Phase 2: Measure actual resolver and mutation work #624
No reviewers
Labels
No labels
bug
cli
core
docs
event
experiment
figure
invariant
metrics
oracle
phase-0
phase-1
phase-2
phase-3
phase-4
phase-5
phase-6
provenance
revocation
tests
workload
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
erikinkinen/AES!624
Loading…
Reference in a new issue
No description provided.
Delete branch "phase2/p2-06-a-resolver-mutation-counters"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Task
Closes #90 (P2-06-A); contributes to #89.
Summary
Measure actual semantic record accesses in all seven legacy resolvers and revocation mutations. Snapshot enumeration, unsuccessful lookups, traversal frontiers, deletion scan positions and repeated lazy markers now produce an explicit measured vector while legacy estimates keep their original values.
Scope
Included
Explicitly excluded
Design intent
A terminal helper owns each access charge. Actual rereads count again; cached fields do not. Whole-record and version-only operations receive one category, and independent index copies count separately. Legacy reporting, staging, oracles and validation do not receive a mechanism meter. Existing resolver interfaces remain available; an external resolver without instrumentation explicitly rejects measured calls instead of silently reporting zero.
Phase discipline
Measured and unmeasured execution use the same local algorithms and preserve identical legacy state/event bytes. Only the new stream exposes actual work. Unsupported measured actions fail explicitly until their implementation is accepted. Failed primitives retain diagnostic meter contents but no accepted event or authority mutation.
Verification
All 50 CTests pass at the source commit in Debug, Release and Clang 22 ASan/UBSan. Analytical chain and star vectors match every strategy, including direct/snapshot differences in enumeration and leaf-first lineage index scans. Tests also cover overlapping and duplicate selectors, empty adjacency, missing roots, repeated/existing markers, failed batches, bucket misuse and uninstrumented resolvers. Per-kind access traces reconstruct the vector and total work. Full snapshot bytes, old target/cost fields and all 31 archived legacy cases remain equal. All hooks and the scope lock pass.
Commands, toolchain identities, logs and JUnit evidence are pinned in
artifacts/backlog/2.2.1/P2-06-A/acceptance.json. All eight exact-head Forgejo checks are required before the single issue checklist item is updated and the PR merged.