c++ - STL vector 的大规模删除导致我的二进制文件崩溃

Question

我的二进制文件崩溃了。在运行核心转储时，我发现了以下内容:

#0  0x00a6a363 in memmove () from /lib/tls/libc.so.6
(gdb) frame 1
#1  0x083a108c in std::__copy_trivial<piola::piolaOrderBook*> (__first=0xb277f2c4, __last=0xb277f2bc, __result=0xb277f2c0)
    at /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../include/c++/3.4.6/bits/stl_algobase.h:258
258           std::memmove(__result, __first, sizeof(_Tp) * (__last - __first));
(gdb) frame 2
#2  0x083a0ad6 in std::__copy_aux2<piola::piolaOrderBook*> (__first=0xb277f2c4, __last=0xb277f2bc, __result=0xb277f2c0)
    at /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../include/c++/3.4.6/bits/stl_algobase.h:279
279         { return std::__copy_trivial(__first, __last, __result); }
(gdb) frame 3
#3  0x083a02d1 in std::__copy_ni2<piola::piolaOrderBook**, __gnu_cxx::__normal_iterator<piola::piolaOrderBook**, std::vector<piola::piolaOrderBook*, std::allocator<emapi::EmapiOrderBook*> > > > (__first=0xb277f2c4, __last=0xb277f2bc, __result=
      {_M_current = 0xb277f2c0})
    at /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../include/c++/3.4.6/bits/stl_algobase.h:296
296           return _OutputIterator(std::__copy_aux2(__first, __last, __result.base(),
(gdb) frame 4
#4  0x0839f1b0 in std::__copy_ni1<__gnu_cxx::__normal_iterator<piola::piolaOrderBook**, std::vector<piola::piolaOrderBook*, std::allocator<piola::piolaOrderBook*> > >, __gnu_cxx::__normal_iterator<piola::piolaOrderBook**, std::vector<piola::piolaOrderBook*, std::allocator<piola::piolaOrderBook*> > > > (__first={_M_current = 0xb277f2c4}, __last=
      {_M_current = 0xb277f2bc}, __result={_M_current = 0xb277f2c0})
    at /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../include/c++/3.4.6/bits/stl_algobase.h:317
317           return std::__copy_ni2(__first.base(), __last.base(),
(gdb) frame 5
#5  0x0839d676 in std::copy<__gnu_cxx::__normal_iterator<piola::piolaOrderBook**, std::vector<piola::piolaOrderBook*, std::allocator<piola::piolaOrderBook*> > >, __gnu_cxx::__normal_iterator<piola::piolaOrderBook**, std::vector<piola::piolaOrderBook*, std::allocator<piola::piolaOrderBook*> > > > (__first={_M_current = 0xb277f2c4}, __last={_M_current = 0xb277f2bc},
    __result={_M_current = 0xb277f2c0})
    at /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../include/c++/3.4.6/bits/stl_algobase.h:358
358            return std::__copy_ni1(__first, __last, __result, __Normal());
(gdb)

其中大部分内容对我来说都很神秘，但是在查找 memmove 时，代码似乎崩溃了，因为它无法处理 vector 中的删除(因为从 vector 中删除对于大型 vector 来说是一项非常繁重的操作)？

我说得对吗？如果是，我该如何解决这个问题(当然除了修复设计之外)？

代码在这里:

for (orderbkIterator = vOrderBook.begin(); orderbkIterator != vOrderBook.end(); orderbkIterator++)
    {

        if (  (*(*orderbkIterator)->getOrderBookId()) == *(TradableInst->getOrderBookId()) )
        {
            long long a = (*(*orderbkIterator)->getOrderBookId());
            ADDVLOG(LOG_INFO, "Removing record (%lld) from vOrderBook", a );
            vOrderBook.erase(orderbkIterator);
        } 

Answer 1

来自 std::vector::erase() :

Iterators and references to the erased elements and to the elements between them and the end of the container are invalidated. Past-the-end iterator is also invalidated.

所以 orderbkIterator 将在下一次递增时无效，如果 erase() 被调用。更改循环的结构，因为 erase() 返回被删除的迭代器之后的下一个迭代器，这意味着只有在 erase() 没有发生时才会递增:

for (orderbkIterator = vOrderBook.begin(); orderbkIterator != vOrderBook.end();)
{
    if (...)
    {
        orderbkIterator = vOrderBook.erase(orderbkIterator);
    }
    else
    {
        ++orderbkIterator;
    }
}