| FETCHMAIL(1) | fetchmail reference manual | FETCHMAIL(1) |
fetchmail - fetch mail from a POP, IMAP, ETRN, or ODMR-capable server
fetchmail [option...] [mailserver...]
fetchmailconf
fetchmail is a mail-retrieval and forwarding utility; it fetches mail from remote mail servers and forwards it to your local (client) machine's delivery system. You can then handle the retrieved mail using normal mail user agents such as mutt(1), elm(1) or Mail(1). The fetchmail utility can be run in a daemon mode to repeatedly poll one or more systems at a specified interval.
The fetchmail program can gather mail from servers supporting any of the common mail-retrieval protocols: POP2 (legacy, to be removed from future release), POP3, IMAP2bis, IMAP4, and IMAP4rev1. It can also use the ESMTP ETRN extension and ODMR. (The RFCs describing all these protocols are listed at the end of this manual page.)
While fetchmail is primarily intended to be used over on-demand TCP/IP links (such as SLIP or PPP connections), it may also be useful as a message transfer agent for sites which refuse for security reasons to permit (sender-initiated) SMTP transactions with sendmail.
For troubleshooting, tracing and debugging, you need to increase fetchmail's verbosity to actually see what happens. To do that, please run both of the two following commands, adding all of the options you'd normally use.
env LC_ALL=C fetchmail -V -v --nodetach --nosyslog
env LC_ALL=C fetchmail -vvv --nodetach --nosyslog
Also see item #G3 in fetchmail's FAQ.
You can omit the LC_ALL=C part above if you want output in the local language (if supported). However if you are posting to mailing lists, please leave it in. The maintainers do not necessarily understand your language, please use English.
Your fetchmail distribution should have come with a README.SSL file, which see. It is recommended to configure all polls with --ssl --sslproto tls1.2+ if supported by the server, which configures fetchmail along recent IETF proposed standards and best current practices, RFC-8314, RFC-8996, RFC-8997.
If fetchmail is used with a POP or an IMAP server (but not with ETRN or ODMR), it has two fundamental modes of operation for each user account from which it retrieves mail: singledrop- and multidrop-mode.
As each message is retrieved, fetchmail normally delivers it via SMTP to port 25 on the machine it is running on (localhost), just as though it were being passed in over a normal TCP/IP link. fetchmail provides the SMTP server with an envelope recipient derived in the manner described previously. The mail will then be delivered according to your MTA's rules (the Mail Transfer Agent is usually sendmail(8), exim(8), or postfix(8)). Invoking your system's MDA (Mail Delivery Agent) is the duty of your MTA. All the delivery-control mechanisms (such as .forward files) normally available through your system MTA and local delivery agents will therefore be applied as usual.
If your fetchmail configuration sets a local MDA (see the --mda option), it will be used directly instead of talking SMTP to port 25.
If the program fetchmailconf is available, it will assist you in setting up and editing a fetchmailrc configuration. It runs under the X window system and requires that the language Python and the Tk toolkit (with Python bindings) be present on your system. If you are first setting up fetchmail for single-user mode, it is recommended that you use Novice mode. Expert mode provides complete control of fetchmail configuration, including the multidrop features. In either case, the 'Autoprobe' button will tell you the most capable protocol a given mail server supports, and warn you of potential problems with that server.
Fetchmail's run-time strings have been translated (localized) to some languages, but the manual is only available in English. In some situations, for comparing output to manual, it may be helpful to switch fetchmail to English output by overriding the locale variables, for instance:
env LC_ALL=C fetchmail # add other options before the hash
env LANG=en fetchmail # other options before the hash
or similar. Details vary by operating system.
The behavior of fetchmail is controlled by command-line options and a run control file, ~/.fetchmailrc, the syntax of which we describe in a later section (this file is what the fetchmailconf program edits). Command-line options override ~/.fetchmailrc declarations.
Each server name that you specify following the options on the command line will be queried. If you do not specify any servers on the command line, each 'poll' entry in your ~/.fetchmailrc file will be queried, unless the idle option is used, which see.
To facilitate the use of fetchmail in scripts and pipelines, it returns an appropriate exit code upon termination -- see EXIT CODES below.
The following options modify the behavior of fetchmail. It is seldom necessary to specify any of these once you have a working .fetchmailrc file set up.
Almost all options have a corresponding keyword which can be used to declare them in a .fetchmailrc file.
Some special options are not covered here, but are documented instead in sections on AUTHENTICATION and DAEMON MODE which follow.
All these alternatives work in basically the same way (communicating with standard server daemons to fetch mail already delivered to a mailbox on the server) except ETRN and ODMR. The ETRN mode allows you to ask a compliant ESMTP server (such as BSD sendmail at release 8.8.0 or higher) to immediately open a sender-SMTP connection to your client machine and begin forwarding any items addressed to your client machine in the server's queue of undelivered mail. The ODMR mode requires an ODMR-capable server and works similarly to ETRN, except that it does not require the client machine to have a static DNS.
Note that this does not magically switch between TLS-wrapped and STARTTLS modes, if you specify a port number or service name here that is TLS-wrapped, meaning it starts to negotiate TLS before sending application data in the clear, you may need to specify --ssl on the command line or ssl in your rcfile.
NOTE: If you use client authentication, the user name is fetched from the certificate's CommonName and overrides the name set with --user.
openssl x509 -in cert.pem -noout -md5 -fingerprint
For details, see x509(1ssl).
--smtphost server1,server2/2525,server3,/var/imap/socket/lmtp
This option can be used with ODMR, and will make fetchmail a relay between the ODMR server and SMTP or LMTP receiver.
Some possible MDAs are "/usr/sbin/sendmail -i -f %F -- %T" (Note: some several older or vendor sendmail versions mistake -- for an address, rather than an indicator to mark the end of the option arguments), "/usr/bin/deliver" and "/usr/bin/maildrop -d %T". Local delivery addresses will be inserted into the MDA command wherever you place a %T; the mail message's From address will be inserted where you place an %F.
Do NOT enclose the %F or %T string in single quotes! For both %T and %F, fetchmail encloses the addresses in single quotes ('), after removing any single quotes they may contain, before the MDA command is passed to the shell.
Do NOT use an MDA invocation that dispatches on the contents of To/Cc/Bcc, like "sendmail -i -t" or "qmail-inject", it will create mail loops and bring the just wrath of many postmasters down upon your head. This is one of the most frequent configuration errors!
Also, do not try to combine multidrop mode with an MDA such as maildrop that can only accept one address, unless your upstream stores one copy of the message per recipient and transports the envelope recipient in a header; you will lose mail.
The well-known procmail(1) package is very hard to configure properly, it has a very nasty "fall through to the next rule" behavior on delivery errors (even temporary ones, such as out of disk space if another user's mail daemon copies the mailbox around to purge old messages), so your mail will end up in the wrong mailbox sooner or later. The proper procmail configuration is outside the scope of this document. Using maildrop(1) is usually much easier, and many users find the filter syntax used by maildrop easier to understand.
Finally, we strongly advise that you do not use qmail-inject. The command line interface is non-standard without providing benefits for typical use, and fetchmail makes no attempts to accommodate qmail-inject's deviations from the standard. Some of qmail-inject's command-line and environment options are actually dangerous and can cause broken threads, non-detected duplicate messages and forwarding loops.
An argument of '-' causes the SMTP batch to be written to standard output, which is of limited use: this only makes sense for debugging, because fetchmail's regular output is interspersed on the same channel, so this is not suitable for mail delivery. This special mode may be removed in a later release.
Note that fetchmail's reconstruction of MAIL FROM and RCPT TO lines is not guaranteed correct; the caveats discussed under THE USE AND ABUSE OF MULTIDROP MAILBOXES below apply. This mode has precedence before --mda and SMTP/LMTP.
An explicit --limit of 0 overrides any limits set in your run control file. This option is intended for those needing to strictly control fetch time due to expensive and variable phone rates.
Combined with --limitflush, it can be used to delete oversized messages waiting on a server. In daemon mode, oversize notifications are mailed to the calling user (see the --warnings option). This option does not work with ETRN or ODMR.
interface/iii.iii.iii.iii[/mmm.mmm.mmm.mmm]
The field before the first slash is the interface name (i.e., sl0, ppp0 etc.). The field before the second slash is the acceptable IP address. The field after the second slash is a mask which specifies a range of IP addresses to accept. If no mask is present 255.255.255.255 is assumed (i.e., an exact match). This option is currently only supported under Linux and FreeBSD. Please see the monitor section for below for FreeBSD specific information.
