lsm / neokai / 25337281501

82%

push

github

Fix template hash to include gate poll and prevent future drift (#1780)

* Refactor template hash to use exhaustive JSON serialization and include gate poll

The template hash function was missing gate `poll` fields, causing the seeder
to not detect drift when poll config was added to built-in templates. Refactor
both gate and channel serialization from hand-crafted pipe-delimited strings
to deterministic JSON.stringify, which automatically captures all structural
fields. Going forward, adding new fields to Gate/GatePoll/Channel types will
not require separate hash function updates.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Canonicalize gate field ordering and check key serialization

Sort gate fields by name before hashing (restoring the sort that was lost
in the refactor) and add explicit serializeCheck() to produce deterministic
key ordering for check objects, avoiding reliance on JS object key
insertion order from imported JSON.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Hash full scripts and normalize single-target channel to

Two fixes from review feedback:

1. Hash full gate and poll script source instead of 64-char prefix, so
   edits beyond the prefix trigger drift detection.
2. Normalize single-element channel `to` arrays to strings, matching
   runtime behavior where `to: "Reviewer"` and `to: ["Reviewer"]` are
   equivalent.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Use locale-independent comparator for gate field sorting

Replace localeCompare with plain less-than/greater-than comparison to
ensure deterministic hash values across environments with different
locale settings.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Add total-order comparator for gate field sorting with tie-breaking

Sort gate fields by name, then type, then serialized check to ensure
deterministic ordering even when field names are duplicated (not enforced
by runtime validation). Extracts compareStrings... (continued)

7853 of 10860 branches covered (72.31%)

Branch coverage included in aggregate %.

51 of 51 new or added lines in 1 file covered. (100.0%)

63494 of 89124 relevant lines covered (71.24%)

252.08 hits per line