🧊 Coldkeep v1.10.7 — Critical-Path Coverage Gates

[franchoy]

Coldkeep v1.10.7 — Critical-Path Coverage Gates

Release status

Pre-release message prepared while PR CI is still running.

Do not publish this release until:

• the pull request has passed required CI checks;
• the critical-coverage-report job has completed successfully;
• Codacy has been reviewed;
• the PR has been merged into main;
• the v1.10.7 tag has been created from the merged main commit.

Release theme

v1.10.7 establishes the critical-path coverage gate foundation for Coldkeep.

This release improves release visibility and correctness governance by adding a report-only critical coverage workflow, formal baseline evidence, invariant mapping, soft threshold policy, regression-prevention design, and local validation records.

This release does not enforce coverage thresholds.

What changed

Critical package inventory

v1.10.7 adds an explicit inventory of packages tracked for critical-path coverage.

The inventory separates packages into three groups:

• Tier 1 — release-critical: storage, verify, maintenance, GC, snapshot, and recovery paths.
• Tier 2 — catalog / graph / storage infrastructure: database, graph, blocks, container, and retention paths.
• Tier 3 — observe-only: CLI, observability, benchmark, scripts, adversarial tests, and integration tests.

The goal is to make coverage meaningful by linking it to correctness-critical areas instead of using a raw global percentage.

Report-only critical coverage script

Added:

scripts/critical_coverage.sh

Supported mode:

scripts/critical_coverage.sh --report
scripts/critical_coverage.sh --report --csv-output <path>

The script generates per-package coverage visibility for the tracked critical packages.

It does not:

• enforce thresholds;
• compare against the baseline;
• add soft gates;
• add hard gates;
• fail because a package is below a target percentage.

Formal coverage baseline

Added formal baseline artifacts:

docs/release/v1.10/v1.10.7-coverage-baseline.csv
docs/release/v1.10/v1.10.7-coverage-baseline.md

The baseline records the current critical-path coverage state and treats it as evidence, not as a pass/fail threshold.

Known low-visibility Tier 1 packages are documented as future hardening priorities:

internal/maintenance: 47.6%
internal/recovery: 44.9%

These are not v1.10.7 release blockers.

Invariant coverage map

Added:

docs/release/v1.10/v1.10.7-invariant-coverage-map.csv
docs/release/v1.10/v1.10.7-invariant-coverage-map.md

The invariant map connects coverage evidence to named correctness invariants, including:

• GC reachability;
• deterministic restore;
• verify fail-closed behavior;
• recovery fail-closed behavior;
• snapshot roots;
• packed / legacy parity;
• physical metadata reuse;
• catalog compatibility;
• container placement;
• retention policy.

This keeps coverage tied to Coldkeep’s correctness-first model.

Soft threshold policy

Added:

docs/release/v1.10/v1.10.7-soft-threshold-policy.md

The policy defines:

• baseline-required evidence;
• regression warning semantics;
• Tier 1 / Tier 2 / Tier 3 interpretation;
• v1.10.7 blockers and non-blockers;
• future hard-gate candidates.

The policy is advisory in v1.10.7. It does not enforce absolute thresholds.

Report-only CI integration

Added a CI job:

critical-coverage-report

The job runs the report-only coverage script and uploads a CSV artifact.

It should fail only if report generation fails, Go coverage execution fails, or the script/runtime fails.

It should not fail because of low package coverage percentages.

Regression-prevention design

Added:

docs/release/v1.10/v1.10.7-regression-prevention-design.md

This document defines future regression-prevention behavior, including:

• what counts as a coverage regression;
• future soft-gate behavior;
• future hard-gate behavior;
• material decrease direction;
• explanation requirements;
• invariant linkage requirements;
• possible future comparison artifact shape.

No regression enforcement is implemented in v1.10.7.

Local validation evidence

Added local validation evidence:

docs/release/v1.10/v1.10.7-local-validation-results.md
docs/release/v1.10/v1.10.7-pre-pr-quality-gate-results.md

Local gates recorded before PR:

go test ./...                 PASS
go test -race ./...           PASS
go vet ./...                  PASS
critical coverage report      PASS, report-only
release evidence validation   PASS
CI/script report-only check   PASS
scope guard                   PASS

What this release does not do

v1.10.7 does not:

• change Coldkeep product behavior;
• change storage behavior;
• change restore behavior;
• change verify behavior;
• change GC behavior;
• change snapshot behavior;
• change recovery behavior;
• change catalog behavior;
• change Go source files;
• change Go test files;
• change dependencies;
• enforce coverage thresholds;
• compare coverage against baseline;
• add --soft-gate;
• add --hard-gate;
• introduce global coverage gates;
• add mutation testing;
• add filesystem fault injection;
• implement engine extraction;
• implement catalog abstraction.

Correctness interpretation

v1.10.7 improves coverage visibility and release governance.

It does not claim that coverage percentages prove correctness.

Coverage is treated as useful evidence only when connected to named Coldkeep correctness invariants.

---

This discussion was created from the release 🧊 Coldkeep v1.10.7 — Critical-Path Coverage Gates.