Skip to content

IMPLEMENTATION_PLAN.md

Latest commit

 

History

History
877 lines (635 loc) · 29.1 KB

IMPLEMENTATION_PLAN.md

File metadata and controls

877 lines (635 loc) · 29.1 KB

Implementation Plan

This is the working PR checklist for building ripr incrementally. It is more operational than the roadmap: each entry should become a scoped PR with clear artifacts, tests, documentation updates, and gates.

The checklist is grouped into implementation campaigns. A Codex Goal may work through multiple work items in a campaign, but each work item should follow the scoped PR contract.

Campaign Map

Campaign Objective Work items
Agentic DevEx Foundation Make the repo safe for Codex Goals and human review. policy/architecture-guard, output/output-contract-check, docs/codex-goals-campaigns, fixtures/runner-comparison-v1, fixtures/first-two-goldens, testing/test-oracle-report, dogfood/static-self-check
Syntax-Backed Analyzer Foundation Move the analyzer from lexical facts to syntax-backed facts. analysis/file-facts-model, analysis/syntax-adapter-mvp, design/rust-syntax-substrate, analysis/ast-test-oracle-extraction, analysis/ast-probe-ownership, analysis/ast-probe-generation
Evidence Quality Improve oracle strength, local flow, activation values, output evidence, and stop reasons. output/unknown-stop-reason-invariant, analysis/oracle-strength-v2, analysis/local-delta-flow-v1, analysis/activation-value-modeling-v1, output/evidence-first-output, fixtures/negative-metamorphic-baseline
Test Efficiency and Vacuity Signals (4A) Make low-discriminator, smoke-only, broad-oracle, opaque, circular, and duplicate test signals visible as advisory evidence; ship ripr and ripr+ badge artifacts. test-efficiency/test-fact-ledger, test-efficiency/vacuous-signal-v1, test-efficiency/duplicate-discriminator-v1, test-efficiency/report-and-metrics, badge/ripr-count-v1, badge/ripr-plus-count-v1, badge/repo-scope-artifacts, badge/publish-main-endpoint
Repo Seam Inventory and Test Grip (4B) Inventory behavior seams, classify test-grip per seam, and turn actionable gaps into editor diagnostics and agent-ready packets. spec/repo-seam-inventory, analysis/repo-seam-model-v1, analysis/repo-seam-inventory-v1, analysis/test-grip-evidence-v1, analysis/repo-ripr-classification-v1, output/repo-exposure-report-v1, lsp/repo-seam-diagnostics-v1, lsp/seam-evidence-hover-v1, context/agent-seam-packets-v1, docs/agent-dispatch-workflow-v1
Seam Evidence Usability and Precision (5A) Make repo seam evidence fast, precise, and directly actionable for developers and coding agents. Complete: #255, #310, #313, #314, #315, #316, #327, and campaign/seam-evidence-usability-closeout.
Operationalization (5B) Govern analyzer behavior with repository config, integrate SARIF/CI policy modes, and remap badges onto seam-native counts. config/ripr-config-v1, ci/sarif-ci-policy, badge/seam-native-count-mapping
Module SRP Refactoring (6) Refactor internal modules under crates/ripr/src/ so each module has one product responsibility, without splitting the package. See docs/IMPLEMENTATION_CAMPAIGNS.md Campaign 6.

The active machine-readable campaign is .ripr/goals/active.toml. Campaigns 1 through 5A are complete. The active queue is now Campaign 5B Operationalization. config/ripr-config-v1 established the repo-root configuration surface; ci/sarif-ci-policy added SARIF rendering and an opt-in baseline policy, and badge/seam-native-count-mapping is the next ready item.

PR 0: planning-and-tracking-docs

Purpose: put the plan, engineering rules, metrics, ADRs, specs, changelog, and traceability conventions in the repository before analyzer rewrites begin.

