Page MenuHomePhabricator

Massive data corruption on es.wiktionary
Closed, ResolvedPublic

Description

Author: mgrojo

Description:
Any prev link on page
http://es.wiktionary.org/w/wiki.phtml?title=Este/Traducci%C3%B3n&action=history
reports the follwing error embedded on the page

Warning [host=]: gzinflate(): data error in
/usr/local/apache/common-local/php-new/includes/Article.php on line 61.

The differences are not shown. the only link working is the prev of the current
version.


Version: unspecified
Severity: critical
URL: http://es.wiktionary.org/w/wiki.phtml?title=ouija&diff=3971&oldid=3825

Details

Reference
bz950

Event Timeline

bzimport raised the priority of this task from to Unbreak Now!.Nov 21 2014, 7:03 PM
bzimport set Reference to bz950.
bzimport added a subscriber: Unknown Object (MLST).

zigger wrote:

Changed URL to one with a smaller character set and no sub-page:
http://es.wiktionary.org/w/wiki.phtml?title=ouija&diff=3971&oldid=3825

en.wiktionary.org & fr.wiktionary.org seem ok.

Many compressed old text entries are corrupt on es.wiktionary.org; it seems to be a side effect of a
botched latin-1 to UTF-8 conversion. Am trying to see if the corruption is reversible or if we have
backups...

The corruption is not reversible, as four byte values are not valid in the CP1252->UTF-8 conversion and
all are changed to the same output value. Will need to restore affected entries from backup.

I've retrieved what seems to be the pre-conversion database dump, and should be able to recover all
damaged data from it. In the meantime I've locked the wiki against further editing; should be back up
within a few hours.

I've recovered the affected revisions from backup, it looks ok.

Please let us know immediately if any further problems like this turn up!