| Back to Fetchmail Home Page | $Date$ |
Before reporting any bug, please read G3 for advice on how to include diagnostic information that will get your bug fixed as quickly as possible.
Note that this FAQ is occasionally updated from the SVN repository and speaks in the past tense ("since") about a fetchmail release that is not yet available. Please try a release candidate for that version in case you need the new option.
If you have a question or answer you think ought to be added to this FAQ list, file it to one of the trackers at our BerliOS project site or post to one of the fetchmail mailing lists (see below).
Fetchmail is a one-stop solution to the remote mail retrieval problem for Unix machines, quite useful to anyone with an intermittent PPP or SLIP connection to a remote mailserver. It can collect mail using any variant of POP or IMAP and forwards via port 25 to the local SMTP listener, enabling all the normal forwarding/filtering/aliasing mechanisms that would apply to local mail or mail arriving via a full-time TCP/IP connection.
Fetchmail is not a toy or a coder's learning exercise, but an industrial-strength tool capable of transparently handling every retrieval demand from those of a simple single-user ISP connection up to mail retrieval and rerouting for an entire client domain. Fetchmail is easy to configure, unobtrusive in operation, powerful, feature-rich, and well documented.
Fetchmail is open-source software. The openness of the sources is the strongest assurance of quality you can have. Extensive peer review by a large, multi-platform user community has shown that fetchmail is as near bulletproof as the underlying protocols permit.
Fetchmail is licensed under the GNU General Public License.
If you found this FAQ in the distribution, see the README for fetchmail's full feature list.
The latest HTML FAQ is available alongside the latest fetchmail sources at the fetchmail home page: http://fetchmail.berlios.de/. You can also usually find both in the POP mail tools directory on Sunsite.
A text dump of this FAQ is included in the fetchmail distribution. Because it freezes at distribution release time, it may not be completely current.
Yes I will, provided you include enough diagnostic information for me to go on. Send bugs to fetchmail-users. When reporting bugs, please include the following:
If you have FTP access to your remote mail account, and you have any suspicion that the bug was triggered by a particular message, please include a copy of the message that triggered the bug.
Often, the first thing I will do when you report a bug is tell you to upgrade to the newest version of fetchmail, and then see if the problem reproduces. So you'll probably save us both time if you upgrade and test with the latest version before sending in a bug report.
If your bug is something that used to work but stopped working when you upgraded, then you can help pin the bug down by trying intermediate versions of fetchmail until you identify the revision that broke your feature. The smart way to do this is by binary search on the version sequence. First, try the version halfway between your last good one and the current one. If it works, the failure was introduced in the upper half of the sequence; if it doesn't, the failure was introduced in the lower half. Now bisect that half in the same way. In a very few tries, you should be able to identify the exact adjacent pair of versions between which your bug was introduced -- and with information like that, I can usually come up with a fix very quickly.
Another useful thing you can do, if you're using POP3, is to test for IMAP4 support on your mailserver using the autoprobe function of fetchmailconf. If you have IMAP4, and fetchmailconf doesn't tell you it's broken, switch immediately. POP3 is a weak, poorly-designed protocol with chronic problems, and the later versions after RFC1725 actually get worse rather than better. Changing over to IMAP4 may well make your problem go away -- and if your ISP doesn't have IMAP4 support, bug them to supply it.
It is helpful if you include your .fetchmailrc file, but not necessary unless your symptom seems to involve an error in configuration parsing. If you do send in your .fetchmailrc, mask the passwords first!
If fetchmail seems to run and fetch mail, but the headers look mangled (that is, headers are missing or blank lines are inserted in the headers) then read the FAQ items in section X before submitting a bug report. Pay special attention to the item on diagnosing mail mangling. There are lots of ways for other programs in the mail chain to screw up that look like fetchmail's fault, but you may be able to fix these by tweaking your configuration.
A transcript of the failed session with -v -v (yes, that's two -v options, enabling debug mode) will almost always be useful. It is very important that the transcript include your POP/IMAP server's greeting line, so I can identify it in case of server problems. This transcript will not reveal your passwords, which are specially masked out precisely so transcripts can be passed around.
If you upgraded your fetchmail and something broke, you should include session transcripts with -v -v of both the working and failing versions. Very often, the source of the problem can instantly identified by looking at the differences in protocol transactions.
If the bug involves a core dump or hang, a gdb stack trace is good to have. (Bear in mind that you can attach gdb to a running but hung process by giving the process ID as a second argument.) You will need to reconfigure with:
CFLAGS=-g LDFLAGS=" " ./configure
Then rebuild in order to generate a version that can be gdb-traced.
Best of all is a mail file which, when fetched, will reproduce the bug under the latest (current) version.
Any bug I can reproduce will usually get fixed very quickly, often within 48 hours. Bugs I can't reproduce are a crapshoot. If the solution isn't obvious when I first look, it may evade me for a long time (or to put it another way, fetchmail is well enough tested that the easy bugs have long since been found). So if you want your bug fixed rapidly, it is not just sufficient but nearly necessary that you give me a way to reproduce it.
Probably not. Most of the feature suggestions I get are for ways to set various kinds of administrative policy or add more spam filtering (the most common one, which I used to get about four million times a week and got really tired of, is for tin-like kill files).
You can do spam filtering better with procmail or maildrop on
the server side and (if you're the server sysadmin) sendmail.cf
domain exclusions. If you really want fetchmail to do it from the
client side, use a preconnect command to call
mailfilter.
You can do other policy things better with the
mda option and script wrappers around fetchmail. If
it's a prime-time-vs.-non-prime-time issue, ask yourself whether a
wrapper script called from crontab would do the job.
I'm not going to do these; fetchmail's job is transport, not policy, and I refuse to change it from doing one thing well to attempting many things badly. One of my objectives is to keep fetchmail simple so it stays reliable.
For reasons fetchmail doesn't have other commonly-requested features (such as password encryption, or multiple concurrent polls from the same instance of fetchmail) see ESR's design notes. Note that this document is partially obsoleted by the updated design notes.
Fetchmail is a mature project, no longer in constant active development. It is no longer my top project, and I am going to be quite reluctant to add features that might either jeopardize its stability or involve me in large amounts of coding.
All that said, if you have a feature idea that really is about a transport problem that can't be handled anywhere but fetchmail, lay it on me. I'm very accommodating about good ideas.
The second-most-requested feature for fetchmail, after
content-based filtering, is the ability to have it remove messages
from a maildrop after N days, typically to be used with the
keep option as a sort of poor man's newsgroup
facility. Microsoft's Outlook Express supports this.
This feature won't be added either. Repeat after me: fetchmail's job is transport, not policy. If you want this, write a Perl or Python script, to be run from a cron job, that deletes old messages off your maildrop. Send it to me and I'll put it in the contrib directory.
There is a fetchmail-users list (fetchmail-users@lists.berlios.de) for bug reports and people who want to discuss configuration issues of fetchmail. It's a Mailman list, see http://lists.berlios.de/mailman/listinfo/fetchmail-users.
There is a fetchmail-devel list (fetchmail-devel@lists.berlios.de) for people who want to discuss fixes and improvements in fetchmail and help co-develop it. It's a Mailman list, which you can sign up for at http://lists.berlios.de/mailman/listinfo/fetchmail-devel. There is also an announcements-only list, fetchmail-announce@lists.berlios.de, which you can sign up for at http://lists.berlios.de/mailman/listinfo/fetchmail-announce.
The fetchmail development was also a sociological experiment, an extended test to see if my theory about the critical features of the Linux development model is correct.
The experiment was a success. I wrote a paper about it titled The Cathedral and the Bazaar which was first presented at Linux Kongress '97 in Bavaria and very well received there. It was also given at Atlanta Linux Expo, Linux Pro '97 in Warsaw, and the first Perl Conference, at UniForum '98, and was the basis of an invited presentation at Usenix '98. The folks at Netscape tell me it helped them decide to give away the source for Netscape Communicator.
If you're reading a non-HTML dump of this FAQ, you can find the paper on the Web with a search for that title.
The short answer: IMAP 2000 running over Unix.
Here's a longer answer:
Fetchmail will work with any POP, IMAP, ETRN, or ODMR server that conforms to the relevant RFCs (and even some outright broken ones like Microsoft Exchange and Novell GroupWise). This doesn't mean it works equally well with all, however. POP2 servers, and POP3 servers without LAST, limit fetchmail's capabilities in various ways described on the manual page.
Most modern Unixes (and effectively all Linux/*BSD systems) come with POP3 support preconfigured (but beware of the horribly broken POP3 server mentioned in D2). An increasing minority also feature IMAP (you can detect IMAP support by running fetchmail in AUTO mode, or by using the 'Probe for supported protocols' function in the fetchmailconf utility).
If you have the option, we recommend using or installing an IMAP4rev1 server; it has the best facilities for tracking message 'seen' states. It also recovers from interrupted connections more gracefully than POP3, and enables some significant performance optimizations. The new IMAP 2000 is particularly nice, as it supports CRAM-MD5 so you don't have to ship your mail password over the net en clair (fetchmail autodetects this capability). Older versions had support for GSSAPI giving a similar effect.
Don't be fooled by NT/Exchange propaganda. M$ Exchange is just plain broken (see item S2) and NT cannot handle the sustained load of a high-volume remote mail server. Even Microsoft itself knows better than to try this; their own Hotmail service runs over Solaris! For extended discussion, see John Kirch's excellent white paper on Unix vs. NT performance.
Source for a high-quality supported implementation of POP is available from the Eudora FTP site. Don't use 2.5, which has a rather restrictive license. The 2.5.2 version appears to restore the open-source license of previous versions.
Fetchmail will work with all popular mail transport programs. It also doesn't care which user agent you use, and user agents are as a rule almost equally indifferent to how mail is delivered into your system mailbox. So any of the popular Unix mail agents -- elm, pine, mh, or mutt -- will work fine with fetchmail.
All this having been said, I can't resist putting in a discreet plug for mutt. My own personal mail setup is sendmail plus fetchmail plus mutt. Mutt's interface is only a little different from that of its now-moribund ancestor elm, but its excellent handling of MIME and PGP put it in a class by itself. You won't need its built-in POP3 support, though; most of the mutt developers will cheerfully admit that fetchmail's is better :-).
Depending on what your mail server you are talking to, this ranges from trivial to impossible. It may even be next to useless.
Most people use fetchmail over phone wires (whether plain old copper or DSL), which are hard to tap. Anybody with the skill and resources to do this could get into your server mailbox with much less effort by subverting the server host. So if your provider setup is phone-company wire going straight into a service box, you probably don't need to worry.
In general there is little point in trying to secure your fetchmail transaction unless you trust the security of the server host you are retrieving mail from. Your vulnerability is more likely to be an insecure local network on the server end (e.g. to somebody with a TCP/IP packet sniffer intercepting Ethernet traffic between the modem concentrator or DSL POP you dial in to and the mailserver host).
Having realized this, you need to ask whether password encryption alone will really address your security exposure. If you think you might be snooped between server and client, it's better to use end-to-end encryption on your whole mail stream so none of it can be read. One of the advantages of fetchmail over conventional SMTP-push delivery is that you may be able to arrange this by using ssh(1); see K3.
Note that ssh is not a complete privacy solution either, as your mail could have been snooped in transit to your POP server from wherever it originated. For best security, agree with your correspondents to use a tool such as GPG (Gnu Privacy Guard) or PGP (Pretty Good Privacy).
If ssh/sshd isn't available, or you find it too complicated for you to set up, password encryption will at least keep a malicious cracker from deleting your mail, and require him to either tap your connection continuously or crack root on the server in order to read it.
You can deduce what encryptions your mail server has available
by looking at the server greeting line (and, for IMAP, the response
to a CAPABILITY query). Do a fetchmail -v to see
these, or telnet direct to the server port (110 for POP3, 143 for
IMAP).
