This allows us to better implement alternative drivers
for devices, such as for device sharing servers or
talking to CoreMini processors within the same device.
The following fixups were added during the squash/merge:
Fix formatting in EthPhyMessage and EthPhyRegPacket
Device: Use std::make_shared when creating the EthPHYControl filter
Network: Create NetID String for EthPHYControl
EthPhyRegPacket: Constants in PascalCase
Because there is now more than one "product name" per device
type, we have a concept of a "generic product name" which
singularly maps onto a device type.
This change comes with a few small breaking changes within
the C++ API:
DeviceType::GetDeviceTypeString has been renamed to
DeviceType::GetGenericProductName to denote that
the returned value is not device specific and
device->getProductName() is preferable.
The member function DeviceType::toString has been renamed
to DeviceType::getGenericProductName for the same reason.
The DeviceType std::ostream& operator<< has been removed
to avoid accidental use of the generic product name.
This allows for disconnections to be detected quickly
where possible.
It also makes sure other driver errors aren't thrown
in the event of a disconnection.
When a device is sending any traffic, the device is considered to be connected. If no traffic if being received from the device, a status is requested. If the device fails to report the status back in a timely manner, it is considered to be disconnected.
If the device fails to reply to the status request, it is important to confirm that the device is not applying settings. While the device is applying settings, it will not be sending heartbeats or able to process a status request.
This also makes it possible for Communication to create more instances of Packetizer
This is necessary because Packetizer is not thread safe,
so when we support more VNETs we will need to create more Packetizers.
We've also decided to leave it out of the message structures since, for most uses going forward, it will not be needed.
Anyone who wants the timestamp resolution can always make the inexpensive device call to get it.
Paul Hollinsky
Kyle Johannes
Kyle Schwarz
EricLiu2000
David Rebbe