[Top] [All Lists]

Re: [Courier-imap] Courier Imap 3.0.7 rpmbuild fails with xfs->lvm2->md1

To: Nathan Scott <nathans@xxxxxxx>
Subject: Re: [Courier-imap] Courier Imap 3.0.7 rpmbuild fails with xfs->lvm2->md1
From: Kai Leibrandt <k_leibrandt@xxxxxxxxxxx>
Date: Sun, 12 Sep 2004 15:06:15 +0200
Cc: Sam Varshavchik <mrsam@xxxxxxxxxxxxxxx>, courier-imap@xxxxxxxxxxxxxxxxxxxxx, linux-xfs@xxxxxxxxxxx
In-reply-to: <20040910012411.GA13692@frodo>
References: <414019EC.8060903@xxxxxxxxxxx> <cone.1094773129.309420.14709.500@xxxxxxxxxxxxxxxxxxxxxxxx> <20040910012411.GA13692@frodo>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mozilla Thunderbird 0.7.3 (Windows/20040803)
Woops, I am _so_ ashamed...

The $HOME/rpm directory was linked to an nfs mounted directory, and I had completely forgotten about that.
Building the rpm locally of course works perfectly.

Sorry for the noise.


Nathan Scott wrote:
Hi there Sam,

On Thu, Sep 09, 2004 at 07:38:49PM -0400, Sam Varshavchik wrote:

From your strace:

rename("confmdtest/tmp/1094656838.M69783P2446_acl.testbed.kailee.net", "confmdtest/courierimaphieracl/foobar") = 0

followed by:

stat64("confmdtest/courierimaphieracl/foobar", {st_mode=S_IFREG|0644, st_size=27, ???}) = 0
unlink("confmdtest/courierimaphieracl/foobar") = 0

The stat/unlink calls come from here:

                        if (stat(p, &stat_buf) == 0 &&
                            stat_buf.st_mtime < now - 60*60)

The unlink should not happen unless the timestamp is one hour old. Except
that this file was created only a few milliseconds ago.  Bzzzzzt.  make
check fails.

I've tried to reduce the problem down to the simple test case
above (see attached script and test prog) - but this seems to
do the right thing ...

[root@bruce test]# ./foo + rm -fr confmdtest
+ mkdir -p confmdtest/tmp confmdtest/courierimaphieracl
+ touch confmdtest/tmp/1094656838.M69783P2446_acl.testbed.kailee.net
+ ./a.out confmdtest/tmp/1094656838.M69783P2446_acl.testbed.kailee.net 
OK: mtime=1094774209 now-60*60=1094770609
[root@bruce test]#

I dimly recall someone mentioning an issue with st_mtime on XFS, a long time

I'm not aware of any - do you remember any of the details there?


ps: does the attached script+test fail for you, Kai?

<Prev in Thread] Current Thread [Next in Thread>