On Fri, 24 Oct 1997 00:22:27 -0500, Philip Guenther
<guenther(_at_)gac(_dot_)edu>
wrote:
nelson(_at_)media(_dot_)mit(_dot_)edu (Nelson Minar) writes:
(The purpose of this code is to delete empty mailboxes. It's not an
(Why?)
It looks like the assignment to serrno is to deliberately *ignore* the
error from fdunlock(). Is that the right thing to do?
I don't think I know enough about the failure modes of kernel-locking
to be able to say what would be safe here. From what I've seen on the
list it seems that this generally occurs when a) kernel locking is
hosed, in which case the message was delivered (but might be
corrupted); or b) the mailbox was removed from under procmail, in which
case the message was delivered correctly, but *some* other process is
not careful enough and mail is being lost. What would be "safe"? Just
writing the message about isn't safe, and neither is truncating and
returning an error. What's that leave?
Bouncing the message, IMHO.
I believe the same thing more or less happened to me here at my
university last Spring when the mail server and NFS went haywire at
the same time (also Digital Alphas running OSF1, incidentally ...).
My sysadmin (who is already rather suspicious about anything like
Procmail) made a comment that some of my mail could have been lost due
to Procmail's ignoring error exits from close calls. I have no idea
what exactly the circumstances would have been, but Nelson's looks
like a competent diagnosis of at least one problem with Procmail.
(Thanks for looking into it!)
/* era */
--
Paparazzi of the Net: No matter what you do to protect your privacy,
they'll hunt you down and spam you. <http://www.iki.fi/~era/spam/>