archetech / archon / 35006261468

92%

main: 95%

Build Type

Pull #1145

github

Committed by macterra

Commit Message

fix(gatekeeper): Close the review's gaps in the confirmation-order guarantee

Ordinals are registry-local. A controller that migrated registries carries
events from its old chain, and an old-chain height can exceed the
operation's new-chain height, so comparing every event against the cutoff
stopped resolution before a rotation confirmed on the old chain -- and the
key it retired authorized again. The ordinal now applies only to events
on the cutoff's registry; any other is ordered by its block time.

Whether a controller can be resolved at a chain position is read from its
events, not inferred from a registry name: it lives on a registry that can
anchor, and every event confirming it there carries the position the chain
assigned. `pin` names a registry and anchors nothing, and a name-based
predicate had classified it, and any custom registry, as a chain.

importDIDs is the restore path for a DID export and went straight to
importBatch, keeping any confirmation the export claimed. It enters as
hints now, like the batch route.

The mediators' head-of-line stop only fired on a throw, but importBatch
reports a retrieval or apply failure in the item instead, and skipped an
item that had failed before. A block that stalled -- errored without
reaching processEvents -- is retried in order and stops the loop while it
stays stalled, in the retry pass too, so no later block is applied ahead
of it. One whose events merely deferred is left to the queue as before.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GYGMfX3foBcVqP7AQNXjWj

Pull Request Pull Request #1145: fix(gatekeeper): Reject a backdated proof once the chain has committed the operation

Coverage Stats

4217 of 4846 branches covered (87.02%)

Branch coverage included in aggregate %.

40 of 42 new or added lines in 2 files covered. (95.24%)

12 existing lines in 1 file now uncovered.

8842 of 9285 relevant lines covered (95.23%)

788.53 hits per line