Note that this option may be removed from a future fetchmail version.
Note that this option may be removed from a future fetchmail version.
This option changes the header fetchmail assumes will carry a copy of the mail's envelope address. Normally this is 'X-Envelope-To'. Other typically found headers to carry envelope information are 'X-Original-To' and 'Delivered-To'. Now, since these headers are not standardized, practice varies. See the discussion of multidrop address handling below. As a special case, 'envelope "Received"' enables parsing of sendmail-style Received lines. This is the default, but discouraged because it is not fully reliable.
Note that fetchmail expects the Received-line to be in a specific format: It must contain "by host for address", where host must match one of the mail server names that fetchmail recognizes for the account in question.
The optional count argument (only available in the configuration file) determines how many header lines of this kind are skipped. A count of 1 means: skip the first, take the second. A count of 2 means: skip the first and second, take the third, and so on.
All modes except ETRN require authentication of the client to the server. Normal user authentication in fetchmail is very much like the authentication mechanism of ftp(1). The correct user-id and password depend upon the underlying security system at the mail server.
If the mail server is a Unix machine on which you have an ordinary user account, your regular login name and password are used with fetchmail. If you use the same login name on both the server and the client machines, you needn't worry about specifying a user-id with the -u option -- the default behavior is to use your login name on the client machine as the user-id on the server machine. If you use a different login name on the server machine, specify that login name with the -u option. E.g., if your login name is 'jsmith' on a machine named 'mailgrunt', you would start fetchmail as follows:
The default behavior of fetchmail is to prompt you for your mail server password before the connection is established. This is the safest way to use fetchmail and ensures that your password will not be compromised. You may also specify your password in your ~/.fetchmailrc file. This is convenient when using fetchmail in daemon mode or with scripts.
If you do not specify a password, and fetchmail cannot extract one from your ~/.fetchmailrc file, it will look for a ~/.netrc file in your home directory before requesting one interactively; if an entry matching the mail server is found in that file, the password will be used. Fetchmail first looks for a match on poll name; if it finds none, it checks for a match on via name. See the ftp(1) man page for details of the syntax of the ~/.netrc file. To show a practical example, a .netrc might look like this:
machine hermes.example.org login joe password topsecret
You can repeat this block with different user information if you need to provide more than one password.
This feature may allow you to avoid duplicating password information in more than one file.
On mail servers that do not provide ordinary user accounts, your user-id and password are usually assigned by the server administrator when you apply for a mailbox on the server. Contact your server administrator if you do not know the correct user-id and password for your mailbox account.
All retrieval protocols can use SSL or TLS wrapping for the transport. Additionally, POP3 and IMAP retrieval can also negotiate SSL/TLS by means of STARTTLS (or STLS).
You can access TLS-encrypted services by specifying the options starting with --ssl, such as --ssl, --sslproto, --sslcertck, and others. You can also do this using the corresponding user options in the .fetchmailrc file. Some services, such as POP3 and IMAP, have different well known ports defined for the SSL encrypted services. The encrypted ports will be selected automatically when SSL is enabled and no explicit port is specified. Also, the --sslcertck command line or sslcertck run control file option should be used to force strict certificate checking with older fetchmail versions - see below.
If TLS or SSL is not configured, fetchmail will usually still try to use STARTTLS somewhat opportunistically. In practice, is it still mandatory because --sslcertck is a default setting and implicitly requires STARTTLS.
STARTTLS can be enforced by using --sslproto auto and defeated by using --sslproto ''. STARTTLS connections use the same port as the unencrypted version of the protocol and negotiate TLS via special command. The --sslcertck command line or sslcertck run control file option should be used to force strict certificate checking - see below.
--sslcertck is recommended: When connecting to an SSL or TLS encrypted server, the server presents a certificate to the client for validation. The certificate is checked to verify that the common name in the certificate matches the name of the server being contacted and that the effective and expiration dates in the certificate indicate that it is currently valid. If any of these checks fail, a warning message is printed, but the connection continues. The server certificate does not need to be signed by any specific Certifying Authority and may be a "self-signed" certificate. If the --sslcertck command line option or sslcertck run control file option is used, fetchmail will instead abort if any of these checks fail, because it must assume that there is a man-in-the-middle attack in this scenario, hence fetchmail must not expose clear-text passwords. Use of the sslcertck or --sslcertck option is therefore advised; it has become the default in fetchmail 6.4.0.
Some SSL encrypted servers may request a client side certificate. A client side public SSL certificate and private SSL key may be specified. If requested by the server, the client certificate is sent to the server for validation. Some servers may require a valid client certificate and may refuse connections if a certificate is not provided or if the certificate is not valid. Some servers may require client side certificates be signed by a recognized Certifying Authority. The format for the key files and the certificate files is that required by the underlying SSL libraries (OpenSSL in the general case).
A word of care about the use of SSL: While above mentioned setup with self-signed server certificates retrieved over the wires can protect you from a passive eavesdropper, it does not help against an active attacker. It is clearly an improvement over sending the passwords in clear, but you should be aware that a man-in-the-middle attack is trivially possible (in particular with tools such as dsniff). Use of strict certificate checking with a certification authority recognized by server and client, or perhaps of an SSH tunnel (see below for some examples) is preferable if you care seriously about the security of your mailbox and passwords.
Early versions of POP3 (RFC1081, RFC1225) supported a crude form of independent authentication using the .rhosts file on the mail server side. Under this RPOP variant, a fixed per-user ID equivalent to a password was sent in clear over a link to a reserved port, with the command RPOP rather than PASS to alert the server that it should do special checking. RPOP is supported by fetchmail (you can specify 'protocol RPOP' to have the program send 'RPOP' rather than 'PASS') but its use is strongly discouraged, and support will be removed from a future fetchmail version. This facility was vulnerable to spoofing and was withdrawn in RFC1460.
RFC1460 introduced APOP authentication. In this variant of POP3, you register an APOP password on your server host (on some servers, the program to do this is called popauth(8)). You put the same password in your ~/.fetchmailrc file. Each time fetchmail logs in, it sends an MD5 hash of your password and the server greeting time to the server, which can verify it by checking its authorization database.
Note that APOP is no longer considered resistant against man-in-the-middle attacks.
fetchmail makes some efforts to make the server believe messages had not been retrieved, by using the TOP command with a large number of lines when possible. TOP is a command that retrieves the full header and a fetchmail-specified amount of body lines. It is optional and therefore not implemented by all servers, and some are known to implement it improperly. On many servers however, the RETR command which retrieves the full message with header and body, sets the "seen" flag (for instance, in a web interface), whereas the TOP command does not do that.
fetchmail will always use the RETR command if "fetchall" is set. fetchmail will also use the RETR command if "keep" is set and "uidl" is unset. Finally, fetchmail will use the RETR command on Maillennium POP3/PROXY servers (used by Comcast) to avoid a deliberate TOP misinterpretation in this server that causes message corruption.
In all other cases, fetchmail will use the TOP command. This implies that in "keep" setups, "uidl" must be set if "TOP" is desired.
Note that this description is true for the current version of fetchmail, but the behavior may change in future versions. In particular, fetchmail may prefer the RETR command because the TOP command causes much grief on some servers and is only optional.
If your fetchmail was built with Kerberos support and you specify Kerberos authentication (either with --auth or the .fetchmailrc option authenticate kerberos_v4) it will try to get a Kerberos ticket from the mail server at the start of each query. Note: if either the pollname or via name is 'hesiod', fetchmail will try to use Hesiod to look up the mail server.
If you use POP3 or IMAP with GSSAPI authentication, fetchmail will expect the server to have RFC1731- or RFC1734-conforming GSSAPI capability, and will use it. Currently this has only been tested over Kerberos 5, so you are expected to already have a ticket-granting ticket. You may pass a username different from your principal name using the standard --user command or by the .fetchmailrc option user.
If your IMAP daemon returns the PREAUTH response in its greeting line, fetchmail will notice this and skip the normal authentication step. This can be useful, e.g., if you start imapd explicitly using ssh. In this case you can declare the authentication value 'ssh' on that site entry to stop .fetchmail from asking you for a password when it starts up.
If you use client authentication with TLS1 and your IMAP
daemon returns the AUTH=EXTERNAL response, fetchmail will notice this
and will use the authentication shortcut and will not send the passphrase.
In this case you can declare the authentication value 'external'
on that site to stop fetchmail from asking you for a password when it
starts up.
If you are using POP3, and the server issues a one-time-password challenge conforming to RFC1938, fetchmail will use your password as a pass phrase to generate the required response. This avoids sending secrets over the net unencrypted.
