Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Corruption of .mh_sequences files?

24 views
Skip to first unread message

pd

unread,
Oct 15, 2008, 11:41:52 PM10/15/08
to
I've been running nmh-1.3 on Mac OS X 5.5 for a month or so. I use
fetchmail and procmail to retrieve messages from a POP3 server, and
all the procmail recipes use rcvstore to file the messages in various
folders. This has been working perfectly until the last couple of
days. Suddenly, I started seeing messages like:

rcvstore: eof encountered in field "7"
rcvstore: /Users/peterdavis/Mail/spam/.mh_sequences is poorly
formatted
rcvstore: eof encountered in field "400"
rcvstore: /Users/peterdavis/Mail/spam/.mh_sequences is poorly
formatted

Any clues about this? If only rcvstore is touching the .mh_sequences
file, what could be corrupting it?

Thank you!

-pd

pd

unread,
Oct 16, 2008, 6:57:12 AM10/16/08
to

Just to add some info:

1) Deleting .mh_sequences doesn't solve the problem. It just gets
recreated, and the messages start again.

2) When I tried to exit the Mac Terminal program, I was prompted about
killing a bunch of background rcvstore processes, so I'm guessing that
procmail is running this asynchronously, and that's what's causing the
problem ... multiple rcvstore processes trying to hit on
the .mh_sequences file at once. My next step is to look at making
procmail run synchronously.

Thanks,
-pd

Kevin the Drummer

unread,
Oct 16, 2008, 11:57:30 AM10/16/08
to

I've seen this a from time to time over the years, and not all that
often. The issue is that the file really does get poorly formated, and
I wish I knew where. Here's a correctly formated example:

cur: 18
pseq: 31
unseen: 31

Here's a bad one:

cur: 18
31
unseen: 31

I always see a number on a line by itself when I get that error message.
If I edit the file and just remove that line, then all is well again.
This has worked for me every time on SunOS, Solaris, and Linux with many
versions of MH, nmh, Xmh and exmh.

I hope that helps...

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.

pd

unread,
Oct 16, 2008, 12:55:53 PM10/16/08
to
On Oct 16, 11:57 am, Kevin the Drummer <nob...@cosgroves.us> wrote:

Thanks. This is similar to what I've seen in some cases, but I've
also seen it fail when the .mh_sequences file is empty.

More and more I suspect the problem lies with multiple child processes
spawned by procmail, all trying to read/write the .mh_sequences file
at once. The procmail lockfile method should prevent that, but I
haven't figured out how to make that work yet.

Thanks!
-pd

pd

unread,
Oct 16, 2008, 4:21:18 PM10/16/08
to
Well, I did get lockfiles working in procmail, so we'll see if that
fixes the problem. As I mentioned, I had been using this setup for a
month and a half with no errors, so I don't know how long to run
before I can consider the problem solved.

Thanks,
-pd

Kevin the Drummer

unread,
Oct 17, 2008, 5:19:26 PM10/17/08
to

How did you get the lockfiles working?

Thanks....

Xavier Maillard

unread,
Jan 23, 2009, 3:50:12 PM1/23/09
to
pd <pfd...@gmail.com> writes:

What was the workaround ? I mean for lockfiles. I have the same
problem with multiple instances of rcvstore.

Regards

0 new messages