Hi Andrew,
On Mon, Apr 28, 2025 at 9:19 AM Andrew Sayers
<kernel.org@xxxxxxxxxxxxxxx> wrote:
>
> pbap_exit() didn't previously unregister itself thoroughly. That
> was fine if it was only called when the service was about to exit,
> because everything was implicitly unregistered when the process ended.
> But we need to be more scrupulous if this can be called throughout
> the program's lifecycle.
>
> Send the UnregisterProfile message directly from pbap_exit(),
> then call unregister_profile().
>
> The UnregisterProfile message can't be sent directly from
> unregister_profile(), because that also needs to be called when
> register_profile() fails halfway through.
>
> Signed-off-by: Andrew Sayers <kernel.org@xxxxxxxxxxxxxxx>
> ---
> obexd/client/pbap.c | 16 ++++++++++++++--
> 1 file changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/obexd/client/pbap.c b/obexd/client/pbap.c
> index 2f234fadf..90f8bdc02 100644
> --- a/obexd/client/pbap.c
> +++ b/obexd/client/pbap.c
> @@ -1485,8 +1485,20 @@ void pbap_exit(void)
> {
> DBG("");
>
> - dbus_connection_unref(conn);
> - conn = NULL;
> + g_dbus_remove_watch(system_conn, listener_id);
> +
> + unregister_profile();
> +
> + if (system_conn) {
> + dbus_connection_close(system_conn);
> + dbus_connection_unref(system_conn);
> + system_conn = NULL;
> + }
> +
> + if (conn) {
> + dbus_connection_unref(conn);
> + conn = NULL;
> + }
>
> obc_driver_unregister(&pbap);
> }
> --
> 2.49.0
Quite a few errors still: https://github.com/bluez/bluez/pull/1208
--
Luiz Augusto von Dentz