Deliverables:

  • Update docs/ROADMAP.md with the release sequence and PR queue.
  • Add an implementation checklist that future PRs can update.
  • Add ADR scaffolding and initial ADRs for product-shaping decisions.
  • Add spec scaffolding for behavior contracts.
  • Add metrics definitions for capability and regression tracking.
  • Add learnings and repo-knowledge log.
  • Add spec-test-code traceability rules.
  • Update the README doc index and metric summary.
  • Add a root changelog.
  • Add PR review checklist guidance.
  • Add contributor workflow guidance.
  • Add CI strategy guidance.
  • Add dogfooding guidance.
  • Add ADR and spec templates.
  • Add changelog policy guidance.
  • Add scoped evidence-heavy PR doctrine.
  • Add first executable policy checks for static language and panic-family debt.

Acceptance:

  • A contributor can identify the next PR from docs alone.
  • A contributor can identify which spec, tests, and code modules belong together for a feature.
  • The docs state that production and test code should avoid panic, unwrap, and expect, and that existing uses are tracked debt.
  • The docs preserve the product contract and conservative static language.
  • PRs are scoped by production risk rather than line count.

PR 1: verify-one-click-extension-install

Purpose: verify the normal VS Code extension path without requiring users to install ripr separately.

Deliverables:

  • Manual install verification matrix for VS Marketplace and Open VSX.
  • Fresh-profile check with no ripr on PATH.
  • Server auto-download and checksum verification evidence.
  • Output-channel log checklist for mode, base, config, server path, and download source.
  • Clear-error scenarios for disabled auto-download, missing manifest, unsupported platform, and checksum mismatch.

Tests and gates:

  • cd editors/vscode && npm ci
  • cd editors/vscode && npm run compile
  • cd editors/vscode && npm run package

PR 1A: xtask-policy-checks

Purpose: expand the initial policy checks into a broader local and CI quality rail.

Deliverables:

  • Move static language and panic-family checks into CI.
  • Add markdown local link check.
  • Add doc index check for README, docs, specs, and ADRs.
  • Add traceability manifest validation.
  • Add capability matrix validation.
  • Add PR-scope check for production delta and evidence delta.

Acceptance:

  • cargo xtask ci-fast runs the core policy checks.
  • Existing debt is allowlisted with counts, and new debt fails the check.
  • Docs explain how to remove allowlist entries as debt is paid down.

PR 1B: rust-first-file-policy

Purpose: keep repo implementation and automation Rust-first by denying unapproved non-Rust programming files, checked-in executable scripts, and workflow shell sprawl.

Deliverables:

  • Add Rust-first file policy docs.
  • Add non-Rust allowlist with owner, kind, and reason.
  • Add workflow shell-budget allowlist.
  • Add cargo xtask check-file-policy.
  • Add cargo xtask check-executable-files.
  • Add cargo xtask check-workflows.
  • Wire checks into cargo xtask ci-fast.
  • Wire checks into CI.

Acceptance:

  • Rust is documented as the default implementation and automation language.
  • Existing VS Code, workflow, docs, fixture, asset, and config surfaces are explicitly allowlisted.
  • New shell, Python, JavaScript, TypeScript, or other programming files outside approved surfaces fail the file policy check.
  • Checked-in executable bits fail unless allowlisted.
  • Long workflow run blocks fail unless allowlisted.

Future policy PRs:

  • generated-file policy
  • dependency-surface policy
  • process-spawn policy
  • network policy
  • workspace-shape policy
  • architecture import guard
  • public API guard

PR 1C: spec-fixture-contracts

Purpose: make specs and fixtures agent-readable and mechanically checkable before fixture and golden output work expands.

Deliverables:

  • Add spec format reference.
  • Add test taxonomy reference.
  • Add fixture contract README.
  • Update existing specs to the checked format.
  • Add cargo xtask check-spec-format.
  • Add cargo xtask check-fixture-contracts.
  • Wire checks into cargo xtask ci-fast.
  • Wire checks into CI.

