关于一个内存泄露的问题

是这样子的
这样子的代码居然能够正常运行

#include 
#include 
using namespace std;
vector<int>::iterator fun(vector<int>  t){
  return t.begin()+2;//返回一个局部对象的迭代器
}
int main(){
  vector<int> test = {1,2,3,4};
  auto p1 = fun(test);
  cout << *p1 << endl;//对之前局部的迭代器进行解引用，应该是不行的
}
// 但是能够正常的输出3

不清楚为什么
使用内存检测工具检查了一下

valgrind --tool=memcheck --leak-check=full --show-reachable=yes ./test.out
==4677== Memcheck, a memory error detector
==4677== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==4677== Using Valgrind-3.12.0 and LibVEX; rerun with -h for copyright info
==4677== Command: ./test.out
==4677== 
==4677== Invalid read of size 4
==4677==    at 0x400BA8: main (t.cpp:13)
==4677==  Address 0x5ab4cd8 is 8 bytes inside a block of size 16 free'd
==4677==    at 0x4C2F160: operator delete(void*) (vg_replace_malloc.c:576)
==4677==    by 0x4016AB: __gnu_cxx::new_allocator::deallocate(int*, unsigned long) (new_allocator.h:110)
==4677==    by 0x4015B8: std::allocator_traits::allocator >::deallocate(std::allocator&, int*, unsigned long) (alloc_traits.h:517)
==4677==    by 0x4013A5: std::_Vector_base, std::allocator >::_M_deallocate(int*, unsigned long) (stl_vector.h:178)
==4677==    by 0x401094: std::_Vector_base, std::allocator >::~_Vector_base() (stl_vector.h:160)
==4677==    by 0x400EB2: std::vector, std::allocator >::~vector() (stl_vector.h:425)
==4677==    by 0x400B9B: main (t.cpp:12)
==4677==  Block was alloc'd at
==4677==    at 0x4C2E1D6: operator new(unsigned long) (vg_replace_malloc.c:334)
==4677==    by 0x4016EF: __gnu_cxx::new_allocator::allocate(unsigned long, void const*) (new_allocator.h:104)
==4677==    by 0x40160D: std::allocator_traits::allocator >::allocate(std::allocator&, unsigned long) (alloc_traits.h:491)
==4677==    by 0x401435: std::_Vector_base, std::allocator >::_M_allocate(unsigned long) (stl_vector.h:170)
==4677==    by 0x401508: std::_Vector_base, std::allocator >::_M_create_storage(unsigned long) (stl_vector.h:185)
==4677==    by 0x40122C: std::_Vector_base, std::allocator >::_Vector_base(unsigned long, std::allocator const&) (stl_vector.h:136)
==4677==    by 0x400F19: std::vector, std::allocator >::vector(std::vector, std::allocator > const&) (stl_vector.h:320)
==4677==    by 0x400B7F: main (t.cpp:12)
==4677== 
3
==4677== 
==4677== HEAP SUMMARY:
==4677==     in use at exit: 72,704 bytes in 1 blocks
==4677==   total heap usage: 4 allocs, 3 frees, 73,760 bytes allocated
==4677== 
==4677== 72,704 bytes in 1 blocks are still reachable in loss record 1 of 1
==4677==    at 0x4C2DBB6: malloc (vg_replace_malloc.c:299)
==4677==    by 0x4EC2EFF: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.21)
==4677==    by 0x40104E9: call_init.part.0 (dl-init.c:72)
==4677==    by 0x40105FA: call_init (dl-init.c:30)
==4677==    by 0x40105FA: _dl_init (dl-init.c:120)
==4677==    by 0x4000CF9: ??? (in /lib/x86_64-linux-gnu/ld-2.23.so)
==4677== 
==4677== LEAK SUMMARY:
==4677==    definitely lost: 0 bytes in 0 blocks
==4677==    indirectly lost: 0 bytes in 0 blocks
==4677==      possibly lost: 0 bytes in 0 blocks
==4677==    still reachable: 72,704 bytes in 1 blocks
==4677==         suppressed: 0 bytes in 0 blocks
==4677== 
==4677== For counts of detected and suppressed errors, rerun with: -v
==4677== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

后来拿了另外一个代码进行测试

#include 
#include 
using namespace std;
vector<int>::iterator fun(vector<int>&  t){//修改了，变成传引用
  return t.begin()+2;
}
int main(){
  vector<int> test = {1,2,3,4};
  auto p1 = fun(test);
  cout << *p1 << endl;
}

这个是肯定没问题的
同样的内存检测工具进行

valgrind --tool=memcheck --leak-check=full --show-reachable=yes ./test.out
==4743== Memcheck, a memory error detector
==4743== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==4743== Using Valgrind-3.12.0 and LibVEX; rerun with -h for copyright info
==4743== Command: ./test.out
==4743== 
3
==4743== 
==4743== HEAP SUMMARY:
==4743==     in use at exit: 72,704 bytes in 1 blocks
==4743==   total heap usage: 3 allocs, 2 frees, 73,744 bytes allocated
==4743== 
==4743== 72,704 bytes in 1 blocks are still reachable in loss record 1 of 1
==4743==    at 0x4C2DBB6: malloc (vg_replace_malloc.c:299)
==4743==    by 0x4EC2EFF: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.21)
==4743==    by 0x40104E9: call_init.part.0 (dl-init.c:72)
==4743==    by 0x40105FA: call_init (dl-init.c:30)
==4743==    by 0x40105FA: _dl_init (dl-init.c:120)
==4743==    by 0x4000CF9: ??? (in /lib/x86_64-linux-gnu/ld-2.23.so)
==4743== 
==4743== LEAK SUMMARY:
==4743==    definitely lost: 0 bytes in 0 blocks
==4743==    indirectly lost: 0 bytes in 0 blocks
==4743==      possibly lost: 0 bytes in 0 blocks
==4743==    still reachable: 72,704 bytes in 1 blocks
==4743==         suppressed: 0 bytes in 0 blocks
==4743== 
==4743== For counts of detected and suppressed errors, rerun with: -v
==4743== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

同样能够正常运行
不过内存检测工具还是检查出问题来了

问题：
不知道为什么之前那个代码能够正常运作