test: assert observed results instead of pins

[ScriptedAlchemy]

Summary

• Review of feat: TraceDecay V2 delivery (PR8–PR13) + daemon performance, memory, and reliability overhaul #707 through test-behavior-not-implementation: tests call the subject and match a literal observed result. Constant pins, absence-only checks, and weak is_ok / is_err / toBeTruthy assertions do not count.
• Merged tip 3e6167b265 (test(code-index): re-pin partitioned codec onto extractor.rust.v8) into this branch. No conflicts.
• Fixed the publication-pointer memo so a same-size rewrite in the same mtime quantum is reread. That was the CompareAndSwap miss in publication_over_an_undecodable_active_generation_refuses_a_moved_pointer after the tip landed NotApplicable / reused. Does not merge feat: TraceDecay V2 delivery (PR8–PR13) + daemon performance, memory, and reliability overhaul #707 to master.

Motivation

Principle: test-behavior-not-implementation.

A constant pin fails when the constant is edited and never observes the mechanism that reads it. An absence-only or toBeTruthy check still passes if the subject returns the wrong non-empty value. A weak is_err passes for the wrong refusal.

Changes

Typed refusals and observed readouts:

• Sealed format compatibility is true/false for literal revisions, not REVISION == 12.
• Clone fingerprint constants are not restated; winnow boundaries stay.
• Sanitizer version is asserted on the receipt. A 1,048,576-byte record is admitted; 1,048,577 bytes is TooLarge.
• Registry workers definition reports schema_revision == 7. Exact workers: 0 is the typed NonCanonical refusal; workers: 2 is Ok(()).
• Identity, handoff, source-edit, CI localization, proximity, topology, query inventory, and cancellation constructors assert the typed error and a concrete accepted value.
• Dashboard text that users see is asserted as that string, not toBeTruthy. Node count is 12.9K (figure plus unit), not the figure fragment alone.

Publication store:

• Tip already covers CodeIndexDemandAdmissionV1::NotApplicable and no longer reads a missing reused field on ChangedCodeChunkSetV1.
• Pointer memo reused on mtime+size alone, so the undecodable-pointer test restored an equal-length pointer and still compared the previous identity. Reuse now requires the file-byte digest.

Test plan

• Dashboard vitest on the rewritten files: 31 files, 289 passed, then ScopedBrain 12 passed after the 12.9K readout assertion
• Domain/contracts/global-db/code-index lib tests named in the first assertion pass
• remote_query_request_enforces_shard_inventory_bounds_and_identity
• generation_manifest_requires_matching_canonical_language_revisions (digest mismatch, then order / set mismatch after restamp)
• source_edit_effect_rejects_zero_catalog_revision, source_edit_effect_rejects_zero_privacy_key_epoch, source_edit_effect_requires_the_exact_current_grant
• publication_store lib tests: 27 passed, including publication_over_an_undecodable_active_generation_refuses_a_moved_pointer
• Full workspace nextest was not run

Checklist

• No secrets
• No changelog: tests plus an internal pointer-read correctness fix
• Breaking changes: none

  

[changeset-bot]

⚠️ No Changeset found

Latest commit: 9d92dc6

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[ScriptedAlchemy]

Integrated or superseded directly on the #707 branch at ff36c4f; follow-up compile/typecheck regressions were fixed on that branch.