Compuserve's RPA authentication is supported. If you compile in the support, fetchmail will try to perform an RPA pass-phrase authentication instead of sending over the password unencrypted if it detects "@compuserve.com" in the host name.
If you are using IMAP, Microsoft's NTLM authentication (used by Microsoft Exchange) is supported. If you compile in the support, fetchmail will try to perform an NTLM authentication (instead of sending over the password unencrypted) whenever the server returns AUTH=NTLM in its capability response. 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.
fetchmail also supports authentication to the ESMTP server on the client side according to RFC 2554. You can specify a name/password pair to be used with the keywords 'esmtpname' and 'esmtppassword'; the former defaults to the username of the calling user.
In daemon mode, fetchmail puts itself into the background and runs forever, querying each specified host and then sleeping for a given polling interval.
There are several ways to make fetchmail work in daemon mode. On the command line, --daemon <interval> or -d <interval> option runs fetchmail in daemon mode. You must specify a numeric argument which is a polling interval (time to wait after completing a whole poll cycle with the last server and before starting the next poll cycle with the first server) in seconds.
Example: simply invoking
will, therefore, poll all the hosts described in your ~/.fetchmailrc file (except those explicitly excluded with the 'skip' verb) a bit less often than once every 15 minutes (exactly: 15 minutes + time that the poll takes).
It is also possible to set a polling interval in your ~/.fetchmailrc file by saying 'set daemon <interval>', where <interval> is an integer number of seconds. If you do this, fetchmail will always start in daemon mode unless you override it with the command-line option --daemon 0 or -d0.
Only one daemon process is permitted per user; in daemon mode, fetchmail sets up a per-user lock file to guarantee this. (You can however cheat and set the FETCHMAILHOME environment variable to overcome this setting, but in that case, it is your responsibility to make sure you are not polling the same server with two processes at the same time.)
Normally, calling fetchmail with a daemon in the background sends a wake-up signal to the daemon and quits without output. The background daemon then starts its next poll cycle immediately. The wake-up signal, SIGUSR1, can also be sent manually. The wake-up action also clears any 'wedged' flags indicating that connections have wedged due to failed authentication or multiple timeouts.
The option -q or --quit will kill a running daemon process instead of waking it up (if there is no such process, fetchmail will notify you). If the --quit option appears last on the command line, fetchmail will kill the running daemon process and then quit. Otherwise, fetchmail will first kill a running daemon process and then continue running with the other options.
The -L <filename> or --logfile
<filename> option (keyword: set logfile) is only effective when
fetchmail is detached and in daemon mode. Note that the logfile must
exist before fetchmail is run, you can use the touch(1)
command with the filename as its sole argument to create it.
This option allows you to redirect status messages into a specified logfile
(follow the option with the logfile name). The logfile is opened for append,
so previous messages are not deleted. This is primarily useful for debugging
configurations. Note that fetchmail does not detect if the logfile is
rotated, the logfile is only opened once when fetchmail starts. You need to
restart fetchmail after rotating the logfile and before compressing it (if
applicable).
The --syslog option (keyword: set syslog) allows you to redirect status and error messages emitted to the syslog(3) system daemon if available. Messages are logged with an id of fetchmail, the facility LOG_MAIL, and priorities LOG_ERR, LOG_ALERT or LOG_INFO. This option is intended for logging status and error messages which indicate the status of the daemon and the results while fetching mail from the server(s). Error messages for command line options and parsing the .fetchmailrc file are still written to stderr, or to the specified log file. The --nosyslog option turns off use of syslog(3), assuming it is turned on in the ~/.fetchmailrc file. This option is overridden, in certain situations, by --logfile (which see).
The -N or --nodetach option suppresses backgrounding and detachment of the daemon process from its control terminal. This is useful for debugging or when fetchmail runs as the child of a supervisor process such as init(8) or Gerrit Pape's runit(8). Note that this also causes the logfile option to be ignored.
Note that while running in daemon mode polling a POP2 or IMAP2bis server, transient errors (such as DNS failures or sendmail delivery refusals) may force the fetchall option on for the duration of the next polling cycle. This is a robustness feature. It means that if a message is fetched (and thus marked seen by the mail server) but not delivered locally due to some transient error, it will be re-fetched during the next poll cycle. (The IMAP logic does not delete messages until they are delivered, so this problem does not arise.)
If you touch or change the ~/.fetchmailrc file while fetchmail is running in daemon mode, this will be detected at the beginning of the next poll cycle. When a changed ~/.fetchmailrc is detected, fetchmail rereads it and restarts from scratch (using exec(2); no state information is retained in the new instance). Note that if fetchmail needs to query for passwords, of that if you break the ~/.fetchmailrc file's syntax, the new instance will softly and silently vanish away on startup.
The --postmaster <name> option (keyword: set postmaster) specifies the last-resort username to which multidrop mail is to be forwarded if no matching local recipient can be found. It is also used as destination of undeliverable mail if the 'bouncemail' global option is off and additionally for spam-blocked mail if the 'bouncemail' global option is off and the 'spambounce' global option is on. This option defaults to the user who invoked fetchmail. If the invoking user is root, then the default of this option is the user 'postmaster'. Setting postmaster to the empty string causes such mail as described above to be discarded - this however is usually a bad idea. See also the description of the 'FETCHMAILUSER' environment variable in the ENVIRONMENT section below.
The --nobounce behaves like the "set no bouncemail" global option, which see.
The --invisible option (keyword: set invisible) tries to make fetchmail invisible. Normally, fetchmail behaves like any other MTA would -- it generates a Received header into each message describing its place in the chain of transmission, and tells the MTA it forwards to that the mail came from the machine fetchmail itself is running on. If the invisible option is on, the Received header is suppressed and fetchmail tries to spoof the MTA it forwards to into thinking it came directly from the mail server host.
The --showdots option (keyword: set showdots) forces fetchmail to show progress dots even if the output goes to a file or fetchmail is not in verbose mode. Fetchmail shows the dots by default when run in --verbose mode and output goes to console. This option is ignored in --silent mode.
By specifying the --tracepolls option, you can ask fetchmail to add information to the Received header on the form "polling {label} account {user}", where {label} is the account label (from the specified rcfile, normally ~/.fetchmailrc) and {user} is the username which is used to log on to the mail server. This header can be used to make filtering email where no useful header information is available and you want mail from different accounts sorted into different mailboxes (this could, for example, occur if you have an account on the same server running a mailing list, and are subscribed to the list using that account). The default is not adding any such header. In .fetchmailrc, this is called 'tracepolls'.
The protocols fetchmail uses to talk to mail servers are next to bulletproof. In normal operation forwarding to port 25, no message is ever deleted (or even marked for deletion) on the host until the SMTP listener on the client side has acknowledged to fetchmail that the message has been either accepted for delivery or rejected due to a spam block.
When forwarding to an MDA, however, there is more possibility of error. Some MDAs are 'safe' and reliably return a nonzero status on any delivery error, even one due to temporary resource limits. The maildrop(1) program is like this; so are most programs designed as mail transport agents, such as sendmail(1), including the sendmail wrapper of Postfix and exim(1). These programs give back a reliable positive acknowledgement and can be used with the mda option with no risk of mail loss. Unsafe MDAs, though, may return 0 even on delivery failure. If this happens, you will lose mail.
The normal mode of fetchmail is to try to download only 'new' messages, leaving untouched (and undeleted) messages you have already read directly on the server (or fetched with a previous fetchmail --keep). But you may find that messages you have already read on the server are being fetched (and deleted) even when you do not specify --all. There are several reasons this can happen.
One could be that you are using POP2. The POP2 protocol includes no representation of 'new' or 'old' state in messages, so fetchmail must treat all messages as new all the time. But POP2 is obsolete, so this is unlikely.
A potential POP3 problem might be servers that insert messages in the middle of mailboxes (some VMS implementations of mail are rumored to do this). The fetchmail code assumes that new messages are appended to the end of the mailbox; when this is not true it may treat some old messages as new and vice versa. Using UIDL whilst setting fastuidl 0 might fix this, otherwise, consider switching to IMAP.
Yet another POP3 problem is that if they cannot make temporary files in the user's home directory, some POP3 servers will hand back an undocumented response that causes fetchmail to spuriously report "No mail".
