[Buildroot] Serious time jump bug found in gpsd 3.20+

[Arnout Vandecappelle]

On 05/08/2021 22:26, Arnout Vandecappelle wrote:
> 
> 
> On 05/08/2021 22:09, Robert Hancock via buildroot wrote:
>> A heads up to the Buildroot community about a significant bug that has been
>> found in versions 3.20 and newer of the gpsd package, and which is fixed in
>> version 3.23 (not yet released, supposed to be shortly):
>>
>> https://gitlab.com/gpsd/gpsd/-/issues/144
>>
>> This will cause the calculated GPS time (and NTP time, if synchronizing to it)
>> to jump from 2021 back to 2002 when GPS week 2180 arrives on October 24, 2021,
>> with potentially disastrous results for applications depending on accurate time
>> keeping.
>>
>> A patch has now been added to the gpsd repository to fix it (note that the
>> commit message incorrectly refers to 2020 rather than 2021 as pointed out by a
>> commenter): 
>> https://gitlab.com/gpsd/gpsd/-/commit/7f30d88d04dc62b8bd6265ad1d09d72d220f97f6
>>
>> I am not sure if the best solution for Buildroot is to just update the package
>> when released, or apply the specific patch to fix the issue? Currently
>> Buildroot master is on 3.21 and the 2021.02.x LTS branch is on 3.20, both of
>> which are affected.
> 
>  For master, I think we can still afford to update to 3.23.
> 
>  For 2021.05.x and 2021.02.x, we'll need to backport the patch. Unless if Peter
> thinks that in this case (since upstream's policy is "always update to latest")
> it's OK to just bump it there as well?

 Forget about bumping the stable branches - the bump to 3.21 deprecates some
options, and we can't do that on the stable branches.


 Regards,
 Arnout