---
title: "Sprint 12 — CI Briefing"
description: "Ban HashMap via clippy, tracing infrastructure, system dependency graph, rng_seed field"
type: sprint
status: archived
sprint: 12
team: "ci"
---

# Sprint 12: Build — CI Tasks

**Goal:** Lay the production-layer foundations — simulation tier system, sound event pipeline, visual grammar, and knowledge boundary enforcement — and complete all v0.1 copy authoring so content packs, opening hooks, and world-building docs are done.

**Branch:** `ci`
**Agents:** Justine (build/deploy)

## Carry-over from Sprint 11

None. Sprint 11 was 7/7 done.

## New Tickets

| # | Title | Blocked by |
|---|-------|------------|
| #343 | Ban HashMap in simulation crate via clippy | — |
| #344 | Set up tracing crate infrastructure | — |
| #346 | System dependency graph debug command | — |
| #527 | Add rng_seed field to ObserverSnapshot for deterministic replay | — |

Use `db/connectors/ticket show <id>` for full details.

## Key Decisions

- `decisions/architecture.md` — D-020 (Godot + Rust IPC, bridge types), D-041 (knowledge graph, observer snapshot contract)
- `decisions/scope.md` — D-030 (testing architecture — D-phase alignment)

## Notes

**#343 — Ban HashMap in simulation crate via clippy**
- Critical priority. Deterministic simulation requires deterministic iteration order; `std::collections::HashMap` does not guarantee this.
- Add `clippy::disallowed_types` to `.clippy.toml` (or `.cargo/config.toml`'s `[target.*.rustflags]` section) scoped to the `simulation` crate only.
- Approved alternatives: `BTreeMap` (ordered) or `IndexMap` from the `indexmap` crate (insertion-ordered, deterministic).
- This lint will surface existing violations. Each violation in `server/src/simulation/` needs a follow-up fix — file separate tickets for each if the count is large, or fix inline if small.
- CI must fail on new violations after this is merged.

**#344 — Set up tracing crate infrastructure**
- Add `tracing` and `tracing-subscriber` to `server/Cargo.toml`.
- Initialize in `server/src/main.rs` with a stdout subscriber in dev (pretty format) and JSON in CI.
- Add structured log points to: tick duration (`tracing::info!(tick_ms = ...)`), per-system timing (wrap heavy systems with `tracing::instrument`), bridge I/O metrics, tier transition events.
- This is prerequisite instrumentation for tier system debugging (Sprint 12 server work in #99).
- In CI (`make ci-server`), the JSON log output should be captured but not asserted on — just confirm the binary runs without panic.

**#346 — System dependency graph debug command**
- Implement a `--dump-schedule` CLI flag (or `make debug-schedule` Makefile target) that prints the bevy_ecs system ordering and component access patterns.
- Use bevy's `World::resource::<Schedules>()` or the `bevy_app::App::render_schedule_graph()` approach (check bevy 0.14/0.15 API — confirm current version in `server/Cargo.toml`).
- Output: plain text or DOT format listing systems in execution order with their component read/write access.
- CI integration: run on each PR, save output as artifact, diff against baseline to catch unintended system reordering.
- Lives in `server/src/main.rs` (flag) or a dedicated `server/src/debug.rs` module.

**#527 — Add rng_seed field to ObserverSnapshot for deterministic replay**
- The WRONG button (#507, done Sprint 11) writes `inputs.jsonl` and `seed.txt` for replay, but `seed.txt` currently writes `"unavailable"` because the server does not include `rng_seed` in `ObserverSnapshot`.
- Fix: add `rng_seed: Option<u64>` to the `ObserverSnapshot` struct in `server/src/bridge/types.rs`.
- Populate it from the simulation's RNG state on each tick. The RNG system lives in `server/src/simulation/rng.rs`.
- This is a server-side change committed from the CI branch. Coordinate with server team to avoid conflict on `bridge/types.rs` — or implement as a separate additive commit that merges cleanly.
- Completes the WRONG button capture loop. After this, replays can fully reproduce observed bugs.

## Dependency Chain

```
#343, #344, #346, #527 — all standalone, all parallel
```

## PR Workflow

When ready to submit, create a PR with `tea` CLI. **All flags are required** to avoid TTY prompts (see CLAUDE.md "Gitea access" section):

```bash
tea pr create --repo jpmschweitzer/settled-reach --login schweitz --title "feat(ci): Sprint 12 — HashMap lint, tracing infra, schedule debug, rng_seed fix" --description "body" --base main --head ci
```