Only the last entry is complete, the others might have been condensed.

1990/12/07: v1.00
1990/12/12: v1.01
1991/02/04: v1.02
1991/02/13: v1.10
1991/02/21: v1.20
1991/02/22: v1.21
1991/03/01: v1.30
1991/03/15: v1.35
	    Started using RCS to manage the source
1991/06/04: v1.99
1991/06/10: v2.00
1991/06/11: v2.01
1991/06/12: v2.02
1991/06/20: v2.03
1991/07/04: v2.10
1991/07/12: v2.11
1991/10/02: v2.20 (never released)
1991/10/18: v2.30
	    Reached the doubtful milestone of having a source file (regexp.c)
	       which provokes a compiler error on an old compiler
	       (if using the optimiser)
1991/10/22: v2.31
1991/12/05: v2.40
1991/12/13: v2.50
1992/01/22: v2.60
1992/01/31: v2.61
1992/04/30: v2.70
1992/07/01: v2.71
	    Gave procmail, formail, lockfile and mailstat a more verbose
	       command line help (called up by -h or -?)
1993/02/04: v2.80
	    Started using CVS to manage the source (god's gift to programmers)
	    Changes to the installation scripts:
	       - the autoconf script now performs a reliability test on kernel
		 locking support
	       - reached the doubtful milestone of consistently crashing the
		 kernel on a Convex by running the locktst program
1993/02/19: v2.81
1993/06/02: v2.82 (never really released, was only available as prerelease 4)
	    Worked my way around the !@#$%^&*() POSIX setgid() semantics (if
	       your OS supports setrgid() or setregid())
1993/07/01: v2.90
	    Condition lines in recipes can now be started with a leading `*',
	       there is no longer a need to count condition lines, simply
	       set the number to zero, and let procmail find out by itself
1993/07/02: v2.91
	    Reached the doubtful milestone to sometimes crash an Ultrix
	       machine (due to the lockingtests, not procmail itself)
1994/06/14: v3.00
	    Changes to procmail:
	       - Changed the semantics of the TRAP keyword.  In order to
		 make procmail accept the exitcode it returns, you now have
		 to set EXITCODE=""
	       - It was still occasionally trying to lock /dev/null, which
		 is of course silly, fixed that
	       - Taught it about `nesting recipes'; they allow parts of
		 an rcfile to be grouped hierarchically
	       - Fixed a discrepancy with /bin/sh backquote expansion in
		 environment assignments (preserving all spaces)
	       - Logs its pid and a timestamp when VERBOSE=on
	       - Caused the regular TIMEOUT to break a `hanging' kernel lock
	       - SIGUSR1 and SIGUSR2 can be used to turn on and off verbose
		 logging
	       - Worked around a bug in the `ANSI'-compiler of Domain/OS
	       - Procmail and lockfile now inherit any ignore status of most
		 regular signals (fixes a problem with some buggy shells)
	       - Optionally reads in a global rcfile (/etc/procmailrc)
		 before doing regular delivery (which includes the new
		 keyword: DROPPRIVS)
	       - Can pipe the mail to stdout on request
	       - Moved the "Reiterating kernel lock" diagnostic into the
		 "extended" (i.e. VERBOSE=on) section
	       - Tightened the loop when skipping comments in rcfiles (for
		 a slight speedup)
	       - Added support for filesystems not capable of creating
		 hardlinks
	       - Tightened the security check on initial absolute rcfiles
		 (they sometimes can't be world writable)
	       - Weighted scoring on conditions
	       - Ability to inline parse ${var-text} and ${var:-text}
	       - Ability to inline parse ${var+text} and ${var:+text}
	       - Skipping spaces after "!" and "$" on condition lines
	       - Implicit delivery somehow got broken: fixed
	       - Default umask is always 077 now for deliverymode
	       - Extended ^FROM_DAEMON and ^FROM_MAILER macro regexps again
	       - The -f option became less strict, everyone can use it now,
		 except that unpriviliged users will get an additional >From_
		 they didn't bargain for (in order to make fakes identifiable)
	       - The date on the From_ line can now be refreshed with -f-
	       - Introduced new recipe flags: E and e (else and error)
	       - Nested blocks clone procmail on a 'c' flag
	       - Introduced the EXITCODE special variable
	       - Implicit delivery mode is now entered if argv[0] doesn't start
		 with the word `procmail'
	       - Fixed the BSD support for kernel-locking only operation
	       - Taught the regexp engine about \< and \>
	       - Fixed bug present on some systems; caused the body to be
		 munged when filtering headers only
	       - Added -o option (makes procmail override the From_ lines, like
		 it used to)
	       - -p and -m together shrink the set of preset variables to the
		 bare minimum
	       - -p is not supported alongside -d anymore
	       - /etc/procmailrcs/ is the place for optional privileged
