Independent **package-manager** harness in `zed-pkg-test` for `zed-pkg`.
**Readiness:** `ready`
**Primary dependency strategy:** `zed`
**Scheduled cadence:** `23 4 * * 2,5` UTC
**Live infrastructure:** None for deterministic pull-request checks.
## Upstream repositories
- `zed-pkg/zed-cli`
-
zed-pkg/zed-interfaces- Verify direct and indirect cycle detection without partial links or corrupt CAS state across the supported happy-path states and canonical fixtures.
-
Verify direct and indirect cycle detection without partial links or corrupt CAS state under retries, interruption, concurrency, offline operation, or partial failure.
-
Verify direct and indirect cycle detection without partial links or corrupt CAS state preserves authorization, idempotency, integrity, observability, and actionable failure classification.
This repository tests the upstream through independent installation paths:
./scripts/bootstrap-upstream.sh git-submodule./scripts/bootstrap-upstream.sh zed./scripts/bootstrap-upstream.sh native-package
The publisher materializes a real Git submodule when authenticated access is available. Zed and native package coordinates are recorded in
dependency-contract.yaml; missing unpublished packages are reported as blocked readiness rather than silently skipped.python3 -m pip install -e '.[test]' pytest -q ./scripts/readiness.py --offline ./scripts/run-live.shPull requests validate the harness and deterministic contract fixtures. Secret-, service-, emulator-, desktop-, database-, provider-, chaos-, scale-, and soak-dependent checks run by schedule or manual dispatch.
A live result must be classified as one of:
- product regression — a behavioral invariant fails after dependencies are ready;
- blocked dependency — an upstream, credential, package, emulator, provider sandbox, or deployment is unavailable;
- harness regression — generated metadata, fixtures, workflow, or runner setup is invalid.
Managed by
github-test-org-factory/1.0.0.