Skip to content

feat(LinuxTentacleE2E): Phase 12.L.E.18 — events.jsonl production-parser symmetry pin

Placeholder ppxd requested to merge phase12.L.E.18-events-production-parser-symmetry into main

Summary

Closes the binding-layer gap between .sh wire format (pinned by J.L.E.16) and what the production parser actually returns to the server-side timeline UI. The two layers can drift independently:

Layer Pinned by
Wire format strict (every line is valid JSON, fields present, kinds in order) J.L.E.16 (E1uEventsContract)
Production parser binds wire format → C# record correctly This PR

What this catches that J.L.E.16 doesn't

The production parser uses [JsonPropertyName] attributes to map:

Wire format C# property
"t" Timestamp (DateTimeOffset?)
"phase" Phase
"kind" Kind
"msg" Message

If those attributes are silently dropped (refactor accident, IDE auto-format reorders annotations off, "modernization" to a JsonNamingPolicy), JsonSerializer falls back to property-name matching → Timestamp/Message/Phase/Kind don't match wire t/msg/phase/kind → every event deserializes to all-defaults (null Timestamp, empty strings).

J.L.E.16's strict parser would NOT catch this — the wire format is still valid JSON; only the C# binding broke. Server UI timeline shows blank cards even though the wire data is fine. This is exactly the regression class this PR pins against.

Test mechanism

  1. Run E1.h-Linux end-to-end.
  2. Strict-parse events file via test-infra parser (J.L.E.16 path).
  3. ALSO production-parse via UpgradeStatusPayload.TryParseEvents.
  4. Assert symmetry: counts match (production parser didn't silently drop a line strict-parse accepted).
  5. Assert binding: every event has bound Phase + Kind (non-empty), AND every event has a non-null Timestamp (catches DateTimeOffset regression).
  6. Cross-check terminal "success" is the last event.

Fidelity tier

🟢 High (Rule 12.4): real prod .sh emits real wire format, real production parser deserializes it. Full round-trip pin.

Expected runtime: ~5s (E1.h + ms for parse).

Test plan

  • Linux E2E workflow runs (manual workflow_dispatch after merge)
  • E1uEventsProductionParser_AfterSuccess_BindsWireFormatToCSharpRecord passes within ~5s
  • No regression on existing 18 Linux E2E tests

🤖 Generated with Claude Code

Merge request reports