If your mailserver is using IMAP 2000, you'll have CRAM-MD5 support built in. Fetchmail autodetects this; you can skip the rest of this section.
The POP3 facility you are most likely to have available is APOP.
This is a POP3 feature supported by many servers (fetchmailconf's
autoprobe facility will detect it and tell you if you have it). If
you see something in the greeting line that looks like an
angle-bracket-enclosed Internet address with a numeric left-hand
part, that's an APOP challenge (it will vary each time you log in).
You can register a secret on the host (using
popauth(8) or some program like it). Specify the
secret as your password in your .fetchmailrc; it will be used to
encrypt the current challenge, and the encrypted form will be sent
back the the server for verification.
Alternatively, you may have Kerberos available. This may require you to set up some magic files in your home directory on your client machine, but means you can omit specifying any password at all.
Fetchmail supports two different Kerberos schemes. One is a POP3 variant called KPOP; consult the documentation of your mail server to see if you have it (one clue is the string "krb-IV" in the greeting line on port 110). The other is an IMAP and POP3 facility described by RFC1731 and RFC1734. You can tell if this one is present by looking for AUTH=KERBEROS_V4 in the CAPABILITY response.
If you are fetching mail from a CompuServe POP3 account, you can use their RPA authentication (which works much like APOP). See I1 for details. If you are fetching mail from Microsoft Exchange using IMAP, you will be able to use NTLM.
Your POP3 server may have the RFC1938 OTP capability to use one-time passwords (if it doesn't, you can get OTP patches for the 2.2 version of the Qualcomm popper from Craig Metz). To check this, look for the string "otp-" in the greeting line. If you see it, and your fetchmail was built with OPIE support compiled in (see the distribution INSTALL file), fetchmail will detect it also. When using OTP, you will specify a password but it will not be sent en clair.
You can get both POP3 and IMAP OTP patches from Craig Metz at http://www.inner.net/opie.
These patches use a SASL authentication method named "X-OTP" because there is not currently a standard way to do this; fetchmail also uses this method, so the two will interoperate happily. They better, because this is how Craig gets his mail ;-)
Finally, you can use SSL for complete end-to-end encryption if you have an SSL-enabled mailserver.
Yes. In order to avoid giving indigestion to certain picky MTAs (notably exim), fetchmail always makes the RCPT TO address it feeds the MTA a fully qualified one with a hostname part. Normally it does this by appending @ and "localhost", but when you are using Kerberos or ETRN mode it will append @ and your machine's fully-qualified domain name (FQDN).
Appending the FQDN can create problems when fetchmail is running in daemon mode and outlasts the dynamic IP address assignment your client machine had when it started up.
Since the new IP address (looked up at RCPT TO interpretation time) doesn't match the original, the most benign possible result is that your MTA thinks it's seeing a relaying attempt and refuses. More frequently, fetchmail will try to connect to a nonexistent host address and time out. Worst case, you could up forwarding your mail to the wrong machine!
Use the smtpaddress option to force the appended
hostname to one with a (fixed) IP address of 127.0.0.1 in your
/etc/hosts. (The name 'localhost' will usually work;
or you can use the IP address itself).
Only one fetchmail option interacts directly with your IP
address, 'interface'. This option can be used to set
the gateway device and restrict the IP address range fetchmail will
use. Such a restriction is sometimes useful for security reasons,
especially on multihomed sites. See C3.
I recommend against trying to set up the interface
option when initially developing your poll configuration -- it's
never necessary to do this just to get a link working. Get the link
working first, observe the actual address range you see on
connections, and add an interface option (if you need
one) later.
You can't use ETRN if you have a dynamic IP address (your ISP changes your IP address occasionally, possibly with every connect). You need to have your own registered domain and a definite IP address registered for that domain. The server needs to be configured to accept mail for your domain but then queue it to forward to your machine. ETRN just tells to server to flush its queue for your domain. Fetchmail doesn't actually get the mail in that case.
You can use On-Demand Mail Relay (ODMR) with a dynamic IP address; that's what it was designed for, and it provides capabilities very similar to ETRN. Unfortunately ODMR servers are not yet widely deployed, as of early 2001.
If you're using a dynamic-IP configuration, one other (non-fetchmail) problem you may run into with outgoing mail is that some sites will bounce your email because the hostname your giving them isn't real (and doesn't match what they get doing a reverse DNS on your dynamically-assigned IP address). If this happens, you need to hack your sendmail so it masquerades as your host. Setting
DMsmarthost.here
in your sendmail.cf will work, or you can set
MASQUERADE_AS(smarthost.here)
in the m4 configuration and do a reconfigure. (In both cases,
replace smarthost.here with the actual name of your
mailhost.) See the sendmail FAQ for
more details.
No. You can use fetchmail with SOCKS, the standard tool for indirecting TCP/IP through a firewall. You can find out about SOCKS, and download the SOCKS software including server and client code, at the SOCKS distribution site.
The specific recipe for using fetchmail with a firewall is at K1
A user asks: but how do we send mail out to the POP3 server? Do I need to implement another tool or will fetchmail do this too?
Fetchmail only handles the receiving side. The sendmail or other preinstalled MTA on your client machine will handle sending mail automatically; it will ship mail that is submitted while the connection is active, and put mail that is submitted while the connection is inactive into the outgoing queue.
Normally, sendmail is also run periodically (every 15 minutes on most Linux systems) in a mode that tries to ship all the mail in the outgoing queue. If you have set up something like pppd to automatically dial out when your kernel is called to open a TCP/IP connection, this will ensure that the mail gets out.
Fetchmail is fully Y2K-compliant.
Fetchmail could theoretically have problems when the 32-bit time_t counters roll over in 2038, but I doubt it. Timestamps aren't used for anything but log entry generation. Anyway, if you aren't running on a 64-bit machine by then, you'll deserve to lose.
No. Fetchmail is a mail transport agent, best understood as a protocol gateway between POP3/IMAP servers and SMTP. Disconnected operation requires an elaborate interactive client. It's a very different problem.
Fetchmail streams message bodies line-by-line; the most core it ever requires per message is enough memory to hold the RFC822 header, and that storage is freed when body processing begins. It is, accordingly, quite economical in its use of memory.
After startup time, a fetchmail running in daemon mode stats its configuration file once per poll cycle to see whether it has changed and should be rescanned. Other than that, a fetchmail in normal operation doesn't touch the disk at all; that job is left up to the MTA or MDA the fetchmail talks to.
Fetchmail's performance is usually bottlenecked by latency on the POP server or (less often) on the TCP/IP link to the server. This is not a problem readily solved by tuning fetchmail, or even by buying more TCP/IP capacity (which tends to improve bandwidth but not necessarily latency).
The vendor-supplied make on FreeBSD systems can only be used within FreeBSD's "scope", e.g. the ports collection. Type "gmake" to run GNU make and better things will happen.
In the immortal words of Alan Cox the last time this came up: "Take the Solaris lex and stick it up the backside of a passing Sun salesman, then install flex and use that. All will be happier."
I couldn't have put it better myself, and ain't going to try now.
(The same problem has been reported under HP-UX v10.20 and IRIX)
If you get errors resembling these
mxget.o(.text+0x35): undefined referenceto '__res_search' mxget.o(.text+0x99): undefined reference to '__dn_skipname' mxget.o(.text+0x11c): undefined reference to '__dn_expand' mxget.o(.text+0x187): undefined reference to '__dn_expand' make: *** [fetchmail] Error 1
then you must add "-lresolv" to the LOADLIBS line in your Makefile once you have installed the 'bind' package.
If you get link errors involving dcgettext, like this:
rcfile_y.o: In function 'yyparse': rcfile_y.o(.text+0x3aa): undefined reference to 'dcgettext__' rcfile_y.o(.text+0x4f2): undefined reference to 'dcgettext__' rcfile_y.o(.text+0x5ee): undefined reference to 'dcgettext__' rcfile_y.o: In function 'yyerror': rcfile_y.o(.text+0xc7c): undefined reference to 'dcgettext__' rcfile_y.o(.text+0xcc8): undefined reference to 'dcgettext__' rcfile_y.o(.text+0xdf9): more undefined references to 'dcgettext__' follow
reconfigure with configure --with-included-gettext. This is due to some brain-damage in the GNU internationalization libraries.
Reconfigure with --disable-nls and recompile.
GNU gettext is an overengineered, fragile pile of crap. I have teetered on the brink of removing support for it entirely several times.
If you were using ETRN mode, change your smtphost option to a fetchdomains option.
The 'via localhost' special case for use with ssh tunnelling is gone. Use the %h feature of plugin instead.
In 5.6.8, the preauth keyword and option were changed back to auth. The preauth synonym will still be supported through a few more point releases.
The imap-gss, imap-k4, and imap-login protocol types are gone. This is a result of a major re-factoring of the authentication machinery; fetchmail can now use Kerberos V4 and GSSAPI not just with IMAP but with POP3 servers that have RFC1734 support for the AUTH command.
When trying to identify you to an IMAP or POP mailserver, fetchmail now first tries methods that don't require a password (GSSAPI, KERBEROS_IV); then it looks for methods that mask your password (CRAM-MD5, X-OTP); and only if it the server doesn't support any of those will it ship your password en clair.
Setting the preauth option to any value other than 'password' will prevent from looking for a password in your .netrc file or querying for it at startup time.
In 5.1.0, the auth keyword and option were changed to preauth.
If the dns option is on (the default), you may need
to make sure that any hostname you specify (for mail hosts or for
an SMTP target) is a canonical fully-qualified hostname). In order
to avoid DNS overhead and complications, fetchmail no longer tries
to derive the fetchmail client machine's canonical DNS name at
startup.
Just after the 'via' option was introduced, I
realized that the interactions between the 'via',
'aka', and 'localdomains' options were
out of control. Their behavior had become complex and confusing, so
much so that I was no longer sure I understood it myself. Users
were being unpleasantly surprised.
Rather than add more options or crock the code, I re-thought it. The redesign simplified the code and made the options more orthogonal, but may have broken some complex multidrop configurations.
Any multidrop configurations that depended on the name just
after the 'poll' or 'skip' keyword being
still interpreted as a DNS name for address-matching purposes, even
in the presence of a 'via' option, will break.
It is theoretically possible that other unusual configurations (such as those using a non-FQDN poll name to generate Kerberos IV tickets) might also break; the old behavior was sufficiently murky that we can't be sure. If you think this has happened to you, contact the maintainer.
The 'remote' keyword has been changed to
'folder'. If you try to use the old keyword, the
parser will utter a warning.
It could be because you're using a .fetchmailrc that's written
in the old popclient syntax without an explicit
'username' keyword leading the first user entry
attached to a server entry.
This error can be triggered by having a user option such as
'keep' or 'fetchall' before the first
explicit username. For example, if you write
poll openmail protocol pop3
keep user "Hal DeVore" there is hdevore here
the 'keep' option will generate an entire user
entry with the default username (the name of fetchmail's invoking
user).
The popclient compatibility syntax was removed in 4.0. It complicated the configuration file grammar and confused users.
The 'interface', 'monitor' and
'batchlimit' options changed after 2.8.
They used to be global options with 'set' syntax
like the batchlimit and logfile options. Now they're per-server
options, like 'protocol'.
If you had something like
set interface = "sl0/10.0.2.15"
in your .fetchmailrc file, simply delete that line and insert 'interface sl0/10.0.2.15' in the server options part of your 'defaults' declaration.
Do similarly for any 'monitor' or
'batchlimit' options.
Either upgrade to a post-5.0.5 fetchmail or put string quotes around it. :-)
The configuration file parser in older fetchmail versions treated any all-numeric token as a number, which confused it when it was expecting a name. String quoting forces the token's class.
