Jeremy Harris [Wed, 28 Jun 2017 14:25:12 +0000 (15:25 +0100)]
DKIM: Enforce any "h" field present in the DNS publickey record. This can be set to require specific
hash types, eg sha256, in signatues. There is an IETF draft in discussion which deprecates sha1 so this
feature may start to be used.
Jeremy Harris [Sun, 6 Aug 2017 14:44:13 +0000 (15:44 +0100)]
Logging: millisecond time on 'no MAIL' lines. Bug 2102
Jeremy Harris [Sat, 5 Aug 2017 14:51:36 +0000 (15:51 +0100)]
tidying
Jeremy Harris [Sat, 5 Aug 2017 14:04:39 +0000 (15:04 +0100)]
tidying
Jeremy Harris [Sat, 5 Aug 2017 13:06:36 +0000 (14:06 +0100)]
Debugging: millisecond timestamps. Bug 2102
Jeremy Harris [Fri, 4 Aug 2017 14:13:17 +0000 (15:13 +0100)]
Logging: millisecond QT and DT. Bug 2102
Jeremy Harris [Tue, 1 Aug 2017 20:49:04 +0000 (21:49 +0100)]
Use retval from sprintf while walking buffers
Jeremy Harris [Sun, 16 Jul 2017 16:21:35 +0000 (17:21 +0100)]
Tidying
Jeremy Harris [Sun, 30 Jul 2017 19:51:10 +0000 (20:51 +0100)]
Logging: millisecond timestamps. Bug 2102
No change to QT logging yet as that will need a spool format update
Jeremy Harris [Sun, 30 Jul 2017 13:37:56 +0000 (14:37 +0100)]
Retire historical build files from Makefile
Jeremy Harris [Sun, 30 Jul 2017 13:21:43 +0000 (14:21 +0100)]
Retire historical build files
Jeremy Harris [Sat, 22 Jul 2017 15:38:08 +0000 (16:38 +0100)]
On callout avoid SIZE every time but noncacheable rcpt main-verify. Bug 2151
Jeremy Harris [Tue, 18 Jul 2017 22:29:54 +0000 (23:29 +0100)]
Do not rely on linker identifying common strings
Jeremy Harris [Tue, 18 Jul 2017 22:10:33 +0000 (23:10 +0100)]
Fix cache-cold random callout verify. Bug 2147
Jeremy Harris [Sun, 16 Jul 2017 17:02:57 +0000 (18:02 +0100)]
Optimise scanning config for macros
Jeremy Harris [Sun, 16 Jul 2017 13:36:38 +0000 (14:36 +0100)]
Fix macro_predef build for auth_cyrus_sasl
Jeremy Harris [Sun, 16 Jul 2017 12:52:57 +0000 (13:52 +0100)]
Fix build for earlier version Berkeley DB
Jeremy Harris [Sat, 15 Jul 2017 18:55:59 +0000 (19:55 +0100)]
tidying
Jeremy Harris [Wed, 5 Jul 2017 11:57:24 +0000 (12:57 +0100)]
Use the BDB "environment" so that a database config file is not looked for. Bug 2141
Jeremy Harris [Tue, 11 Jul 2017 19:52:34 +0000 (20:52 +0100)]
Add directory name as new arg to EXIM_DBOPEN
Jeremy Harris [Wed, 5 Jul 2017 13:30:05 +0000 (14:30 +0100)]
Add $smtp_command_history variable
Jeremy Harris [Sun, 9 Jul 2017 15:12:45 +0000 (16:12 +0100)]
tidying
Jeremy Harris [Sun, 9 Jul 2017 13:29:31 +0000 (14:29 +0100)]
Avoid creating before-buffer pointer. Bug 2145
Found by: ASAN / <ago@gentoo.org>
Jeremy Harris [Sat, 8 Jul 2017 16:54:44 +0000 (17:54 +0100)]
Prebuild the data structure for builtin macros
Jeremy Harris [Sat, 8 Jul 2017 12:24:38 +0000 (13:24 +0100)]
Prebuild the data structure for builtin macros
Jeremy Harris [Sat, 8 Jul 2017 12:59:49 +0000 (13:59 +0100)]
Split macro name storage out from macro definition struct
Heiko Schlittermann (HS12-RIPE) [Tue, 4 Jul 2017 11:33:08 +0000 (13:33 +0200)]
exigrep: use less --quite-if-one-screen --no-init
Jeremy Harris [Sun, 2 Jul 2017 09:30:48 +0000 (10:30 +0100)]
Transform string_append_listele{,_n}() to proper expanding-string triplet interface
(but do always maintain a nul-term string result). This avoids always copying the
previous list version, and should do fewer allocs too.
