20

For this question my answer was apparently accepted twice:

alt text

My https://stackoverflow.com/reputation also confirms this:

 1 4669634 (15) 1 4669634 (15) 2 4669634 (10)

And so does the timeline, at 14:33:25Z and 14:33:47Z, even indicating answer accepted×2:

answer accepted×2

Is it a bug or a glitch in the matrix or a specific handling by the OP which revealed an easter egg?

Improve this question
10
  • 1
    Also stackoverflow.com/posts/4669583/timeline shows this, at 2011-01-12 14:33:25Z and 14:33:47Z with indeed also some odd "answer accepted×2". Commented Jan 12, 2011 at 15:05
  • @Arjan: interesting, there's answer accepted×2, with ×2 already taken into account by the view layer of the webpage. So it's seemingly technically possible to accept an answer more than once? Commented Jan 12, 2011 at 15:06
  • Maybe when two browser screens are open? Commented Jan 12, 2011 at 15:06
  • @Arjan: could be, but I wouldn't expect a race condition like that with this relatively big difference in the timestamps (12 seconds). However, the SO developers are as far the only who can investigate this in more detail. Commented Jan 12, 2011 at 15:09
  • 10
    That must have been a great answer! :) Commented Jan 12, 2011 at 15:14
  • 22 seconds, @Balus! :-) Commented Jan 12, 2011 at 15:16
  • /me sighs and gets himself some more coffee ;) Commented Jan 12, 2011 at 15:17
  • 2
    Hola, it's actually a duplicate, but that has been marked fixed, or no so much! Commented Jan 12, 2011 at 16:09
  • @Arjan: Nice find. Well, I won't expect that it takes at least 22 seconds for the SO backend to mark an answer accepted, allowing room for race conditions, right? There must be more at matter. Commented Jan 12, 2011 at 16:14
  • 2
    Double accepted = Double the awesome :) Commented Jan 12, 2011 at 21:47

2 Answers 2

Reset to default
2

There are about 70 double accepts in the system at the moment, I just added a daily task that ensures they are cleaned up regularly.

The same task alerts us when this happens so we can find and fix the root cause.

Improve this answer
2
  • This just happened to me on this question (which also seems to 500 at the moment): stackoverflow.com/questions/10300063/… Commented Apr 25, 2012 at 19:43
  • @Oded now that you're a dev yourself you can have a look and fix this... ;) Commented Jan 9, 2014 at 12:40
2

Due to the site architecture it's possible to sneak two or more accepts into the database for a given answer. They have to occur at the same time, though, for the race condition to allow it to occur without invalidating one or the other accepted answer.

I'm not sure if this is dealt with during any daily maintenance, though I expect that if the OP chooses a different answer as accepted, both instances would go away.

Improve this answer
8
  • 1
    Would "the same time" include 14:33:25Z and 14:33:47Z? Commented Jan 12, 2011 at 15:15
  • 1
    @Arjan - Each server behind the proxy has its own time, and those times do not match to within one second. It's why we can still (rarely) get an answer that occurs before the question is posted. The race condition can still occur at the same time and result in a 22 second time stamp difference. The proxy attempts to keep a given user on one server, but it doesn't guarantee it. Commented Jan 12, 2011 at 16:13
  • 1
    Didn't they set up a local time source so all the servers were synchronized within the latency of the local network? Commented Jan 12, 2011 at 17:46
  • @Jon - Windows 2008 doesn't do a good job of time synchronization, regardless of the source: serverfault.com/questions/1406/… so unless they are using a third party NTP tool, they are still able to wander off significantly. Commented Jan 12, 2011 at 18:10
  • But the bottom line is that these issues are rare race conditions that SOIS have deemed as acceptable. Commented Jan 12, 2011 at 18:11
  • @Pollyanna, I very often see that my own new (auto-)comments are not visible right away after refreshing a question. Sometimes this also applies to edits I make. If that indicates that I'm often being directed to different servers, and if that's a problem, then let me know if I can help investigate. Commented Jan 13, 2011 at 21:08
  • @Arjan - no, all the servers pull from the same database, so if it really submitted, then any refresh should pull the comment as well. If you are seeing this behavior often, try to document it and post it as a new question here on meta. Commented Jan 14, 2011 at 1:21
  • Okay, Pollyanna, I figured I was looking at some caching when refreshing, which might not hit the database at all? (Will keep an eye on it.) Commented Jan 14, 2011 at 9:54

You must log in to answer this question.