		 rcfiles in -m mailfilter mode
	       - Switched the meanings of SIGUSR1 and SIGUSR2
	       - The 'a' flag didn't work correctly after filter recipes
	       - Changed the permissions on the lockfile, writing zero in it
	       - Check the permissions on the existing system mailbox, correct
		 them if necessary
	       - Clean up zombies more often
	    Changes to formail:
	       - Fixed a sender-determination-weight problem, it mixed up
		 the weights when autoreplying and when regenerating the From_
		 line (and thus didn't always pick the optimal field)
	       - Pays attention to the exitcode of the programs it started
	       - Accepts simultaneous -X and -k options
	       - Fixed a bug introduced in v2.82 in formail when using
		 the -x and the -k options simultaneously
	       - Rearranged the weights for "-rt" (made From: more important)
	       - Parsed return-addresses starting with a \ incorrectly
		 (causing it to coredump on occasion)
	       - Supports the -s option withouth a program argument
	       - Recognise extra UUCP >From_ lines
	       - Introduced the -B option to split up BABYL rmail files
	       - It regards and generates a FILENO variable (for easy
		 numbering)
	       - Moved the idcheck functionality into formail -D (due to
		 popular demand), for eliminating duplicate mails
	       - It terminates early now if it only needs the header
	       - The -n option can now sustain itself by reaping children
		 if it can't fork() immediately
	       - It supports incomplete field specifications which match
		 any field starting similarly
	       - Introduced the -u and -U options
	       - -a Message-ID: and -a Resent-Message-ID: to make it generate
		 new ones
	       - Keep the X-Loop: field when generating autoreplies
	       - Lowered the negative weight for .UUCP reply addresses
	       - Honour Content-Length: fields, also speeds up processing of
		 lengthy messages
	       - Clean up zombies more often
	       - Handle bangpath reconstruction
	       - Made -q the default, use -q- to disable
	    Miscellaneous changes:
	       - Detecting and dodging buggy zshs everywhere
	       - Slightly adjusted autoconf for the new non-standard 386BSD
		 and NeXTStep 3.1 environments
	       - Extended the FAQ
	       - Extended and fixed the procmailex man page
	       - Updated the crontab script recommendation in the procmail
		 man page
	       - Fixed the "procmail"-mailer definition in the procmail man
		 page
	       - Created a new procmailsc man page
	       - Fixed a bug in lockfile, the exitcode was not correct if
		 you used -! with more than one file
	       - Including <limits.h> now, some (old) architectures seem to
		 insist on this
	       - Revamped the library search code
	       - Provided a faster (than most libraries) strstr() routine
	       - Created the setid program (to be used by the SmartList
		 installation)
	       - Checking for fstat() in autoconf
	       - Avoiding i/o-redirection on subshells
	       - Provided for the ability to hotwire the lockingtests
	       - Autoconf asks if you'd like to use the existing autoconf.h
	       - Autoconf determines MAX_argc (for choplist)
1994/06/14: v3.01
	    No changes, version number bump to keep in sync with SmartList
1994/06/16: v3.02
	    Made formail quiet (by default) about Content-Length mismatches
	    The version number in patchlevel.h for this version was incorrect
	       and still displayed v3.01 (yes, silly, I know)
1994/06/30: v3.03
	    Limit the no. of retries on lockfiles if the recipient is over
	       quota (procmail & lockfile)
	    Removed some superfluous "procmail:" prefixes in the middle of
	       an error message
	    Utilise a syslog daemon (if present) to log some critical errors
	       (mostly attempted security violations and errors which are
	       fatal but can't occur (like an unwritable /dev/null))
	    Reconstruct and respect Content-Length: in procmail
	       (if you need the >From lines, you'll have to take any existing
	       Content-Lenght: field out of the header)
	    Reformatted the source code to match the changed conventions
	    Procmail always defaulting the umask to 077 for deliverymode broke
	       some systems, reverting back to the old method of allowing group
	       access on the system mailbox if necessary
1994/08/02: v3.04
	    Changes to procmail:
	       - Support some non-BSD compatible syslog() implementations
	       - Even if the Content-Length is zero, write it out (some
		 programs can't deal with the empty field)
	       - Drop the safety margin on Content-Length calculations, some
		 programs can't deal with those
