**已关闭。**此问题需要debugging details。它目前不接受回答。

编辑问题以包括desired behavior, a specific problem or error, and the shortest code necessary to reproduce the problem。这将帮助其他人回答这个问题。
8天前关闭
Improve this question
我有一个函数，它在C中维护一个动态大小的数组。该程序读取一个文本文件，并创建一个列表，每个独特的词，以及这些词的频率。它有一个搜索功能，看起来像这样：

int vec_lsearch(vector* vec, char* target)
{
    for(int i = 0; i < vec->size; i++) {
        if(strcmp(vec->data[i].word, target) == 0) {
            return i;
        }
    }
    return -1;
}

当我的程序运行时，它有大约60%的时间发生segfaults。然而，它在Valgrind下运行时 * 从不 * 崩溃。使用GDB，我可以找到问题。它也只在我运行两个或更多线程时崩溃。以下是我的GDB日志：

Thread 3 "Evans_Griffin_H" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff6af1700 (LWP 13341)]
__strcmp_ssse3 () at ../sysdeps/x86_64/multiarch/../strcmp.S:173
173    ../sysdeps/x86_64/multiarch/../strcmp.S: No such file or directory.
(gdb) bt
#0  __strcmp_ssse3 () at ../sysdeps/x86_64/multiarch/../strcmp.S:173
#1  0x0000555555555283 in vec_find (vec=0x555555758490, target=0x7ffff72f381b "incomprehensible")
    at Evans_Griffin_HW4_main.c:153
#2  0x00005555555553f1 in count_words (arg=0x555555758500) at Evans_Griffin_HW4_main.c:193
#3  0x00007ffff7bbb6db in start_thread (arg=0x7ffff6af1700) at pthread_create.c:463
#4  0x00007ffff78e461f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
(gdb) select-frame 1
(gdb) p vec->data[i]
$3 = {word = 0x7ffff0002000 "played", freq = 1}
(gdb) p target
$4 = 0x7ffff72f381b "incomprehensible"
(gdb) p vec->data[i].word
$5 = 0x7ffff0002000 "played"
(gdb) p vec->data[i].word[6]
$7 = 0 '\000'
(gdb) p target[16]
$8 = 0 '\000'

据我所知，这两个字符串似乎都有效。我在插入和更新数据时使用互斥锁，但在使用strcmp时不使用（尽管我在这里尝试锁定线程，但没有区别）。有人能帮我弄明白为什么我的程序会崩溃吗？
编辑：由于学术诚信政策，我不能随意提供我的程序的完整源代码。我已经包括了一些更多的片段，希望是有帮助的。

typedef struct {
    char* word;
    int freq;
} word_freq;

typedef struct {
    word_freq* data;
    size_t size;
    size_t capacity;
} vector;

void vec_push(vector* vec, char* e)
{
    word_freq wf;
    wf.freq = 1;
    wf.word = malloc(strlen(e) + 1);
    strcpy(wf.word, e);
    vec->data[vec->size++] = wf;

    // if vector is at capacity, realloc to double its capacity
    if (vec->size == vec->capacity) {
        vec->capacity *= 2;
        vec->data = realloc(vec->data, vec->capacity * sizeof(word_freq));
    }

    if (vec->data[vec->size - 1].word == NULL) {
        printf("Error inserting %s\n", e);
    }
}

void* count_words(void* arg)
{
    job_info* info = arg;
    vector* word_vec = info->words; // vector of word_freq structs
    char* buf = info->buffer;
    int thread_num = info->thread_num;

    char* saveptr;
    char* token = strtok_r(buf, delim, &saveptr);
    while(token != NULL) {
        trim_word(token);

        if(strlen(token) >= MIN_STRING_LENGTH && strlen(token) <= MAX_STRING_LENGTH) {
            // because the size of the vector may change, must lock the thread before
            // accessing it
            if (pthread_mutex_lock(&info->mutex)) {
                perror("mutex lock");
            }
            int index = vec_lsearch(word_vec, token);
            if(index != -1) {

                word_vec->data[index].freq++;   

            } else {
                vec_push(word_vec, token);
            }
            
            vec_health(word_vec);

            if (pthread_mutex_unlock(&info->mutex)) {
                perror("mutex unlock");
            }
        }

        token = strtok_r(NULL, delim, &saveptr);
    }
}

// given input string, convert it to lower case and discard any non-alphabet (a-z and A-Z) characters
void trim_word(char* in)
{
    int i = 0;
    int j = 0;
    char c;

    while (( c = in[i++]) != '\0') {
        if (isalpha(c)) {
            in[j++] = tolower(c);
        }
    }

    in[j] = '\0';
}

int vec_health(vector* vec) 
{
    if (vec->data == NULL) {
        return 1;
    }

    for(int i = 0; i < vec->size; i++) {

        if (vec->data[i].word == NULL) {
            fprintf(stderr, "data[%d] is null!\n", i);
            return 2;
        }

    }

    return 0;
}

我写了一个新的测试函数vec_health，它显示了我的结构体的某些成员是NULL，尽管我仍然不确定这是怎么回事。感谢@solomon-slow的启发性评论。
vec_health将在运行时警告我某些对象具有空值，例如

data[3583] is null!
data[6143] is null!

但是我在广发行检查的时候，好像没什么问题

(gdb) p vec->data[3583]
$2 = {word = 0x7fffe800dfc0 "bucket", freq = 1}
(gdb) p vec->data[6143]
$3 = {word = 0x7fffe80184a0 "defending", freq = 7}