Promote DBusBasicValue and DBus8ByteStruct to be API

In practice, D-Bus bindings end up reinventing DBusBasicValue anyway,
so it might as well be API.

Also stop claiming that all basic-typed values are guaranteed to fit in
8 bytes - this is not true if your platform has more than 8-byte pointers
(I'm not aware of any such platform now, but let's not rule it out).

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=11191
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Will Thompson <will.thompson@collabora.co.uk>

Development version

Prepare 1.5.10

Mark shell-test as modular and installable

This test needs non-public API and so is statically linked, but is OK
to install.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=42811
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Allow a reduced set of installable tests (none yet) to be built without GLib

--enable-modular-tests=auto will build as many as possible, perhaps
excluding the GLib ones. --enable-modular-tests=yes or --enable-tests=yes
will insist on having GLib, to be able to run everything.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=42811
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

dbus-memory: add optional checking for system malloc() (etc.) failing

If tests are enabled and DBUS_MALLOC_CANNOT_FAIL is set, abort on system
malloc() failures (as GLib's g_malloc does). This can be used in
conjunction with a resource limit, to turn runaway memory leaks into a
debuggable core-dump.

Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41048

dbus_realloc: don't crash if realloc() returns NULL while using guards

Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=41048

Turn the non-valgrind code path into inline functions to avoid compiler warnings

Recent gcc will warn if you have a statement that's just a macro
expanding to (0), but not if you have an inline stub function that
always returns 0, so let's do the latter.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

DBusTransport: don't include dbus-marshal-header.h either

No longer needed.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Move _dbus_check_fdleaks_enter to dbus-message-internal.h

This means the bus test can use them without needing
dbus-message-private.h, reducing its view of message internals.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

DBusTransport: don't use dbus-message-private.h

It seems it's no longer needed here.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Zero-initialize DBusCounter at allocation

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Do not AC_SUBST systemd flags from PKG_CHECK_MODULES

They're automatically substituted already.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Invoke PKG_PROG_PKG_CONFIG early in configure.ac

Otherwise, autoconf can generate wrong code, because our first use of
PKG_CHECK_MODULES is conditional.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

internal_bus_get: centralize error handling

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Add a macro to centralize checking for string-like types

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Remove declaration of _dbus_type_is_fixed, no longer exists

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

_dbus_verbose_real: tidy conditional prototype to not break nesting

The mismatched opening parenthesis caused vim syntax highlighting to
consider every subsequent brace in the file to be an error, which was
pretty annoying.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46095
Reviewed-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>

Remove _dbus_condvar_wake_all and both of its implementations

Neither was used, and the Windows version could lead to live-locks.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43744
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=44609
Reviewed-by: Thiago Macieira <thiago@kde.org>

Never use non-libdbus threading primitives

This lets us simplify considerably, by assuming that we always have both
recursive and suitable-for-condition-variable mutexes.

The Windows implementation has been compiled (on 32-bit mingw-w64)
but not tested. Justification for the approach used on Windows,
and in particular, using the existing "non-recursive" locks as if
they were recursive:

* We've been using them in conjunction with condition variables all
  along, so they'd better be suitable

* On fd.o #36204, Ralf points out that mutexes created via CreateMutex
  are, in fact, recursive

* Havoc's admonitions about requiring "Java-style" recursive locking
  (waiting for a condition variable while holding a recursive lock
  requires releasing that lock n times) turn out not to apply to
  either of our uses of DBusCondVar in DBusConnection, because the
  lock is only held for a short time, without calling into user code;
  indeed, our Unix implementation isn't recursive anyway, so if
  the Windows implementation reaches the deadlocking situation
  somehow (waiting for condition variable while locked more than once),
  the Unix implementation would already have deadlocked on the same
  code path (trying to lock more than once)

One possible alternative to a CreateMutex mutex for use with condition
variables would be a CRITICAL_SECTION. I'm not going to implement this,
but Windows developers are welcome to do so.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=36204
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43744
Reviewed-by: Thiago Macieira <thiago@kde.org>

dbus-threads: improve documentation

This reinstates documentation for _dbus_mutex_new() and
_dbus_mutex_free(), and fixes some typos spotted during review.
It also documents the newly-introduced functions.

