Since r61917 (3 February), running the api tests have been creating a user called 'Useruser' which initially had a random password. r72475 (6 September) screwed up by using a hardcoded pasword. Since it wasn't too bad for wikis allowing account creation, r74118 (1 October) made that user a sysop. Finally, r74213 (3 October) split it into two users.
I added in r75588 a feature to block weak passwords, which will disable them. In r75589 I reverted to random password users and renamed them so that they a) Are much more unlikely to conflict with any human account creation. b) It is clear where did such sysop come from.
Sadly, we can't add them to $wgReservedUsernames without breaking the tests.
Any user having run 'make destructive' in the last two months should update to r75589 inmediatly. They are also encouraged to desysop and block the accounts 'Useruser' and 'Useruser1'. They are only at risk if that install is publicly accessible, though.
Users not running MediaWiki from trunk or which haven't run the phpunit tests are unaffected.
On Wed, Oct 27, 2010 at 6:32 PM, Platonides Platonides@gmail.com wrote:
Since r61917 (3 February), running the api tests have been creating a user called 'Useruser' which initially had a random password. r72475 (6 September) screwed up by using a hardcoded pasword. Since it wasn't too bad for wikis allowing account creation, r74118 (1 October) made that user a sysop. Finally, r74213 (3 October) split it into two users.
Why are these tests messing with the live database? For consistency of results, shouldn't they be using a fresh database on each run?
On Thu, Oct 28, 2010 at 7:16 PM, Aryeh Gregor Simetrical+wikilist@gmail.com wrote:
On Wed, Oct 27, 2010 at 6:32 PM, Platonides Platonides@gmail.com wrote:
Since r61917 (3 February), running the api tests have been creating a user called 'Useruser' which initially had a random password. r72475 (6 September) screwed up by using a hardcoded pasword. Since it wasn't too bad for wikis allowing account creation, r74118 (1 October) made that user a sysop. Finally, r74213 (3 October) split it into two users.
Why are these tests messing with the live database? For consistency of results, shouldn't they be using a fresh database on each run?
Well, yes. We also do have the code, but only for parsertests. Somebody should go ahead and generalize this code for all tests.
Bryan
Aryeh Gregor wrote
Why are these tests messing with the live database? For consistency of results, shouldn't they be using a fresh database on each run?
For Selenium tests, it is planned to provide a mechanism that sets up a fresh database and image directory per test suite. I will try to implement this as modular as possible, so I could image this might become a common basis also for the unit tests.
Cheers, Markus
Dear All,
I have used the mwdumper to convert compressed wikipedia dumps. But I have Oracle to use. I don't know enough about databases to know, but if a dump in mysql or postgresql format were generated, could it be converted for use with Orcale? Thanks,
Zeyi
wikitech-l@lists.wikimedia.org