Tickets :: Comment on [#12837] Imap_Client: Leaking IMAP socket on error

* Your Email Address
* Spam protection	Enter the letters below: .__ . \ / __ .___ [__)\| >< / `[__ [__)\|___/ \\__.\|
Comment	>> When an IMAP auth error is thrown, the socket to the IMAP server is >> not closed. > > That is correct. An authorization error shouldn't close the > connection. It just indicates that the transition between > non-authenticated -> authenticated state was not successful. But at > the protocol level you are not disconnected from the IMAP server and, > in fact, the IMAP server remains willing/able to process further > commands. > > For example, in an interactive environment - such as a CLI script - > there is no need to destroy the IMAP connection. If, after an > unsuccessful authentication attempt, the CLI can ask for another > password and we should be reusing the same connection for > authentication the next time we try to authenticate. > >> I've created the attached example script to demonstrate >> the problem. > > I'm confused by this script. First, it's going to continually loop > whether or not authentication is correct or not. So I'm not sure > what the outer while() loop is attempting to show. > > Second, you are not distinguishing between what the error is from > listMailboxes(). listMailboxes() may fail because of an > authentication error. But it also may fail because of invalid input. > Or because of a temporary error on the IMAP server. Or because you > don't have the correct ACL permissions. Simply checking for an > Exception in this case is insufficient. > > It sounds like this is more of a logic error in the Kolab code. For > starters, the Horde_Imap_Client instantiation code should only occur > once and should be disconnected from any of the actions on the object > itself. And if an IMAP action throws an exception, you surely need > to break any loop containing that IMAP action. But this doesn't have > anything to do with authentication errors per se. That just seems to > be deficient exception handling in general. > > This could be a failure of the documentation to explain the different > types of Exceptions. But it doesn't seem to be a design error in > Imap_Client. > >> Could we close the socket on error? > > You can do it manually by calling logout(). But see above: there > shouldn't be any reason to. See, e.g., IMP: we never call logout().
Attachment
Watch this ticket

Saved Queries