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

Use actual recursive pthreads mutexes, rather than NIH'ing them, wrong

Very loosely based on a patch from Sigmund Augdal.

For the moment, we make PTHREAD_MUTEX_RECURSIVE a hard requirement:
it's required by POSIX 2008 Base and SUSv2.

If your (non-Windows) platform doesn't have PTHREAD_MUTEX_RECURSIVE,
please report a bug on freedesktop.org bugzilla with details of the
platform in question.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43744
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Thiago Macieira <thiago@kde.org>

Allow both recursive and non-recursive mutexes to be supplied

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43744
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Thiago Macieira <thiago@kde.org>

Distinguish between two flavours of mutex

dbus-threads.h warns that recursive pthreads mutexes are not compatible
with our expectations for condition variables. However, the only two
condition variables we actually use only have their corresponding
mutexes locked briefly (and we don't call out to user code from there),
so the mutexes don't need to be recursive anyway. That's just as well,
because it turns out our implementation of recursive mutexes on
pthreads is broken!

The goal here is to be able to distinguish between "cmutexes" (mutexes
compatible with a condition variable) and "rmutexes" (mutexes which
are recursive if possible, to avoid deadlocking if we hold them while
calling user code).

This is complicated by the fact that callers are not guaranteed to have
provided us with both versions of mutexes, so we might have to implement
one by using the other (in particular, DBusRMutex *aims to be*
recursive, it is not *guaranteed to be* recursive).

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43744
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Thiago Macieira <thiago@kde.org>

Make _dbus_mutex_new, _dbus_mutex_free static

They're only called within their module.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43744
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Thiago Macieira <thiago@kde.org>

Fixed runtime error caused by using uninitialized variable detected with msvc build.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46335
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Merge branch 'dbus-1.4'

Conflicts:
NEWS

NEWS for 1.4

dbus-protocol.h: compile under C++11

C++11 compilers have a feature called 'user-defined string literals' which
allow arbitrary string suffixes to have user-defined meaning.

This makes code that concatenates macros with string literals without
intervening whitespace illegal under C++11. Fortunately, string literal
concatenation has allowed intervening whitespace since the dawn of time,
so the solution is to simply pad with spaces.

Tested (header) with GCC 4.7 (trunk).

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=46147
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS

dbus-send.1: document --reply-timeout

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=14005
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

dbus-send.1, dbus-send: document --print-reply=literal consistently

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=14005
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

dbus-send.1: use bold and italic type consistently

According to man-pages(7)

* bold is for literal text, the name of the thing being documented,
  or the name of another man page

* italic is for replaceable text, usually in all-caps

* normal type (in the SYNOPSIS) is for special syntax like the []
  indicating optional things

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=14005
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

more NEWS

add and use _dbus_connection_trace_ref

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

add and use _dbus_server_trace_ref

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

add and use _dbus_pending_call_trace_ref

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Add and use _dbus_message_trace_ref

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Provide a hook to allow refcounting to be traced

This is designed to be used from a wrapper function, partly to supply
the same arguments every time for a particular class of object, and partly
to provide a more specific gdb breakpoint. It has several purposes:

* when under gdb, provide a function which can be used in breakpoints
* when not under valgrind and DBUS_MESSAGE_TRACE=1 is set, emit a
  _dbus_verbose when a message's refcount changes
* when under valgrind and DBUS_MESSAGE_TRACE=1 is set, emit a
  VALGRIND_PRINTF_BACKTRACE when a message's refcount changes,
  which lets you see the complete history of each message to track down
  reference leaks

Compile-time support is currently conditional on DBUS_ENABLE_VERBOSE_MODE,
but could be separated out if desired.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

dbus_message_cache_or_finalize: allow message cache to be disabled at runtime

This should make it easier to diagnose message-related ref leaks,
use-after-free, etc. with Valgrind: for optimal results (and pessimal
performance), we want to avoid re-using memory blocks for as long as
possible.

