
    	/home/cll/99_temp/memory_leak/leaktracer/memleak_way1/memleak_test.cpp:83
    	??:0
    	??:?

    解析：49 个重复调用，泄露 100 个字节的内存，其中一个泄漏点在 leaks.out 中的时间是 06219.879013，调用栈从下 往上看。

  • 使用 gdb 处理 leaks.out，生成可读的调用栈及源码对应：

    1
    leak-analyze-gdb {bin} leaks.out > leaks.gdb.txt

    生成的报告示例如下：

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    $ ./helpers/leak-analyze-gdb memleak_test_so  leaks.out 
    found 49 leak(s)
    (gdb) Reading symbols from memleak_test_so...done.
    16 bytes lost in 1 blocks (one of them allocated at 1434.190803), from following call stack:
    main + 364 in section .text
    0x4025e4 is in main() (memleak_test.cpp:80).
    80		new_delete_test *p_new_class_no_free =  new new_delete_test;
    No symbol matches 0x7f05d6333830.
    _start + 41 in section .text

    1 bytes lost in 1 blocks (one of them allocated at 1444.194798), from following call stack:
    No symbol matches 0x7f05d70f9cee.
    main + 893 in section .text
    0x4027f5 is in main() (memleak_test.cpp:111).
    111		leaktracer::MemoryTrace::GetInstance().stopAllMonitoring();
    No symbol matches 0x7f05d6333830.

    无论是 leak-analyze-addr2line，leak-analyze-gdb 二个工具进行解析。或者借助 gdb、objdump 或 map 文件等手段 得到该泄露源的真正文件/行号或函数范围。其中都会碰上各种库引用等问题导致？？？。都会很麻烦，然而 leaktracer 提供了相关源码，我们可以对 leaktracer 进行改造成符合我们的定制化需求。

定制

通过上述文章我们可以轻松理解 leaktracer 使用与 leaktracer 设计与实现。在实际使用过程中，基本上都会碰到一些小问题。需要对 leaktracer 进行定制化修改。以下的定制化请根据各自项目实际需求酌情进行修改。

  • callstack 深度问题
    callstack 深度默认只有 5 个。

    1
    2
    3
    #ifndef ALLOCATION_STACK_DEPTH
    #define ALLOCATION_STACK_DEPTH 5
    #endif

    因此当 callstack 深度较大时无法显示详细的调用关系。如下：

    1
    stack=0x7f05d63801d5 0x7f05d638e594 0x7f05d638d8f8 0x7f05d638c28d 0x7f05d6382738

    将默认 callstack 深度 5 修改为较大值，本文是 50。请根据实际情况调整，一般情况下对整个情况影响不大

    1
    2
    3
    #ifndef ALLOCATION_STACK_DEPTH
    #define ALLOCATION_STACK_DEPTH 100
    #endif

  • callstatck 地址问题
    无论是 leak-analyze-addr2line，leak-analyze-gdb 二个工具进行解析。或者借助 gdb、objdump 或 map 文件等手段 得到该泄露源的真正文件/行号或函数范围。因多个动态库及其他原因，或多或少都会碰上各种库引用等问题导 致？？？。会很麻烦

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    $ ./helpers/leak-analyze-addr2line memleak_test_so leaks.out 
    Processing "leaks.out" log for "memleak_test_so"
    Matching addresses to "memleak_test_so"
    found 49 leak(s)
    100 bytes lost in 1 blocks (one of them allocated at 1444.194222), from following call stack:
    	??:0
    	/home/cll/99_temp/memory_leak/leaktracer/memleak_way1/memleak_test.cpp:110
    	??:0
    	??:?
    400 bytes lost in 1 blocks (one of them allocated at 1444.194550), from following call stack:
    	??:0
    	/home/cll/99_temp/memory_leak/leaktracer/memleak_way1/memleak_test.cpp:110
    	??:0
    	??:?
    328 bytes lost in 1 blocks (one of them allocated at 1434.190960), from following call stack:
    	/home/cll/99_temp/memory_leak/leaktracer/memleak_way1/memleak_test.cpp:83
    	??:0
    	??:?

