Re: glib/gdataset.c (quarky / stringy stuff)

[David Nečas <yeti physics muni cz>]

On Mon, Oct 20, 2014 at 10:39:54AM +0100, John Emmas wrote:
> Hi guys - please forgive me if this isn't the right mailing list for
> glib questions.  I'd like to ask a question about a recent change in
> 'glib/gdataset.c'.  Up until a week or so ago, the code around line
> 1031 looked like this:-
>
>       if (strcmp (g_quark_to_string (data->key), key) == 0)
>       {
>         res = data->data;
>         break;
>       }
>
> Some time in the dim and distant past (to fix a problem which I
> can't remember any more) I changed my local copy to look like this:-
>
>       if (g_quark_to_string (data->key) != 0) /* This line added by
> me !!! */
>         if (strcmp (g_quark_to_string (data->key), key) == 0)
>         {
>           res = data->data;
>           break;
>         }
>
> but after an update this morning I noticed that the official code
> has recently been changed to this:-
>
>         if (g_strcmp0 (g_quark_to_string (data->key), key) == 0)
>         {
>           res = data->data;
>           break;
>         }
>
> Is that change effectively equivalent to what I did locally?

No.  As far as I can tell you can pass NULL key to g_datalist_get_data()
after this change.  If you do so and g_quark_to_string(data->key)
returns NULL (for whatever reason), the two NULLs will be considered
equal by g_strcmp0().  Whereas your change makes a NULL not-equal to
everything, even another NULL.

Regards,

Yeti