For now this is conditional on DBUS_BUILD_TESTS. It could get its own
conditional if desired.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

DBusMemPool: inform valgrind what we're up to

If we tell valgrind what we're doing, it can give better diagnostics.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Add support for inserting (a subset of) Valgrind client requests

If valgrind support is disabled, we define stub versions of the
Valgrind client requests I plan to use, so the actual code doesn't
need #ifdef hell.

[With unnecessary AC_SUBST removed as per Lennart's review -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37286
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Replace a series of booleans, all (apparently) alike, with flags

This makes it a bit clearer what's going on.

Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

NEWS

_dbus_string_append_double, _dbus_string_parse_double: remove

They're unused, except by their own regression tests.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

_dbus_string_append_unichar, _dbus_string_get_unichar: remove

These are unused (except by their regression test!) and not visible to
external callers.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

_dbus_list_pop_last_link: remove, unused

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Merge branch 'dbus-1.4'

Conflicts:
NEWS
configure.ac

Back to development status

Release D-Bus 1.4.18

dbus-daemon: fix forgotten counter increase while copying configured auth mechanisms

Previously, only one auth mechanism was used.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=45106
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Revert "dbus-daemon: fix forgotten counter increase while copying configured auth mechanisms"

This reverts commit 26b57efe43c991616186db5c499f729a900c6544.
It was incorrectly attributed.

_dbus_connection_queue_received_message: remove, unused

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

DBUS_HASH_TWO_STRINGS, DBUS_HASH_POINTER: remove, unused

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

_dbus_list_insert_before: remove, unused

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

_dbus_header_field_to_string: remove, unused

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Remove unused _dbus_string_append_4_aligned, _dbus_string_append_8_aligned

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

_dbus_getgid: remove, unused

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39759
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Lennart Poettering <lennart@poettering.net>

NEWS

spec: document systemd transport

This adds a short section about the systemd transport to the
specification.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=35232
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS for recently-merged branches

Merge branch 'system-search-paths-21620'

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=21620
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Merge branch 'doc-search-paths-35306'

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=35306
Reviewed-by: Lennart Poettering <lennart@poettering.net>

Merge branch 'reversed-dns-37095'

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=37095
Reviewed-by: Lennart Poettering <lennart@poettering.net>

bus: introduce --nopidfile switch to disable writing of PID files

When used with init systems such as systemd (where PID files are
redundant) this allows us to disable PID files even if a path is
configured for them in the normal bus configuration files.

Make use of this new switch in the systemd unit file.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=45520
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Windows compile fix

BUG: https://bugs.freedesktop.org/show_bug.cgi?id=45832

NEWS

_dbus_get_environment: move from shared library to dbus-sysdeps-util

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=34976
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Cosimo Alfarano <cosimo.alfarano@collabora.co.uk>

NEWS

dbus-marshal-recursive: make some internal functions static

Having a disabled type writer, comparing type readers by position, and
performing a single step of copying from a reader to a writer all seem pretty
obscure, and were only used within dbus-marshal-recursive.c (to realign
after insertion).

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=38285
Reviewed-by: Cosimo Alfarano <cosimo.alfarano@collabora.co.uk>

find_field_for_modification: document where the readers are left pointing

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=38285
Reviewed-by: Cosimo Alfarano <cosimo.alfarano@collabora.co.uk>

Reinstate the recursive marshalling test but skip the later parts by default

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=38285
Reviewed-by: Cosimo Alfarano <cosimo.alfarano@collabora.co.uk>

dbus-test: only test for memleaks if we actually ran the test

This reduces output spam if you're only running one part.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=38285
Reviewed-by: Cosimo Alfarano <cosimo.alfarano@collabora.co.uk>

Merge branch 'dbus-1.4'

Conflicts:
NEWS

NEWS

docs: correctly invoke man2html

man2html expects to find its input on stdin, so just passing the
filename will cause it to hang waiting for input.