Heiko Schlittermann (HS12-RIPE) [Fri, 30 Jun 2017 15:22:08 +0000 (17:22 +0200)]
Configure the default opendmarc.tlds file in EDITME
Heiko Schlittermann (HS12-RIPE) [Thu, 29 Jun 2017 10:42:12 +0000 (12:42 +0200)]
Build: set LC_ALL=C for the makefile configuration
Otherwise Perl may complain about missing locales, which
in turn confuses the Configure-Makefile script when parsing
the output from Perl
Heiko Schlittermann (HS12-RIPE) [Wed, 28 Jun 2017 21:42:17 +0000 (23:42 +0200)]
testsuite: don't break on '+' in trusted-config-list path
Heiko Schlittermann (HS12-RIPE) [Wed, 28 Jun 2017 08:45:51 +0000 (10:45 +0200)]
Update Changelog about -p/-oMr
Phil Pennock [Tue, 27 Jun 2017 18:36:53 +0000 (14:36 -0400)]
Fix ChangeLog notice re manualroute option ordering
We do *NOT* permit options before hostnames. It's always hostnames
before options. It's _within_ the options that the use of a transport
name as an option name had been accidentally constrained to come last,
which was then fixed in commit
833c70bc.
I goofed in describing this.
Heiko Schlittermann (HS12-RIPE) [Tue, 27 Jun 2017 15:35:20 +0000 (17:35 +0200)]
Fix appendfile.c errmsg output that broke with d291c7
Phil Pennock [Mon, 26 Jun 2017 22:39:29 +0000 (18:39 -0400)]
Repair manualroute transport name not-last option
Commit
55414b25b did const-ification surgery which removed the old
"temporarily drop a NUL in here" anti-pattern, but one of the
Ustrcmp->Ustrncmp conversions was missed, for transport-name comparison.
As a result, since Exim 4.85, we've had an undocumented
requirement-by-bug that the transport name be the last such option
given. Restore old behaviour.
Still missing a test-case to deter regression.
Heiko Schlittermann (HS12-RIPE) [Sun, 25 Jun 2017 21:35:06 +0000 (23:35 +0200)]
Bail out if a configuration file starts with a byte order mark
Heiko Schlittermann (HS12-RIPE) [Sun, 25 Jun 2017 21:29:18 +0000 (23:29 +0200)]
Heiko Schlittermann (HS12-RIPE) [Sun, 25 Jun 2017 20:54:08 +0000 (22:54 +0200)]
Add quota/quota_filecount transport option modifier "no_check" Bug 1115
This option modifier allows to ignore the quota limits, but update
the maildirsize file.
Heiko Schlittermann (HS12-RIPE) [Sun, 25 Jun 2017 17:57:57 +0000 (19:57 +0200)]
Doc: check for w3m existence
w3m is called as non-last command in pipelines. A missing w3m
creates empty output and Make doesn't get it, because the last command
of the pipe (e.g. Tidytxt) doesn't return failure on empty input.
Jeremy Harris [Tue, 13 Jun 2017 18:01:00 +0000 (19:01 +0100)]
When pipelining under TLS do not treat fd-readability as tls-readability
Jeremy Harris [Fri, 9 Jun 2017 00:20:07 +0000 (01:20 +0100)]
Add compile-time checks on expanding-string function use
Heiko Schlittermann (HS12-RIPE) [Wed, 31 May 2017 21:08:56 +0000 (23:08 +0200)]
Cleanup (prevent repeated use of -p/-oMr to avoid mem leak)
Heiko Schlittermann (HS12-RIPE) [Tue, 13 Jun 2017 15:36:44 +0000 (17:36 +0200)]
Fix log line corruption for DKIM status
Jeremy Harris [Sun, 11 Jun 2017 19:58:21 +0000 (20:58 +0100)]
On a continued-transport delivery, do not check the hostlist for mismatch in the
transport dispatch level when the transport does hosts-override. Instead do the
full trasport process call and let it decide on compatibility with the connection.