  • 文件路径不存在无法生成文件问题
    使用 void writeLeaksToFile(const char* reportFileName); 接口生成文件时，若文件夹路径不存在，则会生成文件失败。然而实际在较大工程中使用时，均会将内存泄露文件统一到某个指定路径。可能存在文件夹路径不存在的情况。

  • 未内存泄露确生成文件问题
    在较大工程中使用时，进程数较多。若未内存泄露确产生文件，对 UI 交互及直观上总有些不协调。

  • 解决办法参考 https://github.com/carlyleliu/LeakTracer/commit/49513580316bf998796b19ae7587b1a3e32bbfac

原理

leaktracer 主要的设计思路为：

  • 实现一组内存的分配/释放函数，这组函数的函数原型与系统的那一组完全一样，让被 trace 的 library 对于内存的分配/释放函数的调用都链接到自己实现的这一组函数中以 override 掉系统的那组内存/分配释放函数；
  • 自己实现的这组函数中的内存分配函数记录分配相关的信息，包括分配的内存的大小，callstack 等，并调用系统本来的内存分配函数去分配内存；
  • 自己实现的这组函数中的内存释放函数则销毁内存分配的相关记录，并使用系统的内存释放函数真正的释放内存；
  • 在 trace 结束时，遍历所有保存的内存分配记录的信息，并把这些信息保存进文件以供进一步的分析

override 系统内存分配/释放函数

LeakTracer 实现的用于 override 系统内存分配/释放函数的那组函数在 AllocationHandlers.cpp 中定义：

1
2
3
4
5
6
7
8
9
10
c++
void* operator new(size_t size)
void* operator new[] (size_t size)
void operator delete (void *p)
void operator delete[]
c
void *malloc(size_t size)
void free(void* ptr)
void* realloc(void *ptr, size_t size)
void* calloc(size_t nmemb, size_t size)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
//
// LeakTracer
// Contribution to original project by Erwin S. Andreasen
// site: <http://www.andreasen.org/LeakTracer/>
//
// Added by Michael Gopshtein, 2006
// mgopshtein@gmail.com
//
// Any comments/suggestions are welcome
//
 
#include "MemoryTrace.hpp"
#include "LeakTracer_l.hpp"
 
void* (*lt_malloc)(size_t size);
void  (*lt_free)(void* ptr);
void* (*lt_realloc)(void *ptr, size_t size);
void* (*lt_calloc)(size_t nmemb, size_t size);
 
void* operator new(size_t size) {
	void *p;
	leaktracer::MemoryTrace::Setup();
 
	p = LT_MALLOC(size);
	leaktracer::MemoryTrace::GetInstance().registerAllocation(p, size, false);
 
	return p;
}
 
void* operator new[] (size_t size) {
	void *p;
	leaktracer::MemoryTrace::Setup();
 
	p = LT_MALLOC(size);
	leaktracer::MemoryTrace::GetInstance().registerAllocation(p, size, true);
 
	return p;
}
 
void operator delete (void *p) {
	leaktracer::MemoryTrace::Setup();
 
	leaktracer::MemoryTrace::GetInstance().registerRelease(p, false);
	LT_FREE(p);
}
 
void operator delete[] (void *p) {
	leaktracer::MemoryTrace::Setup();
 
	leaktracer::MemoryTrace::GetInstance().registerRelease(p, true);
	LT_FREE(p);
}
 
/** -- libc memory operators -- **/
 
/* malloc
 * in some malloc implementation, there is a recursive call to malloc
 * (for instance, in uClibc 0.9.29 malloc-standard )
 * we use a InternalMonitoringDisablerThreadUp that use a tls variable to prevent several registration
 * during the same malloc
 */
void *malloc(size_t size)
{
	void *p;
	leaktracer::MemoryTrace::Setup();
 
