infra: exclude demo-ubi.spec.ts from CI Playwright run (unblock smoke runtime budget)

[SoundMindsAI]

Summary

The smoke job's demo-ubi.spec.ts beforeAll hook drives a full demo reseed whose in-flight wall-clock alone consumes the smoke job's 25-min cap. AC-8 of feat_demo_ubi_study_comparison bounds the reseed at 1140s (~19 min hard ceiling) with §14 estimating ~28 min worst case once the Solr scenario lights up. Adding Playwright + smoke-job setup overhead pushes total wall-clock past 25 min — PR #383 run 26790636716 hit the cap at 25:18.

This PR clears the runtime block with a single-file fix:

• Story 1.1 — extend ui/playwright.config.ts's testIgnore CI-gated branch with '**/demo-ubi.spec.ts'. Single-file edit, matches the established pattern from chore_drop_demo_seed_from_ci + PR chore(ci): optimize smoke + backend jobs (buildx artifact handoff + image cache + pytest-xdist) #291's 4th-run surface (6 pre-existing CI-gated specs; this is the 7th).
• Story 1.2 — new vitest regression guard at ui/src/__tests__/playwright-config-test-ignore.test.ts. Reads playwright.config.ts as text and asserts (a) demo-ubi in CI branch, (b) all 7 expected entries present, (c) demo-ubi NOT outside the CI ternary (local coverage stays intact).
• Story 1.3 — docs/03_runbooks/smoke-solr-stability.md §5 "Reseed runtime (demo-ubi exclusion)" — explains why the exclusion exists, where it lives, the local-coverage promise, the nightly-CI caveat, and the Option C path-forward.
• Story 1.4 — .github/workflows/pr.yml comment blocks (lines 42-58 + 507-523) refreshed: stale "demo-ubi reseed exceeds the budget" framing replaced with "runtime block cleared, flip SMOKE_TEST=true after §16 verification". YAML structure untouched — comments-only diff.
• Story 1.5 — state.md updated: CI-note paragraph stale sentences replaced; Solr-not-CI-ready debt entry rewritten as fully resolved (this PR ships the third sub-task).

Operator decisions locked at idea-preflight:

• D-1/D-2 Option A locked (testIgnore exclusion). Option C (env-var scenario filter, preserves per-PR demo-ubi smoke coverage at multi-file cost) explicitly deferred per operator choice.
• D-3 Option B rejected — bumping timeout-minutes from 25 → 35 leaves <7 min margin against spec §14's ~28 min worst case; erodes with each future demo scenario.
• D-4 sibling coordination — infra_smoke_fork_pr_secret_skip edits the same pr.yml job; resolved by moving the exclusion to playwright.config.ts (sibling now has zero collision risk).

Local make up smoke retains full demo-ubi coverage (CI= unset → testIgnore CI branch doesn't fire). After this merges, operators may flip SMOKE_TEST=true to re-enable per-PR smoke signal after the §16 manual verification below.

Cross-model review trail:

• Spec: 3 GPT-5.5 cycles, 13 findings (1 H + 5 M + 7 L), all accepted and applied.
• Plan: 3 GPT-5.5 cycles, 11 findings (0 H + 4 M + 7 L), all accepted and applied.

§16 Manual verification (recorded for the reviewer)

# AC-1: CI=true excludes demo-ubi from Playwright discovery
$ CI=true pnpm --dir ui exec playwright test --list --reporter=list | grep -c demo-ubi
0
# Total: 86 tests in 30 files

# AC-2: CI=unset includes demo-ubi in Playwright discovery
$ env -u CI pnpm --dir ui exec playwright test --list --reporter=list | grep -c demo-ubi
5  # the spec file + its 4 test cases
# Total: 110 tests in 37 files

# Delta: 7 files / 24 tests removed under CI=true — matches the 7 CI-gated
# entries in playwright.config.ts's testIgnore branch exactly.

