[patch] Fix locking for _dbus_connection_queue_received_message_link

[Havoc Pennington]

On Tue, 2005-01-11 at 14:39 -0500, Joe Shaw wrote:
> On Tue, 2005-01-11 at 11:25 -0500, Havoc Pennington wrote:
> > Fix looks correct to me. If someone without 400K of outstanding patch
> > could commit I'd appreciate it.
> 
> Committed this.
> 
> On an unrelated note, I am planning on doing a tarball release once I
> get a few fixes to the mono bindings in, since the mono runtime has
> changed in 1.1.3 to fix some bugs that the bindings were depending on,
> and it's causing all kinds of problems. :)
> 
> Do you have any objection to that with your stuff as it is now?
> 

There are a couple of other outstanding patches we should probably
commit, there's the "list available services" patch and the DESTDIR
thing and fixing the < vs. > bug in the tests. Maybe some other stuff in
bugzilla.

I'm breaking the hell out of things, and hope to be done any day now
(the code is written, it's just busted...)

The impact is that getting/setting message args API has changed, so all
the bindings have to be ported. My plan is that once I make things
basically work again I will commit the new API and people can start
porting while I finish off the details.

I don't think porting will be hard, since it's just whatever code you
have that marshals/unmarshals the message args. The available types are
a bit different (and recursive) but it should be possible to port by
just bailing out if you see a struct, as a first cut.

If you want to make a new release before this mess lands then that's
fine with me. If not, I definitely want to make a release as soon as the
dust settles.

My goal is to start winding down toward 1.0 since this is the last "big
change" needed to do so. So the approximate roadmap would be:
 - land the new type system
 - chew through all the smaller TODO items while putting out
   betas
 - do a release candidate or two that we think is API-ready and 
   feature-complete
 - 1.0

I figure this will take until July or so ;-) I'm not sure where KDE 4
stands on trying this stuff out. What we may need to do is get to a
point we think is 1.0-worthy, but linger at 0.99 until GNOME and KDE add
a D-BUS dependency and give us the thumbs up.

There are a couple of annoying-breakage changes other than the type
system; the main ones I can think of are the posts I made recently about
dbus_bool_t vs. unsigned char, and renaming "service" - I'd like to push
these through as quickly as possible after the type system changes.

Havoc