The lexical analyzer in 5.0.6 and beyond is smarter and assumes any token following "username" or "password" is a string.
See F2. You're caught in an unfortunate crack between the newer-style syntax for negated options ('no keep', 'no rewrite' etc.) and the older style run-on syntax ('nokeep', 'norewrite' etc.).
Upgrade to a 5.0.6 or later fetchmail, or put string quotes around your token.
The most common cause of mysterious parse errors is putting a server option after a user option. Check the manual page; you'll probably find that by moving one or more options closer to the 'poll' keyword you can eliminate the problem.
Yes, I know these ordering restrictions are hard to understand. Unfortunately, they're necessary in order to allow the 'defaults' feature to work.
Ian T. Zimmerman <itz@rahul.net> asked:
On the machine where I'm the only real user, I run fetchmail as root from a cron job, like this:
fetchmail -u "itz" -p POP3 -s bolero.rahul.net
This used to work as is (with no .fetchmailrc file in root's home directory) with the last version I had (1.7 or 1.8, I don't remember). But with 2.0, it RECPs all mail to the local root user, unless I create a .fetchmailrc in root's home directory containing:
skip bolero.rahul.net proto POP3
user itz is itz
It won't work if the second line is just "user
itz". This is silly.
It seems fetchmail decides to RECP the 'default local user' (i.e. the uid running fetchmail) unless there are local aliases, and the 'default' aliases (itz->itz) don't count. They should.
Answer:
No they shouldn't. I thought about this for a while, and I don't much like the conclusion I reached, but it's unavoidable. The problem is that fetchmail has no way to know, in general, that a local user 'itz' actually exists.
"Ah!" you say, "Why doesn't it check the password file to see if the remote name matches a local one?" Well, there are two reasons.
One: it's not always possible. Suppose you have an SMTP host declared that's not the machine fetchmail is running on? You lose.
Two: How do you know server itz and SMTP-host itz are the same person? They might not be, and fetchmail shouldn't assume they are unless local-itz can explicitly produce credentials to prove it (that is, the server-itz password in local-itz's .fetchmailrc file.).
Once you start running down possible failure modes and thinking about ways to tinker with the mapping rules, you'll quickly find that all the alternatives to the present default are worse or unacceptably more complicated or both.
The easiest way to dispatch fetchmail on logout (which will work reliably only if you have just one login going at any time) is to arrange for the command 'fetchmail -q' to be called on logout. Under bash, you can arrange this by putting 'fetchmail -q' in the file '~/.bash_logout'. Most csh variants execute '~/.logout' on logout. For other shells, consult your shell manual page.
Automatic startup/shutdown of fetchmail is a little harder to arrange if you may have multiple login sessions going. In the contrib subdirectory of the fetchmail distribution there is some shell code you can add to your .bash_login and .bash_logout profiles that will accomplish this. Thank James Laferriere <babydr@nwrain.net> for it.
Some people start up and shut down fetchmail using the ppp-up and ppp-down scripts of pppd.
This depends a lot on your local networking configuration (and right now you can't use it at all except under Linux and the newer BSDs). However, here are some important rules of thumb that can help. If they don't work, ask your local sysop or your Internet provider.
First, you may not need to use --interface at all. If your machine only ever does SLIP or PPP to one provider, it's almost certainly by a point to point modem connection to your provider's local subnet that's pretty secure against snooping (unless someone can tap your phone or the provider's local subnet!). Under these circumstances, specifying an interface address is fairly pointless.
What the option is really for is sites that use more than one provider. Under these circumstances, typically one of your provider IP addresses is your mailserver (reachable fairly securely via the modem and provider's subnet) but the others might ship your packets (including your password) over unknown portions of the general Internet that could be vulnerable to snooping. What you'll use --interface for is to make sure your password only goes over the one secure link.
To determine the device:
To determine the address and netmask:
To illustrate the rule for dynamic IP addresses, let's suppose you're hooked up via SLIP and your IP provider tells you that the dynamic address pool is 255 addresses ranging from 205.164.136.1 to 205.164.136.255. Then
interface "sl0/205.164.136.0/255.255.255.0"
would work. To range over any value of the last two octets (65536 addresses) you would use
interface "sl0/205.164.0.0/255.255.0.0"
This answer covers versions of sendmail from 8.9.3-20 (the version installed in Red Hat 6.2) upwards. If you have an older version, upgrade to sendmail 8.9.
Stock sendmails can now do anti-spam exclusions based on a database of filter rules. The human-readable form of the database is at /etc/mail/access. The database itself is at /etc/mail/access.db.
The table itself uses email addresses, domain names, and network numbers as keys. For example,
spammer@aol.com REJECT cyberspammer.com REJECT 192.168.212 REJECT
would refuse mail from spammer@aol.com, any user from cyberspammer.com (or any host within the cyberspammer.com domain), and any host on the 192.168.212.* network. (This feature can be used to do other things as well; see the sendmail documentation for details)
To actually set up the database, run
makemap hash deny <deny
in /etc/mail.
To test, send a message to your mailing address from that host and then pop off the message with fetchmail, using the -v argument. You can monitor the SMTP transaction, and when the FROM address is parsed, if sendmail sees that it is an address in spamlist, fetchmail will flush and delete it.
Under no circumstances put your mailhost or any host you accept mail from using fetchmail into your reject file. You will lose mail if you do this!!!
Use the interval keyword on the ones that should be checked less often. For example, if you do a poll every 5 minutes, and want to poll some mailboxes every 5 minutes and some every 30 minutes, use something like this:
poll mainsite.example.com proto pop3 user .... poll secondary.example.com proto pop3 interval 6 user ...
Then secondary.example.com will be polled every 6th time that mainsite.example.com is polled, which with a polling interval of every 5 minutes means that secondary.example.com will be polled every 30 minutes.
Often, startup scripts have a different environment than an interactive login shell. For instance, $HOME might point to "/root" when you are logged in as root, but it might be either unset, or set to "/" when the startup scripts are running. That means fetchmail at startup can't find the .fetchmailrc.
Pick a location (such as /etc/fetchmailrc) and use fetchmail's -f option to point fetchmail at it. That should solve the problem.
To forward mail to a host other than the one you are running
fetchmail on, use the smtphost or
smtpname option. See the manual page for details.
For most sendmails, no special configuration is required. Eric
Allman tells me that if FEATURE(always_add_domain) is
included in sendmail's configuration, you can leave the
rewrite option off.
If your sendmail complains "sendmail does not relay", make
sure your sendmail.cf file says Cwlocalhost so that
sendmail recognizes 'localhost' as a name of its host.
If you're mailing from another machine on your local network, also ensure that its IP address is listed in ip_allow or name in name_allow (usually in /etc/mail/)
If you find that your sendmail doesn't like the address
'FETCHMAIL-DAEMON@localhost' (which is used in the bouncemail that
fetchmail generates), you may have to set
FEATURE(accept_unqualified_senders).
Günther Leber reports that Digital Unix sendmails won't
work with fetchmail. The symptom is an error message "553
Local configuration error, hostname not recognized as
local". The problem is that fetchmail normally feeds
sendmail with the client machine's host address in the MAIL FROM
line. These sendmails think this means they're seeing the result of
a mail loop and suppress the mail. You may be able to work around
this by running in --invisible mode.
If you want to support multidrop mode, and you can get access to your mailserver's sendmail.cf file, it's a good idea to add this rule:
H?l?Delivered-To: $h
This will cause the mailserver's sendmail to reliably write the appropriate envelope address into each message before fetchmail sees it, and tell fetchmail which header it is. With this change, multidrop mode should work reliably even when the Received header omits the envelope address (which will typically be the case when the message has multiple recipients). However it will still not distinguish the recipients, your only advantage is that no bounce will be sent if a message is BCC addressed to multiple users at your site. To fix even that problem, you might want to try the following hack, which is however untested and quite experimental:
H?J?Delivered-To: $u
Mmdrop, P=/usr/bin/procmail, F=lsDFMqSPfhnu9J,
S=EnvFromSMTP/HdrFromSMTP, R=EnvToSMTP/HdrToSMTP,
T=DNS/RFC822/X-Unix,
A=procmail -Y -a $u -d $h
For both hacks, you have to declare 'envelope
"Delivered-To:"' on the fetchmail side, to put the virtual
domain (e.g. 'domain.com') with RELAY permission into your access
file and to add a line reading 'domain.com
local:local-pop-user' for the first and 'domain.com
mdrop:local-pop-user' for the second hack to your
mailertable.
You will notice that if the mail already has a Delivered-To header, sendmail will not add another. Further, editing sendmail.cf directly is not very comfortable. Solutions for both problems can be found in Peter 'Rattacresh' Backes' 'hybrid' patch against sendmail. Have a look at it, you can find it in the contrib subdirectory.
Feel free to try Martijn Lievaart's detailed recipe in the contrib subdirectory of the fetchmail source distribution, it attempts to realize multidrop mailboxes with an external script.
If for some reason you are invoking sendmail via the mda option (rather than delivering to port 25 via smtp), don't forget to include the -i switch. Otherwise you will occasionally get mysterious delivery failures with a SIGPIPE as the sendmail instance dies. The problem is messages with a single dot at start of a text line.
Turn on the forcecr option; qmail's listener mode
doesn't like header or message lines terminated with bare
linefeeds.
(This information is thanks to Robert de Bath <robert@mayday.cix.co.uk>.)
If a mailhost is using the qmail package (see http://pobox.com/~djb/qmail.html) then, providing the local hosts are also using qmail, it is possible to set up one fetchmail link to be reliably collect the mail for an entire domain.
One of the basic features of qmail is the 'Delivered-To:' message header. Whenever qmail delivers a message to a local mailbox it puts the username and hostname of the envelope recipient on this line. The major reason for this is to prevent mail loops.
To set up qmail to batch mail for a disconnected site the ISP-mailhost will have normally put that site in its 'virtualhosts' control file so it will add a prefix to all mail addresses for this site. This results in mail sent to 'username@userhost.userdom.dom.com' having a 'Delivered-To:' line of the form:
Delivered-To: mbox-userstr-username@userhost.userdom.dom.com
A single host maildrop will be slightly simpler:
Delivered-To: mbox-userstr-username@userhost.dom.com
The ISP can make the 'mbox-userstr-' prefix anything they choose but a string matching the user host name is likely.
To use this line you must:
So far this reliably delivers messages to the correct machine of the local network, to deliver to the correct user the 'mbox-userstr-' prefix must be stripped off of the user name. This can be done by setting up an alias within the qmail MTA on each local machine. Simply create a dot-qmail file called '.qmail-mbox-userstr-default' in the alias directory (normally /var/qmail/alias) with the contents:
| ../bin/qmail-inject -a -f"$SENDER" "${LOCAL#mbox-userstr-}@$HOST"
Note this does require a modern /bin/sh.
Peter Wilson adds:
"My ISP uses "alias-unzzippedcom-" as the prefix, which means that I need to name my file ".qmail-unzzippedcom-default". This is due to qmail's assumption that a message sent to user-xyz is handled by the file ~user/.qmail-xyz (or ~user/.qmail-default)."
Luca Olivetti adds:
If you aren't using qmail locally, or you don't want to set up
the alias mechanism described above, you can use the option
'qvirtual "mbox-userstr-"' in your fetchmail config
file to strip the prefix from the local user name.
If you have rewrite on:
There is an RFC1123 requirement that MAIL FROM and RCPT TO addresses you pass to it have to be canonical (e.g. with a fully qualified hostname part). Therefore fetchmail tries to pass fully qualified RCPT TO addresses. But exim does not by default accept 'localhost' as a fully qualified domain. This can be fixed.