Test plan

• pnpm vitest run src/__tests__/playwright-config-test-ignore.test.ts — 3/3 pass
• pnpm exec tsc --noEmit — clean
• pnpm lint — clean (0 errors, 136 pre-existing warnings — none on new file)
• §16 manual playwright test --list AC-1 + AC-2 verified (table above)
• git diff -U0 -- .github/workflows/pr.yml | awk '/^[+-]/ && !/^([+-]{3})/ && \$0 !~ /^[+-][[:space:]]*#/' — zero output (comments-only)
• python3 -c "import yaml; yaml.safe_load(open('.github/workflows/pr.yml'))" — YAML parses, all 10 jobs intact
• wc -c state.md — 22 KB (under 60 KB pre-commit gate)
• Post-merge: operator may opt in with gh variable set SMOKE_TEST --body true to re-enable per-PR smoke signal once they're satisfied with the §16 evidence

Diff scope (vs AC-7's strict "5 files")

AC-7 said exactly 5 files. The actual branch diff has 13: the 5 deliverables, plus the standard pipeline-trail artifacts (idea preflight + spec + plan + pipeline_status — every recent feature PR ships these together), plus 4 dashboard regen files emitted by the mvp1-dashboard-regen pre-commit hook. The 5 AC-7 deliverables are still byte-identical to what the spec described:

AC-7 deliverable	File	Change
1	ui/playwright.config.ts	+13 / -0 (single entry + comment block)
2	ui/src/__tests__/playwright-config-test-ignore.test.ts	NEW (118 lines)
3	docs/03_runbooks/smoke-solr-stability.md	+47 / -0 (new §5)
4	.github/workflows/pr.yml	comments-only diff in two known blocks
5	state.md	-2 / +2 (CI note paragraph + debt entry rewrites)

The non-AC-7 files are all metadata or regen artifacts. None of them alter the runtime behavior of the fix.

Notes on the demo-ubi.spec.ts AC-8 citation drift

Both pr.yml:518-523 (now refreshed) and ui/tests/e2e/demo-ubi.spec.ts:11 carry "24 min" / "25-minute ceiling per AC-8" framing that drifts from the spec's actual 1140s/19 min hard ceiling. The pr.yml comment IS refreshed here as part of Story 1.4 (structural to the comment block being rewritten). The demo-ubi.spec.ts:11 drift is explicitly deferred per spec D-6 — the spec file is not edited at all by this PR, and touching it for an unrelated citation fix would violate AC-7's file-shape contract.

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request implements the infra_smoke_reseed_runtime_budget feature, which excludes the heavy demo-ubi.spec.ts Playwright E2E test from the CI smoke test run via the testIgnore configuration in ui/playwright.config.ts to prevent job timeouts. It also adds a Vitest regression test to guard this configuration and updates relevant documentation. The reviewer provided valuable feedback on the new test file, suggesting the use of import.meta.url instead of process.cwd() for more robust path resolution, and pointing out a potential cross-platform failure on Windows due to hardcoded \n line endings in the sliceConfig helper function.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Using process.cwd() to resolve the path to playwright.config.ts makes the test fragile if executed from a different directory (e.g., the repository root in a monorepo setup). It is more robust to resolve the path relative to the test file's directory using import.meta.url.

Suggested change

	import { readFileSync } from 'node:fs';
	import { join } from 'node:path';
	import { describe, expect, it } from 'vitest';
	const CONFIG_PATH = join(process.cwd(), 'playwright.config.ts');
	import { readFileSync } from 'node:fs';
	import { dirname, resolve } from 'node:path';
	import { fileURLToPath } from 'node:url';
	import { describe, expect, it } from 'vitest';
	const __dirname = dirname(fileURLToPath(import.meta.url));
	const CONFIG_PATH = resolve(__dirname, '../../playwright.config.ts');

[gemini-code-assist]

The sliceConfig function is sensitive to line endings because it searches for ]\n : [] and expects \n. On Windows environments where files are checked out with \r\n (CRLF), this test will fail. Normalizing the line endings of the source string to \n before processing will make the test cross-platform robust.

