package building failure irritation

[David Wolfskill]

On Mon, Mar 01, 2010 at 01:58:29PM +0000, xorquewasp at googlemail.com wrote:
> On 2010-03-01 14:34:50, Dag-Erling Smørgrav wrote:
> > % fgrep 'Directory not empty' inkscape.txt 
> > rmdir: /work/ports/devel/boost-libs/work: Directory not empty
> > ...
> > rmdir: /work/ports/x11-toolkits/pangomm/work: Directory not empty
> > 
> > This tells me that either there is another build running in parallel, or
> > /work is mounted from a dodgy NFS server.
> 
> 'Lo,
> 
> There's certainly no parallel building going on, but /work is nullfs
> mounted (from ZFS). Could this cause the above?

Caveat: this may be a "red herring."  But please note that for
stable/7 prior to r190970 (2009-04-12 10:43:41 -0700) or head prior
to r189287 (2009-03-02 12:51:39 -0800; prior to stable/8 branch,
so it's part of stable/8 already), there was a rather nasty (IMO,
as I spent a fair amount of time trying to figure out what was going
on) such that a FreeBSD NFS client would see precisely the above
symptoms if:

* A process on the FreEBSD NFS client performed a chdir() to a
  directory that was NFS-mounted, then started a "recursive descent"
  (e.g., "tar c ..." or "rm -fr") from that directory and

* Some other process on the same FreeBSD NFS client attempted to perform
  an unmount() of the NFS-mounted file system referenced above.

Note that the unmount() is doomed, as the file system is active -- a
directory in it is the $cwd for the first process, after all.

This may seem an unlikely -- possibly even perverse -- combination of
events.  However, it is actually SOP for amd(8): the master amd process
periodically forks a child to perform an attempted unmount of
auto-mounted NFS file systems periodically, and the way amd realizes
that the file system is not eligible for unmounting is if the attempted
unmount() gets EBUSY.


It is *possible* that something akin to this mechanism *might* be
affecting the OP.

Peace,
david
-- 
David H. Wolfskill				david at catwhisker.org
Depriving a girl or boy of an opportunity for education is evil.

See http://www.catwhisker.org/~david/publickey.gpg for my public key.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-hackers/attachments/20100301/dccbb782/attachment.pgp