Add doc to load_ex_ip_helper_procedures().

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66060
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Only take process id of localhost client connection entries.

This patch makes sure, that the process id is fetched only from localhost
 client address connections fetched from the tcp table.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66060
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fixed bug of unsupported GetExtendTcpTable() on Windows XP SP2 and earlier.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66060
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Dump fetched sid on return of _dbus_getsid().

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66060
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Add WINXP <= SP3 support to _dbus_getsid().

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66060
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

test/marshal: Ensure we use suitably aligned buffers

This test was failing on s390; though it could fail
on other platforms too.  Basically we need to be sure
we're passing at least word-aligned buffers to the
demarshalling code.  malloc() will do that for us.

https://bugs.freedesktop.org/show_bug.cgi?id=67279

NEWS for 1.7.x

cmake: do not bind to any particular POSIX C standard

This caused build failures on FreeBSD. Defining _POSIX_C_SOURCE to
a particular version will disable common non-POSIX extensions like
PF_UNIX, and on some systems will also disable features of later
POSIX versions, like IPv6. If we don't ask for a specific version,
we'll get some sort of sensible default.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66257
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
[made the commit message more concise -smcv]
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Merge branch 'dbus-1.6'

Conflicts:
NEWS

NEWS for 1.6.x

Make the test for #53499 more obviously correct

Test: add a test case for escaping byte > 127

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=53499
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fix: a non ascii byte will trigger BadAddress error

If a byte in DBusString *unescaped isn't a ascii byte, which will be
cast to char (signed char on most of platform), so that's the issue
unsigned char cast to signed char. e.g. "\303\266" is a valid unicode
character, if everything goes right, it will be escaped to "%c3%b6".
However, in fact, it escaped to "%<garbage-byte>3%<garbage-byte>6".

_dbus_string_append_byte_as_hex() take an int parameter, so negative
byte is valid, but cause get a negative index in array. So garbage value
will get. e.g. '\303' --> hexdigits[((signed byte)(-61)) >> 4] is
hexdigits[-4].

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=53499
Sgne-off-by: Chengwei Yang <chengwei.yang@intel.com>
[fixed whitespace -smcv]
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

tests to embedded tests: replaced in cmake files

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66291

tests to embedded tests: replaced in automake files

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66291

tests to embedded tests: replaced in libdbus

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66291

tests to embedded tests: replaced in tools

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66291

tests to embedded tests: replaced in dbus-daemon

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66291

DBusString: fix a typo

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
[separated out of a larger patch -smcv]
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66300

NEWS for 33001

cmake: align dir watch backend detection with autotools

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66257

cmake: terminate to generate makefiles due to fatal error

FATAL isn't a valid key for message according to cmake document here.
http://www.cmake.org/cmake/help/v2.8.8/cmake.html#command%3amessage

Due to the real fatal error, FATAL_ERROR should be used to terminate
cmake from continue generating makefiles.

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66257

cmake: get rid of useless commented out code

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66257

dir-watch: remove dnotify backend

dnotify as a dir watch backend is broken since Jan 2010 (almost 3.5
years). According to fd.o: #33001, it's no harm to remove dnotify from
this project.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=33001
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

cmake: clean up libxml2 glue

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66257

NEWS for master

Merge branch 'dbus-1.6'

Conflicts:
NEWS

NEWS

Remove invoke of va_end before va_start

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66300

Fix two small typos in README.win

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66300

kqueue: remove unused variable

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66035

dbus-launch: do not verbose output if build with verbose mode disabled

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66068

dbus-launch: align document

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66068

dbus-launch: fix coding style

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66068

FreeBSD: explicit include signal.h to fix build failure

In Linux envrionment, signal.h included by sys/wait.h, however, this
isn't the case in FreeBSD. So explicit include it to fix build failure.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66197
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fix build failure if build with tests but without verbose mode

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66005
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

