Page MenuHomePhabricator

Check for possible data corruptions caused by T212703
Closed, ResolvedPublic

Description

Any tag logs after 17 December 2018 needs to be checked. Running this query returns zero results on enwiki and commonswiki but lots of results in wikidatawiki:

select * from logging where log_type = 'tag' and log_timestamp  > '20181217' order by log_id desc limit 50;

Event Timeline

The wikis that have changed tags in the given time span are:

  • bnwiki (Possible clashed ids haven't changed in the time span)
  • dewiktionary (no possibility to clash)
  • eswiki (Possible clashed ids haven't changed in the time span)
  • fawiki (One clash case)
  • fiwiki (Possible clashed ids haven't changed in the time span)
  • hewiki (Four clash cases)
  • itwiki (Possible clashed ids haven't changed in the time span)
  • jawiki (Possible clashed ids haven't changed in the time span)
  • svwiki (Possible clashed ids haven't changed in the time span)
  • testwiki (Possible clashed ids haven't changed in the time span)
  • viwiki (Possible clashed ids haven't changed in the time span)
  • wikidatawiki (Possible clashed ids haven't changed in the time span)
  • zhwiki (Possible clashed ids haven't changed in the time span)
Addshore triaged this task as Medium priority.Jan 15 2019, 1:31 PM

Ids that might have clashed (Query: select ct1.ct_rev_id from change_tag as ct1 join change_tag as ct2 on ct1.ct_rev_id = ct2.ct_log_id;)

  • bnwiki (3378 ids)
  • dewiktionary: Empty
  • eswiki:
MariaDB [eswiki_p]> select ct1.ct_rev_id from change_tag as ct1 join change_tag as ct2 on ct1.ct_rev_id = ct2.ct_log_id;
 ----------- 
| ct_rev_id |
 ----------- 
|  12671678 |
|  12671678 |
|  12671678 |
|  12671678 |
 ----------- 
4 rows in set (6.21 sec)
  • fawiki (19998 ids, are you kidding me?)
  • fiwiki: 5 ids:
MariaDB [fiwiki_p]> select ct1.ct_rev_id from change_tag as ct1 join change_tag as ct2 on ct1.ct_rev_id = ct2.ct_log_id;
 ----------- 
| ct_rev_id |
 ----------- 
|   8315654 |
|  10754418 |
|  10784706 |
|  10792545 |
|  10792545 |
 ----------- 
5 rows in set (0.51 sec)
  • hewiki (2519 ids)
  • itwiki (261 ids: P7999)
  • testwiki: 403 ids
  • viwiki (598 ids P8000)
  • zhwiki (one id: 5355705)
  • wikidatawiki: 5870 ids

I wrote this bash script to quick things up:

a=$( sql $1 'select ct1.ct_rev_id from change_tag as ct1 join change_tag as ct2 on ct1.ct_rev_id = ct2.ct_log_id;');
b=$(sql $1 'select log_params from logging where log_type = "tag" and log_timestamp  > "20181217" order by log_id desc;');
for i in $a; do
  echo $b | grep $i
done

With that only one case from Persian Wikipedia stood out. Tags made on log 13848755. I checked and it was vandalism anyway, I reverted the adding tags.

For Hebrew it's four tag additions was found but the match was accidental.

This is done.