leap seconds

[moreati]

Sources

• https://www.ietf.org/timezones/data/leap-seconds.list
• https://pypi.org/project/leapseconds/
• /usr/share/zoneinfo/leapseconds,
  /var/db/timezone/tz_latest/zoneinfo/leapseconds
• /usr/share/zoneinfo/right/UTC
• https://dotat.at/cgi/git/leapseconds.git/blob/HEAD:/syntax.md
• Accessing Windows internal leap second information microsoft/STL#1624

[moreati]

The OS uses shared memory mapping to publish leap second tables to all processes, and it can extend that table without killing active processes or rebooting the machine.
-- dotnet/runtime#50263

[moreati]

apparently recent versions of windows store the leap second table in the registry:
https://github.com/microsoft/STL/blob/7f04137880e1c3df5125c1baf808f16f399dee2e/stl/inc/chrono#L2040
https://github.com/microsoft/STL/blob/b9505b9818c8bd53148922730fc2a426e8f38620/stl/src/tzdb.cpp#L556
-- RustCrypto/formats#675 (comment)

[moreati]

If a package containing the Olson tz database weer used as a source of current leap seconds then I think the update frequency of that package would matter. E.g. if pytz contained leapsecond data, and was used as a source then a release atleast every 6 months would be needed, to validate the TTL field - even if no leapsecond change occurs.

[moreati]

On Linux adjtimex() et al can report TAI offset known to the kernel. On macOS ntp_gettime() may do the same.

• https://www.man7.org/linux/man-pages/man2/adjtimex.2.html
• https://github.com/ziglang/fetch-them-macos-headers/pull/21/files
• https://github.com/moreati/python-adjtimex-experiments