Repository navigation
docs: point stale owner cites and a ci path at the merged refactors - #15645
Conversation
UX Review (Fable 5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing to check. [UX-REVIEWED] d8845d9 Description read[DESCRIPTION-READ] 1009a6592a3ad15d23f3b57e442c5cec073ae3e36ee8aedc742795a60a1bd4e5 That sha256 names everything this verdict read: the PR description, plus the 3 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Design Review (Fable 5) — ✅ PASSDesign-level review of Design-Verdict: PASS One check: [DESIGN-REVIEWED] d8845d9 Description read[DESCRIPTION-READ] 296639aa8f0baa5aebb905dfd7720d003f3328b096b4fa9d16ab6a0b8994e531 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Fable 5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Verify the test's hand-rolled glob matcher against dorny/paths-filter v4's real picomatch semantics (dotfiles, brace sets) — a divergence greens the test while the lane still skips. What this change shipsInventory (8 items) — 7 justifiedIntent: FIX — stale "defined in X" cites left by the merged facade-split refactors, plus the CI gap where update-lane owners skipped Linux Packaging.
[FIRST-PRINCIPLES-REVIEWED] d8845d9 Description read[DESCRIPTION-READ] d46f8d0faf1939208291e47da2582156541f74a546bfd422336117c778505945 That sha256 names the PR description this verdict read. Recompute it with |
GPT 5.6 Review — ✅ no blocking findingsGPT 5.6 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
Opus 5 Review — ✅ human override acceptedReviewed Human judgment by @bolichen97 overrides the Opus 5 finding for [OPUS-OVERRIDE] d8845d9 Verdict recorded from an authorized human decision for commit False positive or not applicable? A repository writer can comment: |
|
Security Scope Review -- ✅ human override accepted@bolichen97 overrode this lane for [SCOPE-REVIEWED] d8845d9 |
The merged component refactors moved code out of facade files into owner modules, but some comments, docstrings, specs, RFCs and one CI path filter still named the facade as the place that defines the code. A reader who followed them landed on a re-export line. A PR that edited only website/electron/runtime/update/* also skipped the Linux Packaging job, because the linux_packaging filter still listed only auto-update.js. Re-point those sentences to the module that now owns the code. Add runtime/update/** to the linux_packaging filter; the job and its gate logic do not change. Add a "where new code goes" paragraph to the specs of the facades that kept most of their file (config, acp-client, the chat sidebar in history.md, the electron README). Record the composer's benign effect reorder and the caption literals' whitespace in the files that own them. Fix the README's auto-update export count, and prune six docs-lint baseline entries that stopped firing when these refactors fixed their cites. test_ci_linux_packaging_filter.py walks the relative-require closure of every Electron entry file the filter lists. It fails when a module that closure loads is outside the filter, which is the case at the base.
2a31aa2 to
d8845d9
Compare
|
Infra rerun on head d8845d9. Nine runs died on
|
|
Rebase probe for green age. Main moved 7460284 -> 7114800. In that range, #15672 edits I cherry-picked this head (d8845d9) onto 7114800 in a throwaway worktree:
So no rebase push for this staleness. |
|
/ai-review override fable d8845d9: Opus was withheld twice by the CI installation API rate limit (runs 36806630831 attempts 1-2); the attempt-2 withheld-verdict artifact (verdict.md) reads 'Opus 5 Review — no blocking findings', and its single advisory (scripts/smoke-linux-packages.sh absent from linux_packaging) is a pre-existing CI-filter gap since #4174, out of this docs-only PR's goal and logged as a CI-maintainer follow-up. |
|
/ai-review override scope d8845d9: Security Scope attempt 1 (run 36806630819) died on the CI installation API rate limit before candidates.json was written; the attempt-2 rerun measured success but is pinned by the per-head monotonic floor to the infra failure; this head changes only comments, docs, one ci.yml path-filter line and a test, with no security-relevant code. |
Human judgment recorded@bolichen97 marked the fable AI finding as false positive, not applicable, or explicitly accepted for
This decision applies only to this commit. A new push requires a new judgment. |
Human judgment recorded@bolichen97 marked the scope AI finding as false positive, not applicable, or explicitly accepted for
This decision applies only to this commit. A new push requires a new judgment. |
iamwhatever
left a comment
There was a problem hiding this comment.
Reviewed. Every hunk in host_auth.py, hooks.py, config/loader.py, config/sections.py and agent_files.py is comment or docstring text; the adf_markdown_safety.json change touches only the description strings; five re-pointed references resolve on main. Two items are not docs and should be said in the title or body: ci.yml:159-161 adds website/electron/runtime/update/** to the linux_packaging paths filter (that job now runs for diffs there, which is correct since the lane files live there), and test/test_ci_linux_packaging_filter.py is a new executable gate that follows the require() chain from each entry file. Both read as right; the macOS shard red is the inherited pair that #15680 fixes.
Problem / Motivation
Goal: Comments, specs and the Linux Packaging CI trigger name the modules that now own the code the merged component refactors moved.
The merged component refactors split large facade files into owner modules. Some sentences in comments, docstrings, specs and RFCs still say the facade defines the code, so a reader who follows one finds only a re-export line. One CI path filter has the same gap:
linux_packaginglistswebsite/electron/auto-update.jsbut not the update lanes that file composes. A PR that edits onlywebsite/electron/runtime/update/feed-lane.js, where the per-format Linux feed is chosen, therefore skips Linux Packaging (build + smoke-install).Why it matters
A stale "defined in X" cite costs every reader a wrong turn. It also invites new code into the facade, which is where most later work on four of these refactors has landed. A Linux packaging regression in an update-lane owner would ship without the one job that installs the packages for real.
