Re: rep-gtk and the TreeModel bindings

[Christopher Roy Bratusek <zanghar freenet de>]

Am Thu, 12 Nov 2009 16:41:21 -0800
schrieb Rodrigo Gallardo <rodrigo nul-unu com>:

> I finally got around to uploading rep-gtk 0.90.0 to Debian
> only to find that the addition of the TreeModel bindings causes a
> build failure in the alpha architecture.
> 
> Specifically, gtk_tree_model_get_valist fails to compile, since the
> glue declares it as taking a string argument, but the actual
> declaration in the GTK+ headers is with a valist. In other arches this
> only causes an invalid pointer cast warning, but in alpha those arg
> types are completely uncompatible and the build dies.
> 
> Now, reading the GTK+ docs, it seems to me that this function
> shouldn't even be exported at all (how would one create a valist to
> pass to it from rep, in the first place), but instead it should be
> used as the backend implementation for the exported version of
> gtk_tree_model_get.
> 
> This doesn't sound like an easy change to make, though.
> 
> I grepped through sawfish's code, and it doesn't seem to use any of
> these functions. Which tells me that it's unlikely they have much
> users anywhere ;) So, as a stopgap measure, could we just remove
> support from them from the gtp-rep bindings?
> 
> Thanks for your comments.

gtk_tree_model_get_valist is commented out in 0.90.1, just do so 
via a patch in debian/patches/bla, that's enough.

Support for whole GtkTreeModel won't be dropped, required for
GtkTreeView.

Regards,
Chris