The IMAP code uses the presence or absence of the server flag \Seen to decide whether or not a message is new. This is not the right thing to do, fetchmail should check the UIDVALIDITY and use UID, but it does not do that yet. Under Unix, it counts on your IMAP server to notice the BSD-style Status flags set by mail user agents and set the \Seen flag from them when appropriate. All Unix IMAP servers we know of do this, though it is not specified by the IMAP RFCs. If you ever trip over a server that does not, the symptom will be that messages you have already read on your host will look new to the server. In this (unlikely) case, only messages you fetched with fetchmail --keep will be both undeleted and marked old.
In ETRN and ODMR modes, fetchmail does not actually retrieve messages; instead, it asks the server's SMTP listener to start a queue flush to the client via SMTP. Therefore it sends only undelivered messages.
Many SMTP listeners allow administrators to set up 'spam filters' that block unsolicited email from specified domains. A MAIL FROM or DATA line that triggers this feature will elicit an SMTP response which (unfortunately) varies according to the listener.
Newer versions of sendmail return an error code of 571.
According to RFC2821, the correct thing to return in this situation is 550 "Requested action not taken: mailbox unavailable" (the draft adds "[E.g., mailbox not found, no access, or command rejected for policy reasons].").
Older versions of the exim MTA return 501 "Syntax error in parameters or arguments".
The postfix MTA runs 554 as an antispam response.
Zmailer may reject code with a 500 response (followed by an enhanced status code that contains more information).
Return codes which fetchmail treats as antispam responses and discards the message can be set with the 'antispam' option. This is one of the only three circumstance under which fetchmail ever discards mail (the others are the 552 and 553 errors described below, and the suppression of multi-dropped messages with a message-ID already seen).
If fetchmail is fetching from an IMAP server, the antispam response will be detected and the message rejected immediately after the headers have been fetched, without reading the message body. Thus, you will not pay for downloading spam message bodies.
By default, the list of antispam responses is empty.
If the spambounce global option is on, mail that is spam-blocked triggers an RFC1892/RFC1894 bounce message informing the originator that we do not accept mail from it. See also BUGS.
Besides the spam-blocking described above, fetchmail takes special actions — that may be modified by the --softbounce option — on the following SMTP/ESMTP error response codes
Other errors greater or equal to 500 trigger bounce mail back to the originator, unless suppressed by --softbounce. See also BUGS.
The preferred way to set up fetchmail is to write a .fetchmailrc file in your home directory (you may do this directly, with a text editor, or indirectly via fetchmailconf). When there is a conflict between the command-line arguments and the arguments in this file, the command-line arguments take precedence.
To protect the security of your passwords, your ~/.fetchmailrc may not normally have more than 0700 (u=rwx,g=,o=) permissions; fetchmail will complain and exit otherwise (this check is suppressed when --version is on).
You may read the .fetchmailrc file as a list of commands to be executed when fetchmail is called with no arguments.
Comments begin with a '#' and extend through the end of the line. Otherwise the file consists of a series of server entries or global option statements in a free-format, token-oriented syntax.
There are four kinds of tokens: grammar keywords, numbers (i.e., decimal digit sequences), unquoted strings, and quoted strings. A quoted string is bounded by double quotes and may contain whitespace (and quoted digits are treated as a string). Note that quoted strings will also contain line feed characters if they run across two or more lines, unless you use a backslash to join lines (see below). An unquoted string is any whitespace-delimited token that is neither numeric, string quoted nor contains the special characters ',', ';', ':', or '='.
Any amount of whitespace separates tokens in server entries, but is otherwise ignored. You may use backslash escape sequences (\n for LF, \t for HT, \b for BS, \r for CR, \nnn for decimal (where nnn cannot start with a 0), \0ooo for octal, and \xhh for hex) to embed non-printable characters or string delimiters in strings. In quoted strings, a backslash at the very end of a line will cause the backslash itself and the line feed (LF or NL, new line) character to be ignored, so that you can wrap long strings. Without the backslash at the line end, the line feed character would become part of the string.
Warning: while these resemble C-style escape sequences, they are not the same. fetchmail only supports these eight styles. C supports more escape sequences that consist of backslash (\) and a single character, but does not support decimal codes and does not require the leading 0 in octal notation. Example: fetchmail interprets \233 the same as \xE9 (Latin small letter e with acute), where C would interpret \233 as octal 0233 = \x9B (CSI, control sequence introducer).
Each server entry consists of one of the keywords 'poll' or 'skip', followed by a server name, followed by server options, followed by any number of user (or username) descriptions, followed by user options. Note: the most common cause of syntax errors is mixing up user and server options or putting user options before the user descriptions.
For backward compatibility, the word 'server' is a synonym for 'poll'.
You can use the noise keywords 'and', 'with', 'has', 'wants', and 'options' anywhere in an entry to make it resemble English. They are ignored, but can make entries much easier to read at a glance. The punctuation characters ':', ';' and ',' are also ignored.
The 'poll' verb tells fetchmail to query this host when it is run with no arguments. The 'skip' verb tells fetchmail not to poll this host unless it is explicitly named on the command line. (The 'skip' verb allows you to experiment with test entries safely, or easily disable entries for hosts that are temporarily down.)
Here are the legal options. Keyword suffixes enclosed in square brackets are optional. Those corresponding to short command-line options are followed by '-' and the appropriate option letter. If option is only relevant to a single mode of operation, it is noted as 's' or 'm' for singledrop- or multidrop-mode, respectively.
Here are the legal global options:
| Keyword | Opt | Mode | Function |
| set daemon | -d | Set a background poll interval in seconds. | |
| set postmaster | Give the name of the last-resort mail recipient (default: user running fetchmail, "postmaster" if run by the root user) | ||
| set bouncemail | Direct error mail to the sender (default) | ||
| set no bouncemail | Direct error mail to the local postmaster (as per the 'postmaster' global option above). | ||
| set no spambounce | Do not bounce spam-blocked mail (default). | ||
| set spambounce | Bounce blocked spam-blocked mail (as per the 'antispam' user option) back to the destination as indicated by the 'bouncemail' global option. Warning: Do not use this to bounce spam back to the sender - most spam is sent with false sender address and thus this option hurts innocent bystanders. | ||
| set no softbounce | Delete permanently undeliverable mail. It is recommended to use this option if the configuration has been thoroughly tested. | ||
| set softbounce | Keep permanently undeliverable mail as though a temporary error had occurred (default). | ||
| set logfile | -L | Name of a file to append error and status messages to. Only effective in daemon mode and if fetchmail detaches. If effective, overrides set syslog. | |
| set pidfile | -p | Name of the PID file. | |
| set idfile | -i | Name of the file to store UID lists in. | |
| set syslog | Do error logging through syslog(3). May be overridden by set logfile. | ||
| set no syslog | Turn off error logging through syslog(3). (default) | ||
| set properties | String value that is ignored by fetchmail (may be used by extension scripts). |
Here are the legal server options:
| Keyword | Opt | Mode | Function |
| via | Specify DNS name of mail server, overriding poll name | ||
| proto[col] | -p | Specify protocol (case insensitive): POP2, POP3, IMAP, APOP, KPOP | |
| local[domains] | m | Specify domain(s) to be regarded as local | |
| port | Specify TCP/IP service port (obsolete, use 'service' instead). | ||
| service | -P | Specify service name (a numeric value is also allowed and considered a TCP/IP port number). | |
| auth[enticate] | Set authentication type (default 'any') | ||
| timeout | -t | Server inactivity timeout in seconds (default 300) | |
| envelope | -E | m | Specify envelope-address header name |
| no envelope | m | Disable looking for envelope address | |
| qvirtual | -Q | m | Qmail virtual domain prefix to remove from user name |
| aka | m | Specify alternate DNS names of mail server | |
| interface | -I | specify IP interface(s) that must be up for server poll to take place | |
| monitor | -M | Specify IP address to monitor for activity | |
| plugin | Specify command through which to make server connections. | ||
| plugout | Specify command through which to make listener connections. | ||
| dns | m | Enable DNS lookup for multidrop (default) | |
| no dns | m | Disable DNS lookup for multidrop | |
| checkalias | m | Do comparison by IP address for multidrop | |
| no checkalias | m | Do comparison by name for multidrop (default) | |
| uidl | -U | Force POP3 to use client-side UIDLs (recommended) | |
| no uidl | Turn off POP3 use of client-side UIDLs (default) | ||
| interval | Only check this site every N poll cycles; N is a numeric argument. | ||
| tracepolls | Add poll tracing information to the Received header | ||