In exim.conf, add 'localhost' to your local_domains declaration if it's not already present. For example, the author's site at thyrsus.com would have a line reading:
local_domains = thyrsus.com:localhost
If you have rewrite off:
MAIL FROM is a potential problem if the MTAs upstream from your
fetchmail don't necessarily pass canonicalized From and Return-Path
addresses, and fetchmail's rewrite option is off. The
specific case where this has come up involves bounce messages
generated by sendmail on your mailer host, which have the
(un-canonicalized) origin address MAILER-DAEMON.
The right way to fix this is to enable the rewrite
option and have fetchmail canonicalize From and Return-Path
addresses with the mailserver hostname before exim sees them. This
option is enabled by default, so it won't be off unless you turned
it off.
If you must run with rewrite off, there is a switch
in exim's configuration files that allows it to accept domainless
MAIL FROM addresses; you will have to flip it by putting the
line
sender_unqualified_hosts = localhost
in the main section of the exim configuration file. Note that this will result in such messages having an incorrect domain name attached to their return address (your SMTP listener's hostname rather than that of the remote mail server).
Smail 3.2 is very nearly plug-compatible with sendmail, and may work fine out of the box.
We have one report that when processing multiple messages from a single fetchmail session, smail sometimes delivers them in an order other than received-date order. This can be annoying because it scrambles conversational threads. This is not fetchmail's problem, it is an smail 'feature' and has been reported to the maintainers as a bug.
Very recent smail versions require an
-smtp_hello_verify option in the smail config file.
This overrides smail's check to see that the HELO address is
actually that of the client machine, which is never going to be the
case when fetchmail is in the picture. According to RFC1123 an SMTP
listener must allow this mismatch, so smail's new behavior
(introduced sometime between 3.2.0.90 and 3.2.0.95) is a bug.
You may also need to say
-smtp_hello_broken_allow=127.0.0.1 in order for smail
to accept the "localhost" that fetchmail normally appends to
recipient addresses.
MMDF itself is difficult to configure, but it turns out that connecting fetchmail to MMDF's SMTP channel isn't that hard. You can read an MMDF recipe that describes replacing a UUCP link with fetchmail feeding MMDF.
The Lotus Notes SMTP gateway tries to deduce when it should convert \n to \r\n, but its rules are not the intuitive and correct-for-RFC822 ones. Use 'forcecr'.
The courier mta doesn't like RCPT addresses that look like
someone@localhost. Work around this with an
smtphost or smtpaddress.
vbmailshield's SMTP interpreter is broken. It doesn't understand RSET.
As a workaround, you can set batchlimit to 1 so RSET is never used.
Qualcomm's qpopper is probably the best-of-breed among POP3 servers, and is very widely deployed. Nevertheless, it has some problems which fetchmail exposes. We recommend using IMAP instead if at all possible. If you must talk to qpopper, here are some problems to be aware of:
Tony Tang <tony@atn.com.hk> reports that there is a bad intercation between fetchmail and qpopper 2.5.3 under Red Hat Linux versions 5.0 to 5.2, kernels 2.0.34 to 2.0.35. When fetching very large messages (over 700K) from 2.5.3, fetchmail will hang with a socket error.
This is probably not a fetchmail bug, but rather a symptom of some problem in the networking stack that qpopper's transmission pattern is tickling, as fetchpop (another Linux POP client) also displays the hang but Netscape running under Win95 does not. The problem can also be banished by upgrading to qpopper 3.0b1.
Versions of fetchmail from 4.4.2 through 4.4.7 had a bad interaction with Eudora qpopper versions 2.3 and later. See X5 for details. The solution is to upgrade your fetchmail.
It's been reliably reported that Exchange 2000's POP3 support is so broken that it's unusable. One symptom is that messages without a terminating newline get the POP3 message termination dot emitted -- you guessed it -- right after the last character of the message, with no terminating newline added. This will hang fetchmail or any other RFC-compliant server. IMAP is alleged to work OK, though.
Older versions of Exchange are semi-usable. They randomly drop attachments on the floor, though. Microsoft acknowledges this as a known bug and apparently has no plans to fix it.
Fetchmail using IMAP supports the proprietary NTLM mode used with M$ Exchange servers. To enable this, configure fetchmail with the --enable-NTLM option and recompile it. Specify a user option value that looks like 'user@domain': the part to the left of the @ will be passed as the username and the part to the right as the NTLM domain.
M$ Exchange violates the POP3 and IMAP RFCs. Its LIST command does not reveal the real sizes of mail in the pop mailbox, but the sizes of the compressed versions in the exchange mail database (thanks to Arjan De Vet and Guido Van Rooij for alerting us to this problem).
Fetchmail works with M$ Exchange, despite this brain damage. Two features are compromised. One is that the --limit option will not work right (it will check against compressed and not actual sizes). The other is that a too-small SIZE argument may be passed to your ESMTP listener, assuming you're using one (this should not be a problem unless the actual size of the message is above the listener's configured length limit).
Somewhat belatedly, I've learned that there's supposed to be a registry bit that can fix this breakage:
HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\MsExchangeIs\Parameters System\Pop3 Compatibility
This is a bitmask that controls the variations from the standard protocol. The bits defined are:
There's another one that may be useful to know about:
KEY_LOCAL_MACHINE\System\CurrentControlSet\Services\MsExchangeIs\Parameters System\Pop3 Performance
The Microsoft pod-person who revealed this information to me admitted that he couldn't find it anywhere in their public knowledge base.
Another specific problem we have seen with Exchange servers has as its symptom a response to LOGIN that says "NO Ambiguous Alias". Grant Edwards writes:
This means that Exchange Server is too f*&#ing stupid to figure out which mailbox belongs to you. Instead of actually keeping track of which inbox belongs to which user, it uses some half-witted, guess-o-matic heuristic to try to guess your mailbox name from your username.
In your case it doesn't work because your username maps to more than one mailbox. For some people it doesn't work because their username maps to zero mailboxes. This is yet another inept, lame, almost criminally negligent design decision from our friends in Redmond.
You've got several options:
But, the best option involves a tactical nuclear weapon (an old ASROC will do), pissing off a lot people who live downwind from Redmond, and your choice of any Linux, NetBSD, FreeBSD, or Solaris CD-ROM.
No special configuration is required, but OpenMail versions prior to 6.0 have an annoying bug similar to the big one in Microsoft Exchange. The message sizes it gives in the LIST are rounded to the nearest 1024 bytes. It also has a nasty habit of discarding headers it doesn't recognize, such as X- and Resent- headers.
As with M$ Exchange, the only real fix for these problems is to get a POP (or preferably IMAP) server that isn't brain-dead. OpenMail's project manager claims these bugs have been fixed in 6.0.
We've had a more recent report (December 2001) that the TOP command fails, returning only one line regrardless of its argument, on something identifying itself as "OpenMail POP3 interface".
The Novell GroupWise IMAP server would be better named GroupFoolish; it is (according to the designer of IMAP) unusably broken. Among other things, it doesn't include a required content length in its BODY[TEXT] response.
Fetchmail works around this problem, but we strongly recommend voting with your dollars for a server that isn't brain-dead. If you stick with code as shoddy as GroupWise seems to be, you will probably pay for it with other problems.
You can't. At least not if you want to be able to see attachments. InterChange has a bug similar to the MailMax server; it reports the message length with attachments but doesn't download them on TOP or RETR.
On Jan 9 2001, the people at InfiniteMail sent me mail informing me that their new 3.61.08 release of InterChange fixes this problem. I don't have any reports one way or the other yet.
You can't. At least not if you want to be able to see attachments. MailMax has a bug; it reports the message length with attachments but doesn't download them on TOP or RETR.
Also, we're told that TOP sometimes fails to retrieve the entire message even when enough lines have been specified. The MailMax developers have acknowledged this bug as of 4 May 2000, but there is no fix yet. If you must use this server, force RETR with the fetchall option.
The FTGate V2 server (and possibly older versions as well) has a
weird bug. It answers OK twice to a TOP request! Use the
fetchall option to force use of RETR and work around
this bug.
First, make sure your fetchmail has the RPA support compiled in.
Stock fetchmail binaries (such as you might get from an RPM) don't.
You can check this by looking at the output of fetchmail
-V; if you see the string "+RPA" after the version ID you're
good to go, otherwise you'll have to build your own from sources
(see the INSTALL file in the source distribution for
directions).
Give your CompuServe pass-phrase in lower case as your password. Add '@compuserve.com' to your user ID so that it looks like 'user <UserID>@compuserve.com', where <UserID> can be either your numerical userID or your E-mail nickname. An RPA-enabled fetchmail will automatically check for csi.com in the POP server's greeting line. If that's found, and your user ID ends with '@compuserve.com', it will query the server to see if it is RPA-capable, and if so do an RPA transaction rather than a plain-text password handshake.
Warning: the debug (-v -v) output of fetchmail will show your pass-phrase in Unicode!
These two .fetchmailrc entries show the difference between an RPA and non-RPA configuration:
# This version will use RPA
poll csi.com via "pop.site1.csi.com" with proto POP3 and options no dns
user "CSERVE_USER@compuserve.com" there with password "CSERVE_PASSWORD"
is LOCAL_USER here options fetchall stripcr
# This version will not use RPA
poll non-rpa.csi.com via "pop.site1.csi.com" with proto POP3 and options no dns
user "CSERVE_USER" there with password "CSERVE_POP3_PASSWORD"
is LOCAL_USER here options fetchall stripcr
You can get fetchmail to download the email for just one user from Demon Internet's POP3 server by giving it a username consisting of your Demon user name followed by your account name, with an at-sign between them.
For example, to download email for the user <philh@vision25.demon.co.uk>, you could use the following .fetchmailrc file:
set postmaster "philh"
poll pop3.demon.co.uk with protocol POP3:
user "philh@vision25" is philh
Demon Internet's SDPS service is an implementation of POP3. All messages have a Received: header added when they enter the maildrop, like this:
Received: from punt-1.mail.demon.net by mailstore for fred@xyz.demon.co.uk
id 899963657:10:27896:0; Thu, 09 Jul 98 05:54:17 GMT
To enable multi-drop mode you need to tell fetchmail that 'mailstore' is the name of the host which accepted the mail, and let it know the hostname part(s) of your E-mail address. The following example assumes that your hostname is xyz.demon.co.uk, and that you have also bought "mail forwarding" for the domain my-company.co.uk (in which case your MTA must also be configured to accept mail sent to user@my-company.co.uk)
poll pop3.demon.co.uk proto pop3 aka mailstore no dns:
localdomains xyz.demon.co.uk my-company.co.uk
user xyz is *
Note that Demon may delete mail on the server which is more than 30 days old; see their POP3 page for details.
There's a different way to do multidrop. It's not necessary on Demon Internet, since fetchmail can parse Received addresses, but the person who implemented this didn't know that. It may be useful if Demon Internet ever changes mail transports.
SDPS includes a non-standard extension for retrieving the envelope of a message (*ENV), which fetchmail optionally supports if compiled with the --enable-SDPS option. If you have it, the first line of the fetchmail -V response will include the string "+SDPS".
Once you have SDPS compiled in, fetchmail in POP3 mode will automatically detect when it's talking to a Demon Internet host in multidrop mode, and use the *ENV extension to get an envelope To address.
The autodetection works by looking at the hostname in the POP3 greeting line; if you're accessing Demon Internet through a proxy it may fail. To force SDPS mode, pick "sdps" as your protocol.
Enable 'fetchall'. A user reports that the 2.2
version of USA.NET's POP server reports that you must use the
'fetchall' option to make sure that all of the mail is
retrieved, otherwise some may be left on the server. This is almost
certainly a server bug.