	       - Truncate folders to their former length if delivery was not
		 successful
	       - Fine-tuned the ^FROM_MAILER and ^FROM_DAEMON macros again
	       - The -v option lists the locking strategies employed
	       - Will create the last member of the mail spool directory if
		 found missing
	    Forgot to define closelog() away if syslog support is missing
	    Worked around the old syslog() interface
	    Worked around a compiler bug old HP compilers (pointer-unsigned),
	       caused the Content-Length: field to be mangled on some older
	       HP/UX systems (not on every mail)
	    Worked around compilation problems on SCO and old versions of IRIX
	    Some fixes to the man pages
	    Changes to formail:
	       - Mistakenly turned X-Loop: fields into Old-X-Loop: when
		 autoreplying
	       - Allow wildcard -i when autoreplying
	       - Renaming short fields to longer fields didn't always work
	       - Renaming with a wildcard source/destination is possible now
	       - -rk didn't behave correctly if a Content-Length: field was
		 present
	    Extended the sendmail directions in examples/advanced, it includes
	       a direct example on how to make use of the -a feature
	    Using EXIT_SUCCESS instead of EX_OK
	    Both procmail and formail take the -Y option, for traditional
	       Berkeley format mailboxes (ignoring Content-Length:)
	    Some NCR machines didn't have WNOHANG defined
1994/08/04: v3.05
	    Formail v3.04 didn't remove the From_ line if given the -I 'From '
	       option, changed that back, allowing for -a 'From '
	    Procmail sometimes didn't reliably count the number of matches on
	       a weighted recipe, fixed
	    Some minor manpage adaptations
1994/08/30: v3.06
	    Groff -mandoc macros managed to display the man pages incorrectly,
	       hacked my way around the .TH dependency to fix it
	    Split up string constant FM_HELP, it exceeded some compiler limits
	    Changes to procmail:
	       - Fixed a bug which was present since v2.30: 'z' was always
		 handled case sensitive (seems like not many people use
		 that letter :-) in regular expression conditions
	       - The ^^ anchor can now also be used to anchor the end of
		 a regular expression
	       - The -m flag will now unset ORGMAIL and will make
		 procmail omit the check for a system mailbox
	       - Allow easy reconfiguration of the default rcfile location
	       - Extend the list of internals displayed with -v
	       - The mail fed to the TRAP command contained some spurious
		 nul characters, fixed
	    Optionally allow the automatic installation of compressed man pages
	    Formail v3.00 and later occasionally seemed to hang if used in
	       a chain of pipes and fed with more text than it needed, fixed
	    Updated the FAQ
	    Updated the man pages (among others: vacation example changed)
	    Sharpened the autoconf const check, AIX 3.2.3 managed to slip past
	       it again
	    Made sure that "make -n" with any make works as expected
1994/10/31: v3.10
	    Changes to procmail:
	       - Minor corrections to the semantics of the 'a' and 'e' flags
	       - Minor correction to the semantics of the -o option
	       - Slight regular expression engine speedup
	       - Regexp matching of environment variables is possible now
	       - Due to popular demand: LOGABSTRACT=all logs *all* successful
		 delivering-recipes executed
	       - Enforce secure permissions on /etc/procmailrcs if used
	       - Take sgid bit in the system mail spool dir into account
		 even if it is world writable
	       - The regexp engine can return matches now (new token "\/",
		 new variable "MATCH")
	       - New recipe flag 'r', raw mode, so procmail doesn't try
		 to ensure the mail ends in an empty line
	       - Success and failure of a filter recipe is well defined now
	       - Procmail v3.06 prepended a bogus "." to explicit rcfile names
		 searched relative to the home directory, fixed
	       - Carved out two subroutines from main() to get it below the
		 optimisation threshold
	       - Eliminated duplicate error messages when procmailrcless
		 delivery fails
	       - Logging "Quota exceeded" messages when appropriate
	       - Truncate notification suppressed when logfile not opened
	       - Truncating didn't always work when delivering across NFS
	       - The $_ special variable was wrong when wasn't set
	    Changes to formail:
	       - New option: -z (zap whitespace and empty fields)
	       - Reading from stdin doesn't require the silly three EOFs
		 anymore
	       - -D with -r cache reply addresses now
	       - Carved out one subroutine from main() to get it below the
		 optimisation threshold
	       - -R with -x didn't work reliably