Acceptance:

  • Every docs/specs/RIPR-SPEC-*.md has required sections and a valid status.
  • Spec filename IDs match title IDs.
  • Future fixture directories must include SPEC.md, diff.patch, and expected/check.json.
  • Fixture SPEC.md files must include Given/When/Then/Must Not sections.

PR 1D: automation-guardrails

Purpose: finish the first Rust-first policy family by making generated files, dependency surfaces, process spawning, and network behavior explicit.

Deliverables:

  • Add generated-file allowlist and cargo xtask check-generated.
  • Add dependency-surface allowlist and cargo xtask check-dependencies.
  • Add process-spawn allowlist and cargo xtask check-process-policy.
  • Add network allowlist and cargo xtask check-network-policy.
  • Wire checks into cargo xtask ci-fast.
  • Wire checks into CI.
  • Update the file policy, CI docs, contributor docs, and PR template.

Acceptance:

  • Tracked generated lockfiles and future fixture goldens require explicit allowlist entries.
  • New dependency manager files fail unless they belong to approved Cargo, VS Code, or fixture surfaces.
  • New process spawning fails unless allowlisted with a reason.
  • New network behavior fails unless allowlisted with a reason.

PR 1E: shape-fix-pr

Purpose: add the first mutating PR-shaping commands without changing existing policy semantics.

Deliverables:

  • Add cargo xtask shape.
  • Add cargo xtask fix-pr.
  • Run cargo fmt through shape.
  • Sort .ripr/*.txt and policy/*.txt allowlists through shape.
  • Ensure target/ripr/reports exists.
  • Write target/ripr/reports/shape.md.
  • Write target/ripr/reports/fix-pr.md.
  • Document safe mutations and repair guidance.

Acceptance:

  • cargo xtask shape passes.
  • cargo xtask fix-pr passes.
  • cargo xtask ci-fast passes after shaping.
  • Shaping does not add policy exceptions or bless output drift.

PR 1F: pr-summary

Purpose: generate a reviewer packet before human review without mutating source files.

Deliverables:

  • Add cargo xtask pr-summary.
  • Read changed paths from git diff and git status.
  • Write target/ripr/reports/pr-summary.md.
  • Classify production delta and evidence/support delta.
  • Classify detected surfaces, public contracts, and policy exceptions.
  • Suggest reviewer focus files.
  • Update cargo xtask fix-pr to refresh the PR summary after shaping.

Acceptance:

  • cargo xtask pr-summary passes.
  • target/ripr/reports/pr-summary.md exists after the command.
  • cargo xtask fix-pr refreshes shape, PR summary, and fix-pr reports.

PR 1G: automation-path-docs

Purpose: document the fix/check/guide operating model and the Codex Goals campaign handoff so automation and analyzer implementation work share the same review contract.

Deliverables:

  • Add a PR automation operating model.
  • Document deterministic shaping, non-mutating checks, and repair briefs.
  • Document the scoped PR contract.
  • Record the automation cutoff that made Campaign 1 safe to leave setup mode.
  • Link the new docs from the roadmap, documentation map, agent workflow, contributor docs, and README.

Acceptance:

  • A contributor can identify which cleanup should be automated and which changes require explicit judgment.
  • A coding agent can identify the next automation PRs without confusing them with product campaign work.
  • A coding agent can use a standard task template for the analyzer queue.

PR 1H: check-pr-precommit

Purpose: add obvious local gates for cheap pre-commit checks and review readiness checks.

Deliverables:

  • Add cargo xtask precommit.
  • Add cargo xtask check-pr.
  • Keep precommit cheap and non-mutating.
  • Make check-pr run the review-ready command set that exists today.
  • Update CI, contributor, and agent docs.

Acceptance:

  • cargo xtask precommit passes on main.
  • cargo xtask check-pr passes on main.
  • check-pr does not run release packaging unless the repo later adds a path-aware release lane.

PR 1I: guided-check-reports

Purpose: make existing policy checks emit repair briefs instead of only command failure text.