The usa.net servers (at least in their 2.2 version, June 1998)
don't handle the TOP command properly, either. Regardless of the
argument you give it, they retrieve only about 10 lines of the
message. Fetchmail normally uses TOP for message retrieval in order
to avoid marking messages seen, but 'fetchall' forces
it to use RETR instead.
Also, we're told USA.NET adds a ton of hops to your messages. You may need to raise the MaxHopCount parameter in your sendmail.cf to avoid having fetched mail rejected.
(Note: Other failure modes have been reported on usa.net's servers. They seem to be chronically flaky. We recommend finding another provider.)
Nathan Cutler reports that the the mail.geocities.com POP3 servers fail to include the first Received line of the message in the send to fetchmail. This can solve problems if your MUA interprets Received continuations as body lines and doesn't parse any of the following headers.
Workaround is to use "mda" keyword or "-mda" switch:
mda "sed -e '1s/^\t/Received: /' | formail | /usr/bin/procmail -d <user>"
Replace \t with exactly one tabulation character.
You should also consider using "fetchall" option because Geocities' servers sometimes think that the first 45 messages have already been read.
Fix: Get an email provider that doesn't suck. The pop-up ads on Geocities are lame, you should boycott them anyway.
You can't directly. But you can use fetchmail with hotmail or lycos webmail with the help of the HotWayDaemon daemon. You don't even need to install hotwayd as a daemon in inetd.conf but can use it as a plugin. Your configuration should look like this:
poll localhost protocol pop3 tracepolls plugin "/usr/local/sbin/hotwayd -l 0 -p yourproxy:yourproxyport" username "youremail@hotmail.com" password "yourpassword" fetchall
As a second option you may consider using gotmail.
You can't. MSN uses something that looks like POP3, except the authentication part is nonstandard. And of course they don't document it, so nobody but their Windows clients can speak it.
This is a customer lock-in tactic; we recommend boycotting MSN as the only appropriate response.
As of 5.0.8, we have support for the client side of NTLM authentication. It's possible this may enable fetchmail to talk to MSN; if so, somebody should report it so this FAQ can be corrected.
The SpryNet POP3 servers mark a message queried with TOP as
seen. This means that if your connection drops in mid-message, it
may end up invisibly stuck on your mail spool. Use the
fetchall flag to ensure that it's recovered on the
next cycle.
Stock fetchmail will work with a comcast.net server...but the Maillennium POP3 server comcat uses seems to have an 80K limit on the length of downloaded messages if you use POP3 TOP to retrieve. Anything larger is silently truncated. Don't mistake this for a fetchmail bug. (Reported July 2003.)
Workaround: use the fetchall option.
Giuseppe Guerini added a --with-socks option that supports linking with socks library. If you specify the value of this option as "yes", the configure script will try to find the Rconnect library and set the makefile up to link it. You can also specify a directory containing the Rconnect library.
Alan Schmitt has added a similar --with-socks5 option that may work better if you have a recent version of the SOCKS library.
To use fetchmail with IPv6, you need a system that supports IPv6, the "Basic Socket Interface Extensions for IPv6" (RFC 2133).
The NRL IPv6+IPsec software distribution can be obtained from: http://web.mit.edu/network/isakmp
More information on using IPv6 with Linux can be obtained from:
Use the plugin option. This is dead simple with IMAP:
plugin "ssh %h /usr/sbin/imapd"
You may have to use a different absolute pathname, whatever the location of imapd on your mailserver is. This option tells fetchmail that instead of opening a connection on the server's port 143 and doing standard IMAP authentication, fetchmail should ssh to the server and run imapd, using the more secure ssh authentication (as well as getting ssh's end-to-end encryption). Most IMAP daemons will detect that they've been called from the command line and assume the connection is preauthenticated.
POP3 daemons aren't quite as smart. They won't know they are preauthenticated in this mode, so you'll actually have to ship your password. It will be under ssh encryption, though, so that shouldn't be a problem.
Fetchmail can use RFC1731 GSSAPI authorization to safely identify you to your IMAP server, as long as you can share Kerberos V credentials with your mail host and you have a GSSAPI-capable IMAP server. UW-IMAP (available via FTP at ftp.cac.washington.edu) is the only one I'm aware of and the one I recommend anyway for other reasons. You'll need version 4.1-FINAL or greater though, and it has to have GSS support compiled in.
Neither UW-IMAP nor fetchmail compile in support for GSS by
default, since it requires libraries from the Kerberos V
distribution (available via FTP at athena-dist.mit.edu).
If you have these, compiling in GSS support is simple: add a
--with-gssapi=[/path/to/krb5/root] option to
configure. For instance, I have all of my Kerberos V libraries
installed under /usr/krb5 so I run configure
--with-gssapi=/usr/krb5
Setting up Kerberos V authentication is beyond the scope of this FAQ (you may find Jim Rome's paper How to Kerberize your site helpful), but you'll at least need to add a credential for imap/[mailhost] to the keytab of the mail server (IMAP doesn't just use the host key). Then you'll need to have your credentials ready on your machine (cf. kinit).
After that things are very simple. Set your protocol to imap-gss in your .fetchmailrc, and omit the password, since imap-gss doesn't need one. You can specify a username if you want, but this is only useful if your mailbox belongs to a username different from your Kerberos principal.
Now you don't have to worry about your password appearing in cleartext in your .fetchmailrc, or across the network.
You'll need to have the OpenSSL libraries installed. Configure with --with-ssl. If you have the OpenSSL libraries installed in the default location (/usr/local/ssl) ths will suffice. If you have them installed in a non-default location, you'll need to specify it as an argument to --with-ssl after an equal sign.
Note that there is a known bug in the implementation of SSL_peek under OpenSSL versions 0.9.5 and older that fetchmail occasionally tripped over, causing hangs. It is recommended that you install 0.9.6 or later.
Fetchmail binaries built this way support ssl,
sslkey, and sslcert options that control
SSL encryption. You will need to have an SSL-enabled mailserver to
use these options. See the manual page for details and some words
of care on the limited security provided.
If your open OpenSSL session dies with a message that complains "PRNG not seeded", update or improve your operating system. This means that the OpenSSL library on your machine has been unable to locate a source of random bits from which to seed its random-number generator; normally these come from the /dev/urandom, and this message probably means your OS doesn't have that device.
An interactive program could seed the random number generator from keystroke timings or some other form of user input. Because fetchmail is primarily designed to run forever as a background daemon, that option is not available in this case.
If you don't have the libraries installed, but do have the OpenSSL utility toolkit, something like this may work:
poll MYSERVER port 993 plugin "openssl s_client -connect %h:%p"
protocol imap username MYUSERNAME password MYPASSWORD
You should note that SSL is only secure against a "man-in-the-middle" attack if the client is able to verify that the peer's public key is the correct one, and has not been substituted by an attacker. fetchmail can do this in one of two ways: by verifying the SSL certificate, or by checking the fingerprint of the peer's public key.
There are three parts to SSL certificate verification: checking that the domain name in the certificate matches the hostname you asked to connect to; checking that the certificate expiry date has not passed; and checking that the certificate has been signed by a known Certificate Authority (CA). This last step takes some preparation, as you need to install the root certificates of all the CA's which you might come across.
The easiest way to do this is using the root CA keys supplied in the OpenSSL distribution, which means you need to download and unpack the source tarball from www.openssl.org. Once you have done that:
mkdir /etc/ssl/certscp *.pem /etc/ssl/certs/$dir="/etc/ssl"Now in .fetchmailrc, set option sslcertpath to point to this directory:
poll pop3.example.com proto pop3 uidl no dns user foobar@example.com password xyzzy is foobar ssl sslcertpath /etc/ssl/certs
If the server certificate has not been signed by a known CA (e.g. it is a self-signed certificate), then this certificate validation will always fail.
Certificate verification is always attempted. If it fails, by default a warning is printed but the connection carries on (which means you are not protected against attack). If your server's certificate has been properly set up and verifies correctly, then add the "sslcertck" option to enforce validation. If your server doesn't have a valid certificate though (e.g. it has a self-signed certificate) then it will never verify, and the only way you can protect yourself is by checking the fingerprint.
To check the peer fingerprint: first use fetchmail -v once to connect to the host, at a time when you are pretty sure that there is no attack in progress (e.g. you are not traversing any untrusted network to reach the server). Make a note of the fingerprint shown. Now embed this in your .fetchmailrc using the sslfingerprint option: e.g.
poll pop3.example.com proto pop3 uidl no dns user foobar@example.com password xyzzy is foobar ssl sslfingerprint "67:3E:02:94:D3:5B:C3:16:86:71:37:01:B1:3B:BC:E2"
When you next connect, the public key presented by the server will be verified against the fingerprint given. If it's different, it may mean that a man-in-the-middle attack is in progress - or it might just mean that the server changed its key. It's up to you to determine which has happened.
Some servers advertise STLS (POP3) or STARTTLS (IMAP), and fetchmail will automatically attempt TLS negotiation if SSL was enabled at compile time. This can however cause problems if the upstream didn't configure his certificates properly.
In order to prevent fetchmail from trying TLS (STLS, STARTTLS) negotiation, add this option:
sslproto ssl23
This restricts fetchmail's SSL/TLS protocol choice from the default "SSLv2, SSLv3, TLSv1" to the two SSL variants, disabling TLSv1. Note however that this causes the connection to be unencrypted unless an encrypting "plugin" is used or SSL is requested explicitly.
Fetchmail itself is probably working, but your SMTP port 25 listener is down or inaccessible.
The first thing to check is if you can telnet to port 25 on your smtp host (which is normally 'localhost' unless you've specified an smtp option in your .fetchmailrc or on the command line) and get a greeting line from the listener. If the SMTP host is inaccessible or the listener is down, fix that first.
In Red Hat Linux 6.x, SMTP is disabled by default. To fix this, set "DAEMON=yes" in your /etc/sysconfig/sendmail file, then restart sendmail by running "/sbin/service sendmail restart".
If the listener seems to be up when you test with telnet, the most benign and typical problem is that the listener had a momentary seizure due to resource exhaustion while fetchmail was polling it -- process table full or some other problem that stopped the listener process from forking. If your SMTP host is not 'localhost' or something else in /etc/hosts, the fetchmail glitch could also have been caused by transient nameserver failure.
Try running fetchmail -v again; if it succeeds, you had one of these kinds of transient glitch. You can ignore these hiccups, because a future fetchmail run will get the mail through.
If the listener tests up, but you have chronic failures trying to connect to it anyway, your problem is more serious. One way to work around chronic SMTP connect problems is to use --mda. But this only attacks the symptom; you may have a DNS or TCP routing problem. You should really try to figure out what's going on underneath before it bites you some other way.
We have one report (from toby@eskimo.com) that you can sometimes
solve such problems by doing an smtp declaration with
an IP address that your routing table maps to something other than
the loopback device (he used ppp0).
We also have a report that this error can be caused by having an /etc/hosts file that associates your client host name with more than one IP address.
It's also possible that your DNS configuration isn't looking at
/etc/hosts at all. If you're using libc5, look at
/etc/resolv.conf; it should say something like:
order hosts,bind
so your /etc/hosts file is checked first. If you're
running GNU libc6, check your /etc/nsswitch.conf file.
Make sure it says something like
hosts: files dns
again, in order to make sure /etc/hosts is seen
first.
If you have a hostname set for your machine, and this hostname does not appear in /etc/hosts, you will be able to telnet to port 25 and even send a mail with rcpt to: user@host-not-in-/etc/hosts, but fetchmail can't seem to get in touch with sendmail, no matter what you set smtpaddress to.
We had another report from a Linux user of fetchmail 2.1 who solved his SMTP connection problem by removing the reference to -lresolv from his link line and relinking. Apparently in some older Linux distributions the libc bind library version works better.
As of 2.2, the configure script has been hacked so the bind library is linked only if it is actually needed. So under Linux it won't be, and this particular cause should go away.
