Artifact d2fe3a0c8e520658b33f05caa2ddab388a67e40be436d5c6a77f392383d9da2e:
- File
internal/horizon/seed/seed.go
-
2026-10-22 15:20:33
— part of check-in
[038e50befb]
on branch feat/horizon-cones
— feat(horizon): write the seed's D card to the second, as zit does
deck.Marshal writes D to the millisecond, so a go-libfossil seed could never match zit's bytes or name even over identical state. Mint now writes D as YYYY-MM-DDTHH:MM:SS, keeping the round-up to the next whole second that Section 5.2.2's "no earlier than the latest tag event" needs, and recomputes Z over the body. Over whole-second tag events the seed is now byte-identical to zit's, which the interop test asserts for the full seed and its name. (user: [email protected], size: 14523) [annotate] [blame] [check-ins using]
-
2026-10-22 15:26:36
— part of check-in
[12dcbdeed6]
on branch feat/horizon-cones
— Merge branch 'wip/horizon-seed' into feat/horizon-cones
The structural horizon's seed (draft-fossil-horizon-00 ยง5.2): an origin mints a control artifact carrying the tags in force at every entry of the held set, a clone holds it outside the artifact store where no sync, clone or rebuild can reach it, and the derivation applies it so every check-in in K derives as it does densely. Byte-identical to zit's seed when the latest summarized event is on a whole second. Conflicts in the horizon suite and table kept both sides; the seed's zit test now probes for a horizon-capable zit like the other horizon interop tests. (user: [email protected], size: 14523) [annotate] [blame] [check-ins using]
-
2026-10-22 15:20:33
— part of check-in
[038e50befb]
on branch feat/horizon-cones
— feat(horizon): write the seed's D card to the second, as zit does