Question

This function runs for a bit, then proc_index variable goes to -1886854513. Is there something wrong with the code?

int parse_words(vector< vector<string> > &temp_word_vec, int num_of_sub_lists)
{
    char word[MAX_WORD_LENGTH+1]; // +1 makes room for a newline character
    int proc_index = 0; //index of word arr for child process "proc_index"
    string word_str;

    cerr << "point1\n";
    while(fscanf (stdin, "%s", &word) != EOF)
    {
        cerr << "point2\n";
        for(int i = 0; i < MAX_WORD_LENGTH; i++)
        {
            word[i] = tolower(word[i]);
            if(word[i] == '\0')
            {
                word_str.push_back('\n');
                word_str.push_back('\0');
                break;
            }
            if(isalpha(word[i]))
            {
               word_str.push_back(word[i]);
            }
        }
        cerr << "point3, proc_index = " << proc_index << ", word is " << word_str << "\n";

        temp_word_vec[proc_index].push_back(word_str);

        ++proc_index;

        if(proc_index == num_of_sub_lists)
            proc_index = 0;
        word_str.clear();
    }

    return 0;
}

Answer 1

It’s almost certainly due to corruption, most likely caused by you reading more bytes into word than you’ve allocated for it.

Easy way to detect, change:

cerr << "point2\n";

to:

cerr << "point2 maxword = " << MAX_WORD_LENGTH <<
    ", strlen = " << strlen (word) << '\n';

As an aside, you never want to do an unbounded *scanf("%s") on data you don’t totally control. Use a bound (like “%20s”) or, better yet since you’re only after character data, use fgets which can prevent buffer overflows.

Or, even better than that, use C++ strings with getline, rather than some weird C/C++ hybrid 🙂