对于strcpy,sprintf,strcat这些不安全的函数的讨论应该算是不少了,大家都知道可以用strncpy,snprintf和strncat来替换它们。但实际工作中似乎这些“安全版”的函数带来的疑惑还是不小,总结了一下,列在这里。
1.sprintf(char * str, const char * format, ...) -> snprintf(char * str, size_t size, const char * format, ...)
vsprintf(char * str, const char * format, va_list ap) -> vsnprintf(char * str, size_t size, const char * format, va_list ap)
按照C/C++标准,snprintf和vsnprintf永远不会往str中输出多于size所指定的字节数的字符(包括结尾的'\0'),它们也保证了一定会在str中写入'\0',所以在使用这个函数后不用担心结尾的问题。
举例:
char buf[5]; snprintf(buf, 5, "This is a test string."); // buf becomes "This", buf[4] is '\0' snprintf(buf, 6, "This is a test string."); // ERROR: buffer overflow snprintf(buf, 5, "abc"); // buf becomes "abc", the value of buf[3] is '\0', buf[4] is undefined.然而,VC中的_snprintf函数并没有按照这样的规定来做,它在输出缓冲区不够大时就不会输出结尾的'\0'(跟strncpy的行为类似)。所以要让上面的程序工作正常,必须做相应的修改。
char buf[5]; _snprintf(buf, 5, "This is a test string.");// buf becomes "This ", buf[4] is ' ' buf[4] = 0; // buf[4] is '\0' now. _snprintf(buf, 6, "This is a test string."); // ERROR: buffer overflow _snprintf(buf, 5, "abc"); // buf becomes "abc", the value of buf[3] is '\0', buf[4] is undefined.如果要保证可移植性,就得按VC的写法,做一次对于标准来说很多余的“填0”的操作,再通过定义宏来区别平台分别选用snprintf或_snprintf。
2. strcat(char * dest, const char * src) -> strncat(char * dest, const char * src, size_t n);
这个函数比较简单,它保证不会写入多于n+1个字符,并且保证最后一定以'\0'结束。
举例:
char dest[5] = "abc"; strncat(dest, "defghijklmn", 5 - 3 - 1); // dest becomes "abcd", dest[4] is '\0', // always minus the buffer length by 1 as the value of n. strncat(dest, "defghijklmn", 5 - 3); // ERROR: buffer overflow 3. strcpy(char * dest, const char * src) -> strncpy(char * dest, const char * src, size_t n);
strncpy是一个比较容易出错的函数,它保证复制src中不多于n字节的内容,但是如果src的前n个字节中没有包含'\0',就会导致 dest没有正常的以'\0'终止。另外,它还保证如果src的长度小于n,则dest剩余的部分都会以'\0'填充。在用这个函数时,有一个最佳实践就 是先把dest以0填充,并传入dest的长度减1的值作为n,这样可以确保安全。
举例:
char buf[5] = {0}; // Always zero-fill the buffer, // always use this form to initialize the stack arrays to get //better performance over memset. char * buf2 = new char[5]; memset(buf2, 0, 5); // Always zero-fill the buffer strncpy(buf, "abcde", 5); // ERROR: buf is not null-terminated. strncpy(buf2, "abcde", 5 - 1); // Right, always minus the buffer length by 1 //as the value of n. buf becomes "abcd", buf[4] //is '\0' by initilization; // This is a common error in the existing code. char buf3[5]; const char * str = "Test"; strncpy(buf3, str, strlen(str)); // Wrong, buf may not null-terminated, // potential buffer-overflow strncpy(buf3, str, strlen(str)+1); // Wrong, potential buffer-overflow. //No difference with using strcpy.