Can't "forget" a scale

[jeffsf]

Disconnect retains the scale's MAC address, making it difficult to change scales

[jeffsf]

Current disconnect behavior is inconsistent:

• DE1 wipes the address (because "the only reason to disconnect would be to change DE1s")
• Scale retains the address (because, well, I don't recall)

Neither of these assumptions or behaviors is right all the time for all people and use cases.

People are potentially disconnecting a DE1 to use it with different control software.

It might be the case that people are wanting to switch scales (without restarting pyDE1), use the scale with different software (pour-over, for example) and, potentially using a second scale for beans/grinds.

---

Potential approach:

• Scan -- Within DE1/Scale, wipe any existing address, find first matching, connect to it (must not be connected to start)
• Connect -- Available if not connected and there is an address, connect to it -- Like "capture" of steam-to-temperature
• Disconnect -- Available if connected or trying to reconnect, just disconnect and retain address -- Like "release" of steam-to-temperature
• Clear -- (better word?) Disconnect if connected, wipe any existing address

[jeffsf]

Working toward adopting capture/release developed for https://github.com/jeffsf/steam-to-temperature

Not quite a drop-in, unfortunately

https://github.com/jeffsf/pyDE1/commit/207a492bcff66b18a4f638e2dc0dc54dce22354a relates to this effort

[a112358132134]

Another consideration, at least with bluez, is that it does not appear (in my practical use) to like making any connection if it considers an ID to be "stale". I don't know whether this is the same on other stacks/platforms.

In this context "stale" would mean has been too long since a scan was performed, or too long since disconnection of the device. I don't know exactly how long this timeframe is, but it seems it could easily be under a minute.

In the GUI layer, I've taken to forcing a scan before all new connections are requested as a result, because the failure rate was otherwise extremely high.

[jeffsf]

Filed as https://github.com/hbldh/bleak/issues/713

The "experimental" flag that they use to try to get around a scan isn't standard, and isn't intended for general use, as per the Bluez mailing lists.

Workaround right now is

                            self._bleak_client._device_path = None
                            self._bleak_client._device_info = None

[jeffsf]

As of bleak v0.18.1 (and perhaps earlier), clearing _device_path should be sufficient

        except bleak.exc.BleakDBusError as e:
            e: bleak.exc.BleakDBusError
            if (e.dbus_error == 'org.freedesktop.DBus.Error.UnknownObject'
                    and e.dbus_error_details.startswith(
                        'Method "Connect" with signature "" on interface')
                    and e.dbus_error_details.endswith("doesn't exist\n")):
                self.logger.info("Seemingly stale device, resetting.")
                self._backend._device_path = None