	       - -r with -i or -I sometimes had unexpected effects (in v3.06)
	       - The nil-Return-Path-override was broken, fixed
	    Updated the man pages, new subsection to procmailrc(5) summarising
	       procmail regexp syntax
	    Expanded on the sendmail.cf $#local example in the
	       examples/advanced file again
	    Revised detection of hard-link incapable filesystems during the
	       installation
	    Fixed bug in lockfile, the exitcode was not correct if
	       you used -! (I hope this finally fixes this -! problem)
	    Using execv() instead of execve()
1995/05/17: v3.11pre3
	    Changes to procmail:
	       - varname ?? < nnn conditions didn't have the expected effect
	       - Regression bug since v3.06, procmail -m /etc/procmailrcs
		 didn't allow any arguments to be passed, fixed
	       - Eliminated a superfluous fork() when processing TRAP
	       - "lockfile ignored" warning was generated inappropriately at
		 times
	       - Renamed testb() into testB() to avoid conflict with Solaris
	       - Eliminated spurious extra / in default MAILDIR value
	       - Whole line comments among the conditions are recognised
	       - Embedded empty lines in a recipe are tolerated
	       - $\name regexp safe variable expansion
	       - Delay searching for bogus From_ lines until writeout time
		 (speeds up filtering and writes to /dev/null)
	       - Finally fixed this mess with transparent backup to kernel
		 locking methods when the spool directory is not writable
	       - Avoid the one second NFS_ATIME_HACK under heavy load
	       - The 'r' flag had some undesirable side effects at times
	       - Dotlocks which fail due to permissions are not retried anymore
	       - Made the USER_TO_LOWERCASE_HACK run-time adapting
	       - /usr/spool/mail perm 1777, procmail setgid mail, procmail
		 could not read .procmailrc files in 700 $HOME dirs, fixed
	       - If called with -d option and not running with enough
		 privileges, procmail will bounce the mail (instead of
		 delivering to the invoker, as it used to)
	       - Severe tweaking on ^FROM_MAILER and ^FROM_DAEMON to reduce
		 false matches
	       - Allow for broken From_ lines with a missing sender address
	    Changes to formail:
	       - Slightly extended the number of known header fields
	       - Eliminated the conflict with the 4.4BSD daemon libidentifier
	       - In an MMDF environment formail -b didn't behave correctly
	       - Extracted another function from main() to make it smaller
	       - Process address groups correctly
	       - Process From_ lines with embedded commas correctly
	    Changes to autoconf:
	       - Catch NeXTstep 3.2 missing DIR definition
	       - Detect & work around Ultrix 4.3 "ANSI" C compiler
	       - A defined DEFsendmail or SYSTEM_MBOX caused some "s to be
		 omitted in autoconf.h
	       - Refined preliminary setsid() checks (2.4 x86/sunpro cc
		 managed to break it)
	       - Worked around a HERE document quoting bug in some shells
	       - Fixed the empty argument "shift" problem
	       - Detect & work around BSD 4.4 brain damaged setrgid()
	    New Makefile variable VISIBLE_BASE
	    Added support for a parallelising make
	    Changed manconf.c to cater for broken systems that have a 100 line
	       limit for sed (instead of a 100 command limit)
	    Fixed some portability problems with the Makefiles for the OSF make
	    Worked around old shells not supporting negated classes
	    Extended the FAQ
	    Updated examples/advanced docs for meta-argument setup in
	       a traditional v5.* sendmail setup
	    Fixed potential memory corruption bug for machines that have
	       sizeof(off_t)>sizeof(off_t*) (has been around for ages)
	    The man pages were remade upon every make, fixed
1995/10/29: v3.11pre4
	    Changes to procmail:
	       - Avoid the NFS delay on directory and MH folders
	       - KEEPENV didn't work reliably for more than one variable
	       - New macro ^TO_, delimits addresses more accurately than ^TO
	       - Don't try to fix the system mailbox permissions too soon,
		 this should put a stop to the numerous confusion reports
	       - SENDMAILFLAGS, new environment variable
	       - Support -y as a substitute kludge for -Y
	       - Fixed parsing of $@' when not doublequoted
	    Changes to formail:
	       - Return failure if the autoreply could not find a proper
		 return address
	       - Multiple -U options sometimes had unfortunate side effects
	       - When splitting and a maximum number of messages was being
		 specified, formail erroneously returned EX_IOERR
	       - Avoid splitting empty messages
	    Changes to autoconf:
	       - If running on a system with good old BSD semantics for