| principal | Set Kerberos principal (only useful with IMAP and kerberos) | ||
| esmtpname | Set name for RFC2554 authentication to the ESMTP server. | ||
| esmtppassword | Set password for RFC2554 authentication to the ESMTP server. | ||
| bad-header | How to treat messages with a bad header. Can be reject (default) or accept. |
Here are the legal user descriptions and options:
| Keyword | Opt | Mode | Function |
| user[name] | -u | This is the user description and must come first after server description and after possible server options, and before user options. It sets the remote user name if by itself or followed by 'there', or the local user name if followed by 'here'. | |
| is | Connect local and remote user names | ||
| to | Connect local and remote user names | ||
| pass[word] | Specify remote account password | ||
| ssl | Connect to server over the specified base protocol using SSL encryption | ||
| sslcert | Specify file for client side public SSL certificate | ||
| sslcertck | Enable strict certificate checking and abort connection on failure. Default only since fetchmail v6.4.0. | ||
| no sslcertck | Disable strict certificate checking and permit connections to continue on failed verification. Discouraged. Should only be used together with sslfingerprint. | ||
| sslcertfile | Specify file with trusted CA certificates | ||
| sslcertpath | Specify c_rehash-ed directory with trusted CA certificates. | ||
| sslfingerprint | Specify the expected server certificate finger print from an MD5 hash. Fetchmail will disconnect and log an error if it does not match. | ||
| sslkey | Specify file for client side private SSL key | ||
| sslproto | Force ssl protocol for connection | ||
| folder | -r | Specify remote folder to query | |
| smtphost | -S | Specify smtp host(s) to forward to | |
| fetchdomains | m | Specify domains for which mail should be fetched | |
| smtpaddress | -D | Specify the domain to be put in RCPT TO lines | |
| smtpname | Specify the user and domain to be put in RCPT TO lines | ||
| antispam | -Z | Specify what SMTP returns are interpreted as spam-policy blocks | |
| mda | -m | Specify MDA for local delivery | |
| bsmtp | Specify BSMTP batch file to append to | ||
| preconnect | Command to be executed before each connection | ||
| postconnect | Command to be executed after each connection | ||
| keep | -k | Do not delete seen messages from server (for POP3, uidl is recommended) | |
| flush | -F | Flush all seen messages before querying (DANGEROUS) | |
| limitflush | Flush all oversized messages before querying | ||
| fetchall | -a | Fetch all messages whether seen or not | |
| rewrite | Rewrite destination addresses for reply (default) | ||
| stripcr | Strip carriage returns from ends of lines | ||
| forcecr | Force carriage returns at ends of lines | ||
| pass8bits | Force BODY=8BITMIME to ESMTP listener | ||
| dropstatus | Strip Status and X-Mozilla-Status lines out of incoming mail | ||
| dropdelivered | Strip Delivered-To lines out of incoming mail | ||
| mimedecode | Convert quoted-printable to 8-bit in MIME messages | ||
| idle | Idle waiting for new messages after each poll (IMAP only) | ||
| no keep | -K | Delete seen messages from server (default) | |
| no flush | Do not flush all seen messages before querying (default) | ||
| no fetchall | Retrieve only new messages (default) | ||
| no rewrite | Do not rewrite headers | ||
| no stripcr | Do not strip carriage returns (default) | ||
| no forcecr | Do not force carriage returns at EOL (default) | ||
| no pass8bits | Do not force BODY=8BITMIME to ESMTP listener (default) | ||
| no dropstatus | Do not drop Status headers (default) | ||
| no dropdelivered | Do not drop Delivered-To headers (default) | ||
| no mimedecode | Do not convert quoted-printable to 8-bit in MIME messages (default) | ||
| no idle | Do not idle waiting for new messages after each poll (IMAP only) | ||
| limit | -l | Set message size limit | |
| warnings | -w | Set message size warning interval | |
| batchlimit | -b | Max # messages to forward in single connect | |
| fetchlimit | -B | Max # messages to fetch in single connect | |
| fetchsizelimit | Max # message sizes to fetch in single transaction | ||
| fastuidl | Use binary search for first unseen message (POP3 only) | ||
| expunge | -e | Perform an expunge on every #th message (IMAP and POP3 only) | |
| properties | String value is ignored by fetchmail (may be used by extension scripts) |
All user options must begin with a user description (user or username option) and follow all server descriptions and options.
In the .fetchmailrc file, the 'envelope' string argument may be preceded by a whitespace-separated number. This number, if specified, is the number of such headers to skip over (that is, an argument of 1 selects the second header of the given type). This is sometimes useful for ignoring bogus envelope headers created by an ISP's local delivery agent or internal forwards (through mail inspection systems, for instance).
The 'folder' and 'smtphost' options (unlike their command-line equivalents) can take a space- or comma-separated list of names following them.
All options correspond to the obvious command-line arguments, except the following: 'via', 'interval', 'aka', 'is', 'to', 'dns'/'no dns', 'checkalias'/'no checkalias', 'password', 'preconnect', 'postconnect', 'localdomains', 'stripcr'/'no stripcr', 'forcecr'/'no forcecr', 'pass8bits'/'no pass8bits' 'dropstatus/no dropstatus', 'dropdelivered/no dropdelivered', 'mimedecode/no mimedecode', 'no idle', and 'no envelope'.
The 'via' option is for if you want to have more than one configuration pointing at the same site. If it is present, the string argument will be taken as the actual DNS name of the mail server host to query. This will override the argument of poll, which can then simply be a distinct label for the configuration (e.g., what you would give on the command line to explicitly query this host).
The 'interval' option (which takes a numeric argument) allows you to poll a server less frequently than the basic poll interval. If you say 'interval N' the server this option is attached to will only be queried every N poll intervals.
Please ensure you read the section titled THE USE AND ABUSE OF MULTIDROP MAILBOXES if you intend to use multidrop mode.
The 'is' or 'to' keywords associate the following local (client) name(s) (or server-name to client-name mappings separated by =) with the mail server user name in the entry. If an is/to list has '*' as its last name, unrecognized names are simply passed through. Note that until fetchmail version 6.3.4 inclusively, these lists could only contain local parts of user names (fetchmail would only look at the part before the @ sign). fetchmail versions 6.3.5 and newer support full addresses on the left hand side of these mappings, and they take precedence over any 'localdomains', 'aka', 'via' or similar mappings.
A single local name can be used to support redirecting your mail when your username on the client machine is different from your name on the mail server. When there is only a single local name, mail is forwarded to that local username regardless of the message's Received, To, Cc, and Bcc headers. In this case, fetchmail never does DNS lookups.
When there is more than one local name (or name mapping), fetchmail looks at the envelope header, if configured, and otherwise at the Received, To, Cc, and Bcc headers of retrieved mail (this is 'multidrop mode'). It looks for addresses with host name parts that match your poll name or your 'via', 'aka' or 'localdomains' options, and usually also for host name parts which DNS tells it are aliases of the mail server. See the discussion of 'dns', 'checkalias', 'localdomains', and 'aka' for details on how matching addresses are handled.
If fetchmail cannot match any mail server usernames or localdomain addresses, the mail will be bounced. Normally it will be bounced to the sender, but if the 'bouncemail' global option is off, the mail will go to the local postmaster instead. (see the 'postmaster' global option). See also BUGS.
The 'dns' option (normally on) controls the way addresses from multidrop mailboxes are checked. On, it enables logic to check each host address that does not match an 'aka' or 'localdomains' declaration by looking it up with DNS. When a mail server username is recognized attached to a matching host name part, its local mapping is added to the list of local recipients.
The 'checkalias' option (normally off) extends the lookups performed by the 'dns' keyword in multidrop mode, providing a way to cope with remote MTAs that identify themselves using their canonical name, while they are polled using an alias. When such a server is polled, checks to extract the envelope address fail, and fetchmail reverts to delivery using the To/Cc/Bcc headers (See below 'Header versus Envelope addresses'). Specifying this option instructs fetchmail to retrieve all the IP addresses associated with both the poll name and the name used by the remote MTA and to do a comparison of the IP addresses. This comes in handy in situations where the remote server undergoes frequent canonical name changes, that would otherwise require modifications to the rcfile. 'checkalias' has no effect if 'no dns' is specified in the rcfile.
The 'aka' option is for use with multidrop mailboxes. It allows you to pre-declare a list of DNS aliases for a server. This is an optimization hack that allows you to trade space for speed. When fetchmail, while processing a multidrop mailbox, grovels through message headers looking for names of the mail server, pre-declaring common ones can save it from having to do DNS lookups. Note: the names you give as arguments to 'aka' are matched as suffixes -- if you specify (say) 'aka netaxs.com', this will match not just a host name netaxs.com, but any host name that ends with '.netaxs.com'; such as (say) pop3.netaxs.com and mail.netaxs.com.
