46

I just found this answer on the "New answers to old questions" list, and looking at the comment I was shocked to see it was coming... from the past!

enter image description here

Is this a bug?

EDIT: Summarizing what's been discussed in the comments below, and looking at the post's revisions it seems the answer was really posted on 2013-11-15, it got the comment and also was deleted the same day.

enter image description here

So the bug seems to be in the answer, not in the comment. Looks like something happened to it to make it appear as if it was answered today (2013-12-13 03:01:30Z). Whatever happened did not reflect in the post revisions page.

Also, as Wooble said in his comment, the original delete votes seemed to be still active on the post.

EDIT 2: In addition to a couple more cases as Flexo♦ pointed out in his comments, there is this post, which was flagged as spam and deleted 3 hours ago (2013-12-17 13:55:59Z), and got an approved suggested edit a couple of hours later (2013-12-17 15:31:07Z). It seems this made the post ressurect.

enter image description here

  • 42
    Time travel is a 10k privilege – Richard Tingle Dec 13 '13 at 10:24
  • The really strange thing is that stewbob hasn't even been on stack overflow since that answer was posted. And nor has alexf1980 who posted the "answer" – Richard Tingle Dec 13 '13 at 10:32
  • 2
    at Tavern, they once discussed time traveling "I could handle flags before they were flagged... I'd be banned for failed audits I didn't even see yet... suspended for voting irregularities committed next week..." etc. Someone apparently really managed to get to the future and comment before the post was submitted – gnat Dec 13 '13 at 10:39
  • 2
    Curiously, looking at it's revisions, it was posted and deleted on Nov 15. No revision about undelete. – hims056 Dec 13 '13 at 11:22
  • 1
    According to stewbob they really did post the comment a long time ago, so it must be the answer timestamp that is wrong – Richard Tingle Dec 13 '13 at 12:27
  • 3
    Presumably it was undeleted at the time it's shown as being posted, although the revisions not showing the undelete is another bug. – Wooble Dec 13 '13 at 12:36
  • Now the answer is gone again.. Which makes me wonder.. – Johannes Kuhn Dec 13 '13 at 12:53
  • The bug seems to be on the answer's timestamp indeed. What @Wooble said below makes sense, not sure why would anyone want to undelete that answer though. – Xavi López Dec 13 '13 at 12:59
  • I was actually able to delete it with a single new delete vote, because the other 2 were still on the question. – Wooble Dec 13 '13 at 13:06
  • 4
    So much for my request yesterday to burninate the [future] tag... – Johnny Bones Dec 13 '13 at 14:56
  • 1
    Happened again: stackoverflow.com/posts/20187420/revisions - based around a failed audit again too – Flexo Dec 15 '13 at 18:20
  • 1
    And again: stackoverflow.com/posts/20468038/revisions - there was a failed audit in the mix again too. – Flexo Dec 17 '13 at 9:53
  • 1
    @Flexo This could be another one: stackoverflow.com/posts/20633838/revisions. This answer was flagged as spam (flag deemed helpful), and now it has resurrected with a suggested edit. Someone somehow was able to suggest an edit to a deleted post (not only deleted but also flagged as spam). – Xavi López Dec 17 '13 at 17:31
  • 1
    @n00b pretty sure it isn't - this is out by weeks not months and there are multiple separate delete events – Flexo Dec 17 '13 at 20:01
  • 1
    @XaviLópez yet again that's had a late answers review audit failed on the post – Flexo Dec 17 '13 at 20:02
17
+50

Shog brought this question's revisions (10K only) to our attention today and I immediately said "LINQ2SQL".

Turns out that /review wasn't loading audit posts detached from the current DataContext, so any changes to its properties (audit posts have most visible properties altered) are tracked and an errant SubmitChanges could persist invalid data.

In these specific cases, that errant SubmitChanges occurred when a user failed an audit and had a UserHistory record recorded of the failure. It only recently started happening (i.e. 5 days ago) when completing a review, because a change was made to validate the audit task server-side - this validation loaded the audit post and overwrote its properties.

I'm pushing the fix out right now and because a post's creation date is backed by separate history records, we can fix the fudged data.

On my "one day todo.txt", I have "burninate SubmitChanges calls".

| improve this answer | |
  • 2
    CreationDate data has been fixed for the 23 affected posts. – Jarrod Dixon Dec 18 '13 at 4:15
  • 1
    Excellent! Also thanks for the details, this is one of the many reasons I love SO :-) I'm more a Java guy so don't know about LINQ2SQL but I could definitely see that happening with JPA/Hibernate too. – Xavi López Dec 18 '13 at 8:12
  • Any reason why this isn't tagged status-completed? – Shadow Wizard Wearing Mask Dec 2 '14 at 14:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .