Re: g_signal_connect() and G_OBJECT

[Emmanuele Bassi <ebassi gmail com>]

The first argument of `g_signal_connect()` is a gpointer (i.e. a
`void*`), so you don't need a cast at all — C will implicitly cast any
pointer to `void*`, and from `void*` to any other pointer.

Additionally, as you discovered, signals are GTypeInstance features —
you can emit signals on anything that inherits from GTypeInstance, not
just GObjects.

The reason why you see a cast macro is a layer of "extra security":
you can check that you're passing the object you meant to be passing,
instead of a NULL pointer or some garbage. This "extra security" is
mostly cargo-culted through tutorials, so people learn the habit and
transmit it.

Personally, I find it pointless; internal state should be checked with
`g_assert()`, and all cast macros can be compiled away with
`G_DISABLE_CAST_CHECKS`, so that buys you almost nothing.

Ciao,
 Emmanuele.

On 30 June 2017 at 11:20, Ingo Brückl <ib wupperonline de> wrote:
> Hi,
>
> it seems that it was common practice to cast the first argument of
> g_signal_connect() to G_OBJECT when I started developing applications with
> GTK+ quite a while ago. At least I've learned it that way and am doing it
> ever since.
>
> I repeatedly see usage of g_signal_connect() without that cast which seems
> to make sense since "instance" only has to be a gpointer and will be checked
> G_TYPE_CHECK_INSTANCE at run time.
>
> So I'm just curious. Is/was there a reason for the G_OBJECT cast? Was it
> necessary back in the days (when GTK+ 2 took over from GTK+ 1)?
>
> Ingo
> _______________________________________________
> gtk-app-devel-list mailing list
> gtk-app-devel-list gnome org
> https://mail.gnome.org/mailman/listinfo/gtk-app-devel-list



-- 
https://www.bassi.io
[@] ebassi [@gmail.com]