gpsd: 1024 week backward time jump due October 2021
Opened this issue · 5 comments
Is work being done on bringing a fix into gpsd of meta-oe to the recently discovered issue in gpsd >=3.20, gpsd <= 3.22 (introduced on 2019-10-08) causing it to unconditionally relay a 1024 week backwards time jump on 2021-10-24 (destination year 2002)?
The bug is triggered by leap seconds not having been inserted for an unusual long (>2180 weeks) period (really: earth spinning faster, maintainer points to global warming as the cause).
A fix has been comitted and was released in gpsd v3.23 on 2021-08-06. Official backports to older releases will not be made available.
Looking at versioning of the gpsd recipe, v3.20 is both in master and the two latest release branches of Yocto (ie. gatesgarth and hardknott) so those will all be exposed to the bug.
Mentions
@Khem, this also affects Master unless we update to the later version 3.23
right ok. Then perhaps we need to update this recipe on master first.
I checked and master is on 3.23. Dunfell is still on 3.19 but as it stands the issue was introduced in 3.20 so dunfell users should not be affected by this. Still it may be a good idea to update dunfell to contain gpsd 3.23