	leaktracer::MemoryTrace::GetInstance().InternalMonitoringDisablerThreadUp();
	p = LT_MALLOC(size);
	leaktracer::MemoryTrace::GetInstance().InternalMonitoringDisablerThreadDown();
	leaktracer::MemoryTrace::GetInstance().registerAllocation(p, size, false);
 
	return p;
}
 
void free(void* ptr)
{
	leaktracer::MemoryTrace::Setup();
 
	leaktracer::MemoryTrace::GetInstance().registerRelease(ptr, false);
	LT_FREE(ptr);
}
 
void* realloc(void *ptr, size_t size)
{
	void *p;
	leaktracer::MemoryTrace::Setup();
 
	leaktracer::MemoryTrace::GetInstance().InternalMonitoringDisablerThreadUp();
 
	p = LT_REALLOC(ptr, size);
 
	leaktracer::MemoryTrace::GetInstance().InternalMonitoringDisablerThreadDown();
 
	if (p != ptr)
	{
		if (ptr)
			leaktracer::MemoryTrace::GetInstance().registerRelease(ptr, false);
		leaktracer::MemoryTrace::GetInstance().registerAllocation(p, size, false);
	}
	else
	{
		leaktracer::MemoryTrace::GetInstance().registerReallocation(p, size, false);
	}
 
	return p;
}
 
void* calloc(size_t nmemb, size_t size)
{
	void *p;
	leaktracer::MemoryTrace::Setup();
 
	leaktracer::MemoryTrace::GetInstance().InternalMonitoringDisablerThreadUp();
	p = LT_CALLOC(nmemb, size);
	leaktracer::MemoryTrace::GetInstance().InternalMonitoringDisablerThreadDown();
	leaktracer::MemoryTrace::GetInstance().registerAllocation(p, nmemb*size, false);
 
	return p;
}

  • 内存分配函数记录分配相关的信息

    inline void registerAllocation(void *p, size_t size, bool is_array); 记录每一次内存分配的相关信息

    1
    2
    3
    4
    5
    6
    7
    /** registers new memory allocation, should be called by the
    	 *  function intercepting "new" calls */
    	inline void registerAllocation(void *p, size_t size, bool is_array);

    	/** registers memory reallocation, should be called by the
    	 *  function intercepting realloc calls */
    	inline void registerReallocation(void *p, size_t size, bool is_array);

    inline void registerReallocation(void *p, size_t size, bool is_array); 记录每一次内存分配的相关信 息，限 realloc

    1
    2
    3
    /** registers memory reallocation, should be called by the
    	 *  function intercepting realloc calls */
    	inline void registerReallocation(void *p, size_t size, bool is_array);

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    27
    28
    29
    30
    31
    32
    33
    34
    35
    36
    37
    38
    39
    40
    41
    42
    43
    44
    45
    46
    47
    48
    // adds all relevant info regarding current allocation to map
    inline void MemoryTrace::registerAllocation(void *p, size_t size, bool is_array)
    {
    	allocation_info_t *info = NULL;
    	if (!AllMonitoringIsDisabled() && (__monitoringAllThreads || getThreadOptions().    monitoringAllocations) && p != NULL) {
    		MutexLock lock(__allocations_mutex);
    		info = __allocations.insert(p);
    		if (info != NULL) {
    			info->size = size;
    			info->isArray = is_array;
    			storeTimestamp(info->timestamp);
    		}
    	}
     	// we store the stack without locking __allocations_mutex
    	// it should be safe enough
    	// prevent a deadlock between backtrave function who are now using advanced dl_iterate_phdr     function
     	// and dl_* function which uses malloc functions
    	if (info != NULL) {
    		storeAllocationStack(info->allocStack);
    	}

    	if (p == NULL) {
    		InternalMonitoringDisablerThreadUp();
    		// WARNING
    		InternalMonitoringDisablerThreadDown();
    	}
    }