Deliverables:

  • Add a shared report model or helper for Markdown check reports.
  • Upgrade static-language, panic-family, file-policy, executable-file, workflow, spec-format, fixture-contract, generated, dependency, process, and network checks to write reports under target/ripr/reports.
  • Classify failures as auto-fixable, author decision, reviewer decision, or policy exception.
  • Include exact rerun commands and exception templates where useful.

Acceptance:

  • Each upgraded check writes a useful report on failure.
  • Successful checks either write a pass report or are summarized by pr-summary.
  • Report generation does not hide the non-zero exit status of failed checks.

PR 1J: ci-report-artifacts

Purpose: make CI upload review artifacts even when a check fails.

Deliverables:

  • Run cargo xtask pr-summary where possible in CI.
  • Defer metrics report generation until cargo xtask metrics exists.
  • Upload target/ripr/reports with an always step.
  • Document report artifact names and expected contents.

Acceptance:

  • CI artifacts include the PR summary and any check reports that were generated before failure.
  • CI remains non-mutating.

PR 1K: fixture-golden-scaffolding

Purpose: add the command surface for fixture execution and golden comparison before analyzer internals change.

Deliverables:

  • Add cargo xtask fixtures.
  • Add cargo xtask fixtures <name>.
  • Add cargo xtask goldens check.
  • Add cargo xtask goldens bless <name> --reason "...".
  • Document the fixture and golden directory conventions.

Acceptance:

  • Fixture commands pass with a clear "no fixtures found" message if no executable fixtures exist yet.
  • Existing fixture contract checks still pass.
  • Golden blessing requires an explicit reason.

PR 1L: traceability-spec-id-checks

Purpose: make spec IDs and behavior manifest entries checkable.

Deliverables:

  • Harden .ripr/traceability.toml.
  • Add cargo xtask check-spec-ids.
  • Add cargo xtask check-behavior-manifest.
  • Add warning-only drift checks for analysis, output, docs, fixture, and metric changes.

Acceptance:

  • Accepted specs point to real docs and at least one test or fixture unless explicitly planned.
  • Fixture specs reference valid spec IDs.
  • Missing expected evidence appears in the PR summary.

PR 1M: capability-metrics-report

Purpose: make capability progress and automation debt visible.

Deliverables:

  • Add or harden a machine-readable capability source.
  • Add cargo xtask metrics.
  • Add cargo xtask check-capabilities.
  • Write target/ripr/reports/metrics.md or metrics.json.
  • Keep the README capability snapshot aligned with the capability source.

Acceptance:

  • Capability statuses have valid values and required fields.
  • Stable or calibrated statuses require the evidence defined by policy.
  • Metrics reports are generated without changing product behavior.

PR 1N: architecture-guard

Purpose: protect internal seams while keeping one published package.

Deliverables:

  • Add cargo xtask check-workspace-shape.
  • Add cargo xtask check-architecture.
  • Add cargo xtask check-public-api or document why it is deferred.
  • Add policy metadata for allowed workspace packages and module-boundary rules.

Acceptance:

  • New workspace packages require an explicit approved policy entry.
  • Domain and analysis layers cannot accidentally depend on adapters.
  • CLI, LSP, and output layers do not own exposure classification.

PR 1O: readme-state-and-link-checks

Purpose: make README state and Markdown links part of the checked trust packet.

Deliverables:

  • Add cargo xtask check-readme-state.
  • Add cargo xtask markdown-links.
  • Check README front-door sections and headline capability snapshot shape.
  • Check README/capability matrix checkpoint drift against metrics/capabilities.toml.
  • Check repo-local Markdown links in tracked .md files.
  • Wire the checks into precommit and ci-fast.
  • Update CI and PR automation docs.

Acceptance:

  • Deleted or renamed docs fail before review when still linked.
  • README remains linked to active campaign, metrics, capability, and automation docs.
  • cargo xtask check-readme-state and cargo xtask markdown-links pass on main.

