------------------
DK-MILTER RELEASE NOTES
This listing shows the versions of the dk-milter package, the date of
release, and a summary of the changes in that release.
Bug and feature request (RFE) numbers that start with "SF" were logged
via Sourceforge (http://www.sourceforge.net) trackers. Those not so labelled
were logged internally at Sendmail, Inc.
1.0.2 2009/02/04
LIBDK: Fix bug in error message generation which could cause
crashes with very large error messages. Code copied
from dkim-milter.
1.0.1 2008/09/04
Set up required callbacks for OpenSSL thread-safety. Code copied
from dkim-milter.
LIBDK: Fix bug in relaxed canonicalization mode when dealing with
very large input lines with cached blank lines.
Problem reported by Mark Martinec.
1.0.0 2008/03/13
Use the current Authentication-Results: format (the -13 draft).
Patch from S. Moonesamy of Eland Systems.
Add improved mlfi_negotiate() function code, copied from dkim-milter,
which does things like symbol requesting and more intelligent
option negotiation. Patch from S. Moonesamy of Eland Systems.
Add hash buffering, copied from libdkim. Patch from S. Moonesamy
of Eland Systems.
Fix bug #SF1736559: in_addr_t is not universal. Problem reported
by Terry White.
Fix bug #SF1763715: Fix string management in mlfi_eoh() with respect
to mctx_domain, which could be left pointing to garbage when
using domain wildcarding. Problem noted by Ronald Huizer.
Activate _FFR_MULTIPLE_KEYS.
LIBDK: Copy the library options structure from libdkim.
LIBDK: Add dk_close().
LIBDK: Feature request #SF1872270: Rename parameters called "new"
in dk.h so that C++ compilers don't complain. Requested
by Paul Macintosh.
LIBAR: Eliminate a possible race condition in ar_dispatcher().
LIBAR: Timeouts passed to select() can't be bigger than 10^8.
Problem noted by S. Moonesamy of Eland Systems.
LIBAR: Fix bug #SF1852618: Handle default case of no "nameserver"
lines in /etc/resolv.conf. Problem noted by Mike Markley
of Bank of America.
LIBAR: Plug descriptor and memory leaks in ar_shutdown().
BUILD: Copy the unit test structure from libdkim and add a few basic
unit tests.
0.6.0 2007/05/31
Fix bug #SF1728696: Repair message corruption occurring when a
message body spams multiple milter writes. Reported
by Eric Singer.
Patch #SF1705006: Fix X-header malformation.
LIBAR: Fix bug #SF1537457: Add proper support for IPv6 nameservers.
Reported by Mark Martinec.
BUILD: Copy the consolidated build system from the dkim-milter
package.
Activate the following FFRs:
_FFR_LOG_SSL_ERRORS
_FFR_QUARANTINE
_FFR_REPORTINFO
0.5.0 2007/04/10
Copy several enhancements from the latest dkim-milter update:
o Support for 8.14 (milter v2) and the leading space patch from
dkim-milter.
o Fixes/enhancements under POPAUTH.
o Pass the correct length variable to RSA_sign() so that
the value returned is sane on 64-bit platforms.
o _FFR_ANTICIPATE_SENDMAIL_MUNGE
o Feature request #SF1497801: _FFR_QUARANTINE
Fix bug #SF1541450: Correct header selection in dk_hdrsigned().
Reported by Mark Martinec.
LIBDK: Fix bug #SF1537918: Add dk_geterror() to retrieve additional
diagnostic data from the API when a function call returns
DK_STAT_INTERNAL or something else whose cause isn't
readily apparent. Copied from libdkim.
LIBAR: Block signals that should be caught and handled elsewhere,
such as in libmilter.
0.4.2 2007/03/13
Fix bug #SF1509093, SF1611082: Set group ID as well as user ID when
"-u" is used on the command line. Patch from Vincent
Rivellino.
Fix bug #SF1514447: Re-query for the job ID in mlfi_eom() to
accomodate postfix's milter implementation. Copied from
dkim-filter; requested by Jakob Schlyter.
Fix bug #SF1541439: Fix mis-canonicalization of skipped headers
in "nofws" mode. Reported by Mark Martinec.
Fix bug #SF1541789: Stop spurious syntax errors on unsigned messages.
Problem reported by S. Moonesamy of Eland Systems.
0.4.1 2006/06/14
LIBDK: Properly handle key and policy records that don't have
whitespace after semicolons.
LIBDK: In dk_eom(), return DK_STAT_SYNTAX if dk_skipbody is set.
Failing to do so means a message with a signature header
below which there is no sender header will report a bogus
success status. Problem noted by Lennert Buytenhek.