    // adds all relevant info regarding current allocation to map
    inline void MemoryTrace::registerReallocation(void *p, size_t size, bool is_array)
    {
    	if (!AllMonitoringIsDisabled() && (__monitoringAllThreads || getThreadOptions().    monitoringAllocations) && p != NULL) {
    		MutexLock lock(__allocations_mutex);
    		allocation_info_t *info = __allocations.find(p);
    		if (info != NULL) {
    			info->size = size;
    			info->isArray = is_array;
    			storeAllocationStack(info->allocStack);
    			storeTimestamp(info->timestamp);
    		}
    	}

    	if (p == NULL) {
    		InternalMonitoringDisablerThreadUp();
    		// WARNING
    		InternalMonitoringDisablerThreadDown();
    	}
    }

  • 内存释放函数则销毁内存分配的相关记录

    inline void registerReallocation(void *p, size_t size, bool is_array); 记录每一次内存释放的相关信息

    1
    2
    3
    /** registers new memory allocation, should be called by the
    	 *  function intercepting "new" calls */
    	inline void registerAllocation(void *p, size_t size, bool is_array);

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    // removes allocation's info from the map
    inline void MemoryTrace::registerRelease(void *p, bool is_array)
    {
    	if (!AllMonitoringIsDisabled() && __monitoringReleases && p != NULL) {
    		MutexLock lock(__allocations_mutex);
    		allocation_info_t *info = __allocations.find(p);
    		if (info != NULL) {
    			if (info->isArray != is_array) {
    				InternalMonitoringDisablerThreadUp();
    				// WARNING
    				InternalMonitoringDisablerThreadDown();
    			}
    			__allocations.release(p);
    		}
    	}
    }

  • 遍历所有保存的内存分配记录的信息，并把这些信息保存

    void writeLeaksToFile(const char* reportFileName); 保存内存分配记录信息到文件。

    1
    2
    /** writes report with all memory leaks */
    	void writeLeaksToFile(const char* reportFileName);

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    // writes all memory leaks to given stream
    void MemoryTrace::writeLeaksToFile(const char* reportFilename)
    {
    	MutexLock lock(__allocations_mutex);
    	InternalMonitoringDisablerThreadUp();

    	std::ofstream oleaks;
    	if (!isFolderExist(reportFilename)) {
    		createDirectory(reportFilename);
    	}
    	if (__allocations.empty()) {
    		return; //no memory leak, not need to create leak file
    	}

    	oleaks.open(reportFilename, std::ios_base::out);
    	if (oleaks.is_open())
    	{
    		writeLeaksPrivate(oleaks);
    		oleaks.close();
    	}
    	else
    	{
    		std::cerr << "Failed to write to \\"" << reportFilename << "\\"\\n";
    	}
    	InternalMonitoringDisablerThreadDown();
    }

  • 遍历自定义 MapMemoryInfo 中所有元素

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    27
    28
    29
    30
    31
    32
    33
    34
    35
    36
    37
    38
    39
    40
    41
    42
    43
    44
    45
    46
    47
    48
    49
    50
    51
    52
    53
    54
    55
    56
    57
    58
    59
    60
    61
    62
    63
    64
    65
    66
    67
    68
    69
    70
    71
    72
    73
    74
    75
    76
    77
    78
    79
    80
    81
    82
    83
    84
    85
    86
    87
    88
    89
    90
    91
    92
    93
    94
    95
    96
    97
    98
    99
    100
    101
    102
    103
    // writes all memory leaks to given stream
    void MemoryTrace::writeLeaksPrivate(std::ostream &out)
    {
    	struct timespec mono, utc, diff;
    	allocation_info_t *info;
    	void *p;
    	double d;
    	const int precision = 6;
    	int maxsecwidth;

    	clock_gettime(CLOCK_REALTIME, &utc);
    	clock_gettime(CLOCK_MONOTONIC, &mono);

    	if (utc.tv_nsec > mono.tv_nsec) {
    		diff.tv_nsec = utc.tv_nsec - mono.tv_nsec;
    		diff.tv_sec = utc.tv_sec - mono.tv_sec;
    	} else {
    		diff.tv_nsec = 1000000000 - (mono.tv_nsec - utc.tv_nsec);
    		diff.tv_sec = utc.tv_sec - mono.tv_sec -1;
    	}

    	maxsecwidth = 0;
    	while(mono.tv_sec > 0) {
    		mono.tv_sec = mono.tv_sec/10;
    		maxsecwidth++;
    	}
    	if (maxsecwidth == 0) maxsecwidth=1;

    	out << "# LeakTracer report";
    	d = diff.tv_sec + (((double)diff.tv_nsec)/1000000000);
    	out << " diff_utc_mono=" << std::fixed << std::left << std::setprecision(precision) << d ;
    	out << "\\n";

