JDK代码的一个小bug

今天看了一点JDK的代码，不巧居然看到了 JDK 的一个小 bug, 呵呵。且看 javax.sql.rowset.serial 包下的 SerialBlob 类，其中有段代码如下

    /** 	
     * Returns the position in this <code>SerialBlob</code> object where	
     * the given pattern of bytes begins, starting the search at the	
     * specified position.	
     *	
     * @param pattern the pattern of bytes for which to search 	
     * @param start the position of the byte in this	
     *              <code>SerialBlob</code> object from which to begin 	
     *              the search; the first position is <code>1</code>;	
     *              must not be less than <code>1</code> nor greater than	
     *              the length of this <code>SerialBlob</code> object	
     * @return the position in this <code>SerialBlob</code> object	
     *         where the given pattern begins, starting at the specified	
     *         position; <code>-1</code> if the pattern is not found
     *         or the given starting position is out of bounds; position
     *         numbering for the return value starts at <code>1</code>	
     * @throws SerialException if an error occurs when serializing the blob	
     * @throws SQLException if there is an error accessing the <code>BLOB</code>
     *         value from the database     
     */
    public long position(byte[] pattern, long start) 
                throws SerialException, SQLException {                
        if (start < 1 || start > len) {
            return -1;
        } 

        int pos = (int)start-1; // internally Blobs are stored as arrays. 
        int i = 0;        
        long patlen = pattern.length;                

        while (pos < len) {     
            if (pattern[i] == buf[pos]) {                
                if (i + 1 == patlen) {
                    return (pos + 1) - (patlen - 1);
                }
                i++; pos++; // increment pos, and i
            } else if (pattern[i] != buf[pos]) {
                pos++; // increment pos only
            }                                    
        }        
        return -1; // not found
    }

这个方法用于查找 blob 中是否存在 pattern 的字符数组，细看之下发现一个小 bug, 且看这段代码

        int pos = (int)start-1; // internally Blobs are stored as arrays. 
        int i = 0;        
        long patlen = pattern.length;                

        while (pos < len) {     
            if (pattern[i] == buf[pos]) {                
                if (i + 1 == patlen) {
                    return (pos + 1) - (patlen - 1);
                }
                i++; pos++; // increment pos, and i
            } else if (pattern[i] != buf[pos]) {
                pos++; // increment pos only
            }                                    
        } 

当发现字符不匹配的情况下，应该将 i 置零，否则匹配结果是不正确的。当改为

        int pos = (int)start-1; // internally Blobs are stored as arrays. 
        int i = 0;        
        long patlen = pattern.length;                

        while (pos < len) {     
            if (pattern[i] == buf[pos]) {                
                if (i + 1 == patlen) {
                    return (pos + 1) - (patlen - 1);
                }
                i++; pos++; // increment pos, and i
            } else if (pattern[i] != buf[pos]) {  // 其实这个判断也没有必要
                pos++;
                i = 0; // 在此处将 i 置 0
            }                                    
        }