I was stunned by this question that I browsed 746000 times

I'd like to share with you one interesting thing I found recently , But with a little knowledge .

Let's get you a program to see ：

public class MainTest {

    public static void main(String[] args) throws Exception {
        SimpleDateFormat simpleDateFormat = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
        Date date = simpleDateFormat.parse("1900-01-01 08:00:00");
        System.out.println(simpleDateFormat.format(date));
    }
}

You said that the above program logic is a simple time format , What do you say is the output result ？

Just need a glance to know , It must output this result ：

1900-01-01 08:00:00

however , You take out the above program , Just run , You will find that the output is like this ：

1900-01-01 08:05:43

I was confused at that time .

I know jet lag 8 Hours , Because of the time zone problem .

I know the time difference 1 Hours , It's because of daylight saving time .

But it's bad here 5 branch 43 second , There are zeros and integers , It makes me a little confused .

The above case was shared to me by a reader , Their default time in the database is 1900-01-01, Plus the time zone , It just became 1900-01-01 08:00:00, So when doing data migration through the program, I stepped on this inexplicable time problem .

At the same time, he also sent me a message about this bug Link to ：

https://bugs.openjdk.java.net/browse/JDK-8266262

At first glance , This bug It's quite new , It belongs to this year's proposal .

After a closer look, I found that it was the same as before bug repeated ：

But here's why ：

He said he could look at this link  https://www.timeanddate.com/time/zone/china/shanghai?year=1900

Inside this , stay 1900 In the year , A change has taken place ：

The timezone offset was UTC +8:05:43 hours all of the period.

Although I didn't quite understand what it meant , But I saw “5 branch 43 second ”：

I understand that it is due to the change of time zone , Caused the time to reset .

Then I followed it , stay stackoverflow I found this on ：

https://stackoverflow.com/questions/6841333/why-is-subtracting-these-two-times-in-1927-giving-a-strange-result

I was shocked .

This 10 The question raised years ago has been browsed 746k Time , A very hot question , I didn't even notice ：

The specific question is like this ：

You just take a glance , I'll translate for you .

Questioner said , He found that 1927-12-31 23:54:07 To 1927-12-31 23:54:08 There's a difference 353 second , It should be 1 Seconds is right ？

But change the time to the following , It's normal again ：

String str3 = "1927-12-31 23:54:07";  
String str4 = "1927-12-31 23:54:08";

I'll stick the program out and you can run , Look at the result. It's amazing ：

public static void main(String[] args) throws ParseException {
    SimpleDateFormat sf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");  
    String str3 = "1927-12-31 23:54:07";  
    String str4 = "1927-12-31 23:54:08";  
    Date sDt3 = sf.parse(str3);  
    Date sDt4 = sf.parse(str4);  
    long ld3 = sDt3.getTime() /1000;  
    long ld4 = sDt4.getTime() /1000;
    System.out.println(ld4-ld3);
}

But when I ran again , I find ： I went to , Well said 353 Seconds? ？

Running out is 1 Seconds? , Nothing wrong ：

I even suspect it's jdk Version problem , So I changed jdk 9,11,15 All ran , All are 1 second .

That's strange .

I feel that this question has a problem .

But after I read the best answer below , I just seem to see a clue .

The answer is long , I'll show you all the screenshots first ：

The reason is that the author has revised his answer several times .

Why did you change the answer ？

And look back , All the answers are hidden here .

I'll choose the key to tell you .

First look at the first paragraph ：

He said （1927 year ） 12 month 31 When day , The time zone in Shanghai has changed .

And about the 1927 Details of Shanghai in , He attached a hyperlink , This hyperlink is the website that appears before , After clicking in, it's like this ：

But this one shows ：

No further time changes in 1927 in Shanghai

Which translates as ：1927 The time of Shanghai did not change further .

This is not the same as what he said below ？

He said below , stay 1927 At midnight at the end of the year , The clock goes back 5 branch 52 second . therefore ,"1927-12-31 23:54:08" It actually happened twice , and Java Take the second time , So there are differences .

In fact, I was stunned to see here , This thing doesn't match , So I went on to search .

Until I found this ：

https://coolshell.cn/articles/5075.html/comment-page-2#comments

This is also an article ten years ago .

Here, the author cut a picture of the website at that time ：

The screenshot of that year shows ：

stay 1927 year 12 month 31 Japan 23:59:59 when , The next second should be 1928 year 1 month 1 Japan 0:0:0, But this time has been adjusted back. 5 branch 52 second , And that's it ,1927 year 12 month 31 Japanese ,23:54:08, therefore , It's done 352 Second crossing .

What does that mean ？

It indicates that the data has been tampered with , Someone tampered with the information on the web page ！

What's going on ？

We go back to stackoverflow So let's look down ：

This is the first time he has revised his answer , because History changes...

History has changed ...

He said here , If you use TZDB Of 2013a Version of the data , The original problem will no longer show exactly the same behavior .

stay 2013a in , The result will be 358 second , The transition time is 23:54:03, instead of 23:54:08.

He mentioned a TZDB, What is this ？

I don't know either , But I searched .

What he should say is this .

https://www.iana.org/time-zones

Look at the name, you know , It is a time zone database , There should be maintained time zone related data .

in other words , In this time zone database , use 2013a Version of the data , The previous code is another kind of output .

In other words, the data has indeed changed .

The key answer is the next edit ：

History has changed again...

History has changed again .

In this time zone database ,2014f In the version , The time for change has moved to 1900-12-31, Now it's just a 343 Second change .

343 second ？

No, it's the one in front of us 5 branch 43 Seconds ？

Okay , Now the jet lag can match ,343 second , But the time is still not right .

Our test time 1900-01-01 08:00:00, The time he wrote here is 1900-12-31.

It's a whole year away ？

good , Let's see what he edited last time ：

I personally understand that what he wants to express is like this .

Java To unify standards in time zones , So a one size fits all policy was adopted .

The unified standard is to let UTC In time zone 1900 Any moment before is standard time .

As for the time difference ...

Make it up at the beginning .

therefore ,1900-01-01 00:00:00 add 8 Time difference in hours , yes 1900-01-01 08:00:00, Add in advance on this basis 27 Years from 1927-12-31 That midnight was brought by time callback 343 second .

1900-01-01 08:05:43, I personally think that's how it came about .

And the front stackoverflow The corresponding program inside , We are now executing output 1, But in 10 Years ago , The output is really 353 .

Just like I changed the program to this ：

The final output is not 1, It is -342.

Time , It happened. “ Backflow ”.

Okay , It's another useless knowledge point .

Last , Add two more cold knowledge .

The first is that I'm jdk bug The list goes back , The earliest time to ask relevant questions is 2005 year ：

https://bugs.openjdk.java.net/browse/JDK-6281408

In this , The official reply is ：

This problem will not be fixed , To avoid any compatibility issues .

It means ： I know the question , But it's not easy to do ,bug First become feature Well , Let's do it first .

Don't ask , There are historical reasons to ask .

The second cold knowledge is , aforementioned , Time zone in 1927 Year changed .

Do you know why ？

I found this description on a website ：

https://zh.wikipedia.org/wiki/%E4%B8%AD%E5%9C%8B%E6%99%82%E5%8D%80

1928 year , The Republic of China 17 year .

That year , The earliest Red Army in the history of the Chinese Communist army , The Fourth Army of the Chinese workers' and peasants' Red Army was established .

That year , Chairman Mao established a rural revolutionary base in Jinggangshan .

That year , sparks of fire , It has become a prairie fire .