_dbus_message_iter_check: compile non-dummy version if asserting, too

Explicitly define macros to get less confusing conditions

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65990
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fix build failure if build with checks but without asserts

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65990
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Convert a{sv} helpers from Stats into generic utility code

Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54445
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>

cmake: remove a duplicate line and fix coding style

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66142
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS, noting build-time configuration change

cmake: fix a typo DBUS_DISABLE_ASSERTS should be DBUS_DISABLE_ASSERT

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66142
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Use Doxyfile.in for cmake build system too.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=64875
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fix build failure if build with "--enable-stats"

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66004
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

DBus Specification: remove incorrect table description

I doubt this line was copied by mistake from section "Summary of types".

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=66115

Let dbus-daemon not print unavailable options on windows.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=42441
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

connection: Add incoming message serials to verbose mode logging

This helps pair up messages with verbose mode output from the other
side of a connection.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=63047
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS

dbus-send: Fix fail to run without "--dest" option

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65923
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Merge branch 'dbus-1.6'

Conflicts:
NEWS

NEWS for #65959

DBusString: fix may crash if try to free an uninitialized str

If the str will be freed hasn't been initialized by _dbus_string_init
correctly, _dbus_string_free may crash due to trying to free an
undefined memory.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65959
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Doc: fix invalid usage of doxygen @param command

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65755

Fix of cmake xmldoc dependencies chain.

This patch fixes an issues that xml documentation is generated on all builds regardless
if related files has been changed or not.

The patch adds a global xmldoc make target to which all generated html or man files are
added as build dependency. Each dependency itself depends on related CMakeLists.txt and
the xml file generated from the related xml.in file.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=64058
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fixed wrong path for generated xml files when creating man pages with cmake.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=64058
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Add a statically-initialized implementation of _dbus_lock() on glibc systems

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Anas Nashif <anas.nashif@intel.com>

Always initialize threading before allocating a dynamic mutex

Dynamic allocation of mutexes can fail anyway, so this is easy.

Justification for not keeping the dummy mutex code-paths, even as an
opt-in thing for processes known to be high-performance and
single-threaded: real mutexes only cut the throughput of
test/dbus-daemon.c by a couple of percent on my laptop (from around
6700 to around 6600 messages per second), and libdbus crashes caused
by not calling dbus_threads_init_default() are sufficiently widespread
that they're wasting a lot of everyone's time.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Anas Nashif <anas.nashif@intel.com>

Make taking a global lock automatically initialize locking if needed

This lets them be thread-safe by default, at the cost that they can
now fail.

init_uninitialized_locks() and init_global_locks() must now both
reimplement the equivalent of _dbus_register_shutdown_func(), by using
_dbus_platform_rmutex_lock() on the same underlying mutex around a call
to _dbus_register_shutdown_func_unlocked().

This is because if they used the usual _DBUS_LOCK() API (as
_dbus_register_shutdown_func() does), it would automatically try to
initialize global locking, leading to infinite recursion.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Anas Nashif <anas.nashif@intel.com>

Replace individual global-lock variables with an array of DBusRMutex *

This means we can use a much simpler code structure in data-slot
allocators: instead of giving them a DBusRMutex ** at first-allocation,
we can just give them an index into the array, which can be done
statically.

It doesn't make us any more thread-safe-by-default - the mutexes will
only actually be used if threads were already initialized - but it's
substantially better than nothing.

These locks really do have to be recursive: for instance,
internal_bus_get() calls dbus_bus_register() under the bus lock,
and dbus_bus_register() can call _dbus_connection_close_possibly_shared(),
which takes the bus lock.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>
Reviewed-by: Anas Nashif <anas.nashif@intel.com>

start spec 0.22 development

Fix Werror=unused-function if build without X11

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65712
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Start on 1.7.5, add release name for 1.7.4

Add release name for 1.6.12

Prepare 1.7.4 for tomorrow

Move libdbus-init-win.cpp to its own convenience library

Otherwise libdbus-1 ends up linked as if it contained C++, even
on Unix, where it doesn't; in turn, that makes it export all the
underscore-prefixed symbols that aren't meant to be exported.

