哈希读/写中 hsearch_r 的内存泄漏

Question

我正在逐行读取一个小文件，并将第一列存储为键，并将后续列存储为不同哈希中的各种值。我正在使用从How to delete element from hsearch 获得的建议，使用 POSIX 哈希函数。

但是，如果没有内存错误，我无法将哈希读入内存。

我正在读取的数据是一个制表符分隔的表格：

sample  detailed_category   primary disease or tissue   _primary_site   _sample_type    _gender _study
TCGA-V4-A9EE-01 Uveal Melanoma  Uveal Melanoma  Eye Primary Tumor   Male    TCGA
TCGA-VD-AA8N-01 Uveal Melanoma  Uveal Melanoma  Eye Primary Tumor   Male    TCGA
TCGA-V4-A9EI-01 Uveal Melanoma  Uveal Melanoma  Eye Primary Tumor   Male    TCGA
TCGA-VD-AA8O-01 Uveal Melanoma  Uveal Melanoma  Eye Primary Tumor   Male    TCGA

而我的C程序如下：

#include <stdio.h>
#include <stdlib.h>
#define GNU_SOURCE
#define __USE_GNU
#include <search.h>//hcreate_r, h*_r
#include <string.h>//strok_r

#define NIL (-1L)
//https://stackoverflow.com/questions/25971505/how-to-delete-element-from-hsearch

void hadd_char(struct hsearch_data *restrict tab, char *restrict key, const char *restrict value) {
    ENTRY item = {key, (char *restrict ) value};
    ENTRY *pitem = &item;

    if (hsearch_r(item, ENTER, &pitem, tab)) {
        pitem->data = (char *restrict ) value;
    }
}

char * hfind(struct hsearch_data *restrict tab, char *restrict key) {
    ENTRY item = {key};
    ENTRY *pitem = &item;

    if (hsearch_r(item, FIND, &pitem, tab)) {
        return (char *) pitem->data;
    }
    return NULL;
}

int main(void) {

    const char PHENOTYPE_FILENAME[] = "head_TcgaTargetGTEX_phenotype.txt";
    FILE *restrict phenotype_fh = fopen(PHENOTYPE_FILENAME, "r");
    if (phenotype_fh == NULL) {
        printf("failed to open %s\n", PHENOTYPE_FILENAME);
        perror("");
        exit(EXIT_FAILURE);
    }

    char *line = NULL;//necessary for reading file
    size_t len = 0;//necessary for reading file
    ssize_t read = 0;//necessary for reading file
    _Bool header = 0;//necessary for reading file
    struct hsearch_data patient_disease = {0};//the hash
    hcreate_r(15, &patient_disease);//bigger than it needs to be
//start reading file
    while ((read = getline ( &line, &len, phenotype_fh)) != -1) {
        if (header == 0) {//skip header
            header = 1;
            continue;
        }
        char *restrict tmp_string = NULL;
        char * tmp_pointer = NULL;
        tmp_string = strtok_r(line, "\t", &tmp_pointer);
        char *restrict patient = strdup(tmp_string);
        unsigned short int column = 1;
        while (tmp_string != NULL) {//read each field in line
            tmp_string = strtok_r(NULL, "\t", &tmp_pointer);
            if (tmp_string == NULL) {
                break;
            }
            column++;
            if (column == 2) {
                printf("patient %s = %s\n", patient, tmp_string);
                hadd_char(&patient_disease, patient, tmp_string);
            }
        }
        free(patient); patient = NULL;
    }

    free(line); line = NULL;
    fclose(phenotype_fh);
//try to get a value back from the hash
    char *restrict x = hfind(&patient_disease, "TCGA-V4-A9EI-01");
    puts(x);
    free(x); x = NULL;
    hdestroy_r(&patient_disease);
    return 0;
}

但是这个段错误会从 valgrind 得到错误：

=17700== HEAP SUMMARY:
==17700==     in use at exit: 0 bytes in 0 blocks
==17700==   total heap usage: 9 allocs, 9 frees, 6,144 bytes allocated
==17700== 
==17700== All heap blocks were freed -- no leaks are possible
==17700== 
==17700== ERROR SUMMARY: 24 errors from 2 contexts (suppressed: 0 from 0)
==17700== 
==17700== 2 errors in context 1 of 2:
==17700== Invalid read of size 1
==17700==    at 0x4C33DA3: strcmp (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==17700==    by 0x4F4BDCA: hsearch_r (hsearch_r.c:171)
==17700==    by 0x108A93: hadd_char (graeme.c:15)
==17700==    by 0x108A93: main (graeme.c:65)
==17700==  Address 0x521d510 is 0 bytes inside a block of size 16 free'd
==17700==    at 0x4C30D3B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==17700==    by 0x108AB8: main (graeme.c:68)
==17700==  Block was alloc'd at
==17700==    at 0x4C2FB0F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==17700==    by 0x4ECFC99: strdup (strdup.c:42)
==17700==    by 0x108A0B: main (graeme.c:55)
==17700== 
==17700== 
==17700== 22 errors in context 2 of 2:
==17700== Invalid read of size 1
==17700==    at 0x4C33DC7: strcmp (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==17700==    by 0x4F4BDCA: hsearch_r (hsearch_r.c:171)
==17700==    by 0x108A93: hadd_char (graeme.c:15)
==17700==    by 0x108A93: main (graeme.c:65)
==17700==  Address 0x521d511 is 1 bytes inside a block of size 16 free'd
==17700==    at 0x4C30D3B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==17700==    by 0x108AB8: main (graeme.c:68)
==17700==  Block was alloc'd at
==17700==    at 0x4C2FB0F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==17700==    by 0x4ECFC99: strdup (strdup.c:42)
==17700==    by 0x108A0B: main (graeme.c:55)
==17700== 
==17700== ERROR SUMMARY: 24 errors from 2 contexts (suppressed: 0 from 0)

我真的很喜欢链接帖子中的作者添加功能的方式，但我无法复制他的成功。

最后，应该测试输出，以便每个键都匹配正确的值。

我怎样才能像这样成功地从这个哈希中读写？

Answer 1

Valgrind 指出你在第 55 行使用strdup，这里：

    char *restrict patient = strdup(tmp_string);

并在第 68 行释放该字符串：

    free(patient); patient = NULL;

但是您将它（第 65 行）作为 key 参数传递给了 hadd_char()：

void hadd_char(struct hsearch_data *restrict tab, char *restrict key, const char *restrict value) {
    ENTRY item = {key, (char *restrict ) value};
    ENTRY *pitem = &item;

    if (hsearch_r(item, ENTER, &pitem, tab)) {
        pitem->data = (char *restrict ) value;
    }
}

指针被复制到ITEM 中，hsearch_r(..., ENTER, ...) 复制该ITEM，包括指针，并将其放入哈希表中。所以当你free(patient) 时，你会在哈希表中留下一个悬空指针。

您需要安排当您将新数据插入哈希表时，只要它存在就保持它的活动状态（这在 C++ 中更容易；在 C 中，管理对象生命周期是最困难的问题之一） .