[PATCH v2][xdm] Implement ConsoleKit support.

[Fernando Lemos]

On Fri, Jul 22, 2011 at 4:49 PM, Julien Cristau <jcristau at debian.org> wrote:
> On Mon, Jun 27, 2011 at 21:04:25 +0200, Cyril Brulebois wrote:
>
>> +static int openCKSession(struct verify_info *verify, struct display *d)
>> +{
>> +    int ret;
>> +    DBusError error;
>> +    char *remote_host_name = "";
>> +    dbus_bool_t is_local;
>> +    char *display_name = "";
>> +    char *display_device = "";
>> +    char devtmp[16];
>> +
>> +    if (!use_consolekit)
>> +     return 1;
>> +
>> +    is_local = d->displayType.location == Local;
>> +    if (d->peerlen > 0 && d->peer)
>> +     remote_host_name = d->peer;
>> +    if (d->name)
>> +     display_name = d->name;
>> +    /* how can we get the corresponding tty at best...? */
>> +    if (d->windowPath) {
>> +     display_device = strchr(d->windowPath, ':');
>> +     if (display_device && display_device[1])
>> +         display_device++;
>> +     else
>> +         display_device = d->windowPath;
>> +     snprintf(devtmp, sizeof(devtmp), "/dev/tty%s", display_device);
>
> Is ck relevant to anything !linux?  If so, that path is probably wrong.
> I'm not quite sure what ck users would do with that information, though,
> and a quick codesearch doesn't really help.  So maybe we should just not
> set that…

Good point. It's ported to FreeBSD, and FreeBSD supports /dev/tty* too
(actually all BSDs do). I have no idea about other platforms, and I'm
not sure there's a portable way to do it. On the other hand, I would
think most (all?) platforms where an user would want to enable CK
support are covered, but that's more of a guess.

>
>> +     display_device = devtmp;
>> +    }
>> +
>> +    connector = ck_connector_new();
>> +    if (!connector) {
>> +     LogOutOfMem("ck_connector");
>> +     return 0;
>> +    }
>> +
>> +    dbus_error_init(&error);
>> +    ret = ck_connector_open_session_with_parameters(
>> +             connector, &error,
>> +             "unix-user", &verify->uid,
>> +             "x11-display", &display_name,
>> +             "x11-display-device", &display_device,
>> +             "remote-host-name", &remote_host_name,
>> +             "is-local", &is_local,
>> +             NULL);
>> +    if (!ret) {
>> +     if (dbus_error_is_set(&error)) {
>> +         LogError("Dbus error: %s\n", error.message);
>> +         dbus_error_free(&error);
>> +     } else {
>> +         LogError("ConsoleKit error\n");
>> +     }
>> +     LogError("console-kit-daemon not running?\n");
>> +     ck_connector_unref(connector);
>> +     connector = NULL;
>> +     return 0;
>
> Is "ck-daemon isn't running" a good reason to prevent the user from
> logging in?  (Why?)

I agree it's better to let the user log in. Without a ck session, the
user might be unable to do some pretty important stuff (configure NM,
shutdown, etc.), but that's still better than not being able to log
in.