    	__allocations.beginIteration();
    	while (__allocations.getNextPair(&info, &p)) {
    		d = info->timestamp.tv_sec + (((double)info->timestamp.tv_nsec)/1000000000);
    		out << "leak, ";
    		out << "time="  << std::fixed << std::right << std::setprecision(precision) <<  std::setfill('0') << std::setw(maxsecwidth+1+precision) << d << ", "; // setw(16) ?
    		out << "stack=";
    		#ifdef BACKTRACE_SYMBOLS_USED
    		unsigned int i_depth = 0;
    		for (i_depth = 0; i_depth < ALLOCATION_STACK_DEPTH; i_depth++) {
    			if (info->allocStack[i_depth] == NULL) break;

    			if (i_depth > 0) out << ' ';
    			out << info->allocStack[i_depth];
    		}
    		out << '\\n';
    		char **trace_symbols = (char **)backtrace_symbols (info->allocStack, i_depth);
    		if (NULL != trace_symbols) {
    			size_t name_size = 64;
    			char *name = (char*)malloc(name_size);
    			for (unsigned int j = 0; j < i_depth; j++) {
    				char *begin_name = 0;
    				char *begin_offset = 0;
    				char *end_offset = 0;
    				for (char *p = trace_symbols[j]; *p; ++p) {
    					if (*p == '(') {
    						begin_name = p;
    					} else if (*p == '+' && begin_name) {
    						begin_offset = p;
    					} else if (*p == ')' && begin_offset) {
    						end_offset = p;
    						break;
    					}
    				}
    				if (begin_name && begin_offset && end_offset ) {
    					*begin_name++ = '\\0';
    					*begin_offset++ = '\\0';
    					*end_offset = '\\0';
    					int status = -4;
    					char *ret = abi::__cxa_demangle(begin_name, name, &name_size, &status);
    					if (0 == status) {
    						name = ret;
    						out << trace_symbols[j] << ":" << name << "+" << begin_offset;
    					} else {
    						out << trace_symbols[j] << ":" << begin_name << "()+" << begin_offset;
    					}
    				} else {
    					out << trace_symbols[j];
    				}
    				out << '\\n';
    			}
    			free(trace_symbols);
    		}
    		#else
    		for (unsigned int i = 0; i < ALLOCATION_STACK_DEPTH; i++) {
    			if (info->allocStack[i] == NULL) break;

    			if (i > 0) out << ' ';
    			out << info->allocStack[i];
    		}
    		out << ", ";
    		#endif
    		out << "size=" << info->size << ", ";

    		out << "data=";
    		const char *data = reinterpret_cast<const char *>(p);
    		for (unsigned int i = 0; i < PRINTED_DATA_BUFFER_SIZE && i < info->size; i++)
    			out << (isprint(data[i]) ? data[i] : '.');
    		out << '\\n';
    	}
    }

整体来看 LeakTracer 的设计与实现都并不复杂，因而能够 trace 的 memory issue 也就有限。比如，LeakTracer 就无法 trace 多次释放等问题。但我们可以通过源码编写更强大的内存相关的 trace 工具。

💡 注意：对于应用，无论是静态链接或动态链接的库，若想能够生成 callstack，都需要添加-funwind-tables 重新编译一次 leak-analyze-addr2line 和 leak-analyze-gdb 在 libleaktracer.tar.gz 中，需要将其中的 addr2line 和 gdb 改成交叉编译工具链中的 arm-xxx-addr2line 和 arm-xxx-gdb。