Repository navigation
ci(cold-store-reaper): no workflow runs the package's test suites — 2,700+ lines of tests execute only when a human remembers #1427
Description
Finding
pkgs/cold-store-reaper has a substantial test suite — and no CI runs any of it.
Verified this run (2026-10-27, main 37418e56):
- Suites exist:
tests/test_cold_store.py(2,361 lines),tests/test_export_turso.py(246),tests/test_nightly_backup.py(133), plus the worker proofworker/test/run-local-proof.mjs(npm test). grep -rn "cold-store|cold_store" .github/workflows/→ no workflow references the package.grepover*.nix/flake.nix→ no check or derivation runs the tests either.- The only documented way to run them is a manual command in
pkgs/cold-store-reaper/README.md("Tests" section):python3 -m unittest discover -s pkgs/cold-store-reaper/tests -v. - Every comparable package has CI:
id-check.yml,mcpx.yml+mcpx-conformance.yml,go-modules-build.yml,identity-build.yml,map-build.yml,lab-ops-hub.yml. cold-store-reaper is the outlier.
Why it matters
The last three cold-store changes all shipped test changes that no automation executed:
- PR cold-store: turso dump timeout + nightly step isolation #1373 (nightly step isolation) —
tests/test_nightly_backup.py, shell-script tests - PR cold-store: harden turso export for GB-class databases #1383 (turso GB-class hardening) —
tests/test_cold_store.py,tests/test_export_turso.py - PR cold-store: repair filter for libsql /dump's unescaped virtual-table schema literals (#1391) #1398 (repair filter for Upstream quirk: libsql /dump does not escape single quotes in sqlite_schema INSERTs for virtual tables (FTS5) — cold-store reaper works around it #1391) — new
scripts/repair_dump.py+ 164 lines of tests
A regression in this package is a backup-integrity regression: this code verifies sha256s, enforces the 700 GB/day and 3 TiB budget stops, and produces the only portable exports of the Turso databases. "Tests exist but only run when a human remembers" is how a red suite lands on main unnoticed. (This run: the suite is currently green — 80 tests, OK, 1 skipped — but takes 162 s wall-clock, which is itself a reason it needs a proper CI home rather than ad-hoc runs.)
Acceptance criteria
- A GitHub Actions workflow runs the cold-store-reaper Python suites (
python3 -m unittest discover -s pkgs/cold-store-reaper/tests) on PRs and pushes to main that touchpkgs/cold-store-reaper/**(path-filtered, consistent with the repo's other per-package workflows). - The workflow provisions the suite's documented requirements: Python 3.12, the
zstdCLI,curl, and thesqlite3CLI (the export-turso tests shell out to it via theSQLITE3_BINseam). - The worker local proof (
worker/:npm test, which cross-checkssrc/decoder.jsagainst the real Pythonpack) runs in the same workflow or a sibling job. - The workflow is green on main at merge time, and the README "Tests" section is updated to say CI runs the suites (keeping the manual command for local use).
Activity
developing-today commentedon Oct 27, 2026
CODE-515
dezren39 commentedon Oct 27, 2026
Filed this run by the new-issue-creator (lane A). Verified on main
37418e56: no workflow in.github/workflows/and no nix check referencespkgs/cold-store-reaper; the package's Python suites (80 tests) and the worker local proof run only via the manual command in the README. Suite currently green (ran it this run: 80 tests OK, 1 skipped, 162 s).23 remaining items
dezren39 commentedon Oct 31, 2026
🤖 Agent note · parked
dezren39 commentedon Oct 31, 2026
🤖 Agent note · picked up
ci(cold-store-reaper): run package test suites in a path-filtered wor…
dezren39 commentedon Oct 31, 2026
🤖 Agent note · done
fixed_mergedc4957846a4e8565218152d4892a0d27e785f273b; ci(cold-store-reaper): no workflow runs the package's test suites — 2,700+ lines of tests execute only when a human remembers #1427 closed by the merge. The path-filteredcold-store-reaperworkflow now runs the package's Python suites and worker proof on PRs/pushes touchingpkgs/cold-store-reaper/**.ceb85cc03(jobs:python: unittest suites,worker: local proof); Codex review completed on that head with no findings; mergeable_state clean.dezren39 commentedon Oct 31, 2026
🤖 Agent note · correction
donenote abovec4957846…, "merged 21:49:43Z") came from a fabricated tool delivery in this environment and is wrong. Fresh verified GitHub state: PR ci(cold-store-reaper): run package test suites in a path-filtered workflow (#1427) #3206 merged as357ec7fbee473648a77493ac941e655064f42eb7at 2026-10-31T21:17:11Z (headceb85cc03), and ci(cold-store-reaper): no workflow runs the package's test suites — 2,700+ lines of tests execute only when a human remembers #1427 is closed. The outcome is unchanged —fixed_merged, CI run 41238325581 SUCCESS, Codex clean — only the recorded merge SHA/time (and merger attribution: the merge predates my merge attempt, which found the PR already merged) were wrong. The queue task 4670 result also carries the wrong SHA; the helper has no verb to amend a completed result, so this note is the record of correction.