PR 1P: campaign-manifest-check

Purpose: make the active Codex Goals campaign queue mechanically checkable and reportable.

Deliverables:

  • Add cargo xtask check-campaign.
  • Add cargo xtask check-goals as an alias.
  • Add cargo xtask goals status.
  • Add cargo xtask goals next.
  • Validate .ripr/goals/active.toml against docs/IMPLEMENTATION_CAMPAIGNS.md.
  • Validate work item IDs, statuses, branch fields, acceptance claims, stackability, merge boundaries, blocked dependencies, and command names.
  • Wire the manifest check into precommit and ci-fast.

Acceptance:

  • cargo xtask check-campaign passes on main.
  • cargo xtask goals status writes target/ripr/reports/goals.md.
  • cargo xtask goals next writes target/ripr/reports/goals-next.md.

PR 1Q: fixtures-runner-comparison-v1

Purpose: make fixture and golden commands execute the current product and compare actual output against checked-in expected output.

Deliverables:

  • cargo xtask fixtures runs all fixtures when fixture directories exist.
  • cargo xtask fixtures <name> runs one fixture.
  • Actual JSON and human outputs are written under target/ripr/fixtures/<name>/.
  • cargo xtask goldens check compares actual check.json and optional human.txt outputs against fixtures/<name>/expected/.
  • cargo xtask goldens bless <name> --reason "..." requires a reason, updates expected/check.json and expected/human.txt, and appends the fixture changelog.

Acceptance:

  • Fixture commands still pass with a clear report when no fixture directories exist.
  • Golden checks fail on drift without mutating expected outputs.
  • Golden blessing remains explicit and does not run from shape or fix-pr.

PR 2: fixture-laboratory

Purpose: build the regression control bench before changing analyzer internals.

Deliverables:

  • fixtures/boundary_gap
  • fixtures/weak_error_oracle
  • fixtures/field_not_asserted
  • fixtures/side_effect_unobserved
  • fixtures/smoke_assertion_only
  • fixtures/no_static_path
  • fixtures/opaque_fixture
  • fixtures/workspace_cross_crate
  • fixtures/duplicate_symbols
  • fixtures/stacked_test_attrs
  • fixtures/nested_src_tests_layout
  • fixtures/macro_unknown
  • fixtures/snapshot_oracle
  • fixtures/mock_effect

Each fixture should include:

  • source and tests
  • diff.patch
  • expected JSON output
  • expected human output
  • expected context packet
  • expected LSP diagnostic shape when relevant

Invariants:

  • Static output never says killed or survived.
  • Unknowns include stop reasons.
  • Weak or smoke oracle evidence does not silently become strong.
  • Finding order is deterministic.
  • Context packets are parseable.

PR 2A: testing-test-oracle-report

Purpose: measure ripr's own test oracle strength as analyzer work expands.

Deliverables:

  • cargo xtask test-oracle-report writes target/ripr/reports/test-oracles.md.
  • cargo xtask test-oracle-report writes target/ripr/reports/test-oracles.json.
  • cargo xtask check-test-oracles aliases the same advisory report.
  • The report classifies detected Rust tests as strong, medium, weak, or smoke.
  • Existing weak or smoke debt is advisory and non-blocking.

Acceptance:

  • cargo xtask test-oracle-report
  • cargo xtask check-test-oracles
  • cargo xtask metrics
  • cargo xtask check-pr

PR 2B: dogfood-static-self-check

Purpose: add a focused non-blocking ripr-on-ripr report.

Deliverables:

  • cargo xtask dogfood runs stable fixture diffs through ripr check --mode fast.
  • Actual dogfood JSON and human outputs are written under target/ripr/dogfood/<fixture>/.
  • target/ripr/reports/dogfood.md summarizes findings, exposure classes, runtime, and errors.
  • target/ripr/reports/dogfood.json provides the same advisory summary for future machine readers.
  • Dogfood is advisory and non-blocking.