Jeremy Harris [Sun, 11 Jun 2017 22:47:08 +0000 (23:47 +0100)]
In tls-proxy process take case to close unused pipe ends, select also for exception conditions on data fds
Jeremy Harris [Sun, 11 Jun 2017 20:03:21 +0000 (21:03 +0100)]
Testsuite: testcase for continued-delivery with transport hosts_override
Jeremy Harris [Fri, 2 Jun 2017 13:22:42 +0000 (14:22 +0100)]
Fix logging of delivery port for callout lazy-close, when port specified by router
Jeremy Harris [Fri, 2 Jun 2017 10:58:28 +0000 (11:58 +0100)]
Testsuite: stabilize output sequence for 2x35/6
Jeremy Harris [Sat, 3 Jun 2017 15:51:05 +0000 (16:51 +0100)]
Docs: tweak "sock" malware interface description
Jeremy Harris [Sat, 3 Jun 2017 12:39:18 +0000 (13:39 +0100)]
Fix crash in transport, on second smtp-connect fail for a list of target hosts
Reported as the sequence:
1MX: 554 on connect (banner)
2MX: TCP conn timeout
Jeremy Harris [Fri, 2 Jun 2017 11:13:06 +0000 (12:13 +0100)]
Docs: add missing dollarsign
Jeremy Harris [Wed, 31 May 2017 16:31:37 +0000 (17:31 +0100)]
Testsuite: stabilize output sequence for case 4530
Jeremy Harris [Tue, 30 May 2017 22:43:09 +0000 (23:43 +0100)]
Testsuite: stabilize output sequence for case 0906
Jeremy Harris [Mon, 29 May 2017 16:23:12 +0000 (17:23 +0100)]
Malware: make "sock" cmdline default usable. Bug 2111
Phil Pennock [Fri, 26 May 2017 20:41:42 +0000 (16:41 -0400)]
doc: string2 not expanded, don't imply otherwise
A simple reading which skips the later paragraph would mislead people;
this was remnant text from before the EXPAND_LISTMATCH_RHS changed the
default behaviour.
A couple of paragraphs later we contradicted this now-misleading
parenthetical, but that's not enough. Nuke it.
Jeremy Harris [Thu, 25 May 2017 15:27:43 +0000 (16:27 +0100)]
typo
Jeremy Harris [Sun, 21 May 2017 15:51:24 +0000 (16:51 +0100)]
TLS: do not use DH_bits for LibreSSL; it is not supported (as of LibreSSL 2.3.2)
Jeremy Harris [Sun, 21 May 2017 15:42:05 +0000 (16:42 +0100)]
Debug: more info during server-side authentication
Jeremy Harris [Mon, 15 May 2017 16:36:37 +0000 (17:36 +0100)]
OpenSSL: add sender_host_address info to certificate verify error log messages when in verify callout
Phil Pennock [Sun, 21 May 2017 03:17:21 +0000 (23:17 -0400)]
Use DH_bits for OpenSSL 1.1.0
Fixes bug 2095
Jeremy Harris [Fri, 19 May 2017 21:55:25 +0000 (22:55 +0100)]
TLS: PIPELINING under OpenSSL
Jeremy Harris [Wed, 17 May 2017 22:19:43 +0000 (23:19 +0100)]
Testsuite: delays in 2x3[567]
Jeremy Harris [Wed, 17 May 2017 20:05:33 +0000 (21:05 +0100)]
Testsuite: add delays for predictable log sequencing in 2x36, log-sorting for 2x38
Jeremy Harris [Wed, 17 May 2017 16:32:30 +0000 (17:32 +0100)]
TLS: do decent PIPELINING under TLS, at least with GnuTLS
I've not found an equivalent in OpenSSL of gnutls_record_cork() nor gnutls_record_check_pending() yet.
Jeremy Harris [Sun, 14 May 2017 16:06:17 +0000 (17:06 +0100)]
Revert "DANE: do not trust a non-dnssec NXDOMAIN return for the TLSA lookup"
This reverts commit
5d6bdf01a921a88030e9baec7ba5f238da90e979.
Phil Pennock [Tue, 9 May 2017 20:00:58 +0000 (16:00 -0400)]
Add option commandline_checks_require_admin
May help with scenarios already so broken that bug report 2118 is
actually an issue (Wordpress vuln).