Reviewed-by: Thiago Macieira <thiago@kde.org>
Reviewed-by: David Zeuthen <davidz@redhat.com>

NEWS

Fix an incorrect comment

_dbus_threads_lock_platform_specific() is implemented on Windows now.

Revert "start spec 0.22 development"

This reverts commit 82b3d94ab10f16e6c6c18eb3bc76594cd63e3854.

_dbus_system_logv: copy the va_list here too

This would crash if HAVE_SYSLOG_H is defined, HAVE_DECL_LOG_PERROR
is false, and the platform calling convention is that va_list is a
struct. Verified on Linux by undefining HAVE_DECL_LOG_PERROR.

Reviewed-by: Colin Walters <walters@verbum.org>

Merge branch 'dbus-1.6'

Conflicts:
NEWS
configure.ac

Start 1.6.13

Prepare embargoed release for tomorrow

Add a test-case for CVE-2013-2168

Reviewed-by: Thiago Macieira <thiago@kde.org>
[build system adjusted to compile it even if we don't have GLib -smcv]

CVE-2013-2168: _dbus_printf_string_upper_bound: copy the va_list for each use

Using a va_list more than once is non-portable: it happens to work
under the ABI of (for instance) x86 Linux, but not x86-64 Linux.

This led to _dbus_printf_string_upper_bound() crashing if it should
have returned exactly 1024 bytes. Many system services can be induced
to process a caller-controlled string in ways that
end up using _dbus_printf_string_upper_bound(), so this is a denial of
service.

Reviewed-by: Thiago Macieira <thiago@kde.org>

NEWS for 1.6.x

Fix dbus-daemon crash due to invalid service file

dbus-daemon will crash due to invalid service file which key/value
starts before section. In that situation, new_line() will try to access
invalid address.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=60853
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS

Also update README to not mention libxml2.

dbus-send: check usage a bit strictly

This commit does several more strictly check for dbus-send as its usage
suggested.

* now --address is an invalid option but --address=, this just like the
  others, say --reply-timeout=, --dest=, --type=
* --print-reply= only take an optional argument "=literal"
* --print-reply= will cause error with missing MSEC and invalid MSEC
  will cause invalid value error
* --dest= will cause error with missing a NAME and also call
  dbus_validate_bus_name to verify the NAME

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65424
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Assign default value to enable compiler coverage

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65424
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fix a typo: enable_x11 -> have_x11

From git history, enable_x11 was used to track have_x11, but it's
useless now.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65443
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

XML: hard depends on expat and delete libxml

[The libxml code path has been broken for at least 2.5 years, and Expat
is tiny, so there seems no point in supporting both. -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=20253
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Fix dbus-daemon crash due to invalid service file

dbus-daemon will crash due to invalid service file which key/value
starts before section. In that situation, new_line() will try to access
invalid address.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=60853
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Do not suggest user to do 'make' if configure failed

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=65415
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

massively simplify run-with-tmp-session-bus.sh by using dbus-run-session

It turns out that if you don't second-guess the system by catching
SIGINT, the right things happen: it's received by every program in the
foreground process group, including dbus-run-session and dbus-daemon.
Neither of those catch SIGINT (unlike dbus-launch) so they'll exit
gracefully without the wrapper script needing to do anything special.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39196
Reviewed-by: Colin Walters <walters@verbum.org>

Recomend dbus-run-session over dbus-launch for starting text-mode sessions

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39196
Reviewed-by: Colin Walters <walters@verbum.org>
[reformatted from roff to Docbook -smcv]

Put dbus-run-session through doclifter and adjust to match other man pages

Add dbus-run-session

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39196
Reviewed-by: Colin Walters <walters@verbum.org>

Fix build error: unused-result

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

When "activating" systemd, handle its special case better

When dbus-daemon receives a request to activate a systemd service before
systemd has connected to it, it enqueues a fake request to "activate"
systemd itself (as a way to get a BusPendingActivationEntry to track the
process of waiting for systemd). When systemd later joins the bus,
dbus-daemon sends the actual activation message; any future activation
messages are sent directly to systemd.

