v0.6.0

New features

• Network settings management has been abstracted into the ControllerApplication.load_network_info and ControllerApplication.write_network_info methods.

Bugfixes

• Fix bug that prevented just the coordinator from permitting joins by @puddly in #86
• Fix broken empty address manager entries after restore on affected devices by @puddly in #92

Changes

• Add "not recommended" note to CC2530 and C2531 by @Hedda in #87
• Bump pyserial-asyncio dependency to 0.6 by @puddly in #91. This release of pyserial-asyncio properly supports Windows.
• Replace internal network state objects with zigpy.state by @puddly in #85
• Add basic clusters to both coordinator endpoints by @puddly in #93

Breaking changes

• The functionality in zigpy_znp.utils has been moved into the zigpy_znp.api.ZNP class.
• Network backups created with zigpy-znp 0.6.0 cannot be restored by older releases.

Full Changelog: v0.5.4...v0.6.0

v0.5.4

Changes

• Syncup with zigpy==0.37.0 changes (#84)

v0.5.3

Bugfixes

• Fix joining error with firmware versions >= 20210707 (ddf79f6)
• Increase startup LED response timeout (c9cb576)

v0.5.2

Bugfixes

• Detect the CC2538 and include the Z-Stack build in the device model string (a571726, #72).
• Properly handle older Z-Stack SYS.SetTxPower responses (#82)
• Permit deserialization failures with the undocumented ZDO.ParentAnnceRsp command (aaf0c31, home-assistant/core#50005).
• Silence "Unhandled command" warnings for ZDO.ParentAnnceRsp, ZDO.ConcentratorInd, and ZDO.MgmtNWKUpdateNotify (#75).
• Joins through specific routers will not also open the coordinator for joining on Z-Stack 20210708 (properly fixes #53)

New features

• Treat messages from unknown devices as implicit joins. This allows for existing networks to be mapped out without explicitly re-joining most devices (#73)
• Normalize command logging for NVRAM reads and writes (0adeb9b)

v0.5.1

Bugfixes

• Command line tools failed to write JSON/binary output to STDOUT and read it from STDIN: 8300994

v0.5.0

New Features

• Huge thank you to @castorw from the Z2M project for working with me to bring concurrent support for the Open Coordinator Backup format to Z2M. This allows for bidirectional migration between any coordinators supported by zigpy-znp. More documentation is available in TOOLS.md#Backup and restore.
• Multicast and broadcast requests are sent immediately
• Log a warning when using the Z2M default network key.

Changes

• Maximum time to wait for a device announcement after a device join before sending the zigpy device_joined signal has been increased from 2 seconds to 5 seconds.
• Write both PRECFGKEYS_ENABLE and EXTENDED_PAN_ID OSAL NVRAM items to ensure Z2M works with adapters commissioned with zigpy-znp.
• Retry using manual source routing first, before route discovery.
• If an exception is encountered when passing a frame from the UART object to ZNP, log the entire frame.
• If an unknown frame is passed to ZNP, log it but do not throw an exception.
• The -c command line flag has been removed from zigpy_znp.tools.nvram_reset and is now the default.

Internal Changes

• Add a force_form keyword argument to ControllerApplication.startup.
• All CC2531-specific types have been removed with the addition of the CStruct class, which computes struct padding and alignment during serialization/deserialization. This is autodetected at startup.
• The ZNP class can now be used more independently of ControllerApplication, especially without a formed network.
  • ControllerApplication._reset has been moved to ZNP.reset.
  • Complete network information is extracted from NVRAM at startup and populates the ZNP instance's network_info attribute.
• New helper methods to read and write NVRAM tables.

