Home » Database » DB_DataObject » Bug #4096

[2005-04-07 16:52 UTC] tw at 6sys dot com

Description:
------------
Using PEAR::DB_DataObject 1.7.13, I put the full DSN in an INI file. The DSN includes the database password, of course. If there is a connection error (in this case, the database server went down), the first use of DB_DataObject results in an error message - raiseError() is getting called, which calls PEAR::raiseError(), which does kill the app (though the module docs seem to say the error isn't actually raised - confusing).

This is all fine. But, when the error is raised, it's printed to the browser, and the error message produced by PEAR::DB contains the DSN, which contains the database password. So, when my database server went down, anyone coming to the site saw the database password.

Reproduce code:
---------------
In the .ini:

[DB_DataObject]
database = "mysql://user:pass@localhost/database"
debug=0
...

In the code:

...
$site = new RDBSites; // a DB_DataObject descendent.
if ($site->get(1) != 1)
die("Can't find the default site.");
...

Expected result:
----------------
Returning FALSE from get() would be best, so I could handle the error myself.

Dying with a descriptive error message output to the PHP log file with error_log(), and perhaps some user-friendly output to the browser like "Error connecting to the database," would also be fine.

Dying with no error message at all would also be preferable.

Actual result:
--------------
Output to browser:

DB_DataObject Error: [db_error: message="DB Error: connect failed" code=-24 mode=return level=notice prefix="" info=" [nativecode=Access denied for user: 'user@localhost' (Using password: YES)] ** mysql://user:pass@localhost/database"]

(note the presence of the password)

[2005-04-07 18:34 UTC] tw at 6sys dot com

Sorry, I misassigned the package - this is against DB_DataObject, not DB, so danielc's comment doesn't apply.

Or if you meant it to apply - DB_DataObject::get() is not returning, so I can't check its return code; it's raising an error that kills the script.

[2005-04-11 14:50 UTC] tw at 6sys dot com

Looks good. Thanks!