Windows

[Tor Lillqvist]

Robert McQueen writes:
 > I'd be interested in helping to pushing any work that people have done
 > on this upstream, if anyone's got any patches lying around. Has anyone
 > got experience building and running D-Bus on Win32 using eg tcp sockets?

I had it running to some degree on Windows in January, but then the
immediate need to have it ported to Win32 from my (employer's) point
of view fizzled out, and the changes never got pushed upstream into
the official sources.

Two versions of the diffs can be found at:
www.iki.fi/tml/dbus.my-diffs.20050121 and
www.iki.fi/tml/dbus.my-diffs.20050518 . (I.e., from January 21 and May
18.) Feel free to pick ideas from them or adopt the code as such.

No, I don't recall any longer what the exact difference between these
two diffs are. It can't be significant, as I didn't do any work on
this after January.

I recall a few things about the diffs:

Instead of Unix domain sockets (which, surprise!, don't exist on
Windows) this port uses TCP sockets boúnd to the loopback interface,
and a listener stores the port number in a file with a pathname that
corresponds to the Unix domain socket pathname that would be used on
Unix. That's not necessarily that good an idea, it might be more
Windows-like to use named pipes or something, I guess. But using TCP
sockets meant less ifdefs in the rest of the code.

File descriptors and sockets aren't at all the same on Win32, but the
D-Bus code (at least back in January) wanted to handle file
descriptors and sockets through the same API. The port uses an
encapsulating layer to hide the fd/socket difference. Not sure whether
that actually was a good idea or not...

--tml