On Wed, 25 Jul 2007 15:35:26 +0100 (BST), "David Carter" <[EMAIL PROTECTED]> said: > On Wed, 25 Jul 2007, Bron Gondwana wrote: > > > See also the attached, our latest patch - it includes a bugfix which stops > > you being able to randomly move some user's home directory to a non-home > > directory elsewhere. Something that you could do only because of the bug > > in mboxname_isusermailbox. > > I think that the problem is specific to virtual domain setups.
I'm pretty sure it's due to unixhierarchysep - if you look at the deleted half of my patch it does a string check for 'user.' specifically - the new one does a check for 'user' and then for start[4] == 'sep'. after walking to that part. The removed code does also search for '!user.' which is the virtual domain support bit - though it doesn't also check that they're enabled. A bit bogus but probably 99.999% safe. > > It also makes mboxname_isusermailbox a fair bit easier to read - though > > it does a config item lookup each use - might be worth making that > > static or even a bit more global. Hmm. What do you think? > > config_getstring() is actually just an array lookup: the hard work is > done > at compile time. > > mboxlist_delayed_delete_isenabled() and mboxlist_in_delete_hierachy() > have > some caching, but this is really just habit on my part. Ahh, premature optimisation. My favourite kind :) It confused me for a bit anyway. I'm tempted to strip them if they're not needed - complexifies the code and makes it harder for a new maintainer to understand. Bron. -- Bron Gondwana [EMAIL PROTECTED]
