[anolisa] test: cover adapter observation contracts #3310

Activity

1. ikunkun-sys

   added

   component:anolisasrc/anolisasrc/anolisa

   type:testTest coverage, acceptance, or validation workTest coverage, acceptance, or validation work

   on Sep 17, 2026
2. github-actions commented on Sep 17, 2026

   

   github-actionsBot

   on Sep 17, 2026 – with GitHub Actions

   Last edited by github-actions

   More actions

   🔀 Issue Router

   • Component: anolisa
   • Confidence: 96%
   • Owners notified: @kongche-jbw @ikunkun-sys
   • No assignee was added; assignment records active implementation ownership.

   问题涉及 ANOLISA CLI 中适配器观测契约的回归测试。

   Evidence: 标题前缀和组件标签均指向 anolisa，正文未出现其他组件的路径或模块线索。

   This routing was applied by the ANOLISA Issue Router through GitHub Actions.
3. ikunkun-sys commented on Sep 17, 2026

   

   ikunkun-sys

   on Sep 17, 2026

   CollaboratorAuthor

   More actions

   Local implementation and validation completed on feature/anolisa/adapter-observation-contract, based on upstream 9fc69e0. Tests only; no production API or behavior changes.

   All requested targeted tests passed in an unprivileged, network-disabled ALinux 4 temporary container with read-only source/vendor and an independent target: core manager 99, managed_files 10, adapter_manager 87, CLI adapter 73, CLI status 104. Full CLI: 1177 passed; full workspace including doctests: 2577 passed. fmt, workspace check, all-target clippy with warnings denied, rustdoc and diff check passed.

   Checks preserve protected state/receipt/resources/framework files and no install lock. OpenClaw status retains its existing plugins list invocation audit. Named component status retains its separate repository-index cache behavior, isolated with a temporary file:// index fixture.

   No commit, push or PR yet, per the requested execution boundary. This issue and R5 tracker #3309 remain open until merge.
4. Forrest-ly commented on Sep 17, 2026

   

   Forrest-ly

   on Sep 17, 2026

   Last edited by Forrest-ly

   Collaborator

   More actions

   A PR for this is up: #3313 (head 0ace70d0). Tests only — two test files, +705/-2, no production evidence model / public API / behavior change.

   Coverage, all driven through the real AdapterManager::scan / status collectors on shared fixtures:

   • source authority — receipt row reports Available; a receipt that outlives its source stays listed but flips to Missing with a reason, and the vanished directory stops being offered as resource_root
   • candidate vs receipt sets — a declared/discovered row with no receipt carries source_status: None and yields zero status rows
   • integrity precedence — SourceAvailable leads, then ManagedBundleMatches, then SourceRevisionMatches in fixed order and never duplicated by the driver's copies; a lost source degrades without faking integrity (Unknown, never False); on-disk tampering is attributed to ManagedBundleMatches=False only
   • read-only boundaries — scan spawns no framework query at all, status issues only read-only verbs; installed.toml stays byte-identical, the package-owned datadir and framework-side trees are unchanged, no install lock is taken, and the queries remain auditable via the existing central log plus the fake CLI's argv log

   Plus three CLI tests that project a real scan result into AdapterObservation instead of a hand-written ScanEntry, so collector/projection drift fails in test.

   CI is green (Test anolisa, AW / required, AW / contracts); the PR is only blocked on a review. Locally: cargo test --workspace --locked green (adapter_manager 86 → 93, CLI bin → 1109), cargo fmt --all --check and all-target clippy clean, and both mutations were re-confirmed to fail exactly the tests that pin them (dropping the projection's source_status mapping; reordering the manager's condition insertion).

   One acceptance criterion I could not meet, flagged rather than papered over: validation in an isolated unprivileged ALinux 4 container. My host has no usable container runtime, so my run was unprivileged ALinux 3 x86_64 with fake framework commands and no network. The suite is host-agnostic (tempdir-rooted layouts, no privileges, no real framework binary), but since you already have the ALinux 4 harness that validated feature/anolisa/adapter-observation-contract, it would close the gap if you could run it against #3313 — happy to respin the branch if anything shows up.

   One thing the tests surfaced, deliberately not changed here: AdapterSourceRevision is built from the recorded package inventory rather than from disk, so SourceRevisionMatches cannot see on-disk tampering — only ManagedBundleMatches can. Coherent, but its doc comment doesn't say which bytes are compared. Worth a doc-only clarification in a later #3309 slice.
5. Forrest-ly

   added 2 commits that reference this issue on Sep 17, 2026

   test(anolisa): cover adapter observation contracts

   ...

   a5f5170

   test(anolisa): cover adapter observation contracts

   ...

   18b7e6f
6. kongche-jbw

   added a commit that references this issue on Sep 18, 2026

   test(anolisa): cover adapter observation contracts

   ...

   5d684af
7. kongche-jbw

   closed this as completedin #3313on Sep 18, 2026
8. Forrest-ly

   added a commit that references this issue on Sep 18, 2026

   Merge branch 'main' into fix/AGE-7186-openclaw-adapter-enable-memory-…

   ...

   7c25aa8
9. haosanzi

   added a commit that references this issue on Sep 20, 2026

   test(anolisa): cover adapter observation contracts

   ...

   4b63611