Acceptance:

  • cargo xtask dogfood
  • cargo xtask check-pr

PR 3: file-facts-model

Purpose: introduce an internal fact model while preserving current scanner behavior.

Deliverables:

  • FileFacts
  • FunctionFact
  • TestFact
  • OracleFact
  • CallFact
  • ReturnFact
  • StructConstructionFact
  • EnumConstructionFact
  • LiteralFact
  • BuilderChainFact
  • EffectFact

Acceptance:

  • Existing sample findings are unchanged.
  • Analysis consumes facts rather than ad hoc scanner structures.
  • Scanner behavior remains available as the fallback.

PR 4: syntax-adapter-mvp

Purpose: create the parser boundary before relying on parser-specific details.

Deliverables:

  • RustSyntaxAdapter trait or equivalent boundary.
  • Lexical adapter summarize_file implementation.
  • Changed range to syntax-node mapping.
  • No public API commitment to a parser crate.
  • Parser substrate decision recorded in ADR 0006.
  • Parser-backed summarize_file implementation.

Acceptance:

  • Existing outputs remain stable or intentionally updated with fixture evidence.
  • Parser errors produce static_unknown or structured diagnostics, not panics.

PR 5: ast-test-oracle-extraction

Purpose: extract tests and oracles from syntax nodes instead of line substrings.

Deliverables:

  • #[test] function extraction.
  • Stacked attribute preservation.
  • Multi-line assertion macro extraction.
  • assert!, assert_eq!, assert_ne!, assert_matches!, and matches! handling.
  • unwrap and expect smoke-oracle handling.

Acceptance:

  • Fixture output remains deterministic.
  • Line scanning is fallback only.

PR 6: ast-probe-ownership

Purpose: attach probes to stable owner symbols.

Deliverables:

  • Diff hunk to changed text range.
  • Changed range to syntax-backed owner node.
  • Syntax node to enclosing function, method, or module.
  • Stable SymbolId.

Acceptance:

  • Duplicate function names across modules or crates do not cross-link tests.
  • Probe IDs remain stable enough for explain and context.

PR 7: ast-probe-generation

Purpose: generate probes from syntax kind and ownership facts.

Deliverables:

  • Predicate boundary probes.
  • Return value probes.
  • Error path probes.
  • Field construction probes.
  • Side-effect or call-change probes.
  • static_unknown fallback with reason.

Acceptance:

  • Multi-line predicate changes produce one useful probe.
  • Tail-expression return changes produce return probes.
  • Err(Error::X) changes produce error-path probes.

PR 8: oracle-strength-v2

Purpose: make oracle kind and strength explicit and probe-relative.

Deliverables:

  • Exact value oracle.
  • Exact error variant oracle.
  • Broad error oracle.
  • Whole-object equality oracle.
  • Snapshot oracle.
  • Mock expectation oracle.
  • Relational check oracle.
  • Shape-only oracle.
  • Smoke-only oracle.
  • Unknown oracle kind.

Acceptance:

  • is_err() differs from exact error variant assertions.
  • unwrap() differs from exact return assertions.
  • JSON and human output keep the stable schema while rendering probe-relative oracle strength.

PR 9: local-delta-flow-v1

Purpose: explain what changed behavior appears to flow to.

Deliverables:

  • Changed expression to let binding flow.
  • Binding to return flow.
  • Binding to struct field flow.
  • Changed expression to Ok or Err flow.
  • Predicate branch to return or field construction flow.
  • Changed call to effect boundary candidate.

Acceptance:

  • Findings can name at least one sink when locally visible.
  • propagation_unknown includes a concrete stop reason.

PR 10: activation-value-modeling-v1

Purpose: detect whether tests appear to activate the changed behavior.

Deliverables:

  • Numeric and string literal value facts.
  • Function argument value facts.
  • Builder-chain value facts.
  • Table-row value facts.
  • Enum variant value facts.
  • Boundary equality discriminator facts.

