# Changelog

## v0.3.0

IMPORTANT: This release moves `Nerves.Time` to the `NervesTime` namespace. If
you are using the API, you will need to rename every instance of `Nerves.Time`
in your code to `NervesTime`.

* Bug fixes
  * `ntpd` crash detection and restart have been simplified by using
    `MuonTrap.Daemon`. This should fix an issue that was seen with `ntpd` not
    getting restarted after it crashed.
  * `NervesTime.synchronized?/0` would return synchronized when changing NTP
    servers and right after restarting `ntpd`. It will return `false` now and
    switch to `true` when time really has been synchronized.

* Improvements
  * Simplified ntpd reporting code. No more regular expressions. Reports are
    sent via a Unix Domain socket and encoded as Erlang terms. This deleted a
    lot of string parsing code that felt brittle.
  * Added a restart delay on unclean `ntpd` restarts to prevent pegging public
    NTP servers. The delay is currently a minute since `ntpd` crashes are not
    well-understood. Luckily, this seems like a rare event.
  * Added more tests to cover NTP use

## v0.2.1

* Improvements
  * Move C build products under `_build`. This lets you switch targets with out
    cleaning builds in between.

## v0.2.0

* Added support for configuring NTP servers at runtime. Thanks to @ConnorRigby
  for sending a PR for this
* Added `Nerves.Time.restart_ntpd/0` for users who know external information
  about good times to bounce the NTP daemon so that it syncs more quickly
* Fixed naming of `is_synchronized/0` to be `synchronized?/0`. Thanks to
  @brodeuralexis for catching my slip.

## v0.1.0

Initial release