Return-Path: MIME-Version: 1.0 In-Reply-To: References: <1438364565-24642-1-git-send-email-adam.moore@savantsystems.com> Date: Sun, 10 Jan 2016 19:37:19 -0300 Message-ID: Subject: Re: [PATCH] gatt-database: allow GattService anywhere in ObjectManager tree From: Luiz Augusto von Dentz To: Adam Moore Cc: "linux-bluetooth@vger.kernel.org" Content-Type: text/plain; charset=UTF-8 Sender: linux-bluetooth-owner@vger.kernel.org List-ID: Hi Adam, > Hi Luiz, > > You are right - though I still like the initial design of having one > service per ObjectManager, as it makes it easier to cleanup or restart (or > D-Bus export/unexport) an individual service at the ObjectManager API > level. > > I only added the pointer to manager_path because it was better to test > that for that prefix than it was to not do any checking. I couldn¹t do > the stronger test for the GattService path because there is no guaranteed > order to the proxy callbacks. Please check the patch-set Ive sent last week, the idea is that you would register the application only once, later on we might implement support for InterfacesAdded/InterfacesRemoved allowing services to be added/removed without having to unregistering and registering again. > How strict do you think we need to be here in validating the parent/child > relationships in the object tree? Right now it expects the objects to appear in order, service -> characteristic -> descriptor, so that we can established proper relationship between the objects. I did not add any check for multiple applications within the same connection id except if it is the very same object path given for the object manager.