		 setrgid(), use the extra features offered
	    Changed the Mprocmail example, use $g instead of $f
1997/04/28: v3.11pre7
	    Changes to procmail:
	       - Cater for a race condition that occurs if two procmails
		 try to create an empty system mailbox (bogus BOGUS.* files)
	       - SysV autoforwarding mailboxes didn't work, regression bug in
		 v3.10
	       - Autocreating the last dirmember of the spooldir didn't
		 (always?) work due to the trailing /
	       - Kernel lockf() method doesn't change the position of the
		 filepointer anymore (results in more accurate lockingtests)
	       - Multiple directory folders are assigned to LASTFOLDER
	       - Don't strip trailing \n in a $MATCH
	       - Refuse to open directories for INCLUDERC files
	       - Syslog failed -o attempts
	       - Don't log non-delivering recipes, even with 'c' flag
	    Changes to formail:
	       - Skip leading spaces when checking for duplicates (will break
		 checks with old id-databases)
	    Worked around an nroff-coredumping problem with IRIX
	    Corrected the last(?) "make -n" glitch
	    Fixed library detection loop for some Solaris 2.[3-5] setups
	    Changes to procmail and lockfile: use the authenticate library
	       for easier integration with custom authentication and mailbox
	       locations
1999/03/02: v3.12
	    Changes to procmail:
	       - Use BOGUS.$LOGNAME.inode for bogus files to ease recovery
	       - Define RESTRICT_EXEC to restrict execution of programs
	       - Perform continuous checks on heap overflow, everywhere
		 If overflow is occurs then new variable PROCMAIL_OVERFLOW
		 is set
	       - Catch overly long rcfile names
	       - New variable PROCMAIL_VERSION
	       - LOGABSTRACT=all no longer logs filtering or variable capture
		 actions
	       - Don't strip leading \n in a $MATCH
	       - Worked around a compiler bug in Sun C compiler 4.2 (fdefault
		 cached past function calls)
	       - Tempfile names would grow on retry
	       - Open or reopen rcfiles as the user to prevent peeking when
		 not in privileged mailfilter mode
	       - Don't use $HOME/.procmailrc if it's group-writable or in a
		 group-writable directory, unless it's the user's default group
		 and GROUP_PER_USER is set in config.h
	       - hardlink in a an NFS-resistant manner
	    Worked around a compiler bug old HP compilers (pointer-unsigned),
	       caused the Content-Length: field to be mangled on some older
	       HP/UX systems (not on every mail)
	    Changes to formail:
	       - Generated Message-IDs don't contain "s anymore
	       - Fix off-by-one error when zapping whitespace
	       - -z option allows for leading tab instead of space
	    Changes to formail and lockfile:
	       - -v option displays version information
	    Changes to autoconf:
	       - Detect & work around inefficient realloc() implementations
	    Mailstat returns grand totals as well now
	    Update FAQ and docs to reflect default placing of procmail
	       in /usr/bin instead of /usr/local/bin
1999/03/31: v3.13
	    Mailstat was too loose in its awk syntax
	    Changes to formail:
	       - Formail was ignoring the exitcode of all but the last
		 invocation (or last several, if -n was in effect)
	    Changes to procmail:
	       - Variable expansion of builtin numeric variables in
		 conditions could overwrite the condition (broke SmartList)
	       - weights<1 didn't work if floats changed accuracy when stored
	    Worked around a bug in the Dunix 4.0e compiler (pointer addition
	       not commutative)
1999/11/22: v3.14
	    Changes to procmail:
	       - Some zero-length extractions using \/ could core dump
	       - Missed a couple possible overflows
	       - Eliminated the conflict with the C9x `restrict' keyword
	       - Support delivery to maildir mailboxes
	       - Support all styles of mailbox for the mail spool
	       - Don't use a locallockfile on $DEFAULT if it's a directory
	       - Set LINEBUF in the environment on startup
	       - Avoid renaming over old messages in directory folders
	       - New variable SWITCHRC performs `tail call'
	       - Refuse to open anything but regular files with INCLUDERC
		 and SWITCHRC
	       - Indicate whether GROUP_PER_USER was defined in the -v output
	       - Stopped depending on parens to stop function macros (they
		 don't under SunOS 4.x cc)
	       - Small heap compilation would fail on nomemerr() in pipes.c
	       - Worked around Tru64 UNIX V4.0E and V4.0F compilers
		 (strcpy() builtin doesn't always return pointer type)
	       - Warn about using 'c' flag with 'f' flag or on variable
		 capture recipes
	       - Warn about using 'h', 'b', 'i', or 'r' flags on nested
		 block recipes.
