On 2009-10-21 I (Tor Lillqvist <tml@iki.fi>) cherry-picked a bunch of
commits from the dbus4win-noncetcp branch of the dbus4win repository
to (my clone of) dbus upstream's master branch.

The dbus4win repository is at least currently at
git://repo.or.cz/dbus4win.git

I went through the commits done to the dbus4win-noncetcp branch since
the dbus-1.2.14 tag, i.e. commit f719d454329a1a54ff2d569feaf10ceee7cead50.

The following commits were left out because they either 0) were
already upstream, 1) seemed broken, or 2) seemed pointless to me,
especially if the commit message didn't say much. For each commit left
out, below is three lines: its id, the first log comment line, and an
explanation why I left it out.

f719d454329a1a54ff2d569feaf10ceee7cead50
configure.in: not all gccs support -Wno-pointer-sign
already upstream

6eddb6e1229b617ad44c8d61107983229dc7d589
configure.in: fail abstract socket test gracefully when cross-compilin
already upstream

c54dd9eefa5ed0d54be3b09cf30c12f05130fda1
r783: one ifdef lesser
probably pointless

93c288ca3e86e7760d3ac3fa6490257c1b6dc544
r783: compile if ENOMEM or EINTR is undefined.
huh, ENOMEM and EINTR is defined in the Microsoft C library

4f4ba13357da15c35a9d4ad111bf972a5d0d5db0
r783: introduced _dbus_daemon_release() function called in _dbus_loop_quit()...
this can't be right, there can be nested dbus loops afaik

124eb334324472a10de2c7e0a1455d353431f08f
unix build fixes against windbus-svn
can't be right to make a char* const when the code below then assigns through it

594bd19a0c45822d8763871e3491d99beb4d22cb
introduce DBUS_DIR_SEPERATOR and make use of it. On windows, disable some system service tests
"forward" slashes work fine on Windows, no need to uglify code with DBUS_DIR_SEPARATOR

cd2cdb3457f72cf058d5b19796ee691ceca6b52c
r783: introduced DBUS_CLEANUP_OLD_SERVICES that does cleanup prev zombie-instances...
unsure about this. something specific to KDE-on-Windows?

eeedba5b2f5b008819619764943caddd52441adf
build fixes for OS X, xcode 3.1.2, gcc 4.2.1
doesn't apply, seems to be upstream?

7dc7f71cf3003e006f6e177b5460d14a7cdbf2de
configure.in: fix pthread detection
doesn't apply

05b37fa87b1f6aa670de9007879f53a8198a7a89
configure.in: suppress -fPIC, -fPIE for W32
doesn't apply

b0d14fed92d9f539cd3478e72dc2f3ea75f3941a
configure.in: only check for wspiapi.h on Windows
nah, no harm checking for it on UNIX too

1724fc10e516e2d570537d522d8608abba90be5e
prototypes cleanup
related to cd2cdb3457f72cf058d5b19796ee691ceca6b52c which was skipped above