test(module): drive a source sync over the bus, not just its routing

[YellowSnnowmann]

Summary

every_declared_method_is_actually_routed walks the module's declared method list and checks each one dispatches. Six sync members are in that list and, until now, none was ever called. Routing is not behaviour, and the sync pipeline is what a connector-backed install spends its time doing.

Measured before writing anything: 143 unique declared members, 96 driven over the bus, 47 listed-only — ten of them sync.

This adds one isolated loader case that drives RunSourceSync end to end through a real dlopened module and asserts its effect, then drives SourceSyncState and pins how it refuses.

Why bus-level coverage is not redundant with the ~397 in-process sync tests

sources/sync_tests.rs, sync/audit_tests.rs, sync/workspace/periodic_tests.rs and the tinymemory-sync post-processor suites are thorough, and every one of them runs in-process. The module is a separately compiled cdylib with its own statics, so a task-local set host-side reads as absent inside it — and the two at this boundary (source_scope, the self-echo exclusion) are permission checks that fail open. The host's sync entry points are exactly where scope and credentials cross.

module_e2e.rs is the only file in the repo that loads a real module (sole match for TINYMEMORY_TEST_MODULE), so this is the only place that gap can be closed.

What the case does

Syncs a folder source over a temp directory holding one file — no network, no credentials, no fixture beyond tempfile — then asks the store, over the same bus, whether the chunks arrived.

That second call is the point. records_ingested is the run's own report; chunks_synced is the store's. A member that answered a plausible SyncRunOutcome without writing anything satisfies the first assertion and fails the second.

Mutation-tested: removing the fixture file gives records_ingested: 0 and takes the case red. Restoring it passes. An assertion that cannot fail is not coverage.

Three findings the implementation turned up, now pinned

The source registry is not admit_module's memory_sources config key. That key is the serialized registry snapshot operations read; run_source_sync resolves ids through get_source_in against the host's config.toml. Registering in the wrong one answers NotFound for every source — openhuman#5820's "no memory source registered as src_…" strand, which config.rs's own doc comment predicts, and which this test hit on its first run.

RunSourceSync writes no audit row. It goes through engine::run_source_pipeline; the audit appends live in sources::sync::sync_source, the periodic path. Asserting an audit row here — the obvious reading of the issue's "assert the audit row" — would have been asserting something the member does not do. The absence is pinned instead, with a note to invert the assertion if that ever changes: the two entry points look interchangeable from the wire, and a caller polling the audit log after a manual sync would wait forever.

SourceSyncState refuses on this engine, with Invalid and a message naming the connector module as the owner. Invalid rather than Unsupported is the distinction worth keeping — the family is served, this member's subject is not — and a refusal that says where the answer lives beats a None that reads as "no state yet".

Scope

Two members driven properly, per the issue: RunSourceSync and SourceSyncState. BackfillConnectorTrees, SyncAuditLog (beyond reachability), IsToolkitSyncable and RunConnectionSync stay out — each needs a Composio connection, which is not hermetic. The NotFound contract for an unregistered id is asserted too; it came free and nothing checked it.

Test plan

• cargo fmt --all --check and the module manifest's own
• cargo clippy --manifest-path crates/tinymemory-module/Cargo.toml --all-targets -- -D warnings
• cargo build --workspace / cargo test --workspace
• Module lane by hand: TINYMEMORY_TEST_MODULE=… cargo test --manifest-path crates/tinymemory-module/Cargo.toml --test module_e2e — 13 isolated cases (12 existing + this one), harness green
• Mutation-tested: fixture removed → red, restored → green

Closes #149

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 3288f7a6-e1bc-45ea-9731-bbc240674e41

---

_{Comment @coderabbitai help to get the list of available commands.}

[tinysweeper]

tinysweeper found nothing blocking. Approving.

             $0.0024 · 32,809 in / 326 out · 0 cached (0%) · openrouter/openai/text-embedding-3-small, deepseek/deepseek-v4-flash · 490 embedded
critique:    $0.0009 · 13,214 in / 73 out  · 0 cached (0%) · deepseek/deepseek-v4-flash
security:    $0.0009 · 13,193 in / 148 out · 0 cached (0%) · deepseek/deepseek-v4-flash
description: $0.0005 · 6,402 in  / 105 out · 0 cached (0%) · deepseek/deepseek-v4-flash

[tinysweeper]

How this change flows

1 changed behaviour across 13 relationships. 6 surrounding behaviours are shown (60 graph nodes walked). 40 further behaviours left out to keep the diagram readable.

flowchart LR
  n0["...s_its_provider_registry_inside_the_module<br/>changed"]:::changed
  n1["call"]:::impacted
  n2["assert"]:::impacted
  n3["admit_module"]:::impacted
  n4["Result"]:::impacted
  n5["proxy"]:::impacted
  n6["..._runs_over_the_bus_and_lands_in_the_store"]:::impacted
  n0 -->|calls| n1
  n0 -->|tests| n1
  n0 -->|calls| n2
  n0 -->|calls| n3
  n0 -->|uses| n4
  n0 -->|calls| n5
  n1 -->|uses| n4
  n6 -->|calls| n1
  n6 -->|tests| n1
  n6 -->|calls| n2
  n6 -->|calls| n3
  n6 -->|uses| n4
  n6 -->|calls| n5
  classDef changed fill:#0d4429,stroke:#238636,color:#e6edf3
  classDef impacted fill:#161b22,stroke:#6e7681,color:#c9d1d9
  classDef flagged fill:#5a1e02,stroke:#d93f0b,color:#ffffff
  classDef blocking fill:#67060c,stroke:#f85149,color:#ffffff

Loading

_{Green: changed behaviour. Grey: surrounding behaviour. Arrows name the call, use, implementation, or test relationship. Orange: has findings. Red: has a finding that blocks the merge.}