On Wed, 29 Sept 2021 at 10:06, Paul Eggert <eggert@cs.ucla.edu> wrote:
On 9/28/21 4:56 PM, Stephen Colebourne via tz wrote:
What I discovered was that the primary damage to Joda-Time would have come from the commit that move the Links from the main files to backward (2a18a625a0171e5c54249a54179eda3409b1b838).
Thank you for reporting the problem. The commit you mention is independent of the alike-since-1970 merge and so it's not that important. I reverted it by installing the attached patch into the tzdb development repository.
I can't see it here https://github.com/eggert/tz/commits/main
If you'll compare the resulting 'backward' file to what you derived in your copy, you should find only a couple of minor differences, which I'm also attaching to this email separately, as a convenience. I hope this serves as an effective merge of the two versions.
I think what the "correct" answer for those two locations is probably depends on the larger discussion, so that is OK for now AFAICT.
It sounds like we should release tzdb 2021c shortly, since this is a significant problem downstream.
This would be very helpful for those Joda-Time users that install the data set manually, thanks.
Previous merges were not consistent. Sometimes they moved the resulting Link into 'backward' (e.g., America/Coral_Harbour, Europe/Belfast), but often (as you noticed) they did not. Although (as others have noted in recent emails to this list) it would be nice to fix Link inconsistencies,
I agree that having a consistent set of rules and data is a good aim. thanks Stephen