Description: ------------ in lines 216-222 in mysql.php it says: ------------------------------------------ if (isset($dsn['new_link']) && ($dsn['new_link'] == 'true' || $dsn['new_link'] === true)) { $params[] = true; } else { $params[] = false; } ------------------------------------------ The new_link parameter became available in PHP 4.2.0 for mysql_connect. But it isn't a parameter that is available for mysql_pconnect. I suggest this code fix: ------------------------------------------ if (!$persistent && version_compare(phpversion(), $this->features['new_link'], '>=')) { if (isset($dsn['new_link']) && ($dsn['new_link'] == 'true' || $dsn['new_link'] === true)) { $params[] = true; } else { $params[] = false; } } ------------------------------------------ For reference compare http://www.php.net/mysql_pconnect with http://www.php.net/mysql_connect and search for new_link

Yep, it became available in 4.2.0. And 4.2.0 is an explicit dependency for the DB package and the minimum supported by PEAR in general.

Hmm... Now, I see you're talking about pconnect. Sorry about that... Let me examine.

Fixed in CVS and will be in a new release (1.7.5) in a few minutes. Thanks.

Thanks for your quick response, danielc. Unfortunately, the fix you've added in the CVS doesn't resolve this bug. The way it's coded now, there will be a parameter added to $params, regardless of it's value (either true or false). What I was argumenting for was that this parameter shouldn't be added if mysql_pconnect is used. So AROUND the if-branch should be a check that ensures that mysql_connect is used. My suggestion is to remove !$persistent from the inner if-branch and make it an outer if-branch like so: if (!$persistent) { if (isset($dsn['new_link']) && ($dsn['new_link'] == 'true' || $dsn['new_link'] === true)) { $params[] = true; } else { $params[] = false; } } This way the parameter doesn't end up in mysql_pconnect, thus shifting the value of client_flags out of reach. I also added the version_compare because in the $features array (in line 76) the item 'new_link'=>'4.2.0' was available. So if that value is there, why not use it? :-) So that's were my first code fix suggestion came from.

Crap. You're right. I wasn't paying attention. I'll fix it.

thanks danielc