指针数组只有在打印时才正确

Array of pointer correct only if printed?

本文关键字:打印 数组 指针      更新时间:2023-10-16

函数用于将文本拆分为不同的(可能是平衡的)块(我在这里已经详细讨论过了)。下面是代码(查看下面的问题描述):

pair<off_t,off_t>* getSplits() {
    struct stat st;
    off_t size;
    if (stat(file_name.c_str(), &st) == 0)
        size = st.st_size;
    int nMappers = size > nWorkers ? nWorkers : size; //if workers are greater than file size
    pair<off_t,off_t> splits [nMappers];
    double split_size = (double) size / nMappers;
    off_t acc = 0 ;
    ff::ParallelFor pf( ff_realNumCores() );
    string prova = file_name;
    pf.parallel_for(0,nMappers,[&splits,split_size, prova, size](const long i) {
        ifstream ifs (prova , ifstream::in);
        off_t begin = ceil((double) i*split_size);
        off_t end = ceil((double) (i+1)*split_size-1);
        char c;
        string s;
        if(begin>0){
            //if char before the first one is different from ' ' or 'n'
            //then the split begins in the middle of a word (bad)
            ifs.seekg(begin-1,ios::beg);
            ifs.get(c);
            if(c!=' ' && c!='n'){
                getline(ifs,s,' ');
                begin+=s.length();
            }
            if(begin>end)
                end=begin;
        }
        ifs.seekg(end,ios::beg);
        ifs.get(c);
        if(c!=' ' && c!='n' && end != size){
            getline(ifs,s,' ');
            end+=s.length();
        }
        splits[i] = {begin, end};
    });
    pair<off_t,off_t> *p = splits;
    for(int i=0;i<nWorkers;i++){
        cout<<"begin="<<p[i].first<<" end="<<splits[i].second<<endl;
    }
    return p;
}

这就是我称之为并打印其内容的方式

pair<off_t,off_t> *splits = input_format->getSplits();
for(int i=0; i<nWorkers; i++){
    cout<<"outside split begin="<<splits[i].first<<" second="<<splits[i].second<<endl;
    this->ff_send_out(new MapTask<MIK,MIV,MOK,MOV> (record_reader->clone(),splits[i],map_func));
}

问题是,如果我在getSplits()(最后一个for周期)内打印p的内容,那么结果是正确的:

begin=0 end=13
begin=14 end=14
begin=15 end=21
begin=22 end=28
outside split begin=0 second=13
outside split begin=14 second=14
outside split begin=15 second=21
outside split begin=22 second=28

但是如果我没有(所以我删除了打印for),那么结果是错误的(只有第一对是正确的)!

outside split begin=0 second=13
outside split begin=140152066182136 second=140152054622976
outside split begin=140152066227112 second=29521758
outside split begin=140152054622960 second=2564825869

这怎么可能?

您返回一个局部变量的地址,因此您有一个悬空指针。

我建议退回std::vector<std::pair<off_t, off_t>

您将获得一个指向堆栈的指针(函数定义的拆分位置)。

使用全局/静态变量,或者更好的方法是传递指针数组并将其填充到函数中。