Re: GLib 64 bit integer problem revisited

[Tim Janik <timj gtk org>]

On Mon, 18 Nov 2002, Owen Taylor wrote:

> Thse machines have 64-bit longs, not 64-bit ints.
>
> The problem you are running into is that gint64 and 'long int' ae
> the same.
>
> I'm not really sure that there is much you can do here .. you'll
> probably just have to punt the issue and go with:
>
>  Inti::G::Value::get64 (gint64 &)
>  Inti::G::Value::set64 (gint64)
>
> The problem is that the C++ equivalent of:
>
>  g_value_set_int64 (value, num);
>
> Needs to leave the value with a type of G_TYPE_INT64, so it has
> to be distinct from the C++ equivalent of g_value_set_long (value, num);
>
> In hindsight, having both int/long and int64 for GValue is probably
> a bad idea; it probably would have been better to just have int32
> and int64... but not something we can change at this point.

right, using longs as properties or signal arguments is a bad idea anyways,
so language bindings like C++ should simply use set/get (gint&) and (gint64&)
and skip longs entirely (internally, these may be passed on as TYPE_INT or
TYPE_INT64 values to g_object_set_property(), the property system performs
the necessary conversions if required).
that is, unless there're compat reasons set/get (glong&) have to stay in the
C++ bindings.

>
> Regards,
>                                         Owen

---
ciaoTJ