Last modified: 2012-04-20 20:03:02 UTC
The first revision is missing and I can't delete this revision. Moving the whole file also fails. If it would not be in use, I would try to temporarily delete the whole file and restore only the non-corrupt version. But Commons Delinker is very fast :P http://commons.wikimedia.org/wiki/File:Нева.jpg The person requesting the move wanted one of the following new names: * File:Neva River and Palace Embankment, 2009-07-24.jpg * File:Вид на Неву и Дворцовую наб. с Троицкого моста, 2009-07-24.jpg
Another example: http://commons.wikimedia.org/wiki/File:Viestura_ordenis.jpg (uploads 2009-02-14)
See also full list of files with empty oi_archive_name http://toolserver.org/~kalan/commons-broken.txt (tab-separated, 1864 row) (select was "select * from oldimage where oi_archive_name=''")
next file that exists but is not shown http://commons.wikimedia.org/wiki/File:Khajuraho_India_-_Lakshman_Temple_-_Sculpture_15.JPG http://upload.wikimedia.org/wikipedia/commons/archive/f/f4/20120324100019!Khajuraho_India_-_Lakshman_Temple_-_Sculpture_15.JPG And here it turns out a real problem of the timestamp based storage: https://commons.wikimedia.org/w/api.php?action=query&prop=imageinfo|info&iiprop=size|timestamp|url&iiurlwidth=365&titles=File:%D0%9D%D0%B5%D0%B2%D0%B0.jpg&iilimit=10&format=jsonfm { "timestamp": "2011-09-23T21:32:56Z", "size": 3352071, "width": 3072, "height": 2304, "thumburl": "https:\/\/upload.wikimedia.org\/wikipedia\/commons\/thumb\/archive\/a\/a5\/20110923213256%21%D0%9D%D0%B5%D0%B2%D0%B0.jpg\/365px-%D0%9D%D0%B5%D0%B2%D0%B0.jpg", "thumbwidth": 365, "thumbheight": 274, "url": "https:\/\/upload.wikimedia.org\/wikipedia\/commons\/archive\/a\/a5\/20110923213256%21%D0%9D%D0%B5%D0%B2%D0%B0.jpg", "descriptionurl": "https:\/\/commons.wikimedia.org\/wiki\/File:%D0%9D%D0%B5%D0%B2%D0%B0.jpg" }, { "timestamp": "2011-09-23T21:32:56Z", "size": 3310009, "width": 3072, "height": 2304, "thumburl": "https:\/\/upload.wikimedia.org\/wikipedia\/commons\/thumb\/archive\/a\/a5\/\/365px-%D0%9D%D0%B5%D0%B2%D0%B0.jpg", "thumbwidth": 365, "thumbheight": 274, "url": "https:\/\/upload.wikimedia.org\/wikipedia\/commons\/archive\/a\/a5\/", "descriptionurl": "https:\/\/commons.wikimedia.org\/wiki\/File:%D0%9D%D0%B5%D0%B2%D0%B0.jpg" } Yes, the same timestamp -> result: comment #1
*comment #0 *
I ran a cleanup script on all orphaned commons files and didn't fix this particular file row. I'll need to see how many empty oi_archive_name rows are left and consider just deleting those rows.
Directly deleting the rows? At least you should keep a backup before...
(In reply to comment #5) > and consider just deleting those rows. This will lead to inconsistency with OploadLog, is not?
*UploadLog
Note what are two difference cases: Example #1 - http://commons.wikimedia.org/wiki/File:Нева.jpg - empty 'oi_archive_name' value Example #2 (comment #1) - http://commons.wikimedia.org/wiki/File:Viestura_ordenis.jpg - valid 'oi_archive_name' value, but "File not found" (corrupt file storage)
Moving this down in priority only because Aaron is working on what he believes is the root cause (bug 36132), and I want to move this down below that on his list of stuff to do. With any luck, this will automatically get fixed as a result of that.
(In reply to comment #6) > Directly deleting the rows? At least you should keep a backup before... Of course. Though I don't plan on doing that anymore after looking at some of the examples more. It nice to at least see some of the history more clearly (without consulting action=history), so I'd prefer to leave the stub rows. What can be more easily fixed is just changing the doDBUpdates() function of LocalFileMoveBatch. The line "$status->failCount += $total - $affected;" results in negative values in this case, causing the 'imageinvalidfilename' error.
Fixed in https://gerrit.wikimedia.org/r/#change,5488.
