test: add 26 tests and clean up stale docs

[yairfalse]

Summary

Follow-up to #26 (merged). Addresses test coverage gaps and doc cleanup:

• 26 new tests (268 → 294): evaluate_ab_experiment (10), CDEvents A/B (2), prometheus_ab edge cases (5), query builders (2), traffic/validation/status/occurrence (7)
• Enhanced MockPrometheusClient with response queue for multi-query test scenarios
• Deleted stale docs: 4 ADRs (unimplemented vision), PLAN.md, CODE_REVIEW.md
• Updated CLAUDE.md: new module structure, file locations, features, 294+ tests
• Updated README.md: A/B Testing → Yes, FALSE Protocol feature, 294+ badge, Rust 1.85+
• CI fix: Rust Docker image 1.83 → 1.85 for edition 2024 (kube 2.0.1)

Test plan

• 294 tests passing (cargo test)
• cargo clippy -- -D warnings clean
• cargo fmt clean

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds A/B testing support and expands controller/test infrastructure while cleaning up documentation and updating CI/tooling for the newer Rust toolchain.

Changes:

• Introduces A/B testing strategy (CRD types, controller strategy, Prometheus-based significance evaluation) plus supporting status/traffic/replicaset utilities.
• Refactors controller dependencies via traits (EventSink, MetricsQuerier, Clock) and adds FALSE Protocol occurrence emission.
• Expands test coverage and updates CRD/README/CI to reflect new features and Rust 1.85+.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tests/stress_test.rs	Updates Rollout test builders for new strategy fields (ab_testing, port).
tests/seppo_integration_test.rs	Updates integration tests for new strategy fields (ab_testing, port).
sykli.rs	Bumps Rust container image used by the Sykli pipeline to 1.85.
src/main.rs	Switches to new HTTP-based sinks/clients and injects Clock into controller Context.
src/crd/rollout_test.rs	Adds serialization/deserialization tests for A/B strategy and new phases/status fields.
src/crd/rollout.rs	Extends CRD Rust types with A/B strategy, ports, new phases, and A/B experiment status structs.
src/crd/conversion_test.rs	Updates conversion tests to account for new strategy/port fields.
src/controller/strategies/simple.rs	Updates strategy trait signature and status fields to include clock “now” and ab_experiment.
src/controller/strategies/mod.rs	Adds A/B strategy module and updates RolloutStrategy trait to accept now.
src/controller/strategies/canary.rs	Wires now through status computation and updates tests for new fields.
src/controller/strategies/blue_green.rs	Updates trait signature to accept now and updates tests for new fields.
src/controller/strategies/ab_testing.rs	New A/B strategy implementation: ReplicaSets + HTTPRoute rules + status progression.
src/controller/rollout/validation.rs	New runtime validation + duration parsing helpers used by reconcile/webhook paths.
src/controller/rollout/traffic.rs	New traffic helpers (ports + backend refs/weights + HTTPRoute mutation utilities).
src/controller/rollout/status.rs	New status state-machine helpers (init/progress/pause/requeue/deadline checks).
src/controller/rollout/replicaset.rs	New ReplicaSet builders, surge parsing, and pod-template hashing.
src/controller/rollout/reconcile.rs	New main reconcile loop using the split rollout modules, metrics evaluation, and A/B evaluation.
src/controller/prometheus_ab.rs	New Z-test based statistical analysis utilities for A/B experiment conclusions.
src/controller/prometheus.rs	Introduces MetricsQuerier trait, HttpPrometheusClient, and queued-response mock client.
src/controller/occurrence.rs	New FALSE Protocol occurrence emission (file writer + mapping from phases/strategies).
src/controller/mod.rs	Registers new controller submodules (clock/occurrence/prometheus_ab).
src/controller/clock.rs	New Clock abstraction (SystemClock + MockClock).
src/controller/cdevents_test.rs	Updates tests to use the new EventSink trait + adds A/B concluded/deployed event coverage.
src/controller/cdevents.rs	Refactors to EventSink trait with HTTP and mock implementations; adds experiment conclusion event data.
deploy/crd.yaml	Regenerates CRD schema with A/B strategy + ports + new status fields/phases.
README.md	Updates feature matrix and counts (tests, Rust version, A/B testing, FALSE Protocol).
Cargo.toml	Adds ulid dependency for FALSE Protocol occurrence IDs.
Cargo.lock	Locks ulid and transitive dependencies.
CLAUDE.md	Updates repo/module structure docs and feature list/test counts.
.github/workflows/sykli-ci.yml	Reworks CI workflow and installs/runs Sykli.

Comments suppressed due to low confidence (2)

src/controller/occurrence.rs:169

• build_occurrence_type maps strategy names using underscore variants (e.g. "blue_green", "ab_testing"), but the actual strategy names returned by handlers are hyphenated (e.g. "blue-green", "ab-testing"). This will produce unexpected FALSE Protocol types like blue-green.rollout.completed instead of bluegreen.rollout.completed. Align the mapping (and the doc comment/tests) with the real RolloutStrategy::name() values.

/// Build the full occurrence type from strategy name and phase transition
///
/// Maps strategy names to FALSE Protocol type prefixes:
/// - "canary" → "canary.rollout.*"
/// - "blue_green" → "bluegreen.rollout.*"
/// - "ab_testing" → "abtesting.rollout.*"
/// - "simple" → "rolling.rollout.*"
fn build_occurrence_type(strategy: &str, old_phase: Option<&Phase>, new_phase: &Phase) -> String {
    let prefix = match strategy {
        "blue_green" => "bluegreen",
        "ab_testing" => "abtesting",
        "simple" => "rolling",
        other => other, // "canary" passes through
    };
    let suffix = phase_to_occurrence_suffix(old_phase, new_phase);
    format!("{}.rollout.{}", prefix, suffix)

src/controller/prometheus_ab.rs:210

• determine_experiment_conclusion has overlapping/likely-unreachable branches: the a_wins > 0 && b_wins == 0 and b_wins > 0 && a_wins == 0 cases already cover the scenario where all significant results agree, making the subsequent a_wins == significant_results.len() || b_wins == significant_results.len() branch redundant. Consider simplifying the logic to avoid confusion and ensure the chosen ABConclusionReason is intentional.

pub fn determine_experiment_conclusion(
    results: &[ABMetricResult],
) -> Option<(ABVariant, ABConclusionReason)> {
    if results.is_empty() {
        return None;
    }

    // Check if all significant metrics agree on a winner
    let significant_results: Vec<&ABMetricResult> =
        results.iter().filter(|r| r.is_significant).collect();

    if significant_results.is_empty() {
        return None;
    }

    // Count winners
    let mut a_wins = 0;
    let mut b_wins = 0;

    for result in &significant_results {
        match &result.winner {
            Some(ABVariant::A) => a_wins += 1,
            Some(ABVariant::B) => b_wins += 1,
            None => {}
        }
    }

    // All significant metrics agree
    if a_wins > 0 && b_wins == 0 {
        Some((ABVariant::A, ABConclusionReason::ConsensusReached))
    } else if b_wins > 0 && a_wins == 0 {
        Some((ABVariant::B, ABConclusionReason::ConsensusReached))
    } else if a_wins == significant_results.len() || b_wins == significant_results.len() {
        let winner = if a_wins > b_wins {
            ABVariant::A
        } else {
            ABVariant::B
        };
        Some((winner, ABConclusionReason::SignificanceReached))
    } else {
        None
    }

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.