(I hear this one from people who have run into the blank-line problem in X1.)
Try sending yourself test mail and retrieving it using the
command-line options '-k -m cat'. This will dump
exactly what fetchmail retrieves to standard output (plus the
Received line fetchmail itself adds to the headers).
If the dump doesn't match what shows up in your mailbox when you configure an MDA, your MDA is mangling the message. If it doesn't match what you sent, then fetchmail or something on the server is broken.
This is usually reported from AIX or Ultrix, but has even been
known to happen on Linuxes without a recent version of
flex installed. The problem appears to be a result of
building with an archaic version of lex.
Workaround: fix the syntax of your .fetchmailrc file.
Fix: build and install the latest version of flex from the Free Software Foundation. An FSF mirror site will help you get it faster.
We've had this reported to us under Linux using libc-5.4.17 and gcc-2.7.2. It does not occur with libc-5.3.12 or earlier versions.
Workaround: link with GNU malloc rather than the stock C library malloc.
We're told there is some problem with the malloc() code in that version which makes it fragile in the presence of multiple free() calls on the same pointer (the malloc arena gets corrupted). Unfortunately it appears from doing gdb traces that whatever free() calls producing the problem are being made by the C library itself, not the fetchmail code (they're all from within fclose, and not an fclose called directly by fetchmail, either).
We have one report from a SunOS 4.1.4 user that trying to run fetchmail in detached daemon mode doesn't work, but that using the same options with -N (nodetach) is OK. We have another report of similar behavior from one Linux user, but many other Linux users reportt no problem.
If this happens, you have a specific portability problem with the code in daemon.c that detaches and backgrounds the daemon fetchmail. The isolated Linux case has been chased down to a failure in dup(2) that may reflect a glibc bug.
As a workaround, you can start fetchmail with -N and an ampersand to background it. A Sun user recommends this:
(fetchmail --nodetach <other params> &)
The extra pair of parens is significant --- it makes sure that the process detaches from the initial shell (one more shell is started and dies immediately, detaching fetchmail and making it child of PID 1). This is important when you start fetchmail interactively and than quit interactive shell. The line above makes sure fetchmail lives after that!
Check the MTU value in your PPP interface reported by
/sbin/ifconfig. If it's over 600, change it in your
PPP options file. (/etc/ppp/options on my box). Here
are option values that work:
mtu 552 mru 552
Another circumstance that can trigger this is if you are polling a virtual-mail-server name that is round-robin connected to different actual servers, so you get different IP addresses on different poll cycles. To work around this, change the poll name either to the real name of one of the servers in the ring or to a corresponding IP address.
In RH 6.0, the HOME value in the boot-time root environment changed from /root to / as the result of a change in init. Move your .fetchmailrc or use a -f option to explicitly point at the file. (Oddly, a similar problem has been reported from Debian systems.)
There's a TCP/IP stalling problem under Redhat 6.0 (and possibly other recent Linuxes) that can cause this symptom. Brian Boutel writes:
TCP timestamps are turned on on my Linux boxes (I assume it's now the default). This uses 12 extra bytes per segment. When the tcp connection starts, the other end agrees a MSS of 1460, and then fragments 1460 byte chunks into 1448 and 12, because is is not allowing for the timestamp.
Then, for reasons I can't explain, it waits a long time (typically 2 minutes) after the ack is sent before sending the next (fragmented) packet. Turning off tcp timestamps avoids the fragmentation and restores normal behaviour. To do this, [execute]
echo 0 > /proc/sys/net/ipv4/tcp_timestamps
I'm still unclear about the details of why this is happening. At least [now] I am now getting good performance and no queue blocking.
This is probably a general networking issue. Sending a "RETR" command will cause the server to start sending large amounts of data, which means large packets. If your networking layer has a packet-fragmentation problem, that's where you'll see it.
This probably means you have an mda option. Your
MDA is croaking while being passed a message. Best fix is to remove
the mda option and pass mail to your port 25 SMTP
listener.
If for some reason you are invoking sendmail via the mda option (rather than delivering to port 25 via smtp), don't forget to include the -i switch. Otherwise you will occasionally get mysterious delivery failures with a SIGPIPE as the sendmail instance dies. The problem is messages with a single dot at start of a text line.
Your POP3 server is broken. You can work around this with the declaration auth password in your .fetchmailrc.
/etc/services file (or other
services database) contains the necessary service entries. If you
cannot fix the services database, use the --service option and give the
numeric port address. Common port addresses are:| service | port |
|---|---|
| IMAP | 143 |
| IMAP+SSL | 993 |
| POP3 | 110 |
| POP3+SSL | 995 |
Your problem may be with pppd's 'demand' option. We have a report that fetchmail doesn't play well with it, but works with pppd if 'demand' is turned off. We have no idea why this is.
The symptom: 'fetchmail -v' retrieves mail fine, but appears to hang after sending the MAIL FROM command
SMTP> MAIL FROM: <someone@somewhere>
The hang is actually occuring when sendmail looks up a sender's address in DNS. The problem isn't in fetchmail but in the configuration of sendmail. You must enable the 'nodns' and 'nocanonify' features of sendmail.
Here was my fix for RedHat 7.2:
FEATURE(nodns) FEATURE(nocanonify)
# m4 sendmail-mine.cf > /etc/sendmail.cf
For more details consult the file /usr/share/sendmail-cf/README.
The symption: 'fetchmail -v' retrieves the first few messages, but hangs returning:
fetchmail: SMTP< 550 5.0.0 Access denied fetchmail: SMTP> RSET fetchmail: SMTP< 250 2.0.0 Reset state .......fetchmail: flushed fetchmail: POP3> DELE 1 fetchmail: POP3< +OK marked deleted
Check and see if you're allowing sendmail connections through TCP wrappers.
Adding 'sendmail : 127.0.0.1' to /etc/hosts.allow could solve this problem.
Maybe you have a .forward or alias set up that you've forgotten about. You should probably remove it.
Or maybe you're trying to run fetchmail in multidrop mode as root without a .fetchmailrc file. This doesn't do what you think it should; see question C1.
Or you may not be connecting to the SMTP listener. Run fetchmail -v and see R1.
Or you may have your local user set incorrectly. In the following line
user 'remoteuser' there with password '*' is 'localuser' here
make sure that 'localuser' does exist and can receive mail.
One POP3 daemon used in the Berkeley Unix world that reports itself as POP3 version 1.004 actually throws the queue away. 1.005 fixed that. If you're running this one, upgrade immediately. (It also truncates long lines at column 1024)
Many POP servers, if an interruption occurs, will restore the whole mail queue after about 10 minutes. Others will restore it right away. If you have an interruption and don't see it right away, cross your fingers and wait ten minutes before retrying.
Some servers (such as Microsoft's NTMail) are mis-designed to
restore the entire queue, including messages you have deleted. If
you have one of these and it flakes out on you a lot, try setting a
small --fetchlimit value. This will result in more IP
connects to the server, but will mean it actually executes changes
to the queue more often.
Qualcomm's qpopper, used at many BSD Unix sites, is better behaved. If its connection is dropped, it will first execute all DELE commands as though you had issued a QUIT (this is a technical violation of the POP3 RFCs, but a good idea in a world of flaky phone lines). Then it will re-queue any message that was being downloaded at hangup time. Still, qpopper may require a noticeable amount of time to do deletions and clean up its queue. (Fetchmail waits a bit before retrying in order to avoid a 'lock busy' error.)
Fetchmail only sends a delete mail request to the server when
either (a) it gets a positive delivery acknowledgment from the SMTP
listener, or (b) it gets one of the spam-filter errors (see the
description of the antispam> option) from the
listener. No interrupt can cause it to lose mail.
However, IMAP2bis has a design problem in that its normal fetch command marks a message 'seen' as soon as the fetch command to get it is sent down. If for some reason the message isn't actually delivered (you take a line hit during the download, or your port 25 listener can't find enough free disk space, or you interrupt the delivery in mid-message) that 'seen' message can lurk invisibly in your server mailbox forever.
Workaround: add the 'fetchall' keyword to your
fetch options.
Solution: switch to an IMAP4 server.
Somehow your fetchmail is never recognizing the hostname part of recipient names it parses out of To/Cc/envelope-header lines as matching the name of the mailserver machine. To check this, run fetchmail in foreground with -v -v on. You will probably see a lot of messages with the format "line rejected, %s is not an alias of the mailserver" or "no address matches; forwarding to %s."
These errors usually indicate some kind of DNS configuration problem either on the server or your client machine.
The easiest workaround is to add a 'via' option (if
necessary) and add enough aka declarations to cover all of your
mailserver's aliases, then say 'no dns'. This will
take DNS out of the picture (though it means mail may be
uncollected if it's sent to an alias of the mailserver that you
don't have listed).
It would be better to fix your DNS, however. DNS problems can hurt you in lots of ways, for example by making your machines intermittently or permanently unreachable to the rest of the net.
Occasionally these errors indicate the sort of header-parsing problem described in M7.
A lot of people want to use fetchmail as a poor man's internetwork mail gateway, picking up mail accumulated for a whole domain in a single server mailbox and then routing based on what's in the To/Cc/Bcc lines.
In general, this is not really a good idea. It would be smarter to just let the mail sit in the mailserver's queue and use fetchmail's ETRN or ODMR modes to trigger SMTP sends periodically (of course, this means you have to poll more frequently than the mailserver's expiration period). If you can't arrange this, try setting up a UUCP feed.
If neither of these alternatives is available, multidrop mode
may do (though you are going to get hurt by some mailing
list software; see the caveats under THE USE AND ABUSE OF MULTIDROP
MAILBOXES on the man page). If you want to try it, the way to do it
is with the 'localdomains' option.
In general, if you use localdomains you need to make sure of two other things:
1. You've actually set up your .fetchmailrc entry to invoke multidrop mode.
Many people set a 'localdomains' list and then
forget that fetchmail wants to see more than one name (or the
wildcard '*') in a 'here' list before it will do
multidrop routing.
2. You may have to set 'no envelope'.
Normally, multidrop mode tries to deduce an envelope address from a message before parsing the To/Cc/Bcc lines (this enables it to avoid losing to mailing list software that doesn't put a recipient address in the To lines).
Some ways of accumulating a whole domain's messages in a single
server mailbox mean it all ends up with a single envelope address
that is useless for rerouting purposes. You may have to set
'no envelope' to prevent fetchmail from being
bamboozled by this.
Check also answer T1 on a reliable way to do multidrop delivery if your ISP (or your mail redirection provider) is using qmail.
This isn't fetchmail's fault. Check your mailing list. If the list expansion includes yourself or anybody else at your mailserver (that is, not on the client side) you've created a mail loop. Just chop the host part off any local addresses in the list.
If you use sendmail, you can check the list expansion with
sendmail -bv.
We have one report from a Linux user (not the same one as in R1!) who solved this problem by removing the reference to -lresolv from his link line and relinking. Apparently in some older Linux distributions the libc5 bind library version works better.
As of 2.2, the configure script has been hacked so the bind library is linked only if it is actually needed. So under Linux it won't be, and this problem should go away.
Use the 'aka' option to pre-declare as many of your
mailserver's DNS names as you can. When an address's host part
matches an aka name, no DNS lookup needs to be done to check
it.
If you're sure you've pre-declared all of your mailserver's DNS
names, you can use the 'no dns' option to prevent
other hostname parts from being looked up at all.
Sometimes delays are unavoidable. Some SMTP listeners try to call DNS on the From-address hostname as a way of checking that the address is valid.
In order for sendmail to execute the command strings in the majordomo alias file, it is necessary for sendmail to think that the mail it receives via SMTP really is destined for a local user name. A normal virtual-domain setup results in delivery to the default mailbox, rather than expansion through majordomo.
