-1

我有一个程序,它遍历一个集合并替换一个元素,调用自身直到它无法继续,然后撤消它所做的并搜索下一个分支。

for(set<int>::iterator it=set1.begin();it!=set1.end();)
    {
        if(condition)
        {
            int l=*it;
            if(condition) set1.insert(l-rails[inuse]).first;
            set<int>::iterator it1=it;
            it++;
            set1.erase(it1);//this line has the problem
            //do other things, including a recursive call

            if(l>rails[inuse+1]+rails[inuse]) set1.erase(l-rails[inuse]);
            set1.insert(l);
        }
        else ++it;
    }

我的程序似乎运行良好,并且在我的系统上正常运行,但它在另一个系统上因分段错误而崩溃。valgrind检测到分段违规:

==3610== Process terminating with default action of signal 11 (SIGSEGV)
==3610==  Access not within mapped region at address 0x18
==3610==    at 0x4EA8039: std::_Rb_tree_rebalance_for_erase(std::_Rb_tree_node_base*, std::_Rb_tree_node_base&) (in /usr/lib/libstdc++.so.6.0.17)
==3610==    by 0x402018: std::_Rb_tree<int, int, std::_Identity<int>, std::less<int>, std::allocator<int> >::_M_erase_aux(std::_Rb_tree_const_iterator<int>) (stl_tree.h:1497)
==3610==    by 0x401A4A: std::_Rb_tree<int, int, std::_Identity<int>, std::less<int>, std::allocator<int> >::erase(std::_Rb_tree_const_iterator<int>) (stl_tree.h:787)
==3610==    by 0x401586: std::set<int, std::less<int>, std::allocator<int> >::erase(std::_Rb_tree_const_iterator<int>) (stl_set.h:517)
==3610==    by 0x40104A: test() (fence8.cpp:32)
==3610==    by 0x40105E: test() (fence8.cpp:34)
==3610==    by 0x40105E: test() (fence8.cpp:34)
==3610==    by 0x40105E: test() (fence8.cpp:34)
==3610==    by 0x40105E: test() (fence8.cpp:34)
==3610==    by 0x40105E: test() (fence8.cpp:34)
==3610==    by 0x40105E: test() (fence8.cpp:34)
==3610==    by 0x40105E: test() (fence8.cpp:34)

但我不明白是什么原因造成的。我认为这与我如何使用迭代器有关,但我找不到它。可能出了什么问题?

4

1 回答 1

1

请注意在迭代时更改集合 - 如果其中一个递归调用删除了迭代器it引用的元素,那么它将不再有效。考虑到每个递归调用从一开始就迭代整个集合,这是一种可能性。

于 2013-03-16T00:30:37.193 回答