Hi Raymond,
this is a bug in gdbus then. The exactly same code is used in
bluetoothd and connmand so it seems weird. Did you try to debug it
where it actually fails to get this wrong? Are we missing some
"invalidate parent" calls?
I found the reason. During interface register object_path_unref did
call invalidate_parent_data. But for our case: we have path "/" at
first, and we register interface at "/org/openobex/Session0", there
are no data for "/org/openobex", so no any parent introspect data is
invalidated. But the whole /org/openobex path is created for the
first time. So actually the "/"'s introspect data need to be
invalidated.
I suspected that this might be the reason. I fixed a similar bug some
time ago. Funny that BlueZ didn't run into this since it might suffer
the same with the new modified adapter and device object paths.
I modify the code as below to invalidate the parent data up to "/",
so that it will fix this problem.
-------------------------------------------------------------------
diff --git a/gdbus/object.c b/gdbus/object.c
old mode 100644
new mode 100755
index a417ab9..074f40b
--- a/gdbus/object.c
+++ b/gdbus/object.c
@@ -299,25 +299,28 @@ static void
invalidate_parent_data(DBusConnection *conn, const char *child_path)
char *parent_path, *slash;
parent_path = g_strdup(child_path);
- slash = strrchr(parent_path, '/');
- if (!slash)
- goto done;
-
- *slash = '\0';
- if (!strlen(parent_path))
- goto done;
+ do{
+ slash = strrchr(parent_path, '/');
+ if (!slash)
+ break;
- if (!dbus_connection_get_object_path_data(conn, parent_path,
- (void *)
&data))
- goto done;
+ if (slash == parent_path) {
+ *(slash+1) = '\0';
+ } else {
+ *slash = '\0';
+ }
- if (!data)
- goto done;
+ if (!dbus_connection_get_object_path_data(conn,
parent_path,
+
(void *) &data))
+ continue;
- g_free(data->introspect);
- data->introspect = NULL;
+ if (!data)
+ continue;
-done:
+ g_free(data->introspect);
+ data->introspect = NULL;
+ }while(slash != parent_path);
+
g_free(parent_path);
}
-------------------------------------------------------------------------
However, It got some drawback here: if part of the parent path did
exist before, and the parent's parents data do not need to change.
This fix will do much more thing then it should. I guess to found
out all the change up to the necessay level will need more code. So
I just make it simple here.
so first, please fix the coding style here. I can tell from just the
pure email that you are mixing up tabs and spaces and also that you
are missing whitespaces here and there. Check that your mail client
doesn't mangle it. If in doubt send it as attachment.
Regards
Marcel
--
To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