Not a goal
What changed (motivation → approach → change)
.github/workflows/ci.ymllinux_packagingalso triggers onwebsite/electron/runtime/update/**(audit M1 / F10-1).hooks.py,history_search.py,dep_sync.py,agent_files.py,agent_sdk/host_auth.py,providers/mirrors/registry.py,doctor_checks/agents.py,config/loader.py,config/sections.py,apps/routes.py,apps/bridges.py,apps/builtins/{meetings,issue_radar,auto_research}, and inwebsite/src/components,website/src/hooks,website/src/lib,website/src/pages,website/src/store,website/src/types,website/src/utils,website/src/appsandwebsite/electron.app-kit-platform.md, four RFCs and.kiro/specs/pod-cli/design.md. A "where new code goes" paragraph inconfig.md,acp-client.md,history.md(the chat-sidebar owner map) andwebsite/electron/README.md. The README's auto-update export count is now 16.ChatInput.refactor.ownership.test.tsrecords the composer's benign effect reorder;runtime/window/linux-captions.jsrecords that its injected literals are whitespace-normalized..github/docs-lint-baseline.txttest/fixtures/adf_markdown_safety.json,test/docstrings_commentand docstring re-points only.Each new sentence was checked against the definition at the base commit: the cited file defines the named symbol. The old text was checked to be false (or, for line cites, to point at other code). No test asserts any of the changed text: no test file contains any removed fragment.
Audit items closed (IDs from the quality audit of the merged refactors and its doc follow-up list): M1/F10-1; D1 (hooks, history_search, ADF fixture and test); D2; D3; D4; D5 (six RFC sentences, plus two more in the same files and for the same cause); D6 (
api.ts); D7 (five F05 cites,AboutPanel.downloadStates.test.tsx, README count); D8 (L369, L382, L385, L388, L585, L606); D9 (all four); D11 (both); the B03, B05, B08, B09, I06, F02, F03, F06 and F07 doc follow-ups that still held at the base.Skipped, with reasons:
sandbox.py, D7index.css, D6gatewayd.py:30docstring: excluded until their owners (B10, F09, QF2) merge.aws-control.mdbackup composition sentence): the text is delegated to the open scope that holds that spec's lease and is rewriting the same sentence.notify_out.py:67-68: still true at the base (routes.pystill imports it).SettingsPanel.tsx:798and itsopsMissionControl.test.tscomment twin: the surrounding claim ("dispatch.pyzero", "the only caller") has been false since before the refactor. A true rewrite reverses a user-facing string that a test asserts, so it is left as a follow-up for the ops-mission-control owner, not a text sync.rfc-security-conductor.md:88: a fact record measured at a named older commit, which is still true for that commit.rfc-i18n-measurement.md:238-239: still true at the base.onToolCall(L48): it stops firing only because a refactor added a same-named code identifier. Pruning it would turn a later rename into a failure on an unrelated PR.resource-protection.md,mcp-gateway-daemon-lifecycle.md,service/apparmor.py,pod/runtime_boot.py,memory-skills-hooks.md,check_sync_io_in_async.py,memory-benchmark.yml,bundle.py.backend_mcp_ability.py:60,dashboard/urls.py,mcp_cleanup.py:322,issue_radar/backend/watch.py:17,rfc-issue-radar-crews.md:129,test_review_findings_prb.py:24.security.md,platform-context.md,taskq.md,injected-messages.md,learn-cron-dashboard.md,docs/feature-map/README.md,website/docs/extension-seams.md,ChatPage.tsx,pages/chat/ActivityViewer.tsx,cli_doctor.py:857,pod/runtime.py,test_core_path_redact_before_bound.py,test_routes_composition_contract.py.test_harness_parity._RUNTIME_PATH_MODULESand the I02 characterization rows.knowledge.md:409(model and tools claim), theconfig.md"28 sections" count,build-windows.yml:27-28,macos-on-demand.yml:148(a deleted test file),agent_files.py:6-9.Backwards compatibility
Compatible: the diff changes comments, docstrings, docs and one CI trigger list. The added path makes Linux Packaging run on more PRs; it never skips a PR the filter selected before.
Tests
test/test_ci_linux_packaging_filter.py(new). It walks the relative-require closure of each Electron.jsentry thelinux_packagingbucket lists, and asserts that every module in the closure matches the bucket. At the base it fails, namingruntime/update/{feed-lane,managed-lane,state-reporter}.js. With the filter line it passes. Two companion tests pin the glob matcher's*/**semantics and that the walk reaches past the entries.ci.yml, plustest_app_mcp_scoping.py,test_docs_lint_fact_checks.py,test_mirrored_owner_storage.pyandtest_prepare_pr_profiles.py(10044 passed, 6 skipped). Electron unit tests (2305 passed), the 31 related vitest files (612 passed),tsc -p tsconfig.app.json, eslint on the touched files, the i18n literal check, docs-lint (34 stale baseline entries left, down from 40), and the comment-history, brand, black, flake8, isort and mypy gates.scripts/local-gate.py --dry-runselects 2104 backend and 1295 frontend files, so the broad set is left to CI.Manual verification
N/A: no runtime behaviour changes. The CI change was checked by parsing
ci.ymlwith the bucket's glob semantics, before and after.Screenshots / video
Why no screenshot: the website files change only in comments; nothing renders differently.
Related Issues
no linked issue: a follow-up to the audit of the merged component refactors.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)