The 'localdomains' option allows you to declare a list of domains which fetchmail should consider local. When fetchmail is parsing address lines in multidrop modes, and a trailing segment of a host name matches a declared local domain, that address is passed through to the listener or MDA unaltered (local-name mappings are not applied).
If you are using 'localdomains', you may also need to specify 'no envelope', which disables fetchmail's normal attempt to deduce an envelope address from the Received line or X-Envelope-To header or whatever header has been previously set by 'envelope'. If you set 'no envelope' in the defaults entry it is possible to undo that in individual entries by using 'envelope <string>'. As a special case, 'envelope "Received"' restores the default parsing of Received lines.
The password option requires a string argument, which is the password to be used with the entry's server.
The 'preconnect' keyword allows you to specify a shell command to be executed just before each time fetchmail establishes a mail server connection. This may be useful if you are attempting to set up secure POP connections with the aid of ssh(1). If the command returns a nonzero status, the poll of that mail server will be aborted.
Similarly, the 'postconnect' keyword similarly allows you to specify a shell command to be executed just after each time a mail server connection is taken down.
The 'forcecr' option controls whether lines terminated by LF only are given CRLF termination before forwarding. Strictly speaking RFC821 requires this, but few MTAs enforce the requirement so this option is normally off (only one such MTA, qmail, is in significant use at time of writing).
The 'stripcr' option controls whether carriage returns are stripped out of retrieved mail before it is forwarded. It is normally not necessary to set this, because it defaults to 'on' (CR stripping enabled) when there is an MDA declared but 'off' (CR stripping disabled) when forwarding is via SMTP. If 'stripcr' and 'forcecr' are both on, 'stripcr' will override.
The 'pass8bits' option exists to cope with Microsoft mail programs that stupidly slap a "Content-Transfer-Encoding: 7bit" on everything. With this option off (the default) and such a header present, fetchmail declares BODY=7BIT to an ESMTP-capable listener; this causes problems for messages actually using 8-bit ISO or KOI-8 character sets, which will be garbled by having the high bits of all characters stripped. If 'pass8bits' is on, fetchmail is forced to declare BODY=8BITMIME to any ESMTP-capable listener. If the listener is 8-bit-clean (as all the major ones now are) the right thing will probably result.
The 'dropstatus' option controls whether nonempty Status and X-Mozilla-Status lines are retained in fetched mail (the default) or discarded. Retaining them allows your MUA to see what messages (if any) were marked seen on the server. On the other hand, it can confuse some new-mail notifiers, which assume that anything with a Status line in it has been seen. (Note: the empty Status lines inserted by some buggy POP servers are unconditionally discarded.)
The 'dropdelivered' option controls whether Delivered-To headers will be kept in fetched mail (the default) or discarded. These headers are added by qmail and Postfix mail servers in order to avoid mail loops but may get in your way if you try to "mirror" a mail server within the same domain. Use with caution.
The 'mimedecode' option controls whether MIME messages using the quoted-printable encoding are automatically converted into pure 8-bit data. If you are delivering mail to an ESMTP-capable, 8-bit-clean listener (that includes all of the major MTAs like sendmail), then this will automatically convert quoted-printable message headers and data into 8-bit data, making it easier to understand when reading mail. If your e-mail programs know how to deal with MIME messages, then this option is not needed. The mimedecode option is off by default, because doing RFC2047 conversion on headers throws away character-set information and can lead to bad results if the encoding of the headers differs from the body encoding.
The 'idle' option is intended to be used with IMAP servers supporting the RFC2177 IDLE command extension, but does not strictly require it. If it is enabled, and fetchmail detects that IDLE is supported, an IDLE will be issued at the end of each poll. This will tell the IMAP server to hold the connection open and notify the client when new mail is available. If IDLE is not supported, fetchmail will simulate it by periodically issuing NOOP. If you need to poll a link frequently, IDLE can save bandwidth by eliminating TCP/IP connects and LOGIN/LOGOUT sequences. On the other hand, an IDLE connection will eat almost all of your fetchmail's time, because it will never drop the connection and allow other polls to occur unless the server times out the IDLE. It also does not work with multiple folders; only the first folder will ever be polled.
The 'properties' option is an extension mechanism. It takes a string argument, which is ignored by fetchmail itself. The string argument may be used to store configuration information for scripts which require it. In particular, the output of '--configdump' option will make properties associated with a user entry readily available to a Python script.
The words 'here' and 'there' have useful English-like significance. Normally 'user eric is esr' would mean that mail for the remote user 'eric' is to be delivered to 'esr', but you can make this clearer by saying 'user eric there is esr here', or reverse it by saying 'user esr here is eric there'
Legal protocol identifiers for use with the 'protocol' keyword are:
auto (or AUTO) (legacy, to be removed from future release)
pop2 (or POP2) (legacy, to be removed from future release)
pop3 (or POP3)
sdps (or SDPS)
imap (or IMAP)
apop (or APOP)
kpop (or KPOP)
Legal authentication types are 'any', 'password', 'kerberos', 'kerberos_v4', 'kerberos_v5' and 'gssapi', 'cram-md5', 'otp', 'msn' (only for POP3), 'ntlm', 'ssh', 'external' (only IMAP). The 'password' type specifies authentication by normal transmission of a password (the password may be plain text or subject to protocol-specific encryption as in CRAM-MD5); 'kerberos' tells fetchmail to try to get a Kerberos ticket at the start of each query instead, and send an arbitrary string as the password; and 'gssapi' tells fetchmail to use GSSAPI authentication. See the description of the 'auth' keyword for more.
Specifying 'kpop' sets POP3 protocol over port 1109 with Kerberos V4 authentication. These defaults may be overridden by later options.
There are some global option statements: 'set logfile' followed by a string sets the same global specified by --logfile. A command-line --logfile option will override this. Note that --logfile is only effective if fetchmail detaches itself from the terminal and the logfile already exists before fetchmail is run, and it overrides --syslog in this case. Also, 'set daemon' sets the poll interval as --daemon does. This can be overridden by a command-line --daemon option; in particular --daemon 0 can be used to force foreground operation. The 'set postmaster' statement sets the address to which multidrop mail defaults if there are no local matches. Finally, 'set syslog' sends log messages to syslogd(8).
There are various ways in that fetchmail may "crash", i. e. stop operation suddenly and unexpectedly. A "crash" usually refers to an error condition that the software did not handle by itself. A well-known failure mode is the "segmentation fault" or "signal 11" or "SIGSEGV" or just "segfault" for short. These can be caused by hardware or by software problems. Software-induced segfaults can usually be reproduced easily and in the same place, whereas hardware-induced segfaults can go away if the computer is rebooted, or powered off for a few hours, and can happen in random locations even if you use the software the same way.
For solving hardware-induced segfaults, find the faulty component and repair or replace it. The Sig11 FAQ may help you with details.
For solving software-induced segfaults, the developers may need a "stack backtrace".
By default, fetchmail suppresses core dumps as these might contain passwords and other sensitive information. For debugging fetchmail crashes, obtaining a "stack backtrace" from a core dump is often the quickest way to solve the problem, and when posting your problem on a mailing list, the developers may ask you for a "backtrace".
1. To get useful backtraces, fetchmail needs to be installed without getting stripped of its compilation symbols. Unfortunately, most binary packages that are installed are stripped, and core files from symbol-stripped programs are worthless. So you may need to recompile fetchmail. On many systems, you can type
file `which fetchmail`
to find out if fetchmail was symbol-stripped or not. If yours was unstripped, fine, proceed, if it was stripped, you need to recompile the source code first. You do not usually need to install fetchmail in order to debug it.
2. The shell environment that starts fetchmail needs to enable core dumps. The key is the "maximum core (file) size" that can usually be configured with a tool named "limit" or "ulimit". See the documentation for your shell for details. In the popular bash shell, "ulimit -Sc unlimited" will allow the core dump.
3. You need to tell fetchmail, too, to allow core dumps. To do this, run fetchmail with the -d0 -v options. It is often easier to also add --nosyslog -N as well.
Finally, you need to reproduce the crash. You can just start fetchmail from the directory where you compiled it by typing ./fetchmail, so the complete command line will start with ./fetchmail -Nvd0 --nosyslog and perhaps list your other options.
