[ Converted text/html to text/plain ]
Point taken. I was looking at it from the standpoint that on v3 it worked and
on v4 it did not (bug??). However, as it was pointed out, there was a change
in v4 to support better data integrity (a very good thing), it would have been
nice if somewhere in the docs it had said that it would break NFS spools. For
better or worse, I think NFS spools need to tested (not necessarily supported
with the same data integrity) since there are a lot of people using NFS for
this and many other purposes. Thanks for all the imput on this!
EA
>From: David Woodhouse
>To: exim-users@???
>CC: Edgar Lovecraft
>Subject: Re: [Exim] upgrade 3.36 -> 4.04, "sync error" ??
>Date: Tue, 14 May 2002 11:08:55 +0100
>
>
>ph10@??? said:
> > The difference between Exim 3 and Exim 4 is that the latter contains
> > some special code for improving the integrity of file handling under
> > Linux. This is the comment in the code:
>
> > /* Linux (and maybe other OS?) does not automatically sync a directory
> > after an operation like rename. We therefore have to do it forcibly
> > ourselves in these cases, to make sure the file is actually accessible
> > on disk, as opposed to just the data being accessible from a file in
> > lost+found. Linux also has O_DIRECTORY, for opening a directory. */
>
> > It seems (from your test) that the code to do this does not work when
> > the directory is NFS mounted. The call to sync the directory fails.
>
> > Are there any Linux file system experts out there? Care to comment on
> > this? Should Exim ignore the "invalid argument" error, and assume that
> > that it means "directory is NFS-mounted, you can do no more"?
>
>fsync() on an NFS directory should be a NOP under Linux -- I don't even want
>to think about the possibility of it being otherwise. Hence it should
>probably return zero instead of -EINVAL.
>
>In general, I would think that it's best for Exim to assume that an error
>return from fsync() meant that something went wrong, not that operations are
>synchronous and the fsync() was unnecessary. Although the latter is
>(currently) the case for the linux NFS client, it should probably be fixed.
>
>In the meantime, you could argue that refusing to use a spool on NFS is a
>feature rather than a bug ;)
>
>--
>dwmw2
>
>
>
>--
>
>## List details at http://www.exim.org/mailman/listinfo/exim-users Exim
details at
http://www.exim.org/ ##
------------------------------------------------------------------------------
Join the worlds largest e-mail service with MSN Hotmail. Click Here[1]
===References:===
1.
http://g.msn.com/1HM105401/47
