On Thu, Sep 27, 2007 23:25:31 PM +0200, Dallman Ross
(dman(_at_)nomotek(_dot_)com) wrote:
My guess is there is no native locking on msgid.cache. I won't
get into whether the OS is supposed to be writing to the file
atomically -- I do not know.
I understand the comment, but I don't know the answer. FWIW, this is a
VPS (Centos 4.5 on OpenVZ; I don't remember how to check what the file
system is): may it matter?
A few more info. Here are the examples of corrupt msgids I get, those
which probably create false duplicates:
# strings msgid.cache | grep -v '^<'
<!~!UE relly long string removed, see the initial space? ....com>
2.vbhcs.org>
# strings msgid.cache | grep -v '>$'
<46B6FB5800DBCF76(_at_)ims2(_dot_)aprovider(_dot_)com> (added by
postmaster(_at_)ivs2(_dot_)aprovider(_dot_)com)
and the only recipe before this, one which makes a cache of msgids of
uninteresting messages, has the same problems:
#strings ignore.cache | grep -v '^<'
p(_dot_)ty8ap9hdvk8tcq(_at_)localhost>
But if you really want an uncorrupted cache file, you might need to
go to using a global lockfile around the formail recipe.
I am going to bed in two minutes, otherwise I'll fell asleep anyway...
but I will try tomorrow and report. Comments from other posters are
extremely welcome, of course.
Thanks,
Marco
--
Help *everybody* love Free Standards and Free Software
http://digifreedom.net/
____________________________________________________________
procmail mailing list Procmail homepage: http://www.procmail.org/
procmail(_at_)lists(_dot_)RWTH-Aachen(_dot_)DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/procmail