GObject signals and hash tables

[Kalle Vahlman]

On 6/27/05, Colin Walters <walters at verbum.org> wrote:
> On Mon, 2005-06-27 at 19:11 +0300, Kalle Vahlman wrote:
> > On 6/22/05, Kalle Vahlman <kalle.vahlman at gmail.com> wrote:
> > > Are (string-string) hash tables supposed to work with the GObject
> > > signal exporting at the moment?
> > >
> > > And if so, do I need to use some black magic to get them marshalled?
> > > The signals are delivered fine locally, but fail when marshalling for
> > > sending to the bus:
> > >
> > > ** (process:28816): WARNING **: failed to marshal parameter 1 for
> > > signal protocol_error
> >
> > I tried to run the glib-tests and encountered the same error:
> >
> > ** WARNING **: No marshaller for signature of signal 'Sig1'
> 
> We weren't converting parameter types to fundamental types; fixed.

Cool.
 
> > So obviously something's wrong with marshalling somewhere. Are the
> > glib-tests supposed to run cleanly at the moment?
> 
> They should, yes.

They do now :)
 
> > (I'm using cvs from last week I suppose)
> 
> ...as long as you're running the latest CVS :)

Oh, yeah, I was using an *ancient* version (at least five days old ;)
 
> > And if so, does anybody have pointers to what might be wrong with my setup?
> 
> I did fix one unrelated bug in the testsuite; we were failing to add
> G_TYPE_INVALID to terminate a varargs list.  This might have caused
> random testsuite failures.
> 
> Anyways, I added a test for a{ss} signals to the test suite, it seems to
> work.

So it does, and after fixing my code to do it properly, so does my
code too. I can finally throw hashtables over dbus like small rocks
over something that isn't that high.

\o/

Colin, you are my hero for the work you've done here :)

-- 
Kalle Vahlman, zuh at iki.fi