After the crash, run your debugger to obtain the core dump. The debugger will often be GNU GDB, you can then type (adjust paths as necessary) gdb ./fetchmail fetchmail.core and then, after GDB has started up and read all its files, type backtrace full, save the output (copy & paste will do, the backtrace will be read by a human) and then type quit to leave gdb. Note: on some systems, the core files have different names, they might contain a number instead of the program name, or number and name, but it will usually have "core" as part of their name.
When trying to determine the originating address of a message, fetchmail looks through headers in the following order:
Return-Path:
Resent-Sender: (ignored if it does not contain an @ or !)
Sender: (ignored if it does not contain an @ or !)
Resent-From:
From:
Reply-To:
Apparently-From:
The originating address is used for logging, and to set the MAIL FROM address when forwarding to SMTP. This order is intended to cope gracefully with receiving mailing list messages in multidrop mode. The intent is that if a local address does not exist, the bounce message will not be returned blindly to the author or to the list itself, but rather to the list manager (which is less annoying).
In multidrop mode, destination headers are processed as follows: First, fetchmail looks for the header specified by the 'envelope' option in order to determine the local recipient address. If the mail is addressed to more than one recipient, the Received line will not contain any information regarding recipient addresses.
Then fetchmail looks for the Resent-To:, Resent-Cc:, and Resent-Bcc: lines. If they exist, they should contain the final recipients and have precedence over their To:/Cc:/Bcc: counterparts. If the Resent-* lines do not exist, the To:, Cc:, Bcc: and Apparently-To: lines are looked for. (The presence of a Resent-To: is taken to imply that the person referred by the To: address has already received the original copy of the mail.)
Note that although there are password declarations in a good many of the examples below, this is mainly for illustrative purposes. We recommend stashing account/password pairs in your $HOME/.netrc file, where they can be used not just by fetchmail but by ftp(1) and other programs.
The basic format is:
Example:
poll pop.provider.net protocol pop3 username "jsmith" password "secret1"
Or, using some abbreviations:
poll pop.provider.net proto pop3 user "jsmith" password "secret1"
Multiple servers may be listed:
poll pop.provider.net proto pop3 user "jsmith" pass "secret1" poll other.provider.net proto pop2 user "John.Smith" pass "My^Hat"
Here is the same version with more whitespace and some noise words:
poll pop.provider.net proto pop3
user "jsmith", with password secret1, is "jsmith" here; poll other.provider.net proto pop2:
user "John.Smith", with password "My^Hat", is "John.Smith" here;
If you need to include whitespace in a parameter string or start the latter with a number, enclose the string in double quotes. Thus:
poll mail.provider.net with proto pop3:
user "jsmith" there has password "4u but u cannot krak this"
is jws here and wants mda "/bin/mail"
You may have an initial server description headed by the keyword 'defaults' instead of 'poll' followed by a name. Such a record is interpreted as defaults for all queries to use. It may be overwritten by individual server descriptions. So, you could write:
defaults proto pop3
user "jsmith" poll pop.provider.net
pass "secret1" poll mail.provider.net
user "jjsmith" there has password "secret2"
It is possible to specify more than one user per server. The 'user' keyword leads off a user description, and every user specification in a multi-user entry must include it. Here is an example:
poll pop.provider.net proto pop3 port 3111
user "jsmith" with pass "secret1" is "smith" here
user jones with pass "secret2" is "jjones" here keep
This associates the local username 'smith' with the pop.provider.net username 'jsmith' and the local username 'jjones' with the pop.provider.net username 'jones'. Mail for 'jones' is kept on the server after download.
Here is what a simple retrieval configuration for a multidrop mailbox looks like:
poll pop.provider.net:
user maildrop with pass secret1 to golux 'hurkle'='happy' snark here
This says that the mailbox of account 'maildrop' on the server is a multidrop box, and that messages in it should be parsed for the server user names 'golux', 'hurkle', and 'snark'. It further specifies that 'golux' and 'snark' have the same name on the client as on the server, but mail for server user 'hurkle' should be delivered to client user 'happy'.
Note that fetchmail, until version 6.3.4, did NOT allow full user@domain specifications here, these would never match. Fetchmail 6.3.5 and newer support user@domain specifications on the left-hand side of a user mapping.
Here is an example of another kind of multidrop connection:
poll pop.provider.net localdomains loonytoons.org toons.org
envelope X-Envelope-To
user maildrop with pass secret1 to * here
This also says that the mailbox of account 'maildrop' on the server is a multidrop box. It tells fetchmail that any address in the loonytoons.org or toons.org domains (including sub-domain addresses like 'joe@daffy.loonytoons.org') should be passed through to the local SMTP listener without modification. Be careful of mail loops if you do this!
Here is an example configuration using ssh and the plugin option. The queries are made directly on the stdin and stdout of imapd via ssh. Note that in this setup, IMAP authentication can be skipped.
poll mailhost.net with proto imap:
plugin "ssh %h /usr/sbin/imapd" auth ssh;
user esr is esr here
Use the multiple-local-recipients feature with caution -- it can bite. All multidrop features are ineffective in ETRN and ODMR modes.
Also, note that in multidrop mode duplicate mails may be suppressed. A piece of mail is considered duplicate if it does not have a discernible envelope recipient address, has the same header as the message immediately preceding and more than one addressee. Such runs of messages may be generated when copies of a message addressed to multiple users are delivered to a multidrop box. (To be precise, fetchmail 6.2.5 through 6.4.X use an MD5 hash of the raw message header, and only fetchmail 6.4.16+ document this properly. Fetchmail 5.0.8 (1999-09-14) through 6.2.4 used only the Message-ID header. 5.0.7 and older did not suppress duplicates.)
Note that this duplication killer code checking the entire header is very restrictive and may not suppress many duplicates in practice - for instance, if some X-Original-To or Delivered-To header differs. This is intentional and correct in such situations: wherever envelope information is available, it should be used for reliable delivery of mailing list and blind carbon copy (Bcc) messages. See the subsection Duplicate suppression below for suggestions.
The fundamental problem is that by having your mail server toss several peoples' mail in a single maildrop box, you may have thrown away potentially vital information about who each piece of mail was actually addressed to (the 'envelope address', as opposed to the header addresses in the RFC822 To/Cc headers - the Bcc is not available at the receiving end). This 'envelope address' is the address you need in order to reroute mail properly.
Sometimes fetchmail can deduce the envelope address. If the mail server MTA is sendmail and the item of mail had just one recipient, the MTA will have written a 'by/for' clause that gives the envelope addressee into its Received header. But this does not work reliably for other MTAs, nor if there is more than one recipient. By default, fetchmail looks for envelope addresses in these lines; you can restore this default with -E "Received" or 'envelope Received'.
As a better alternative, some SMTP listeners and/or mail servers insert a header in each message containing a copy of the envelope addresses. This header (when it exists) is often 'X-Original-To', 'Delivered-To' or 'X-Envelope-To'. Fetchmail's assumption about this can be changed with the -E or 'envelope' option. Note that writing an envelope header of this kind exposes the names of recipients (including blind-copy recipients) to all receivers of the messages, so the upstream must store one copy of the message per recipient to avoid becoming a privacy problem.
Postfix, since version 2.0, writes an X-Original-To: header which contains a copy of the envelope as it was received.
Qmail and Postfix generally write a 'Delivered-To' header upon delivering the message to the mail spool and use it to avoid mail loops. Qmail virtual domains however will prefix the user name with a string that normally matches the user's domain. To remove this prefix you can use the -Q or 'qvirtual' option.
Sometimes, unfortunately, neither of these methods works. That is the point when you should contact your ISP and ask them to provide such an envelope header, and you should not use multidrop in this situation. When they all fail, fetchmail must fall back on the contents of To/Cc headers (Bcc headers are not available - see below) to try to determine recipient addressees -- and these are unreliable. In particular, mailing-list software often ships mail with only the list broadcast address in the To: header.
Note that a future version of fetchmail may remove To/Cc parsing!
When fetchmail cannot deduce a recipient address that is local, and the intended recipient address was anyone other than fetchmail's invoking user, mail will get lost. This is what makes the multidrop feature risky without proper envelope information.
A related problem is that when you blind-copy a mail message, the Bcc information is carried only as envelope address (it is removed from the headers by the sending mail server, so fetchmail can see it only if there is an X-Envelope-To header). Thus, blind-copying to someone who gets mail over a fetchmail multidrop link will fail unless the mail server host routinely writes X-Envelope-To or an equivalent header into messages in your maildrop.
In conclusion, mailing lists and Bcc'd mail can only work if the server you are fetching from
Multiple local names can be used to administer a mailing list from the client side of a fetchmail collection. Suppose your name is 'esr', and you want to both pick up your own mail and maintain a mailing list called (say) "fetchmail-friends", and you want to keep the alias list on your client machine.
