Stabilize workspace app-boot smoke timing and diagnostics #354

New issue

Closed

opened 2026-06-07 08:18:52 +00:00 by steve · 0 comments

steve commented

2026-06-07 08:18:52 +00:00

Owner

Stabilize the workspace browser app-boot smoke path so CI failures are easier to diagnose and less likely to fail on cold Linux runners while preserving the current boot-order contract.

Problem

Recent public Forgejo checks.yml failures are not failing in pnpm check:all; they are failing later in pnpm test:smoke during packages/workspace/e2e/app-boot-smoke.spec.ts. The first smoke test times out waiting for app.workspace.activeLeaf after creating a fresh Browser vault, even though the same path often passes locally and in the containerized CI repro.

That leaves two gaps:

The boot-smoke helper waits on one narrow ready signal (activeLeaf plus .workspace) and provides little context when it times out.
The smoke harness does not surface enough startup-phase detail to distinguish a true boot regression from a slow-but-progressing cold CI boot.

Goal

Make workspace app-boot smoke failures produce actionable diagnostics tied to the current startup phase and visible failure UI when the shell does not finish booting in time.
Reduce avoidable flakiness in the browser smoke helper without masking real boot regressions.

Scope

Update the workspace smoke helper(s) and/or smoke spec so timeout failures capture the most relevant startup state before failing.
Tighten the readiness predicate or timeout handling only where that improves CI stability without weakening the regression signal.
Keep the current app boot contract intact: vault creation still leads into the normal App.svelte startup path, and the smoke test should still fail on startup errors.

Non-goals

Reworking the runtime boot order in App.svelte or VaultBootstrap.svelte unless the investigation proves a real product bug.
Relaxing the smoke test so far that it stops detecting broken startup states.
Changing unrelated daily-use or desktop smoke behavior.

Acceptance Criteria

Workspace smoke timeout failures report enough state to see the current boot phase, any visible startup error panel, and relevant console/page errors without rerunning blindly.
The browser app-boot smoke path is less sensitive to cold CI timing while still failing on real startup regressions.
packages/workspace/e2e/app-boot-smoke.spec.ts continues to assert successful shell boot and core app-service availability.
Relevant workspace validation passes, including package checks/tests and the workspace smoke suite.

Implementation Notes

The public Forgejo run that motivated this follow-up is checks.yml run 246 for commit 36851532759a on 2026-06-07; pnpm check:all and pnpm test both passed there, and the failure occurred in pnpm test:smoke.
The same timeout signature also appeared in earlier public checks.yml failures, so this looks like a recurring smoke-stability problem rather than one isolated commit.
Keep any new diagnostics local to the smoke harness or targeted helper paths unless a product-level bug is identified.

Suggested Files or Specs To Inspect

packages/workspace/e2e/app-boot-smoke.spec.ts
packages/workspace/e2e/boot-smoke.helpers.ts
packages/workspace/playwright.smoke.config.ts
packages/workspace/scripts/run-smoke.mjs
packages/workspace/src/lib/components/app/App.svelte
spec/src/20-packages/workspace/index.md
spec/src/40-testing/current-state.md

Validation Commands

pnpm --filter @lapis-notes/workspace check:all
pnpm --filter @lapis-notes/workspace test
pnpm --filter @lapis-notes/workspace test:e2e:smoke

Follow-up to #23 — Broaden regression coverage for workspace layout restore, community plugin flows...

Follow-up Tasks

Implementation Summary

Reworked workspace app-boot smoke readiness waits to report shell and startup diagnostics instead of timing out on activeLeaf.

Stabilize the workspace browser app-boot smoke path so CI failures are easier to diagnose and less likely to fail on cold Linux runners while preserving the current boot-order contract. ## Problem Recent public Forgejo `checks.yml` failures are not failing in `pnpm check:all`; they are failing later in `pnpm test:smoke` during `packages/workspace/e2e/app-boot-smoke.spec.ts`. The first smoke test times out waiting for `app.workspace.activeLeaf` after creating a fresh Browser vault, even though the same path often passes locally and in the containerized CI repro. That leaves two gaps: - The boot-smoke helper waits on one narrow ready signal (`activeLeaf` plus `.workspace`) and provides little context when it times out. - The smoke harness does not surface enough startup-phase detail to distinguish a true boot regression from a slow-but-progressing cold CI boot. ## Goal - Make workspace app-boot smoke failures produce actionable diagnostics tied to the current startup phase and visible failure UI when the shell does not finish booting in time. - Reduce avoidable flakiness in the browser smoke helper without masking real boot regressions. ## Scope - Update the workspace smoke helper(s) and/or smoke spec so timeout failures capture the most relevant startup state before failing. - Tighten the readiness predicate or timeout handling only where that improves CI stability without weakening the regression signal. - Keep the current app boot contract intact: vault creation still leads into the normal `App.svelte` startup path, and the smoke test should still fail on startup errors. ## Non-goals - Reworking the runtime boot order in `App.svelte` or `VaultBootstrap.svelte` unless the investigation proves a real product bug. - Relaxing the smoke test so far that it stops detecting broken startup states. - Changing unrelated daily-use or desktop smoke behavior. ## Acceptance Criteria - [ ] Workspace smoke timeout failures report enough state to see the current boot phase, any visible startup error panel, and relevant console/page errors without rerunning blindly. - [ ] The browser app-boot smoke path is less sensitive to cold CI timing while still failing on real startup regressions. - [ ] `packages/workspace/e2e/app-boot-smoke.spec.ts` continues to assert successful shell boot and core app-service availability. - [ ] Relevant workspace validation passes, including package checks/tests and the workspace smoke suite. ## Implementation Notes - The public Forgejo run that motivated this follow-up is `checks.yml` run `246` for commit `36851532759a` on 2026-06-07; `pnpm check:all` and `pnpm test` both passed there, and the failure occurred in `pnpm test:smoke`. - The same timeout signature also appeared in earlier public `checks.yml` failures, so this looks like a recurring smoke-stability problem rather than one isolated commit. - Keep any new diagnostics local to the smoke harness or targeted helper paths unless a product-level bug is identified. ## Suggested Files or Specs To Inspect - `packages/workspace/e2e/app-boot-smoke.spec.ts` - `packages/workspace/e2e/boot-smoke.helpers.ts` - `packages/workspace/playwright.smoke.config.ts` - `packages/workspace/scripts/run-smoke.mjs` - `packages/workspace/src/lib/components/app/App.svelte` - `spec/src/20-packages/workspace/index.md` - `spec/src/40-testing/current-state.md` ## Validation Commands - `pnpm --filter @lapis-notes/workspace check:all` - `pnpm --filter @lapis-notes/workspace test` - `pnpm --filter @lapis-notes/workspace test:e2e:smoke` ## Related Issues - Follow-up to #23 — [Broaden regression coverage for workspace layout restore, community plugin flows...](../closed/23-broaden-regression-coverage-for-workspace-layout-restore-community-plugin-flows.md) ## Follow-up Tasks ## Implementation Summary Reworked workspace app-boot smoke readiness waits to report shell and startup diagnostics instead of timing out on activeLeaf.

steve

2026-06-07 08:18:52 +00:00