Michael <michael@bizsystems.com> gave us a recipe for dealing with this case that pairs a run control file like this:
poll your.pop3.server proto pop3:
no envelope no dns
localdomains virtual.localdomain1.com virtual.localdomain2.com ...
user yourISPusername is root * here,
password yourISPpassword fetchall
with a hack on your local sendmail.cf like this:
############################################# # virtual info, local hack for ruleset 98 # ############################################# # domains to treat as direct mapped local domain CVvirtual.localdomain1.com virtual.localdomain2.com ... --------------------------- in ruleset 98 add ------------------------- # handle virtual users R$+ <@ $=V . > $: $1 < @ $j . > R< @ > $+ < @ $=V . > $: $1 < @ $j . > R< @ > $+ $: $1 R< error : $- $+ > $* $#error $@ $1 $: $2 R< $+ > $+ < @ $+ > $: $>97 $1
This ruleset just strips virtual domain names off the addresses of incoming mail. Your sendmail must be 8.8 or newer for this to work. Michael says:
I use this scheme with 2 virtual domains and the default ISP user+domain and service about 30 mail accounts + majordomo on my inside pop3 server with fetchmail and sendmail 8.83
It may happen that you're getting what appear to be well-formed sendmail Received headers, but fetchmail can't seem to extract an envelope address from them. There can be a couple of reasons for this.
First, fetchmail might be looking at the wrong Received header. Normally it looks only on the first one it sees, on the theory that that one was last added and is going to be the one containing your mailserver's theory of who the message was addressed to.
Some (unusual) mailserver configurations will generate extra
Received lines which you need to skip. To arrange this, use the
optional skip prefix argument of the 'envelope' option; you may
need to say something like 'envelope 1 Received' or
'envelope 2 Received'.
When fetchmail parses a Received line that looks like
Received: from send103.yahoomail.com (send103.yahoomail.com [205.180.60.92])
by iserv.ttns.net (8.8.5/8.8.5) with SMTP id RAA10088
for <ksturgeon@fbceg.org>; Wed, 9 Sep 1998 17:01:59 -0700
it checks to see if 'iserv.ttns.net' is a DNS alias of your mailserver before accepting 'ksturgeon@fbceg.org' as an envelope address. This check might fail if your DNS were misconfigured, or if you were using 'no dns' and had failed to declare iserv.ttns.net as an alias of your server.
It's a consequence of multidrop. What's happening is that you have N users subscribed to the same list. The list software sends N copies, not knowing they will end up in the same multidrop box. Since they are both locally addressed to all N users, fetchmail delivers N copies to each user.
Fetchmail tries to eliminate adjacent duplicate messages in a multidrop mailbox. However, this logic depends on the message-ID being identical in both copies. It also depends on the two copies being adjacent in the server mailbox. The former is usually the case, but the latter condition sometimes fails in a timing-dependent way if the server was processing multiple incoming mail streams.
I could eliminate this problem by keeping a list of all message-IDs received during a poll so far and dropping any message that matches a seen mail ID. The trouble is that this is an O(N**2) operation that might significantly slow down the retrieval of large mail batches.
What's probably happening is that the POP/IMAP daemon on your mailserver is inserting a non-RFC822 header (like X-POP3-Rcpt:) and something in your delivery path (most likely an old version of the deliver program, which sendmail often calls to do local delivery) is failing to recognize it as a header.
This is not fetchmail's problem. The first thing to try is
installing a current version of deliver. If this doesn't
work, try to figure out which other program in your mail path is
inserting the blank line and replace that. If you can't do either
of these things, pick a different MDA (such as procmail) and
declare it with the 'mda' option.
First, see X1. This is quite probably the same problem (X-POP3-Rcpt header or something similar being inserted by the server and choked on by an old version of deliver).
The O'Reilly sendmail book does warn that IDA sendmail doesn't process X- headers correctly. If this is your problem, all I can suggest is replacing IDA sendmail, because it's broken and not RFC822 conformant.
If you know the messages aren't split in your server mailbox, then this is a problem with your POP/IMAP server, your client-side SMTP listener or your local delivery agent. Fetchmail cannot split messages.
Some POP server daemons ignore Content-Length headers and split messages on From lines. We have one report that the 2.1 version of the BSD popper program (as distributed on Solaris 2.5 and elsewhere) is broken this way.
You can test this. Declare an mda of 'cat' and send yourself one piece of mail containing "From" at start of a line. If you see a split message, your POP/IMAP server is at fault. Upgrade to a more recent version.
Sendmail and other SMTP listeners don't split RFC822 messages either. What's probably happening is either sendmail's local delivery agent or your mail reader are not quite RFC822-conformant and are breaking messages on what it thinks are Unix-style From headers. You can figure out which by looking at your client-side mailbox with vi or more. If the message is already split in your mailbox, your local delivery agent is the problem. If it's not, your mailreader is the problem.
If you can't replace the offending program, take a look at your sendmail.cf file. There will likely be a line something like
Mlocal, P=/usr/bin/procmail, F=lsDFMShP, S=10, R=20/40, A=procmail -Y -d $u
describing your local delivery agent. Try inserting the 'E' option in the flags part (the F= string). This will make sendmail turn each dangerous start-of-line From into a >From, preventing programs further downstream from acting up.
The first thing you need to do is pin down what program is doing the mangling. We don't like getting bug reports about fetchmail that are actually due to some other program's malfeasance, so please go through this diagnostic sequence before sending us a complaint.
There are five possible culprits to consider, listed here in the order they pass your mail:
mda.Often it happens that fetchmail itself is OK, but using it exposes pre-existing bugs in your downstream software, or your downstream software has a bad interaction with POP/IMAP. You need to pin down exactly where the message is being garbled in order to deduce what is actually going on.
The first thing to do is send yourself a test message, and retrieve it with a .fetchmailrc entry containing the following (or by running with the equivalent command-line options):
mda "cat >MBOX" keep fetchall
This will capture what fetchmail gets from the server, except
for (a) the extra Received header line fetchmail prepends, (b)
header address changes due to rewrite, and (c) any
end-of-line changes due to the forcecr and
stripcr options. MBOX will in fact contain what
programs downstream of fetchmail see.
The most common causes of mangling are bugs and misconfigurations in those downstream programs. If MBOX looks unmangled, you will know that is what is going on and that it is not fetchmail's problem. Take a look at the other FAQ items in this section for possible clues about how to fix your problem.
If MBOX looks mangled, the next thing to do is compare it with
your actual server mailbox (if possible). That's why you specified
keep, so the server copy would not be deleted. If your
server mailbox looks mangled, programs upstream of your server
mailbox are at fault. Unfortunately there is probably little you
can do about this aside from complaining to your site postmaster,
and nothing at all fetchmail can do about it!
More likely you'll find that the server copy looks OK. In that case either the POP/IMAP server or fetchmail is doing the mangling. To determine which, you'll need to telnet to the server port and simulate a fetchmail session yourself. This is not actually hard (both POP3 and IMAP are simple, text-only, line-oriented protocols) but requires some attention to detail. You should be able to use a fetchmail -v log as a model for a session, but remember that the "*" in your LOGIN or PASS command dump has to be replaced with your actual password.
The objective of manually simulating fetchmail is so you can see exactly what fetchmail sees. If you see a mangled message, then your server is at fault, and you probably need to complain to your mailserver administrators. However, we like to know what the broken servers are so we can warn people away from them. So please send us a transcript of the session including the mangling and the server's initial greeting line. Please tell us anything else you think might be useful about the server, like the server host's operating system.
If your manual fetchmail simulation shows an unmangled message, congratulations. You've found an actual fetchmail bug, which is a pretty rare thing these days. Complain to us and we'll fix it. Please include the session transcript of your manual fetchmail simulation along with the other things described in the FAQ entry on reporting bugs.
This may happen in versions of fetchmail after 4.4.1 and before 4.4.8. Versions after 4.4.1 use POP3's TOP command rather than RETR, in order to avoid marking the message seen (leaving it unseen is helpful for later recovery if you lose your connection in the middle of a retrieval).
Versions of fetchmail from 4.4.2 through 4.4.7 had a bad interaction with Eudora qpopper versions 2.3 and later. The TOP bounds check was fooled by an overflow condition in the TOP argument. Decrementing the TOP argument in 4.4.7 fixed this.
Fix: Upgrade to a later version of fetchmail.
Workaround: set the fetchall option. Under POP3
this has the side effect of forcing RETR use.
Fetchmail doesn't discard attachments; fetchmail doesn't have any idea that attachments are there. Fetchmail treats the body of each message as an uninterpreted byte stream and passes it through without alteration. If you are not receiving attachments through fetchmail, it is because your mailserver is not sending them to you.
The fix for this is to replace your mailserver with one that works. If its operating system makes this difficult, you should replace its operating system with one that works. Windows- and NT-based POP servers seem especially prone to mangle attachments. If you are running one of these, replacing your server with a Unix machine is probably the only effective solution.
We've had sporadic reports of problems with Microsoft Exchange and Outlook servers. These sometimes randomly fail to ship attachments to your client. This is a known bug, acknowledged by Microsft.
They may also mangle the attachments they do pass through. If you see unreadable attachments with a ContentType of "application/x-tnef", you're having this problem. The TNEF utility may help.
The Mail Max POP3 server and the InterChange and Imail IMAP servers are known to simply drop MIME attachments when uploading messages.
We've also had a report that Lotus Notes sometimes trashes the MIME type of messages. In particular, it seems to modify MIME headers of type application/pdf, mangling the type to application/octet-stream. It may corrupt other MIME types as well.
The IMAP service of Lotus Domino has a known bug in the way it generates MIME Content-type headers (observed on Lotus Domino 5.0.2b). It's a subtle one that doesn't show up when Netscape Messenger and other clients use a FETCH BODY[] to grab the whole message. When fetchmail uses FETCH RFC822.HEADER and FETCH RFC822.TEXT to get first the header and then the body, Domino generates different Boundary tags for each part, .e.g. one tag is declared in the Content-type header and another is used to separate the MIME parts in the body. This doesn't work. (I have heard a rumor that this bug is scheduled to be fixed in Domino release 6; you can find a workaround at contrib/domino.)
Rob Funk explains: Unfortunately there also remain many mail user agents that don't write correct MIME messages. One big offender is Sun MailTool attachments, which are formatted enough like MIME that some programs could get confused; these are generated by the mailtool and dtmail programs (the mail programs in Sun's OpenWindows and CDE environments).
One solution to problems related to misformatted MIME attachments is the emil program; see its tutorial file at that site for details on emil. It is useful for converting character sets, attachment encodings, and attachment formats. At this writing, emil does not appear to have been maintained since a patch to version 2.1.0beta9 in late 1997, but it is still useful.
One good way of using emil is from within procmail. You can have procmail look for signs of problematic message formatting, and pipe those messages through emil to be fixed. emil will not always be able to fix the problem, in which case the message is unchanged.
A possible rule to be inserted into a .procmailrc file for using emil would be:
:0HB
* 1^1 ^Content-Type: \/X-sun[^;]*
* 1^1 ^Content-Type: \/application/mac-binhex[^;]*
* 1^1 ^Content-Transfer-Encoding: \/x-binhex[^;]*
* 1^1 ^Content-Transfer-Encoding: \/x-uuencode[^;]*
{
LOG="Converting $MATCH
"
:0fw
| emil -A B -T Q -B BA -C iso-8859-1 -H Q -F MIME \
| gawk '{gsub(/\r\n?/,"\n");print $0}'
}
The "1^1" in the conditions is a way of specifying to procmail that if any one of the four listed expressions is found in the message, the total condition is considered true, and the message gets passed into emil. These four subconditions check whether the message has a Sun attachment, a binhex attachment, or a uuencoded attachment; there are others that could be added to check these things better and to check other relevant conditions. The "LOG=" line writes a line into the procmail log; the lone double-quote beginning the following line makes sure the log entry gets an end-of-line character. The call to gawk (GNU awk) is for fixing end-of-line conventions, since emil sometimes leaves those in the format of the originating machine; it could probably be replaced with a sed subsitution.
