RFC: Splitting Qt4 packages and implications of this

[David Faure]

On Wednesday 06 December 2006 14:32, Josef Spillner wrote:
> Am Mittwoch, 6. Dezember 2006 11:31 schrieb David Faure:
> > I think it's ok to make kdelibs compileable without session management
> > support, this is probably useful on Windows/MacOsX as well.
> >
> > However making it compileable without DBUS support seems like madness to
> > me... Everything relies on DBUS.
> 
> An example I think I talked to you about already some time ago is something 
> like kwebdesktop. That is, small useful applications which make use of KDE 
> libraries. Having some additional functionality compiled in is not the end of 
> the world (even though it better isn't), but relying on other processes 
> running makes development of such applications less attractive. I would like 
> to be able to run kwebdesktop on the command line. On an abstract level, it 
> is a function f:a->b with a being a stupid web page and b being a stupid 
> image file. For some reason it is (a+x)->b in KDE with x being additional 
> run-time dependencies like an X server and other processes.

There are no such things as stupid web pages. Web pages involve passwords, 
cookies, and support for loading N things from M protocols, i.e. launching kioslaves.

> Now it happens that it wants to load protocol information for the URL for 
> which it uses ksyscoca and so even if one sets 
> KApplication::disableAutoDcopRegistration() it will try to start a dcopserver 
> (and work fine without succeeding it seems). 
That's a KDE 3 problem, yes. It's fixed in kde4.

> In KDE 4 there doesn't seem to be  
> a method to disable the dbus registration.
There is: just don't use KApplication. But KUrl doesn't need ksycoca anymore.

> In my view many pieces in kdelibs are components which should be reusable in 
> contexts other than "desktop application". This is not about bringing a 
> stripped-down version of KDE to people but to make the software more 
> versatile.
Well we are indeed trying to make things modular, and the problem you mention
above is already fixed, for one. But we can't make everything modular to the point
of no integration anymore.

> P.S. Beside kwebdesktop shouldn't return an exit code of 0 if it cannot load 
> the page I think

 --- kwebdesktop.cpp     (revision 596408)
+++ kwebdesktop.cpp     (working copy)
@@ -79,7 +79,7 @@ void KWebDesktopRun::slotFinished( KIO::
     if (job->error())
     {
         kdDebug() << job->errorString() << endl;
-        kapp->quit();
+        kapp->exit(1);
     }
 }



-- 
David Faure, faure at kde.org, sponsored by Trolltech to work on KDE,
Konqueror (http://www.konqueror.org), and KOffice (http://www.koffice.org).