test(review): block #1819/#1915 acceptance until the evidence foundation ships - #276
Open
Alan-TheGentleman wants to merge 2 commits into
Open
test(review): block #1819/#1915 acceptance until the evidence foundation ships#276Alan-TheGentleman wants to merge 2 commits into
Alan-TheGentleman wants to merge 2 commits into
Conversation
…no-action absence Wave 1's final work unit (W3). Adds tests/native-review-parity-issue-evidence.test.ts and its pure support module tests/support/provider-behavior-evidence.ts: - evaluateProviderBehaviorAssertion: a pure evaluator proving Pi asserts only transport and outcome for #1819 (corrected-delivery topology) and #1915 (retry-successor authority) -- both provider-owned. A development/bootstrap evidence class never reaches "pass"; absent evidence always reports blocked, naming consume-gentle-ai-release-artifacts as the missing dependency; a mutable-build evidence class is refused at runtime, not only by the TS union type. - The #1819/#1915 final-acceptance test against the real foundation harness (tests/evidence/**, not present in this repository yet) is DELIBERATELY FAILING. This is the correct fail-closed state for "Qualifying evidence absent at verification" -- there is no test.skip, no early return, and no swallowed exception. Its failure message explains why the red result is correct and exactly what has to ship before it turns green. - scanForReconstructionSymbols: a static guard proving lib/ and extensions/ declare no delivery-topology or authority-graph symbol. It scans declaration sites only (function/class/interface/type/const/let/enum names), not bare substrings, so it does not false-positive on the two pre-existing legitimate occurrences of "squash" (a --squash git commit flag) and "receipt_discovery" (an echoed provider capability name). - #2074 (Claude user-registry migration) and #910 (PowerShell fallback ladder) are proven no-action by asserting the surfaces genuinely do not exist in lib/, extensions/, or scripts/, that no other test file names either issue, and that the Wave 1 coverage enumeration lists both as no-action rather than as passing fixtures. - The RDD boundary is asserted at both the start and the end of this file's run against the real installed binary: global/effective mode stays off, and review validate's delivery evidence reports disabled/unmanaged. pnpm test: 1067 tests, 1062 pass, 4 fail (the 3 pre-existing RDD-off failures in tests/native-review-parity-runtime.test.ts, unchanged from baseline, plus this file's one deliberately-failing acceptance test).
All 11 W3 tasks (W3.1-11) checked off. Wave 1 (#2028 Pi-host behavior) is now complete: W1 (capture/admission/transport), W2 (diagnostics/relaunch/ lost-output/cleanup), and W3 (evidence consumption + no-action assertions) are all implemented.
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Repository UI Review profile: ASSERTIVE Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Final slice of Wave 1, stacked on #275. Completes #2028 host behavior.
This PR intentionally adds one failing test. Please do not silence it.
The final-acceptance check for #1819 and #1915 runs against the foundation's evidence harness, which lives in the sibling change
consume-gentle-ai-release-artifactsand has not shipped into this repository yet.Absent qualifying evidence must block. A silent skip would record a false success for behavior nobody verified — the exact failure this spec exists to prevent. So the test fails, loudly, with a message that explains it is the correct fail-closed state, names the two conditions that turn it green, and asks the reader not to add a skip or delete it.
It fails in both directions: if the harness is missing it fails saying so, and if the harness appears but the real check has not been wired, it fails demanding that work. It cannot pass by accident.
Its failure is scoped to itself via a dynamic import rather than a top-level one, so it does not abort the rest of the file.
Confirmed the only intentional failure Wave 1 adds: the suite reports 1062 pass and 4 fail — this one plus the three pre-existing receipt-driven-development-disabled failures, byte-identical to the pre-W3 baseline.
Pi asserts outcomes. It never rebuilds provider algorithms.
#1819 and #1915 are provider-owned: corrected-delivery topology and receipt discovery, retry-successor authority and evidence-graph validation. A static guard proves Wave 1 introduced no delivery-topology or authority-graph symbols into
lib/orextensions/.The guard scans declaration sites only, not bare substrings, because a naive scan false-positived on two legitimate strings: a
--squashgit-commit flag in an allowlist, and an echoed provider capability name. Matching those would have made the guard noise.#2074 and #910 are proven by asserting absence
Prose cannot fail, and a fixture would manufacture the false coverage claim the plan warns about. So the tests assert the surfaces genuinely do not exist: no Claude user-registry write path, no PowerShell fallback ladder, no fixture or journey named for either issue, and the coverage enumeration lists both as no-action rather than as covered.
All passed on first run. The surfaces really are absent.
The receipt-driven-development boundary is asserted, not assumed
Global and effective mode is checked as
offbefore and after the file runs, against the real installed binary rather than a fake, and delivery evidence reportsdisabled/unmanaged.Rollback
Delete the two new files. Nothing depends on them, and the foundation harness is untouched because it does not exist here.