	       - Test for allowing rcfiles in sticky directories iff chown
		 is restricted was reversed
	       - LASTFOLDER wasn't correctly set when delivering to multiple
		 folders
	       - -f- couldn't find the timestamp if the address contained a
		 space
	       - SENDMAIL and SENDMAILFLAGS are now split in forwarding actions
	       - Variable capture actions now see the variable's current value
		 and restore it if the action fails.  Previously unset
		 variables will remain unset.
	       - fsync() mailboxes before closing them
	       - Actually suppress the 'E' and 'a' flags when combined with
		 the 'e' flag instead of just saying so
	       - Avoid some calls to alarm()
	       - Overflows at certain times would confuse procmail
	       - dyna_long code now meets strict ANSI restrictions
	       - 'W' flag changes "Program failure" to "Non-zero exitcode"
	       - Nested blocks must open and close within the same rcfile
	       - Root owned lockfiles aren't bogus
	       - A lone trailing '$' wasn't terminated properly when expanded
	    Changes to formail:
	       - Replies without the -t option go to the envelope sender
	       - Replies without "-a Resent-" and -t option ignore the
		 Resent-* headers
	       - Prevent corrupt idcaches by suppressing the -n option when
		 splitting with the -D option
	       - Accept and strip whitespace between the fieldname and colon
	       - Renaming from a wildcard to nothing now works
	    Changes to mailstat:
	       - Work around the detab done on checkin to CVS
	       - Recognize maildir mailboxes
	       - Don't use a tempfile
	    Changes to autoconf:
	       - Don't assume realloc(0,size) works (doesn't under SunOS 4)
	    Stopped using `implicit int' (for C9x)
	    Cache gethostname() and uname() output
	    Changed the form of tempfile names to make them `more' unique
	     and deal with filename length limits more gracefully
	    Updated the FAQ and the list of mirrors in the README
	    Documented the exact behavior of lockfile's -! option
	    Documented the suggested usage of -r vs -rt
2000/08/25: v3.15
	    Changes to procmail:
	       - v3.14 broke compilation on systems without lstat() or
		 that didn't declare it
	       - Rewrite folder type parsing: corrects handling of MH and
		 maildir style spools
	       - v3.14 changed '!' actions too much: revert to v3.13 behavior
		 but continue to split SENDMAILFLAGS
	       - Contents of skipped nested blocks could affect 'E', 'e', 'a',
		 and 'A' flags when they shouldn't have
	       - Prevent peeking into buffers on "Out of memory" errors
	       - Unquoted $\var expansions could alter the interpretation of
		 the following whitespace
	       - Prevent attempts to set LINEBUF to really huge values
	       - Optimize SWITCHRC = $_
	       - Use a secure PATH when processing /etc/procmailrc
	       - Prevent attempts to exercise a Linux kernel security hole
	       - Use 2^31-1 as the maximum score, even if LONG_MAX is larger
	    Changes to formail:
	       - Allow -n with -D and -s again -- corruption couldn't happen
		 after all
	       - Don't strip pre-colon whitespace until header is identified
	       - Properly handle NULs in the body when generating an autoreply
		 that keeps the body (could coredump)
	    Changes to autoconf:
	       - Avoid coredump on systems where sprintf() calls getenv()
	    Documented that $\var expansions are never split on whitespace
	    More manpage tweaks
	    Worked around linkers that don't support compile-time stripping
	       (for MacOS X)
	    Removed ':' and '@' from list of characters that can appear in
	       tempfile names
	    Called nice() when shouldn't have
	    Workaround SunOS 4.x compiler again
2001/06/28: v3.20
	    Changes to procmail:
	       - SECURITY: don't do unsafe things from signal handlers:
		  - ignore TRAP when terminating because of a signal