The emil call itself tries to ensure that the message uses:
Most of these (the primary exceptions being the character set and the Apple binary format) are as they should be for good internet interoperability.
Some mail servers (Lotus Domino is a suspect here) mangle Sun-formatted messages, so the conversion to MIME needs to happen before such programs see the message. The ideal is to rid the world of Sun-formatted messages: don't use mailtool for sending attachments (it doesn't understand MIME anyway, and most of the world doesn't understand its attachments, so it really shouldn't be used at all), and make sure dtmail is set to use MIME rather than mailtool's format.
This isn't fetchmail's problem either; fetchmail doesn't know anything about mail attachments and doesn't treat them any differently from plain message data.
The most usual cause of this problem seems to be bugs in your network transport layer's capability to handle the very large TCP/IP packets that attachments tend to turn into. You can test this theory by trying to download the offending message through a webmail account; using HTTP for the message tends to simulate large-packet stress rather well, and you will probably find that the messages that seem to be choking fetchmail will make your HTTP download speed drop to zero.
This problem can be caused by subtle bugs in the packet-reassembly layer of your TCP/IP stack; these often don't manifest at normal packet sizes. It may also be caused by malfunctioning path-MTU discovery on the mailserver. Or, if there's a modem in the link, it may be because the attachment contains the Hayes mode escape "+++".
Blame it on that rancid pile of dung and offal called Microsoft Exchange. Due to the problem described in S2, the IMAP support in fetchmail cannot follow the IMAP protocol 100%. Most of the time it doesn't matter, but if you combine it with an SMTP server that behaves unusually, you'll get a spurious ) at message end.
One piece of software that can trigger this is the Interchange mail server, as used by, e.g., mailandnews.com. Here's what happens:
1. Someone sends mail to your account. The last line of the message contains text. So at the SMTP level, the message ends with, e.g. "blahblah\r\n.\r\n"
2. The SMTP handler sees the final "\r\n.\r\n" and recognizes the end of the message. However, instead of doing the normal thing, which is tossing out the ".\r\n" and leaving the first '\r\n' as part of the email body, Interchange throws out the whole "\r\n.\r\n", and leaves the email body without any line terminator at the end of it. RFC821 does not forbid this, though it probably should.
3. Fetchmail, or some other IMAP client, asks for the message. IMAP returns it, but it's enclosed inside parentheses, according to the protocol. The message size in bytes is also present. Because the message doesn't end with a line terminator, the IMAP client sees:
....blahblah)...
where the ')' is from IMAP.
4. Fetchmail only deals with complete lines, and can't trust the stated message size because Microsoft Exchange fscks it up.
5. As a result, fetchmail takes the final 'blahblah)' and puts it at the end of the message it forwards on. If you have verbosity on, you'll get a message about actual != expected.
There is no fix for this. The nuke mentioned in S2 looks more tempting all the time.
This is a feature, not a bug. It's in line with normal practice for system daemons and allows you to suppress logging by removing the log, without hacking potentially fragile startup scripts. To get around it, just touch(1) the logfile before you run fetchmail (this will have no effect on the contents of the logfile if it already exists).
Fetchmail uses the local sendmail to perform final delivery, which Netscape and other clients doesn't do; the announcement of new messages is done by a daemon that sendmail pokes. There should be a "biff" command to control this. Type
biff n
to turn it off. If this doesn't work, try the command
chmod -x $(tty)
which is essentially what biff -n will do. If this
doesn't work, comment out any reference to "comsat" in your
/etc/inetd.conf file and restart inetd.
In Slackware Linux distributions, the last line in /etc/profile is
biff yChange this to
biff nto solve the problem system-wide.
No, but versions 5.2.2 and later will notice when you modify your rc file and restart, reading it.
Because you're using a POP3 other than Qualcomm qpopper, or an IMAP with a long expunge interval.
According to the POP3 RFCs, deletes aren't actually performed until you issue the end-of-session QUIT command. Fetchmail cannot fix this, because doing it right takes cooperation from the server. There are two possible remedies:
One is to switch to qpopper (the free POP3 server from Qualcomm, the Eudora people). The qpopper software violates the POP3 RFCs by doing an expunge (removing deleted messages) on a line hangup, as well as on processing a QUIT command.
The other (which we recommend) is to switch to IMAP. IMAP has an explicit expunge command and fetchmail normally uses it to delete messages immediately after they are downloaded.
If you get very unlucky, you might take a line hit in the window between the delete and the expunge. If you've set a longer expunge interval, the window gets wider. This problem should correct itself the next time you complete a successful query.
Because logging is done based on the address indicated by the sending SMTP's MAIL FROM, and some listeners are picky about that address.
Some SMTP listeners get upset if you try to hand them a MAIL FROM address naming a different host than the originating site for your connection. This is a feature, not a bug -- it's supposed to help prevent people from forging mail with a bogus origin site. (RFC 1123 says you shouldn't do this exclusion...)
Since the originating site of a fetchmail delivery connection is localhost, this effectively means these picky listeners will barf on any MAIL FROM address fetchmail hands them with an @ in it!
Versions 2.1 and up try the header From address first and fall back to the calling-user ID. So if your SMTP listener isn't picky, the log will look right.
Sendmail does a hostname lookup when it first starts up, and also each time it gets a HELO in listener mode.
Your resolver configuration may be causing one of these lookups
to fail and time out. Check /etc/resolv.conf and
/etc/hosts file. Make sure your hostname and
fully-qualified domain name are both in /etc/hosts,
and that hosts is looked at before DNS is queried. You probably
also want your remote mail server(s) to be in the hosts file.
You can suppress the startup-time lookup if need to by
reconfiguring with FEATURE(nodns).
Configuring your bind library to cache DNS lookups locally may help, and is a good idea for speeding up other services as well. Switching to a faster MTA like qmail or exim might help.
Because that's not the order the server hands it to fetchmail in.
Fetchmail getting mail from a POP server delivers mail in the order that your server delivers mail. Fetchmail can't do anything about this; it's a limitation of the underlying POP protocol.
In theory it might be possible for fetchmail in IMAP mode to sort messages by date, but this would be in violation of two basics of fetchmail's design philosophy: (a) to be as simple and transparent a pipe as possible, and (b) to hide, rather than emphasize, the differences between the remote-fetch protocols it uses.
Re-ordering messages is a user-agent function, anyway.
There is a combination of circumstances that can confuse fetchmail. If you have set up demand dialing with pppd, and pppd has an idle timeout, and you have lcp-echo-interval set, then the lcp-echo-interval time must be longer than the pppd idle timeout. Otherwise it is going keep increasing the packet counters that fetchmail relies upon, triggering fetchmail into polling after its own delay interval and thus preventing the pppd link from ever reaching its inactivity timeout.
First, check to see that you haven't enabled the keep and fetchall option. If you have, turn keep off.
There are various forms of lossage involving the POP3 UIDL feature that can lead to all your old messages being seen again after a line drop. I have given up trying to fix these, as the UIDL code breaks worse every time I touch it. The problem is fundamental; maintaining and garbage-collecting the right kind of client-side state is just hard. Whoever put UIDLs in RFC1725 and removed LAST should be hung up by his thumbs and whipped with scorpions. The right answers are either (a) live with the occasional breakage, (b) switch to IMAP4, or (c) fix the code yourself and send me a patch. Unless you choose (c), I don't want to hear about it.
This can also happen when some other mail client is logged in to your mail server, if it uses a simple exclusive-locking scheme (and many, especially most POP3 servers, do exactly that). Your fetchmail is able to retrieve the messages, but because the mailbox is write-locked by the other instance yours can neither mark messages seen or delete them. The solution is to either (a) wait for the other client to finish, or (b) terminate it.
James Stevens <James.Stevens at kyzo.com> writes:
We had a Linux box dialing the Net and collecting mail from an NT POP3 server. Fetchmail was correctly collecting and deleting each e-mail one by one. However,the dial-up connection was very unreliable and would often just drop out in the middle of a session.
Interestingly, unless the TCP POP3 connection was terminated normally (I guess with a POP3 "QUIT" command) NT would then roll back all the deletes !!!
This meant if the first e-mail was very large it might just end up continuously collecting it, basically jamming the queue. Or, if the queue became very full itmight never get a long enough phone connection to retrieve the entire mailbox, and NT would roll back any deletes, so it would end up collecting (and delivering) the first few e-mails again and again. As the POP3 mailbox became fuller and fuller the chances of getting a connection long enough to collect theentire mailbox became smaller and smaller.
Our solution was to make fetchmail only collect a few (say 5 or 10) e-mails at atime, thus trying to ensure that the POP3 connection is terminated correctly.
Unfortunately, this is exactly the way POP3 servers are supposed to behave on a line drop, according to the RFCs. I recommend switching to IMAP and using a short expunge interval.
This is a design choice in your MTA, not fetchmail. It's taking the received date from the last Received header.
This is your server barfing on the CAPA probe that fetchmail sends.
If you run fetchmail in daemon mode (say "set daemon 600"), you will get the message only once per run.
If you set an authentication method explicitly (say, with
auth password), you will never get the message.
This is a feature, not a bug.
Any 4xx response (like 451) indicates a transient (temporary) error. This means that the mail could be accepted if retried later. Lookup failures are normally transient errors as a mail should not get rejected if a dns server is unreachable or down.
A permanent reject response is of the form 5xx (like 550).
You could tell your SMTP server to not lookup any addresses if you are not keen on checking the sender addresses. This problem typically occurs if your mail server is not checking the sender addresses, but your local server is.
Or you could declare antispam 451.
Or, you could check your nameserver configuration and query logs for dns errors.
All these issues are not related to fetchmail directly.
Write a preconnect command in your configuration file that
does something like "date >> $HOME/Procmail/fetchmail.log".
Use --limitflush (available since release 6.3.0) to
delete oversized mails along with the --limit option. If
you are already having flush in your rcfile to delete
oversized mails, replace it with limitflush to
avoid losing mails unintentionally.
The --flush option is primarily designed to delete
mails which have been read/downloaded but not deleted yet. This option
cannot be overloaded to delete oversized mails as it cannot be guessed
whether the user wants to delete only read/downloaded mails or only
oversized mails or both when a user specifies both
--limit and --flush. Hence, a separate
--limitflush has been added to resolve the ambiguity.
This happens when fetchmail sees an "X-IMAP:" header in the very first message in your mailbox. This usually stems from a message like the one shown below, which is automatically created on your server. This message shows up if the University of Washington IMAP or PINE software is used on the server together with a POP2 or POP3 daemon that is not aware of these messages, such as some versions of Qualcomm Popper (QPOP):
From MAILER-DAEMON Wed Nov 23 11:38:42 2005 Date: 23 Nov 2005 11:38:42 +0100 From: Mail System Internal Data <MAILER-DAEMON@imap.example.org> Subject: DON'T DELETE THIS MESSAGE -- FOLDER INTERNAL DATA Message-ID: <1132742322@imap.example.org> X-IMAP: 1132742306 0000000001 Status: RO This text is part of the internal format of your mail folder, and is not a real message. It is created automatically by the mail system software. If deleted, important folder data will be lost, and it will be re-created with the data reset to initial values.
As this message does not contain useful information, fetchmail is not retrieving it. And deleting it might slow down the server if you are keeping messages on the server, and the server would recreate it anyways, that's why fetchmail does not bother to delete it either.
| Back to Fetchmail Home Page | $Date$ |