function sliceConfig(source: string): {
  beforeCi: string;
  ciBranch: string;
  afterCi: string;
} {
  const normalizedSource = source.replace(/\r\n/g, '\n');
  const ciIdx = normalizedSource.indexOf('process.env.CI');
  if (ciIdx === -1) {
    throw new Error(
      'Could not locate `process.env.CI` in playwright.config.ts — testIgnore CI-ternary shape changed?',
    );
  }
  // Find the `? [` after process.env.CI.
  const openBracketIdx = normalizedSource.indexOf('? [', ciIdx);
  if (openBracketIdx === -1) {
    throw new Error(
      'Could not locate `? [` after `process.env.CI` — testIgnore CI-ternary shape changed?',
    );
  }
  // Find the matching `] : []` close that ends the ternary.
  const closeIdx = normalizedSource.indexOf(']\n      : []', openBracketIdx);
  if (closeIdx === -1) {
    throw new Error(
      'Could not locate `] : []` closing the CI ternary — testIgnore CI-ternary shape changed?',
    );
  }
  return {
    beforeCi: normalizedSource.slice(0, ciIdx),
    ciBranch: normalizedSource.slice(openBracketIdx, closeIdx),
    afterCi: normalizedSource.slice(closeIdx),
  };
}

[SoundMindsAI]

Review adjudication (Gemini Code Assist + GPT-5.5 final review)

Commits landing fixes: 17b34619 (Gemini), fd70f6df (GPT-5.5).

Gemini Code Assist (2 findings — both accepted)

#	Sev	Location	Verdict	Notes
1	Medium	ui/src/__tests__/playwright-config-test-ignore.test.ts:37	Accepted	Swapped process.cwd() → import.meta.url for path resolution. Plan D-7 approved both shapes; Gemini's robustness point is real for ad-hoc operator runs (pnpm vitest run <path> from repo root, where cwd ≠ ui/). Fixed in 17b34619.
2	Medium	ui/src/__tests__/playwright-config-test-ignore.test.ts:103	Accepted	Added CRLF normalization in sliceConfig() before the \n-anchored indexOf searches. Zero-cost cross-platform defense. Fixed in 17b34619.

GPT-5.5 final review (3 findings — 2 accepted, 1 rejected)

#	Sev	Location	Verdict	Notes
1	High	branch diff scope	Rejected (re-raise without new evidence)	Counter-evidence: PR body's "Diff scope (vs AC-7's strict '5 files')" section already documents this. Every recent merged PR (#383, #416, #421, #422) ships the pipeline-trail (idea/spec/plan/pipeline_status) per project convention; the 4 dashboard regen files are emitted by the mvp1-dashboard-regen pre-commit hook (forbidden to skip per CLAUDE.md Rule #7). AC-7's strict literal "5 files" predates the project-convention consideration of pipeline-trail co-shipping; the spec's intent (the 5 deliverables described in FR-1..FR-5) is satisfied byte-identically.
2	Medium	ui/playwright.config.ts:73	Accepted	Real broken pointer: comment said "See ... §4 for the lever cascade context" but runbook §4 is "Why each lever is GHA-only" — the lever cascade is §3 and my new section is §5. Updated to point at §5 (which itself carries the reseed-runtime-vs-Solr-stability relationship table). Fixed in fd70f6df.
3	Low	docs/03_runbooks/smoke-solr-stability.md §5	Accepted	FR-3 required actual "cross-link"s, not inline-code mentions. Converted three mentions to relative-path markdown links (verified all 3 targets resolve). Fixed in fd70f6df.

Outcomes

• Applied fixes (4): import.meta.url path resolution + CRLF normalization (Gemini docs: reorganize into numbered IA, add pipeline skills and feature templates #1+docs: MVP1 architecture + 12 feature specs + GPT-5.5 review fixes #2) + §4 → §5 pointer fix + markdown links in runbook §5 (GPT-5.5 docs: MVP1 architecture + 12 feature specs + GPT-5.5 review fixes #2+chore(pipeline): add project-wide priority-ordered status mode #3).
• Rejected with cited counter-evidence (1): AC-7 file-shape contract — addressed in PR body before review; project convention + pre-commit hook output drives the 9-file shape.
• Deferred as non-regression follow-ups (0): none.

CI is re-running on fd70f6df. Ready for human review + merge once green.