Skip to content

v1.2.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 30 Sep 04:50
· 21 commits to main since this release
v1.2.0
5322fab
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v1.2.0 - 2026-09-30

Baseline v1.0.6. openHop Modems and SPI radio hats join KISS firmware, the MQTT status now uses
the firmware's own names and meanings, and brokers stay connected through outages and token
refreshes.

Config adds the openhop:// and spi:// connections and the spiBoard, modemToken and
dutyCycle settings. nodeType is no longer needed.

Added

  • openHop Modems. Connect with openhop:///dev/ttyUSB0 over USB, or openhop://host:port
    over the network with modemToken as its password. The modem's radio is set to MeshCore's
    settings on every connect, and a dropped link reconnects on its own.
  • SPI radio hats on a Raspberry Pi. Connect with spi:// and choose the hat by name with
    spiBoard, from 15 SX1262 hats, the same list OwlShack has. A tx above the hat's rating is
    refused rather than turned down. To add a hat or correct one, put a boards.json in the
    directory the bot runs from.
  • Packets the bot sends appear on MQTT. Each one is published with "direction": "tx", as
    firmware nodes publish theirs. Signal fields are left off, since there is no received signal to
    report.
  • Received packets carry score, duration and path. score and duration match the
    firmware's packet log. path is the source and destination prefixes, as the firmware prints them,
    not the route.
  • The status reports more of what the firmware does. mcu_temp_c, last_snr, last_rssi,
    rx_air_secs, tx_air_secs, flood_tx and direct_tx are new in stats, and repeat is new
    at the top level. mcu_temp_c is left out on a board without a temperature sensor.
  • The status reports transmit and receive faults. New counters in stats show packets
    retried, dropped or failed, and whether the channel was busy or the bot sent faster than the radio
    could keep up. They count everything this process sends and receives, not one bot.
  • SPI hats and openHop Modems report chip-level errors. crc_errors, driver_errors and
    recv_recoveries appear in stats where the radio can measure them, and are left out on KISS.
  • dutyCycle limits how much of the time the radio transmits. It is a percentage, like the
    firmware's set dutycycle, and takes fractions such as 0.1 for EU868 sub-bands. The default is
    unchanged, and the duty cycle in use is logged at startup.
  • Slow message handling is logged and counted. One slow reply holds up every bot's receiving,
    so any that takes over half a second is logged and counted in stats.handler_slow.
  • The bot reconnects a radio that is lost. An unplugged modem, a dropped link, or a radio that
    stops answering for about 90 seconds is reconnected on its own, with retries up to every 30
    seconds, and the bots and MQTT status come back with it. Before, the bot kept running but heard
    and sent nothing until it was restarted.

Changed

  • recv_errors now means what it means on a firmware node. It counts packets the radio
    failed to receive. It used to count packets that arrived intact but did not decode, which is now
    packet_parse_errors. Firmware nodes publish recv_errors to the same brokers, so the two now
    agree.
  • battery_percent is replaced by stats.battery_mv. The firmware reports millivolts, and the
    old percentage showed 100% on boards with no battery. battery_mv is left out when the board has
    no battery or has stopped answering.
  • noise_floor moved into stats. It sits with the other radio readings, as the firmware
    reports it.
  • stats.packets_received is now stats.recv, with stats.packets_recv beside it. These are
    the names the firmware and CoreScope use. stats.sent is published alongside
    stats.packets_sent.
  • SNR and RSSI are left out when a packet arrives without them. They used to read 0, which
    looked like a real measurement.
  • The connection chooses the radio. nodeType is ignored, so an existing config still loads
    and it can be deleted.

Fixed

  • A broker that is down at startup is retried. It used to be dropped until the bot was
    restarted or reloaded. It is now retried with a growing delay, up to 5 minutes.
  • Startup no longer waits for brokers. Each unreachable broker could hold up startup for 10
    seconds. The bot now starts at once and connects in the background.
  • Shutdown and reload no longer hang on an unresponsive broker. The last status sent on the way
    down now gives up after 5 seconds rather than waiting forever.
  • A failed token refresh no longer leaves a broker disconnected. The broker used to stay
    down for 6 minutes or more. It now reconnects straight away.
  • Refreshing a token no longer makes the connection flap. The old connection is closed before
    it can push the new one off.
  • packets_sent and queue_len report real values. Both read 0 on every status since v1.0.6.
  • Every counter in a status is taken at the same moment. A packet arriving mid-report could
    show recv: 2 beside a last_rssi of 0.
  • Sends that meet a busy radio are retried again. They had been given up on at the first try.
  • Long packets at slow radio settings are no longer counted as failed. The bot now waits as
    long as the packet takes on air, rather than a fixed time that the largest packets could outlast.
  • A modem that stops answering no longer shows its last battery and temperature. They are left
    out once it has been silent for 45 seconds.
  • A packet that failed to send no longer appears on MQTT as sent.
  • A config mistake on reload keeps the running config. An unknown connection used to stop the
    bot. It is now refused when the config loads, as is spi:// without spiBoard.

Upgrading

  • If you read this node's MQTT status, check recv_errors first. It keeps its name but changes
    meaning, and nothing on the wire marks the change. A rate or delta over it will jump at the
    upgrade. Use packet_parse_errors if you wanted the old count.
  • Other MQTT fields to check: battery_percent is gone, noise_floor has moved into stats,
    stats.packets_received is now stats.recv, and SNR, RSSI, battery_mv and mcu_temp_c
    can be absent.
  • Building from source needs Go 1.26.7.

What's Changed

Full Changelog: v1.0.6...v1.2.0

Contributors

tuzzmaniandevil
Assets 13
Loading