On your server, you can alias 'fetchmail-friends' to 'esr'; then, in your .fetchmailrc, declare 'to esr fetchmail-friends here'. Then, when mail including 'fetchmail-friends' as a local address gets fetched, the list name will be appended to the list of recipients your SMTP listener sees. Therefore it will undergo alias expansion locally. Be sure to include 'esr' in the local alias expansion of fetchmail-friends, or you will never see mail sent only to the list. Also be sure that your listener has the "me-too" option set (sendmail's -oXm command-line option or OXm declaration) so your name is not removed from alias expansions in messages you send.
This trick is not without its problems, however. You will begin to see this when a message comes in that is addressed only to a mailing list you do not have declared as a local name. Each such message will feature an 'X-Fetchmail-Warning' header which is generated because fetchmail cannot find a valid local name in the recipient addresses. Such messages default (as was described above) to being sent to the local user running fetchmail, but the program has no way to know that this is actually the right thing.
Multidrop mailboxes and fetchmail serving multiple users in daemon mode do not mix. The problem, again, is mail from mailing lists, which typically does not have an individual recipient address on it. Unless fetchmail can deduce an envelope address, such mail will only go to the account running fetchmail (probably root). Also, blind-copied users are very likely never to see their mail at all.
If you are tempted to use fetchmail to retrieve mail for multiple users from a single mail drop via POP or IMAP, think again (and reread the section on header and envelope addresses above). It would be smarter to just let the mail sit in the mail server'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 mail server's expiry period). If you cannot arrange this, try setting up a UUCP feed.
If you absolutely must use multidrop for this purpose, make sure your mail server writes an envelope-address header that fetchmail can see. Otherwise you will lose mail and it will come back to haunt you.
Normally, when multiple users are declared fetchmail extracts recipient addresses as described above and checks each host part with DNS to see if it is an alias of the mail server. If so, the name mappings described in the "to ... here" declaration are done and the mail locally delivered.
This is a convenient but also slow method. To speed it up, pre-declare mail server aliases with 'aka'; these are checked before DNS lookups are done. If you are certain your aka list contains all DNS aliases of the mail server (and all MX names pointing at it - note this may change in a future version) you can declare 'no dns' to suppress DNS lookups entirely and only match against the aka list.
If fetchmail's duplicate suppression code does not kick in for your multidrop mail account, other options is using sieve, or for instance Courier's maildrop package (and in particular, its reformail program with the -D option) as the delivery agent (either from fetchmail, or from your local mail server that fetchmail injects into).
Support for socks4/5 is a compile time configuration option. Once compiled in, fetchmail will always use the socks libraries and configuration on your system, there are no run-time switches in fetchmail - but you can still configure SOCKS: you can specify which SOCKS configuration file is used in the SOCKS_CONF environment variable.
For instance, if you wanted to bypass the SOCKS proxy altogether and have fetchmail connect directly, you could just pass SOCKS_CONF=/dev/null in the environment, for example (add your usual command line options - if any - to the end of this line):
env SOCKS_CONF=/dev/null fetchmail
To facilitate the use of fetchmail in shell scripts, an exit status code is returned to give an indication of what occurred during a given connection.
The exit codes returned by fetchmail are as follows:
|| [ $? -eq 1 ]
to the end of the fetchmail command line, note that this leaves 0 untouched, maps 1 to 0, and maps all other codes to 1. See also item #C8 in the FAQ.
When fetchmail queries more than one host, return status is 0 if any query successfully retrieved mail. Otherwise the returned error status is that of the last host queried.
Fetchmail's behavior can be altered by providing it with environment variables. Some may alter the operation of libraries that fetchmail links against, for instance, OpenSSL. Note that in daemon mode, you will need to quit the background daemon process and start a new fetchmail daemon for environment changes to take effect.
If HOME_ETC and FETCHMAILHOME are both set, FETCHMAILHOME takes precedence and HOME_ETC will be ignored.
If a fetchmail daemon is running as root, SIGUSR1 wakes it up from its sleep phase and forces a poll of all non-skipped servers. For compatibility reasons, SIGHUP can also be used in 6.3.X but may not be available in future fetchmail versions.
If fetchmail is running in daemon mode as non-root, use SIGUSR1 to wake it (this is so SIGHUP due to logout can retain the default action of killing it).
Running fetchmail in foreground while a background fetchmail is running will do whichever of these is appropriate to wake it up.
Please check the NEWS file that shipped with fetchmail for more known bugs than those listed here.
Fetchmail cannot handle user names that contain blanks after a "@" character, for instance "demonstr@ti on". These are rather uncommon and only hurt when using UID-based --keep setups, so the 6.X.Y versions of fetchmail will not be fixed.
Fetchmail cannot handle configurations where you have multiple accounts that use the same server name and the same login. Any user@server combination must be unique.
The assumptions that the DNS and in particular the checkalias options make are not often sustainable. For instance, it has become uncommon for an MX server to be a POP3 or IMAP server at the same time. Therefore the MX lookups may go away in a future release.
The mda and plugin options interact badly. In order to collect error status from the MDA, fetchmail has to change its normal signal handling so that dead plugin processes do not get reaped until the end of the poll cycle. This can cause resource starvation if too many zombies accumulate. So either do not deliver to a MDA using plugins or risk being overrun by an army of undead.
The --interface option does not support IPv6 and it is doubtful if it ever will, since there is no portable way to query interface IPv6 addresses.
The RFC822 address parser used in multidrop mode chokes on some @-addresses that are technically legal but bizarre. Strange uses of quoting and embedded comments are likely to confuse it.
In a message with multiple envelope headers, only the last one processed will be visible to fetchmail.
Use of some of these protocols requires that the program send unencrypted passwords over the TCP/IP connection to the mail server. This creates a risk that name/password pairs might be snaffled with a packet sniffer or more sophisticated monitoring software. Under Linux and FreeBSD, the --interface option can be used to restrict polling to availability of a specific interface device with a specific local or remote IP address, but snooping is still possible if (a) either host has a network device that can be opened in promiscuous mode, or (b) the intervening network link can be tapped. We recommend the use of ssh(1) tunnelling to not only shroud your passwords but encrypt the entire conversation.
Use of the %F or %T escapes in an mda option could open a security hole, because they pass text manipulable by an attacker to a shell command. Potential shell characters are replaced by '_' before execution. The hole is further reduced by the fact that fetchmail temporarily discards any set-uid privileges it may have while running the MDA. For maximum safety, however, do not use an mda command containing %F or %T when fetchmail is run from the root account itself.
Fetchmail's method of sending bounces due to errors or spam-blocking and spam bounces requires that port 25 of localhost be available for sending mail via SMTP.
If you modify ~/.fetchmailrc while a background instance is running and break the syntax, the background instance will die silently. Unfortunately, it cannot die noisily because we do not yet know whether syslog should be enabled. On some systems, fetchmail dies quietly even if there is no syntax error; this seems to have something to do with buggy terminal ioctl code in the kernel.
The -f - option (reading a configuration from stdin) is incompatible with the plugin option.
The 'principal' option only handles Kerberos IV, not V.
Interactively entered passwords are truncated after 63 characters. If you really need to use a longer password, you will have to use a configuration file.
A backslash as the last character of a configuration file will be flagged as a syntax error rather than ignored.
The BSMTP error handling is virtually nonexistent and may leave broken messages behind.
Send comments, bug reports, gripes, and the like to the fetchmail-devel list
An fetchmail FAQ (in HTML form) is available at the fetchmail home page, it should also accompany your installation.
Fetchmail is currently maintained by Matthias Andree and Rob Funk with major assistance from Sunil Shetye (for code) and Rob MacGregor (for the mailing lists).
Most of the code is from Eric S. Raymond. Too many other people to name here have contributed code and patches.
This program is descended from and replaces popclient, by Carl Harris; the internals have become quite different, but some of its interface design is directly traceable to that ancestral program.
This manual page has been improved by Matthias Andree, R. Hannes Beinert, and Héctor García.
README, README.SSL, README.SSL-SERVER, The Fetchmail FAQ, mutt(1), elm(1), mail(1), sendmail(8), popd(8), imapd(8), netrc(5), the fetchmail home page, (alternative URI); the maildrop home page.
Note that this list is just a collection of references and not a statement as to the actual protocol conformance or requirements in fetchmail.
| 2022-07-16 | fetchmail 6.4.37 |