在 C 中连接字符串的未定义行为

Question

我正在编写一个 C 程序，它将连接所有行（包括“\n”），同时将指针保存到最终字符串的最后一个字符。但是，我没有得到我预期的结果。可能是什么问题？

#include <stdio.h>
#include <stdlib.h>
#include <string.h>

struct Node {
  struct Node *next;
  struct Node *prev;
};

struct Node *CreateNewNode() {
  struct Node *newNode = malloc(sizeof(struct Node));
  return newNode;
}

struct PieceTable {
  char *buffer;
  char *ptr_to_last_character;
} PT;

void strconcatenate(char *source) {
  size_t source_len = strlen(source);
  size_t buffer_len = strlen(PT.buffer);
  PT.buffer = realloc(PT.buffer, buffer_len + source_len + 1);
  while (*source)
    *PT.ptr_to_last_character++ = *source++;
  *PT.ptr_to_last_character = '\0';
}

int main(int argc, char *argv[]) {

  char input_line[1024];

  PT.buffer = malloc(sizeof(char) * 2);
  *PT.buffer = '\0';
  PT.ptr_to_last_character = PT.buffer;

  struct Node *new_node = CreateNewNode();
  new_node->next = NULL;
  new_node->prev = NULL;

  strconcatenate("Lorem ipsum\n");
  strconcatenate("dolor sit amet\n");
  strconcatenate("consectetur adipiscing elit\n");

  printf("%s", PT.buffer);

  return 0;
}

预期输出：

Lorem ipsum
dolor sit amet
consectetur adipiscing elit

输出：

etur adipiscing elit

Answer 1

对于初学者来说，不清楚为什么在这个语句中分配了两个字节而不是一个字节。

PT.buffer = malloc(sizeof(char) * 2);
*PT.buffer = '\0';

函数内this语句

size_t buffer_len = strlen(PT.buffer);

考虑到您已经有一个指向终止零的指针，这没有意义。所以它只是效率低下。

拨打realloc之后

PT.buffer = realloc(PT.buffer, buffer_len + source_len + 1);

指针PT.ptr_to_last_character 可能无效。

您还应该检查内存分配是否成功。否则指针PT.buffer可以设置为NULL以防万一失败。

该功能可以通过以下方式实现，如下面的演示程序所示。

#include <stdio.h>
#include <stdlib.h>
#include <string.h>

struct PieceTable {
  char *buffer;
  char *ptr_to_last_character;
} PT;

int strconcatenate( const char *source ) 
{
    int success = 1;
    
    size_t n = PT.ptr_to_last_character - PT.buffer;
    
    size_t source_len = strlen( source );
    
    if ( source_len != 0 )
    {
        char *tmp = realloc( PT.buffer, n + source_len + 1 );
        
        if ( ( success = tmp != NULL ) )
        {
            PT.buffer = tmp;
            PT.ptr_to_last_character = PT.buffer + n;

            while ( ( *PT.ptr_to_last_character = *source++ ) ) ++PT.ptr_to_last_character;
        }
    }
    
    return success;
}   

int main(void) 
{
    PT.buffer = malloc( sizeof(char) );
    *PT.buffer = '\0';
    PT.ptr_to_last_character = PT.buffer;
    
    strconcatenate( "Lorem ipsum\n" );
    strconcatenate( "dolor sit amet\n" );
    strconcatenate( "consectetur adipiscing elit\n" );

    puts( PT.buffer );
    
    return 0;
}

程序输出是

Lorem ipsum
dolor sit amet
consectetur adipiscing elit

还要考虑到不是这个手动编写的循环

while ( ( *PT.ptr_to_last_character = *source++ ) ) ++PT.ptr_to_last_character;

您可以使用标准的 C 字符串函数 strcat 或 strcpy，这样效率会更高。

在这种情况下，您的功能可以如下所示

int strconcatenate( const char *source ) 
{
    int success = 1;
    
    size_t n = PT.ptr_to_last_character - PT.buffer;
    
    size_t source_len = strlen( source );
    
    if ( source_len != 0 )
    {
        char *tmp = realloc( PT.buffer, n + source_len + 1 );
        
        if ( ( success = tmp != NULL ) )
        {
            PT.buffer = tmp;
            PT.ptr_to_last_character = PT.buffer + n;

            strcpy( PT.ptr_to_last_character, source );
            
            PT.ptr_to_last_character += source_len;
        }
    }
    
    return success;
}   

Answer 2

警告： Paul 已经发现了问题，但是既然你说你需要一个指针......

您可以通过保存/恢复偏移量来调整“最后一个”指针：

ptrdiff_t offset = PT.ptr_to_last_character - PT.buffer;
PT.buffer = realloc(PT.buffer, buffer_len + source_len + 1);
PT.ptr_to_last_character = PT.buffer + offset;

---

我在下面的完整代码中注释了其他一些清理问题。 #if 0 包装旧/原始代码。 #if 1 包装新的/重构的代码：

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stddef.h>

struct Node {
    struct Node *next;
    struct Node *prev;
};

struct Node *
CreateNewNode()
{
    struct Node *newNode = malloc(sizeof(struct Node));

    // NOTE/WARN: this was moved from main to be more general
#if 1
    newNode->next = NULL;
    newNode->prev = NULL;
#endif

    return newNode;
}

struct PieceTable {
    char *buffer;
    char *ptr_to_last_character;
} PT;

void
strconcatenate(char *source)
{
    size_t source_len = strlen(source);
    size_t buffer_len = strlen(PT.buffer);

    // NOTE/BUG: after the realloc, PT.buffer can change but
    // PT.ptr_to_last_character is pointing within the _old_ buffer
    // we need to recalc using an offset
#if 0
    PT.buffer = realloc(PT.buffer, buffer_len + source_len + 1);
#else
    ptrdiff_t offset = PT.ptr_to_last_character - PT.buffer;
    PT.buffer = realloc(PT.buffer, buffer_len + source_len + 1);
    PT.ptr_to_last_character = PT.buffer + offset;
#endif

    while (*source)
        *PT.ptr_to_last_character++ = *source++;

    *PT.ptr_to_last_character = '\0';
}

int
main(int argc, char *argv[])
{

    char input_line[1024];

    // NOTE/BUG: sizeof(char) is _always_ 1 by definition
#if 0
    PT.buffer = malloc(sizeof(char) * 2);
#else
    PT.buffer = malloc(2);
#endif
    *PT.buffer = '\0';
    PT.ptr_to_last_character = PT.buffer;

    // NOTE/WARN: this isn't used anywhere
    struct Node *new_node = CreateNewNode();

    // NOTE/WARN: this would be better inside CreateNewNode
#if 0
    new_node->next = NULL;
    new_node->prev = NULL;
#endif

    strconcatenate("Lorem ipsum\n");
    strconcatenate("dolor sit amet\n");
    strconcatenate("consectetur adipiscing elit\n");

    printf("%s", PT.buffer);

    return 0;
}

Answer 3

你重新分配了PT.buffer，但你没有更新PT.ptr_to_last_character。当realloc 不能只是扩展当前分配而是在不同地址返回一个更大的新区域时，这会导致未定义的行为，从而使ptr_to_last_character 指向旧内存。

可能更好的是存储缓冲区的长度而不是指向最后一个字符的指针，这样您就不必担心它会失效。