0.4.0 2006/05/19
Remove spurious CRLFs injected into canonicalization when multiple
body chunks arrive from the MTA. Patch from Suzuki Takahiko
of Internet Initiative Japan, Inc.
If _FFR_REPORTINFO is enabled, don't call dkf_report() if no DomainKeys
context was ever created for a message.
Simplify dk_sterilize() a little, and handle failures from it.
Problem reported by Fredrik Pettai.
RFC2822 doesn't require any recipient headers, so remove those checks
inside _FFR_REQUIRED_HEADERS.
Fix bug #SF1485119: Canonicalize in the correct order when not using
"-H" on the command line. Problem noted by S. Moonesamy of
Eland Systems.
Activate _FFR_MACRO_LIST and _FFR_EXTERNAL_IGNORE_LIST.
LIBDK: New flag DK_OPTS_HDRLIST for dk_options().
0.3.4 2006/05/02
If _FFR_REPORTINFO is enabled, send reports on all failures, not just
those which aren't in test mode.
Ignore unknown tags in keys and policies, rather than returning an
error.
LIBDK: Return an error if the signing function returned success but
also reported a zero-length signature. Reported by
S. Moonesamy of Eland Systems.
LIBAR: Add a timeout to the I/O wait so that retransmissions
actually get done while waiting for activity.
0.3.3 2006/03/13
Fix test mode check at the end of mlfi_eom(), which was overriding
any configuration settings from the command line.
Reported by Arkadi Poliakevitch of Invidi Technologies.
Copy the value of -C before parsing it so the output of "ps" doesn't
get munged. Reported by Arkadi Poliakevitch of Invidi
Technologies.
Fix "-o", which wasn't actually working at all. Reported by Ben
Lentz.
Add _FFR_LOG_SSL_ERRORS which sends to syslog errors reported by
the OpenSSL libraries.
0.3.2 2005/12/12
Patch a small but definite memory leak. Reported by Ray Krebs
of eBay.
0.3.1 2005/12/02
Tolerate "b=" in signature headers at places other than the end
of the signature (and, in fact, other things at the end
of the signature). Reported by Jason Long. (Bug SF1234164)
Don't reject or report about messages which fail verification
when the sending domain advertises that it's in test mode.
Patch from Adrian Havill.
Fixes to POPAUTH compilation from S. Moonesamy of Eland Systems.
Allow "-d" to specify a list from which domains should be read,
and allow wildcarding in domain names. Requested by
Ray Krebs of eBay. (Feature request SF1312453)
Add "-o" command line option to allow certain headers to be omitted
from signing operations. Suggested by Ben Lentz.
(Feature request SF1314350)
LIBAR: Fix a build issue introduced in the last release.
- patch-aa was used to edit Makefile. Now it edits Makefile.in instead
- patch-a[cefh] are removed, equivalent fix applied
- patch-dns.c is added to silent prototype warning
(upstream)
- Update 1.20 to 1.23
-------------------
1.23: Feb.17,2015:
dns.c, smtp.c:
typo fixed: INET6 should be AF_INET6
1.22: Jun.9,2012:
smtp.c:
logging whole message in multi-lined SMTP response.
1.21: Jun.7,2012:
smtp.c, common.h, main.c:
'-2' option added to avoid piggybacking among domains.
(to support spec change on GoogleApps service)
dns.c:
definitions of functions with "static" are moved outside
to avoid "invalid strage class" errors.
*.c, extern.h:
log() is renamed to logg() to avoid conflict.
---------------------
0.040 2015-03-13 16:11:14-07:00 America/Los_Angeles
0.039 2015-02-23 19:27:47-08:00 America/Los_Angeles (TRIAL RELEASE)
Various POD updates (B. Estrade)
Exclude failing test on Mac OS with older Git (#41) (John Anderson)
POD and typo fixes (Doug Bell)
-------------------
1.12 2015-03-07
- The constructor now accepts named params. In addition to a magic_file param,
it now accepts follow_symlinks and uncompress params. These do what you'd
expect, tell the object to follow symlinks and/or uncompress files when
getting info from them. The old single parameter constructor is still
supported, but is no longer documented.
- The old APIs are now called "discouraged" rather than "deprecated" in the
documentation. I have no plan to remove these APIs any time soon or to make
them warn, so "deprecated" was too strong a word.
- Changed a test to hopefully skip rather than fail on OpenBSD systems.
1.11 2015-03-01
- The info_from_handle() method did not handle binary data correctly, which
could lead to incorrect results.
- The info_from_*() methods returned a full mime type with encoding in the
"mime_type" key rather than just the mime type alone.
- Allow File::LibMagic->new() to accept an array reference containing multiple
magic file paths.