Jeremy Harris [Sun, 7 May 2017 16:40:41 +0000 (17:40 +0100)]
Testsuite: add DANE cases for DNS secure no-TLSA lookups
Jeremy Harris [Sun, 7 May 2017 14:37:18 +0000 (15:37 +0100)]
Testsuite: add DANE testcase for TLSA lookup SERVFAIL
Jeremy Harris [Sun, 7 May 2017 16:49:58 +0000 (17:49 +0100)]
DANE: do not trust a non-dnssec NXDOMAIN return for the TLSA lookup
Jeremy Harris [Sun, 7 May 2017 16:46:49 +0000 (17:46 +0100)]
DANE: avoid info leak by checking TLSA dnssec before connecting to MX
Jeremy Harris [Sun, 7 May 2017 13:41:49 +0000 (14:41 +0100)]
tidying
Jeremy Harris [Sun, 7 May 2017 15:55:12 +0000 (16:55 +0100)]
Testsuite: fix fakens to not claim that an unsupported NXDOMAIN is dnssec-AD
Jeremy Harris [Sat, 6 May 2017 20:57:46 +0000 (21:57 +0100)]
Enable use of sendfile on FreeBSD
Jeremy Harris [Sat, 6 May 2017 19:19:31 +0000 (20:19 +0100)]
Enable use of sendfile on FreeBSD
Jeremy Harris [Sat, 6 May 2017 20:01:45 +0000 (21:01 +0100)]
Teach SMTP input sync check ("input sent too soon") about SMTP input buffering
Jeremy Harris [Sat, 6 May 2017 17:51:02 +0000 (18:51 +0100)]
Docs: add note on Received-By: header creation under cutthrough
Jeremy Harris [Thu, 4 May 2017 17:39:44 +0000 (18:39 +0100)]
testsuite: tidying
Jeremy Harris [Thu, 4 May 2017 15:59:46 +0000 (16:59 +0100)]
CHUNKING / wire-format spool: use block-copies for receiption
Jeremy Harris [Mon, 1 May 2017 13:13:57 +0000 (14:13 +0100)]
testsuite output changes
Jeremy Harris [Mon, 1 May 2017 11:42:32 +0000 (12:42 +0100)]
testsuite: tidying
Jeremy Harris [Mon, 1 May 2017 11:08:17 +0000 (12:08 +0100)]
tidying: coverity
Jeremy Harris [Sun, 30 Apr 2017 21:11:27 +0000 (22:11 +0100)]
Fix continue_more on TLS connection. Bug 2104
Jeremy Harris [Sun, 30 Apr 2017 13:39:37 +0000 (14:39 +0100)]
Docs: more info on wire-format spoolfiles
Jeremy Harris [Sat, 29 Apr 2017 12:28:38 +0000 (13:28 +0100)]
tidying: coverity fixes
Jeremy Harris [Fri, 28 Apr 2017 23:19:45 +0000 (00:19 +0100)]
Docs: describe operation of continued-connection TLS
Jeremy Harris [Fri, 28 Apr 2017 22:54:35 +0000 (23:54 +0100)]
Support wire-format spoolfiles
Jeremy Harris [Wed, 26 Apr 2017 14:28:22 +0000 (15:28 +0100)]
Enable use of sendfile
Jeremy Harris [Wed, 26 Apr 2017 19:18:31 +0000 (20:18 +0100)]
Tweak debug output
Jeremy Harris [Mon, 24 Apr 2017 12:41:09 +0000 (13:41 +0100)]
Docs: add 8BITMIME status field to log line info
Jeremy Harris [Wed, 19 Apr 2017 15:39:23 +0000 (16:39 +0100)]
Docs: note that dkim_domain can take a list for signing
Jeremy Harris [Sun, 23 Apr 2017 12:52:15 +0000 (13:52 +0100)]
Fix DISABLE_DKIM biuld
Jeremy Harris [Sun, 23 Apr 2017 11:20:43 +0000 (12:20 +0100)]
DKIM: avoid use of temporary file for signing
Jeremy Harris [Tue, 18 Apr 2017 14:13:20 +0000 (15:13 +0100)]
CHUNKING: pipeline data right after the BDAT command
Jeremy Harris [Mon, 17 Apr 2017 19:43:23 +0000 (20:43 +0100)]
Reduce number of places knowing about filename-construction for mbox file-for-scanning
Jeremy Harris [Sun, 16 Apr 2017 15:32:06 +0000 (16:32 +0100)]
Rework detection of multiple ports on a given IP, for the daemon status line
Previously only spotted adjacent cases in the address list, now a full scan
Jeremy Harris [Sun, 16 Apr 2017 16:19:23 +0000 (17:19 +0100)]
tidying
Jeremy Harris [Sat, 15 Apr 2017 15:22:52 +0000 (16:22 +0100)]
Transport: fix smtp under combo of mua_wrapper and limited max_rcpt
Jeremy Harris [Fri, 14 Apr 2017 13:29:40 +0000 (14:29 +0100)]
Builtin macros: note config trigger line in debug output
Phil Pennock [Thu, 13 Apr 2017 03:22:36 +0000 (23:22 -0400)]
$SOURCE_DATE_EPOCH support for build date stamp
If, and _only_ if, $SOURCE_DATE_EPOCH is found in environ during build,
use it to set the timestamp embedded in the binary instead of using
__DATE__ and __TIME__ cpp directives.
This per <https://reproducible-builds.org/specs/source-date-epoch/>
spec. It's sane and sensible, without removing date stamping which
matters. The examples encourage packagers to use timestamps which
do change when they backport patches, so that the date remains useful
for distinguishing builds instead of claiming one date forever across
multiple patchlevels.
This change written so that the old behaviour and code is used if the
environment variable is not found, to better continue to support ancient
platforms with other variants of date(1).
Built with and without an override, on macOS.
Jeremy Harris [Sun, 9 Apr 2017 14:42:49 +0000 (15:42 +0100)]
compiler quietening