需要有关 valgrind 输出理解的帮助

Question

看到我只是使用 valgrind 来检测代码中的内存泄漏。所以我已经使用了

valgrind -v --leak-check=full --show-reachable=yes  ./my_binary

，我看到了以下我无法理解的内容，

`==16545== 
==16545== HEAP SUMMARY:
==16545==     in use at exit: 20,171 bytes in 647 blocks
==16545==   total heap usage: 993 allocs, 346 frees, 30,090 bytes allocated
==16545== 
==16545== Searching for pointers to 647 not-freed blocks
==16545== Checked 124,548 bytes
==16545== 
==16545== 1 bytes in 1 blocks are still reachable in loss record 1 of 253
==16545==    at 0x400677E: malloc (vg_replace_malloc.c:195)
==16545==    by 0x80AAB31: xmalloc (in /bin/bash)
==16545==    by 0x80796D0: make_variable_value (in /bin/bash)
==16545==    by 0x80798C8: ??? (in /bin/bash)
==16545==    by 0x807BA40: initialize_shell_variables (in /bin/bash)
==16545==    by 0x805E897: ??? (in /bin/bash)
==16545==    by 0x805FA10: main (in /bin/bash)
==16545== 
==16545== 1 bytes in 1 blocks are still reachable in loss record 2 of 253
==16545==    at 0x400677E: malloc (vg_replace_malloc.c:195)
==16545==    by 0x80AAB31: xmalloc (in /bin/bash)
==16545==    by 0x80A578B: set_locale_var (in /bin/bash)
==16545==    by 0x80A5912: set_default_lang (in /bin/bash)
==16545==    by 0x805FA15: main (in /bin/bash)
==16545== 
==16545== 1 bytes in 1 blocks are still reachable in loss record 3 of 253
==16545==    at 0x400677E: malloc (vg_replace_malloc.c:195)
==16545==    by 0x80AAB31: xmalloc (in /bin/bash)
==16545==    by 0x8061B9B: ??? (in /bin/bash)
==16545==    by 0x8062D95: ??? (in /bin/bash)
==16545==    by 0x8065481: ??? (in /bin/bash)
==16545==    by 0x806801A: yyparse (in /bin/bash)
==16545==    by 0x8060580: parse_command (in /bin/bash)
==16545==    by 0x806064F: read_command (in /bin/bash)
==16545==    by 0x80608AE: reader_loop (in /bin/bash)
==16545==    by 0x805FEFB: main (in /bin/bash)`e

我只是想了解这一点。？ 为什么这没有给我ma代码中尚未释放的特定行号？

Answer 1

这些错误来自 /bin/bash 而不是您自己的代码。您是否在代码中运行任何 bash 脚本？ 这里解释了它的含义。我认为你可以忽略它。

“仍然可达”意味着你的程序可能没问题——它不是免费的
它可能有一些记忆。这是很常见的，而且往往是合理的。
如果您不想看到这些报告，请不要使用 --show-reachable=yes。

Answer 2

看起来这些地址属于您的程序代码。也许使用调试信息进行编译可以让 valgrind 看到行号。 （提示：-g 开启 gcc）

Answer 3

您检查过 http://valgrind.org/docs/manual/快速启动.html#quick-start.intro ？那里有一个很好的文档。您在编译时不应该使用任何优化。

Answer 4

如果您想检查自动工具驱动的编译的结果（它实际上在您的构建目录中创建了一个 shell 脚本），您可能想尝试以下操作：

libtool --mode=execute valgrind -v --leak-check=full --show-reachable=yes  ./my_binary