All-
On Monday, 3 Jun 1996, I posted a question asking if others had encountered
problems with mailboxes (ie /var/spool/mail/$LOGNAME) having out-of-band
characters in the headers of the enclosed messages (the original message I
sent to this list is included below, as a refresher). These characters
cause problems for certain MUAs, including pine 3.91 and elm.
As I indicated in my initial message, we're using sendmail 8.7.x (shouldn't
make a difference) on the 3.2c system that's having this problem. Most of
our users access their email through pine 3.91. Some users use POP-mail;
we're currently running the Qualcomm popper, 2.1.4-R3.
My essential question was, "Has anyone else had this happen, or is it
particular to our site?".
I received responses from:
Qiegang Long <qlong_at_osf.org>
Jo Knox <fxjwk_at_aurora.alaska.edu>
Eric <Eric.Rostetter_at_utoledo.edu>
Kent Adams <Kent.Adams_at_jcu.edu.au>
mark_at_qualcomm.com (Mark Erikson)
Chua Koon Teck <koonteck_at_singnet.com.sg>
"Michael A. Crowley" <mcrowley_at_mhc.mtholyoke.edu>
Veselin Terzic <terzic_at_mda.ca>
Andreas Schuenemann schuenem_at_rz5.rz.fh-wilhelmshaven.de
Many of the people that responded indicated that they too had seen the same
problem, so it's not something that's unique to my site. A number of people
indicated that they had problems with ASCII 0 (null) characters in their
mailboxes. We haven't seen this problem; for us it's always ASCII 255.
After all the responses, I'm more suspicious than ever that it's a problem
with /bin/mail, the local delivery agent. I do not think the particular
problem we're seeing is being caused by the Qualcomm popper because some
of the people that have had this problem with their mailbox apparently do
not use POP mail. Qiegang Long indicated that using `procmail' for local
delivery would alleviate the problem (which again points a finger at
/bin/mail), but unfortunately that's probably not a good solution for us
considering the volume of mail we push through the machine.
Kent Adams was kind enough to get Mark Erikson, from Qualcomm, involved
in the chase. Again, I'm pretty sure the problem we're seeing is not due
to the Qualcomm popper, but many people will probably be interested to note
that version 2.2.0 of the Qualcomm popper is no longer in beta. Mark has made
some improvements to the popd and has more planned for the future. He has
this to say regarding the Qualcomm popper:
>) There has been a problem in previous versions of the popper which
>) have inserted NULLs at the beginning of the mailspool. I believe I have
>) fixed this in version 2.2. At least, the number of reports is much less
>) that it was before the change.
In a more recent message, Mark also said:
I'm also going to be rewriting the
IO to use only unbuffered reads and writes. The original code uses
both buffered and unbuffered IO. This will be in a release of 2.2.1
hopefully in beta in a month or so.
Chua Koon Teck indicated that pine 3.93 is capable of working around the
control characters in the header. Several other people mentioned that mailx
is also not affected by the control characters in the header.
A couple of people also indicated that they suspected a correlation between
filling /var/tmp and/or /tmp and problems with out-of-band characters in the
mail headers in people's spool files.
As of right now, I have no definitive fix for the problem. We've had no
problems since I sumitted the original question. My boss has written a
Perl script to run through all the mailboxes looking for hex ff and report
any that are found.
Because pine 3.93 can apparently work around the characters we'll probably
upgrade to that sooner than expected to help alleviate the effects. We'll
also upgrade to the latest version of the Qualcomm popper sometime this
summer. Hopefully those two upgrades will make the problem less noticable to
users, even if the problem persists.
Until I have some type of definitive proof (instead of just a strong belief)
that it's a problem with /bin/mail, I won't open a call with Digital's CSC.
Sometime this week I hope to have time to try reproduce the problem on my
personal workstation. Hopefully then I'll have enough information to open a
useful support call.
Thanks to all the people that responded! You've been very helpful.
Tim
--
Tim Mooney mooney_at_toons.cc.ndsu.NoDak.edu
Information Technology Services (701) 231-1076 (Voice)
Room 242-J6, IACC Bldg. (701) 231-8541 (FAX)
North Dakota State University, Fargo, ND 58105
-------------original message-----------------
All-
We've been having sporadic problems with users' mailboxes having control
characters left in the `From ' area of the header of a message. I suspect
that the problem is with binmail (the local delivery program) but I'm not
certain.
The machine is running Digital Unix 3.2c. We're using a recent version of
sendmail (8.7.x, not Digital's 5.65 sendmail), with the following stanza in
the sendmail.cf for the local delivery agent:
Mlocal, P=/bin/mail, F=lsDFMAw5:/|_at_rmn, S=10/30, R=20/40,
T=DNS/RFC822/X-Unix,
A=mail -d $u
We're also running the Qualcomm pop server, version 2.1.4-R3, but I'm fairly
certain that the problem with control characters has occurred even for people
that don't use pop mail. Most of our users use pine 3.91 to access their
email.
The control character that is left in the mailbox is always ASCII 255 (octal
377). It generally ends up preceding the `From ' of a mail header, so that
pine won't see any messages after (and including) the message with the
corrupted header.
We haven't had any problems of this nature with other machines at our site
running other flavors of Unix. Those other machines are also running pine
3.91 and the 2.1.4r3 qualcomm popper, so the problem seems to be particular
to our Digital Unix machines.
Has anyone else had this happen, or is it particular to our site?
--
Tim Mooney mooney_at_toons.cc.ndsu.NoDak.edu
Information Technology Services (701) 231-1076 (Voice)
Room 242-J6, IACC Bldg. (701) 231-8541 (FAX)
North Dakota State University, Fargo, ND 58105
Received on Tue Jun 11 1996 - 21:43:16 NZST