From 1e33415f670b68efde7d5d6bd546b78953c4f8a6 Mon Sep 17 00:00:00 2001 From: Dotta <34892728+cryppadotta@users.noreply.github.com> Date: Fri, 13 Nov 2026 12:55:31 -0600 Subject: [PATCH] test(runner): run Kimi layout suites once in prep (#20487) ## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The Runner package has node:test suites that verify the pinned Kimi Code native runtime layout. > - `test:typescript:prep` selects each Kimi layout suite directly. The same command also selects the `build-verified-provider-entrypoints` aggregator. The aggregator imported the same four Kimi suites. > - One command therefore registered every Kimi layout test twice. Commit b866fa4bd (#18028) added both routes in the same change. The duplicate has no second contract: both routes run the same files, assertions, and fixtures in the same required run. > - This pull request removes the aggregator imports. The direct prep selection stays. Each Kimi suite now runs once per command. > - The benefit is less test work and less CI time in each prep run, with the same assertions. ## Linked Issues or Issue Description No public issue exists. Problem description: **What existing behavior does this improve?** The `test:typescript:prep` command in `packages/paperclip-runner/package.json`. It selects test files for `pnpm test:typescript` and `check:protocol` (release verification). **Current behavior** The command passes `build-verified-provider-entrypoints.test.mjs` and the four Kimi suites to one `node --test` run. The aggregator file also imports the four Kimi suites. Each Kimi test file then registers its tests twice: once inside the aggregator process, and once in its own process. Local measurement on the exact prep selection: 81 test registrations, 71 pass, 10 skip, about 6.6 seconds. **Proposed behavior** The aggregator keeps its own two tests and its imports of suites that prep does not select (`provision-pi-package`, the four Antigravity suites). The four Kimi suites run only through their direct prep selection. Same selection of the aggregator, same files, same assertions. Local measurement after the change: 62 test registrations, 56 pass, 6 skip, 0 fail, about 4.6 seconds. The removed 19 registrations are exact duplicates; each Kimi test still runs once. **Reason and benefit** Double registration wastes CI and developer time on each prep run and makes test counts misleading. One run of each suite keeps the same regression proof. **Breaking changes** None. No production code changes. The suites and their assertions are unchanged. ## What Changed - Removed the four Kimi suite imports from `packages/paperclip-runner/scripts/build-verified-provider-entrypoints.test.mjs` (`materialize-kimi-distribution`, `provision-kimi-package`, `patch-kimi-runtime`, `kimi-metadata-probe`). - Added a comment that says the Kimi suites are selected directly by `test:typescript:prep`, so this aggregator must not import them again. - No changes to `package.json`, routing, or the suites themselves. ## Verification - Checked the routing: `test:typescript:prep` in `packages/paperclip-runner/package.json` lists all four Kimi suites directly, so each suite stays selected with multiplicity one after this change. The aggregator is only run through that same command; no workflow or script runs it alone. - Baseline (master `233f61564`): `node --test` with the exact prep selection of the aggregator plus the four Kimi suites -> 81 tests, 71 pass, 0 fail, 10 skipped, 6.6 s. - After the change (this head): same command -> 62 tests, 56 pass, 0 fail, 6 skipped, 4.6 s. 19 exact duplicate registrations removed; the Kimi suites still run and pass once each. - Timing is a single local sample; treat the speed difference as indicative, not proven. The structural change (19 duplicate registrations removed) is exact. - Not run: full `pnpm -r typecheck`, full workspace test suite, Rust checks. A filtered dependency install was used because the workspace disk is 10 GB and a full workspace install exceeds it. Required PR CI supplies the typecheck, build, and vitest coverage for this head. ## Risks Low risk. The aggregator still loads and its own two tests pass. If a future lane runs the aggregator alone, the Kimi suites would not run there; today no such lane exists, and the prep command still selects them directly. ## Model Used - Provider: Z.ai, model GLM (assisted by Paperclip's Test Gardener agent). - Exact model ID: `openrouter/z-ai/glm-5.3`. - Reasoning/thinking mode: standard chain-of-thought; no extra thinking mode. - Capabilities: tool use (shell, file edit, GitHub CLI), repo test execution. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [ ] All Paperclip CI gates are green - [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [ ] I will address all Greptile and reviewer comments before requesting merge --- .../scripts/build-verified-provider-entrypoints.test.mjs | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/packages/paperclip-runner/scripts/build-verified-provider-entrypoints.test.mjs b/packages/paperclip-runner/scripts/build-verified-provider-entrypoints.test.mjs index 70e2a1e2c53..7b7fbf3edeb 100644 --- a/packages/paperclip-runner/scripts/build-verified-provider-entrypoints.test.mjs +++ b/packages/paperclip-runner/scripts/build-verified-provider-entrypoints.test.mjs @@ -40,12 +40,10 @@ test("written provider entrypoints satisfy qualified launch permissions", async }); // Package-layout regression runs in the existing verified-entrypoint test lane. +// Kimi layout suites are selected directly by test:typescript:prep; do not +// import them here or they register twice in that required run. import "./provision-pi-package.test.mjs"; -import "./materialize-kimi-distribution.test.mjs"; -import "./provision-kimi-package.test.mjs"; -import "./patch-kimi-runtime.test.mjs"; -import "./kimi-metadata-probe.test.mjs"; import "./antigravity-provisioner-layout.test.mjs"; import "./materialize-antigravity-distribution.test.mjs"; import "./probe-antigravity-metadata.test.mjs";