Re: Things for unstable cycle

[Jens Georg <mail jensge org>]

On Mo, 2011-09-26 at 17:53 +0300, Zeeshan Ali (Khattak) wrote:
> On Sun, Sep 25, 2011 at 10:23 AM, Jens Georg <mail jensge org> wrote:
> > Hi,
> >
> > some things I'd like to (see) fixed in the upcoming unstable cycle:
> >
> > Further improve introspection support:
> >
> > Some issues I've experienced when playing with gupnp + gi + gjs:
> >  - ServiceProxyAction needs to be boxed. this also affects Vala.
> >   According to Luca Bruno (Lethalman) the current Vapi is rather
> >   hackish and not guaranteed to work on further versions.
> >  - All the non-varargs version of begin_action_* don't properly
> >   introspect because the GError * argument is not the last one, making
> >   g-i not marking it as throws (This could also be a g-i bug)
> 
>   Better be fixed in g-i, no?

The second part? Yes. The first one - most likely not. 

> 
> > Bindability or usability from other toolkits: (Qt/"gupnpmm"):
> >  - It would be great to have an abstract ResourceFactory or interface
> >   which exports the functions from gupnp-resource-factory-private.h to
> >   hook up more complex factory functions than the ones that create a
> >   new GType (e.g. think of creating a new QUPnPResourceFactory that
> >   allows you to create new QUPnPServiceProxys).
> 
>   Looking at these API/‌ABI break and the ones we had to do in every
> minor release cycle, I think GUPnP API/ABI isn't as stable as we
> thought it is. Because of that, we really should also version the
> pkg-config files, our binaries and shared dir
> (/usr/share/gupnp-tools-0.x) etc to allow for parallel install of
> different gupnp minor releases.
> 
Yes, true.

Another thing: We should check and if possible remove the use of nested
mainloops inside of gupnp. Those call for problems.