Bug 58493

Comment on attachment 71777 [details] [review]
user: Fix x-session -> xsession property notification

Review of attachment 71777 [details] [review]:
-----------------------------------------------------------------

::: src/user.c
@@ +315,4 @@
>          if (s != NULL) {
>                  g_free (user->x_session);
>                  user->x_session = s;
> +                g_object_notify (G_OBJECT (user), "xsession");

can you describe a little more the reasoning behind this?

we do g_object_notify (..., "x-session") elsewhere in the code and the class_init function has:

        g_object_class_install_property (gobject_class,•
                                         PROP_X_SESSION,•
                                         g_param_spec_string ("x-session",•
                                                              "X session",•
                                                              "User's X session.",•
                                                              NULL,•
                                                              G_PARAM_READABLE));•

I'm assuming this patch was written to address a specific problem you ran into, but I think if there's a problem, the fix might be more elaborate then just that change.

(In reply to comment #1)
> Comment on attachment 71777 [details] [review] [review]
> user: Fix x-session -> xsession property notification
> 
> Review of attachment 71777 [details] [review] [review]:
> -----------------------------------------------------------------
> 
> ::: src/user.c
> @@ +315,4 @@
> >          if (s != NULL) {
> >                  g_free (user->x_session);
> >                  user->x_session = s;
> > +                g_object_notify (G_OBJECT (user), "xsession");
> 
> can you describe a little more the reasoning behind this?

Booted latest gnome-ostree, saw a g_warning from accountsservice.  Created patch.  No more warning.

> we do g_object_notify (..., "x-session") elsewhere in the code

User != ActUser

It looks like src/user.c has a lot of leftover bits from before it was switched to using gdbus-codegen.

This whole thing though is really a mess.  Right now gdbus-codegen is being run twice.  Probably a side effect of your use of recursive automake (as opposed to nonrecursive) - recursive easily creates broken cross-folder dependency ordering problems.

So why did you say "Introduced by commit acb79f31770318f461f2dec95d0ba971cbfbc177" if we're not talking about act-user.c?

oh nevermind, I get it, just wrong commit id. pushed.

(as http://cgit.freedesktop.org/accountsservice/commit/?id=5b2b8e554d19a37ea5b5e0a1ba2d1b67328863f5 )