Issue with _dbus_write_socket on windows

Peter Kümmel syntheticpp at
Fri Nov 17 11:58:55 PST 2006

Havoc Pennington wrote:
> Ralf Habacker wrote:
>> Havoc Pennington schrieb:
>>> I don't know exactly, since I don't know how this works on Windows.
>>> What is the "fd" you pass to the daemon's --print-addr-fd function?
>>> Is it a pipe or a socket or what is it? On UNIX autolaunch does a
>>> fork/exec, what does it do on Windows?
> btw what is the answer here? (where do you guys have the windows code
> btw, can we look at the work in progress?)

We have a cvs checkout in a subversion repository at sourceforge:

all changes to the original D-Bus code are in the DBus-win32.patch:

And there is a commit list for the changes:

The home is at:

>> We are thinking to store the recently used dbus address in the
>> registry, which allows the autolaunch code to detect a running
>> dbus-daemon  and to be able to connect to it.
> Maybe Windows should only have "autolaunch" and never start the daemon
> on login; unlike unix there might be a more reliable way to implement
> autolaunch. How does the registry solution work in terms of locking,
> i.e. say two buses start up simultaneously in the same session, one
> should exit and not be used. Is there a way to lock a registry entry or
> something that allows that to work?
> For the mugshot client we did single instance COM object with
> RegisterActiveObject() - that seems like one option. I'm not sure how it
> works if the user logs in multiple times (is that even possible?) though.
> Havoc
> _______________________________________________
> dbus mailing list
> dbus at

Peter Kümmel

More information about the dbus mailing list