		  - resolve the host and protocol of COMSAT when it is set
		  - save the absolute path form of $LASTFOLDER for the comsat
		    message when it is set
		  - only use the log buffer if it's safe
	       - Support LMTP for delivery mode (not enabled by default)
	       - Preliminary support for using mmap() for `large' messages
		 (this doesn't work yet)
	       - SWITCHRC=zero-length-file didn't always abort the current
		 rcfile
	       - A race to create the mailspool would bounce one of the
		 messages due to an internal error
	       - LC_ in KEEPENV would preserve only the first LC_foo variable
	       - Strip runtime linker variables (LD_*) from environment on
		 all platforms
	       - Drop duplicate and malformed environment entries
	       - Multiple -a options will now set $2, $3, etc
	       - Command line assignments to INCLUDERC and SWITCHRC no longer
		 have any effect
	       - Worked around AIX 4.3.3 xlc compiler (incorrect file-scope
		 variable initialization)
	       - When delivering to a maildir, don't force the message to end
		 with an empty line
	       - Be more paranoid about leaking information between recipients
	       - Unset LOCKFILE if we can't actually lock it
	       - Set MAILDIR to '.' if the chdir fails
	       - LASTFOLDER was sometimes set by '?' conditions
	       - Buffer the log more efficiently
	       - Use the `standard' format for maildir filenames and retry
		 on name collision
	       - Rename by linking to prevent lossage
	       - Avoid dangling pointers when variable capture actions fail
	    Changes to autoconf:
	       - Check for enum support
	       - Warn people that non-ISO/ANSI C compilers might not be
		 supported in the future
	       - IRIX compiler (7.3.1) failed the const check from warnings
	    Changes to lockfile:
	       - Include the system mailbox lockfile path in the -v output
	       - Resist attempts to use a setuid lockfile command
	       - Fix infinite loop on -l, -r, or -s option with no value
	    Documented formail's treatment of >From_ lines as continuations
	       of the From_ line and warned of problems caused by non-RFC822
	       field names like 'Old-From '
	    Clarified procmail's treatment of $@ and $#
	    Fixed a man page formatting problem
	    Use long, not off_t, with fseek()/ftell()
	    Increase our paranoia: start to use strlcat()
	    The default MAILDIR is now configurable separately from the
	       default rcfile location
	    Include an RPM spec file in the examples directory for automated
	       builds
	    Include and use mkinstalldirs
	    Called nice() when shouldn't have (this time for sure!)
2001/06/29: v3.21
	    Changes to procmail:
	       - Incorrect prototype broke compilation on Tru64 UNIX
	       - INCLUDERC was broken by trying to be fancy and not fully
		 succeeding when support for multiple -a options was added
2001/09/10: v3.22
	    Changes to procmail:
	       - Regression bugs from 3.20:
		  - Broke compilation with K&R compilers
		  - procmail -p -m was overridding PATH
		  - maildir delivery included garbage in filenames
		  - Mismatched HOST in last rcfile didn't discard the message
		  - COMSAT wasn't turned off by an rcfile on the command line
	       - Catch overly long command line variable assignments
	       - If a command expansion is truncated, set PROCMAIL_OVERFLOW
		 and don't trim trailing (really middle) newlines
	       - If the comsat host can't be resolved, set COMSAT to "no"
	    Some fixes to the man pages
	    More paranoia: start to use strlcpy()
	    Generate safe temp and maildir filenames when the hostname
	       contains / or : by mapping them to \ooo
2001/09/13: v3.23pre
	    Changes to procmail:
	       - Don't coredump in comsat code if interrupted early
	       - Correctly handle COMSAT=on
	       - Once used the 'H' flag would never be cleared