Unix fd passing, second try

[Lennart Poettering]

Heya!

I have now reworked my patchset considerably. My changes should
address all issues raised by Havoc and others (except for the async
eror synthesizing in dbus_connection_send_with_reply() and the unix fd
array handling as described in those other mails I just sent).

Beyond that I made some substantial changes how the unix fds are
unserialized from the transport stream. Most of the code now moved to
DBusMessageLoder where it probably fits much better. The loader now
processes two independant streams in very similar ways: the actual
byte stream, and the stream of unix fds. This makes the code much
nicer to read and should also ease porting the whole stuff to
transports that need to pass unix fds via a completely independant
channel.

The repo now also includes the spec updates and the handling of limits
on unix fds similar to those on message size. I believe the patchset
is complete now.

I also added a handful of other unrelated changes, such as atomic int
support based on gcc's __sync and suchlike.

http://git.0pointer.de/?p=dbus.git;a=summary
git://git.0pointer.de/dbus.git

I'd like to ask you for a second review!

If needed I can sent this patch series to the ML for review. 

Also, I can split off the patches that are unrelated to unix fd
passing, if anyone needs that.

Thanks,

Lennart

-- 
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net
http://0pointer.net/lennart/           GnuPG 0x1A015CC4