Breaking changes

• Serialization and deserialization of complex types is strongly encouraged to be done via znp.nvram.serialize/deserialize instead of the object's serialize/deserialize methods, which takes into account structure alignment and padding.
• item_type is now a required keyword argument for all NVRAM reads and no trailing data after deserialization is permitted. Old behavior can be preserved by passing item_type=t.Bytes.

v0.4.0

New Features

• The default LED mode is now off.
• Support for ZDOCmd.Unbind_req (#54).
• Very verbose TRACE-level debug logging is now hidden behind the logger object in zigpy_znp.logger. To enable it again in Home Assistant, add the following to your logging configuration:

  logger:
    logs:
      zigpy_znp.logger: debug

• Periodically pings Z-Stack to detect soft lockups and treat this failure as a disconnect. For sticks that reset upon reconnect, like slaesh's CC2652RB stick, this will make recovery automatic.
• Log network settings on startup under the INFO log level. For security, the network key will be logged under the DEBUG level.
• During request retrying, always attempt source routing, even if a device's relays are not present.

Bugfixes

• permit_ncp is now a no-op and all calls to ControllerApplication.permit will open joins on the coordinator, even when joins are permitted only through another router (#53). A full solution will require changes to Z-Stack.
• Do not attempt manual child aging on older coordinators, as their firmware does not have the required command and MT serialization inconsistencies will cause errors to be logged.
• Prevent the installation of pyserial_asyncio==0.5 on Windows, which breaks writes (pyserial/pyserial-asyncio#69).

v0.3.0

Security Fixes

• CC2531 coordinators running Z-Stack Home 1.2 firmware erroneously permit joins on startup, even after being reset. Joins are now explicitly disabled every time the coordinator is started (home-assistant/core#43245).

New Features

• Handle errors internal to Z-Stack by retrying requests within zigpy-znp. Missing routes are explicitly re-discovered using a few different methods, including manual source routing as a last resort. Thanks zigbee-herdsman!
• Check for and re-discover missing routes before sending ZDO requests.
• Upon receipt of a Trust Center key exchange notification, handle_join will be delayed for at most two seconds, until a device sends a ZDO Parent Announce broadcast.

Bugfixes

• NVRAM backup and restore did not read/write tables except for the first and last items. This is now properly handled but the backup format is now backwards-incompatible.
• ControllerApplication.force_remove does not throw exceptions (zigpy/zigpy#563).
• Fix communication with devices using nonstandard profiles by always defaulting to endpoint 1 (#43)
• Fix ZDOCmd.Mgmt_Rtg_req ZDO converter incorrectly sending ZDO.MgmtLqiReq instead of ZDO.MgmtRtgReq.

Internal Changes

• NwkNvIds and OsalNvIds were improperly named. NwkNvIds are now OsalNvIds and OsalNvIds are now ExNvIds. The new-style and legacy NVRAM operations are now exposed via ZNP.nvram.read and ZNP.nvram.legacy_read, respectively. The ZNP.nvram object transparently handles items that require multiple reads/writes to fully populate and bypasses NVRAM read protection for secure items when used with certain builds of Z-Stack Home 1.2.
• Shield request/response/callback coroutines from cancellation in order to reduce the number of logged unhandled command warnings.
• Remove handling of "auto" as a device path.

v0.2.2

Bugfixes

• Fix LED control for CC2531s running Z-Stack Home 1.2 (reported by @JeffResc).

Tools

• Support zigpy_znp.tools.network_scan on Z-Stack Home 1.2.
• Include complete beacon information in zigpy_znp.tools.network_scan, including LQI.
• Allow duplicate beacons to be printed with zigpy_znp.tools.network_scan by passing the --allow-duplicates (-a) flag.

v0.2.1

Bugfixes

• Fix tests being erroneously installed as a Python package along with zigpy_znp.
• Support Python 3.8.6 by fixing enum incompatibilities (https://bugs.python.org/issue41889) and requiring at least zigpy 0.25.0.
• Explicitly handle device announcement and source routing messages generated by devices on the network that are not known to zigpy, preventing unnecessary tracebacks from being logged (reported by @Samantha-uk).
• Reduce the logging level of missing NVRAM item messages in zigpy_znp.tools.nvram_reset from WARNING to DEBUG (reported by @Samantha-uk).