Changing threading semantics from init early to init before second thread

[Alexander Larsson]

On Mon, 2006-08-21 at 18:33 -0400, Havoc Pennington wrote:
> Tor Lillqvist wrote:
> > But on the other hand it's (to the best of my knowledge) robust,
> > tested, existing code.
> > 
> > OK, if the threading needs of dbus are *really* trivial (just starting
> > a thread?), maybe it's overkill, and just using the Microsoft C
> > library's _beginthreadex() is enough.
> > 
> 
> Doesn't even start threads - all we need is locks and cond variables.

As a matter of fact, all the internal API is in dbus already, in the
dbus-threads-internal.h header. All we need is a pthreads and a win32
implementation of those calls. Shouldn't be a lot of work.

> I tend to think that sticking to the current setup is better than either 
> requiring an extra dependency or maintaining our own sucked-in copy of a 
> largish dependency, though - the whole idea here is to simplify things 
> for developers.

I don't think we need any extra dependencies. And the current setup
isn't exactly simplifying things. In fact, it greatly complicates things
that I have to link to glib-dbus which is API unstable just so i can
init threads in dbus for gnome-vfs. Furthermore the whole thread
initialization did cause a lot of problems with the gaim thread
initialization time issue that would just not be a problem if we used
internal threads.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 Alexander Larsson                                            Red Hat, Inc 
                   alexl at redhat.com    alla at lysator.liu.se 
He's a suave soccer-playing card sharp on the edge. She's a violent 
cigar-chomping research scientist who believes she is the reincarnation of an 
ancient Egyptian queen. They fight crime!