[PATCH] Don't use positional arguments to initialize Gtk objects

David Shea dshea at redhat.com
Tue Feb 18 12:21:05 UTC 2014


On 02/18/2014 05:05 AM, Vratislav Podzimek wrote:
> I think that instead of doing this "fix", we should file a bug on
> pygobject-introspection because we really do nothing wrong or
> non-standard. What exactly causes the error? The empty constructor or
> calling those set_* methods? I believe the latter is okay and the
> constructor is the same as it is called in the C except we don't pass
> the object as the first instance. This really looks to me like a bug in
> some other place not in our code.

The set_* methods are still fine, and I could put those back. The 
complaint is about using unnamed arguments (something like 
Gtk.Adjustment(1000,1000,1,10,0), or more often in our case 
Gtk.Label("text")), and I figured if we're moving to lists of properties 
for those calls, we might as well stick the half dozen or so function 
calls right after the constructor into the property list, too.

The reason for the warning, I suppose, is that with 
gobject-introspection the constructor call isn't really the same as it 
is in C. Most of the time the introspected call doesn't call the *_new 
function (or doesn't call it when you'd expect it to) and skips right to 
_init and a series of set_property calls.


More information about the anaconda-patches mailing list