[man2html 1.6g as shipped in Debian seems to be fine with files on the
command line, but apparently other versions aren't? -smcv]

Signed-off-by: Jack Nagel <jacknagel@gmail.com>
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=43875

Explicitly don't search XDG_DATA_DIRS for system services, and document it

In practice, it never works, because the activation helper doesn't
respect environment variables for security reasons.
If you want to vary the search path, alter system.conf instead, to
replace or augment <standard_system_servicedirs/> with your preferred
search path.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=21620
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

spec: document how to install session .service files

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=35306
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

spec: document the search path for session services

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=35306

systemd: drop machine UUID generation for unit file

To optimize startup a bit, drop machine UUID generation from the unit
file. Since a while D-Bus was already capable of making use of the
/etc/machine-id file, which is guaranteed to exist on systemd systems
(since the first released versions already), hence there's no need to
ever generate a D-Bus specific machine ID.

In times where userspace takes < 800ms to boot up getting rid of this
process optimizes boot time further.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=45520
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS

Check for a new enough systemd that sd_uid_get_seats (., ., NULL) is valid

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39609

at_console: ask systemd whether a user is at the console

systemd manages seats and users. This patch optionally asks systemd
whether a user is at the console. It used libsystemd-login for that, a
low-level library that allows querying this kind of information without
expensive round trips.

In order to be nice to the Debian folks this patch falls back to
traditional modes of operation if systemd is not found to be around.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=39609
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

NEWS

Merge branch 'socket-set-33337'

Reviewed-by: Will Thompson <will.thompson@collabora.co.uk>
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=33337

NEWS

talk about stable branches a bit more

Merge branch 'dbus-1.4' rejecting commit 4ebb275ab7b6f71d5

Conflicts:
configure.ac

Don't warn about deprecated declarations in this stable branch

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=44413

Port to glib 2.31.x g_thread API

g_thread_init() is deprecated since glib 2.24, call g_type_init() instead.
Bump glib requirement accordingly.

g_thread_create is deprecated since 2.31, use g_thread_new() instead. When
building with a glib earlier than 2.31, provide a backwards compatibility shim.

[Added a comment about why we're using g_type_init() in a test that
doesn't otherwise use GObject -smcv]

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=44413
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Merge branch 'dbus-1.4', rejecting commit 08dd53c14b53b88c

Revert addition of files which were only meant to exist on master, too

This completes the reversion started in 5df8c3db12590edd68e968.

update NEWS

dbus-daemon: fix forgotten counter increase while copying configured auth mechanisms

Previously, only one auth mechanism was used.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=45106
Reviewed-by: Simon McVittie <simon.mcvittie@collabora.co.uk>

Revert "windows fix: use install root as base path for relative pathes in dbus service file"

This reverts commit 10fae2352742d6dfd2ab27545048939f10f4922a and
reopens <https://bugs.freedesktop.org/show_bug.cgi?id=44030>.

Revert "keep cmake in sync with automake related to default bus addresses"

This reverts commit fdab4af0f4748f1fb5d23ca41e201cca4adb0b2f.
Further discussion required.

Related to https://bugs.freedesktop.org/show_bug.cgi?id=38201

Revert "made session service dirs customizable: cmake part"

This reverts commit 89e453216c02bb85c5e53d459997f8bc3b2c73d4.
It makes the tests fail under autotools.

Revert "made session service dirs customizable: autotools part"

This reverts commit 61760cf7de5af3452423244a8b6f5f1b5d00512b.

Merge branch 'dbus-1.4'

Merge commit '5df8c3db12590edd68e968975a335da9d0415e5a' into new-master

Reject its changes so that git merge-tracking will not want to revert the
changes that were unsuitable for 1.4.