In the "pending" code path, the activation messages are currently
dispatched as though they had been sent by the same process that sent
the original activation request, which is wrong: the bus security
policy probably doesn't allow that process to talk to systemd directly.
They should be dispatched as though they had been sent by the
dbus-daemon itself (connection == NULL), the same as in the non-pending
code path.

In the worst case, if the attempt to activate systemd timed out, the
dbus-daemon would crash with a (fatal) warning, because in this special
case, activation_message is a signal with no serial number, whereas the
code to send an error reply is expecting a method call with a serial
number.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=50199
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Tested-by: Ma Yu <yu.ma@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS

Fix build error: unused-result

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

doc: fix a little bit for dbus-send

Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

When "activating" systemd, handle its special case better

When dbus-daemon receives a request to activate a systemd service before
systemd has connected to it, it enqueues a fake request to "activate"
systemd itself (as a way to get a BusPendingActivationEntry to track the
process of waiting for systemd). When systemd later joins the bus,
dbus-daemon sends the actual activation message; any future activation
messages are sent directly to systemd.

In the "pending" code path, the activation messages are currently
dispatched as though they had been sent by the same process that sent
the original activation request, which is wrong: the bus security
policy probably doesn't allow that process to talk to systemd directly.
They should be dispatched as though they had been sent by the
dbus-daemon itself (connection == NULL), the same as in the non-pending
code path.

In the worst case, if the attempt to activate systemd timed out, the
dbus-daemon would crash with a (fatal) warning, because in this special
case, activation_message is a signal with no serial number, whereas the
code to send an error reply is expecting a method call with a serial
number.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=50199
Signed-off-by: Chengwei Yang <chengwei.yang@intel.com>
Tested-by: Ma Yu <yu.ma@intel.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Remove unused global mutexes for win_fds, sid_atom_cache

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>

Turn a runtime assertion into a compile-time assertion

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>

NEWS for part 1 of fd.o#54972

dbus_threads_init_default, dbus_threads_init: be safe to call at any time

On Unix, we use a pthreads mutex, which can be allocated and
initialized in global memory.

On Windows, we use a CRITICAL_SECTION, together with a call to
InitializeCriticalSection() from the constructor of a global static
C++ object (thanks to Ralf Habacker for suggesting this approach).

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>

Fix compiler warnings when X11 autolaunch and launchd are both disabled

From the department of "if it isn't tested, it doesn't work". I tried
compiling dbus without an assortment of optional features:

    in_builddir ~/build/dbus/legacy ${MR_REPO}/configure \
        --enable-developer --enable-maintainer-mode --enable-tests \
        dbus_cv_sync_sub_and_fetch=no \
        --disable-selinux \
        --disable-inotify \
        --disable-dnotify \
        --disable-epoll \
        --disable-kqueue \
        --disable-launchd \
        --disable-systemd \
        --disable-libaudit \
        --without-valgrind \
        --disable-x11-autolaunch \
        && ...

and it resulted in -Wunused warnings.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=64362
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Thiago Macieira <thiago@kde.org>

dbus_threads_init: call _dbus_threads_init_platform_specific()

This reverses the relationship between these two functions.
Previously, dbus_threads_init() wouldn't allocate dbus_cond_event_tls
on Windows, call check_monotonic_clock on Unix, or call
_dbus_check_setuid on Unix.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=54972
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Alban Crequy <alban.crequy@collabora.co.uk>
Reviewed-by: Ralf Habacker <ralf.habacker@freenet.de>