Re: [patch] constify g_simple_async_result_set_from_error

[Steve Frécinaux <nudrema gmail com>]

David Zeuthen wrote:
> On Wed, 2009-09-02 at 19:17 +0100, Richard Hughes wrote:
> > In PackageKit, I have to explicitly cast in my finish function:
> >
> > static void
> > pk_package_sack_merge_state_finish (PkPackageSackState *state, const
> > GError *error)
> > {
> > ...
> > 		g_simple_async_result_set_from_error (state->res, (GError*) error);
> > ...
> > }
> >
> > The finish function is passed a const GError (as the error can't be
> > freed), hence the cast is required for the compiler not to warn.
> >
> > Of course, the correct thing to do would be to trivially fix gio, and
> > a patch is attached that does exactly that. Okay to commit?
>
> Mmmm, wouldn't this break the API? E.g. cause compilation of existing
> apps to spew warnings. If so, we can't do it.

There is no problem providing a non-const value to a function which 
expects a const one, hence the previously working code should behave 
just fine.

> Unfortunately there's a lot of const incorrectness in the GLib and GTK+
> stack - maybe we should fix this in GLib3/GTK3 since it's going to be a
> new ABI.

+1, but I don't think it would make the ABI change...