When adding or removing a page from a group on [[m:Special:AggregateGroups]], the change is saved (as I can check by refreshing the page), but I get the error "Exception Caught: Error in numRows(): Table 'metawiki.revtag_type' doesn't exist (10.0.6.36)" which is a bit ugly at least.
Another error with revtag_type was bug 39287, provisionally adding it in dependencies.
Version: 1.21.x
Severity: major
URL: https://meta.wikimedia.org/wiki/Special:AggregateGroups
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Nikerabbit | T44430 Exception Caught: Error in numRows(): Table 'metawiki.revtag_type' doesn't exist (10.0.6.36) | |||
| Resolved | None | T41287 calling tableExists breaks fetchObject |
This is an issue in the core database abstraction layer.
Minimal test case:
$db = wfGetDB( DB_SLAVE ); $res = $db->select( 'page', '*', array( 'page_id' => 1 ) ); $db->tableExists( 'foo' ); var_dump( $res->numRows() );
Error in numRows(): Table 'mediawiki.bw_foo' doesn't exist (localhost)
Dependency for this issue appears to have been resolved (bug 39287), but the issue still exists.
(In reply to comment #8)
- Bug 45879 has been marked as a duplicate of this bug. ***
Workaround for this specific case in translate: https://gerrit.wikimedia.org/r/#/c/52769/
Minimal test case confirms it's ok in vanilla PHP.
<?php
$sql = "SELECT * FROM wikidb.mw_user LIMIT 1";
$sql1 = "SELECT 1 FROM foobar LIMIT 1";
$con=mysql_connect("","","");
$result = mysql_query($sql, $con);
$result1 = mysql_query($sql1, $con);
var_dump( (bool)$result1 ); echo "\n";
echo mysql_num_rows( $result ) . "\n";
mysql_close($con);
/**
*/
function numRows( $res ) {
if ( $res instanceof ResultWrapper ) {
$res = $res->result;
}
wfSuppressWarnings();
$n = mysql_num_rows( $res );
wfRestoreWarnings();
if( $this->lastErrno() ) {
throw new DBUnexpectedError( $this, 'Error in numRows(): ' . htmlspecialchars( $this->lastError() ) );
}
return $n;}
It's strange, as the code in numRows() is very minimal and doesn't depend on any class member variables
If this is not backported, I need to consider Reedy's workaround for Translate as BC code.
(In reply to comment #15)
(In reply to comment #14)
Still needs manual cherry-pick.
To 1.21?
Apparently... Unless someone rejects the backport request, that is.