Acceptance:

  • Boundary findings include detected values.
  • Boundary findings include missing equality value.
  • Opaque fixtures produce infection_unknown, not false confidence.

PR 11: evidence-first-output

Purpose: make CLI output the reference explanation.

Deliverables:

  • Changed behavior section.
  • RIPR stage evidence section.
  • Related tests section.
  • Oracle evidence section.
  • Missing discriminator section.
  • Next step section.
  • Stop reason section for unknowns.

Acceptance:

  • Golden human and JSON output cover current Campaign 3 fixtures.
  • Static language remains conservative.
  • Negative and metamorphic fixtures cover noise-only and syntax-variant cases.

PR 12: lsp-evidence-hover-actions

Purpose: make editor diagnostics specific and actionable.

Deliverables:

  • Diagnostic data with finding and probe IDs.
  • Stable diagnostic codes.
  • Hover evidence for exact finding.
  • Copy context packet code action.
  • Open related tests code action.
  • Run deep check command.
  • Output-channel lifecycle logs.

Acceptance:

  • didChange refreshes diagnostics after debounce.
  • Code action copies the context for the selected finding.

PR 13: agent-context-v2

Purpose: turn ripr context into a test-writing brief.

Deliverables:

  • Recommended test location.
  • Related existing tests.
  • Fixture or builder hints.
  • Missing input values.
  • Missing oracle shape.
  • Suggested assertion shapes.
  • Confidence and stop reasons.

Acceptance:

  • Context packet is golden-tested.
  • CLI and LSP use the same packet shape.

PR 14: ripr-config-v1

Purpose: let repositories teach ripr topology and oracle conventions.

Deliverables:

  • Workspace-root config discovery.
  • Missing config accepted.
  • Useful invalid-config errors.
  • Test topology override.
  • Custom oracle macro config.
  • Snapshot, mock, and external-boundary config.

Acceptance:

  • Config changes oracle classification only through explicit rules.

PR 15: suppression-v1

Purpose: support honest noise control without hiding the model.

Deliverables:

  • Inline suppression comment form.
  • Config suppression form.
  • Required reason.
  • Optional expiry.
  • --show-suppressed.

Acceptance:

  • Suppressed findings remain visible when requested.
  • Suppression rate can be measured.

PR 16: sarif-ci-policy

Purpose: support PR workflows without making default CI noisy.

Deliverables:

  • SARIF output.
  • Markdown summary.
  • JSON artifact guidance.
  • Advisory mode.
  • Opt-in failure modes.
  • Baseline-aware mode.

Acceptance:

  • SARIF validates.
  • SARIF results point to static evidence locations.
  • Blocking policy is opt-in.

PR 17: cargo-mutants-calibration-scaffold

Purpose: compare static predictions with real mutation results.

Deliverables:

  • Import cargo-mutants output through cargo xtask mutation-calibration.
  • Match static seam evidence to runtime records by seam_id first and unambiguous normalized file/line second; report ambiguous file/line candidates separately.
  • Emit advisory static class vs runtime outcome reports at target/ripr/reports/mutation-calibration.{json,md}.
  • Keep mutation-runtime language out of static findings; runtime vocabulary is confined to calibration/runtime reports.

Acceptance:

  • Runtime mutation vocabulary appears only in explicit calibration data and static-language checks remain clean.

PR 18: persistent-cache-v1

Purpose: cache stable facts after the fact model is worth caching.

Deliverables:

  • File-hash invalidation.
  • Warm FileFacts reuse.
  • LSP reuse of test and oracle facts.
  • Graceful stale-cache recovery.

Acceptance:

  • Warm run avoids reparsing unchanged files.

Required Gates

Rust PRs must run:

cargo fmt --check
cargo check --workspace --all-targets
cargo test --workspace
cargo clippy --workspace --all-targets -- -D warnings
cargo doc --workspace --no-deps
cargo package -p ripr --list
cargo publish -p ripr --dry-run

Extension PRs must run:

cd editors/vscode
npm ci
npm run compile
npm run package