Last modified: 2012-08-04 20:49:06 UTC
Report on IRC: "I've just installed a new instance of mediawiki. I created a test user called "Test" by registering with that name. Now, I've logged in as the administrator user and I'm trying to block the test user, but the blocking interface says the user is already blocked. However, the block list says there's nobody there, and indeed the user doesn't seem to be blocked in any way. I'm using mediawiki 1.12 and PostgreSQL 8.3, anyone know what might be wrong?" Sounds like a DBMS issue to me.
== SQL: INSERT /* Block::insert OverlordQ */ INTO ipblocks (ipb_id,ipb_address,ipb_user,ipb_by,ipb_by_text,ipb_reason,ipb_timestamp,ipb_auto,ipb_anon_only,ipb_create_account,ipb_enable_autoblock,ipb_expiry,ipb_range_start,ipb_range_end,ipb_deleted,ipb_block_email) VALUES ('4','Over1ordQ','15','1','OverlordQ','test','2008-08-13 06:38:50 GMT','0','0','1','','2008-08-13 08:38:50 GMT','','','0','') SQL ERROR (ignored): ERROR: invalid input syntax for integer: "" == Looks to be the query causing that.
Apparently, it doesn't like that empty strings are inserted into boolean columns ipb_enable_autoblock and ipb_block_email.
Max: those columns were changed from boolean to smallint: if they are still boolean for you, run updaters.php. Argh, I wish Mysql has a real boolean type. I'll see about having the query not assume that the database backend will convert an empty string to the number 0.
Brent, can you try this on the latest revision? Looking over Block.php in subversion, I'm not seeing any place that would cause mEnableAutoblock to NOT be 0, so I'm curious how the statement you showed would have got generated unless it's from an older version.
Problem does still exist in SVN. Here's from debug log and postgres log: http://pastebin.ca/QGazBKyC password is: mw.
Ah, I think I found it. Line 64 of specials/SpecialBlockip.php: $this->BlockEnableAutoblock = $wgRequest->getBool( 'wpEnableAutoblock', $byDefault ); I'm guessing getInt would be more appropriate here, but I'm not going to change that until someone can weight in from the mysql sie of things.
In MySQL we use tinyint(1) columns for booleans. It should definitely be inserting numbers into them, not empty strings. If PostgreSQL is using integers consistently, then perhaps we could fix this for both DBMSes in Database::addQuotes()? if ( is_null( $s ) ) { return 'NULL'; } elseif ( is_bool( $s ) ) { return intval( $s ); } else { return "'" . $this->strencode( $s ) . "'"; }
Tim: I really like that solution. Ideally we'd have Postgres using true BOOLS and map things to 'true' and 'false', but I think it's early days for that, as there are lots of other places in the code that assumes bool==tinyint. Brent, I made the above change to r39763, please give it a whirl.
It's kinda fixed, the query works but there's still oddities going on. As can be seen here, we're starting with an empty blocklist: http://toolserver.org/~overlordq/ipblocks.txt Now let me block the test account. From the debug log: http://pastebin.ca/1180966 However, after form submission, Special:BlockIP says: "TestUser" is already blocked. But according to the ipblocks table, it's the first block: http://toolserver.org/~overlordq/ipblocks2.txt
Fixed in r40515, backported to 1.13 in r40517.
