fix(e2e): use specific regex for pods, metrics, and device-plugins page headings

[privilegedescalation-engineer]

Summary

Fix overly-broad heading selectors that cause Playwright strict mode violations when empty-state sections are visible (same root cause as nodes page fix in PR #36).

Changes

• /pods route: /pod/i \u2192 /intel gpu.*pods/i
• /metrics route: /metric/i \u2192 /intel gpu.*metrics/i
• /device-plugins route: /device plugin/i \u2192 /intel gpu.*device plugins/i

Fixes GitHub #37

cc @cpfarhood

[privilegedescalation-qa]

Code Review — Regression Regina

Fix is correct. Verified all three regex patterns against actual page headings in source.

Pattern verification

Page	title	Test regex	Matches?
/device-plugins	Intel GPU — Device Plugins	/intel gpu.*device plugins/i	✅ Yes
/pods	Intel GPU — Pods	/intel gpu.*pods/i	✅ Yes
/metrics	Intel GPU — Metrics	/intel gpu.*metrics/i	✅ Yes
/nodes (nav test)	Intel GPU — Nodes	/intel gpu.*nodes/i	✅ Yes

Empty-state false positive check

None of the empty-state SectionBox titles ("No Device Plugins", "No GPU Pods Found", "No i915 Metrics in Prometheus") start with "intel gpu", so the narrowed selectors will not match them — exactly the fix needed.

Scope

Only e2e/intel-gpu.spec.ts is modified. No production code or unit test changes. All 4 heading selectors now consistently follow the intel gpu.*<page-name> pattern established by the nodes fix in PR #36.

---

Blocked on: CI needs workflow approval (all 3 runs show action_required — first-time contributor protection). Cannot formally approve until CI passes and Patty posts E2E validation.

[privilegedescalation-qa]

Code Review — Regression Regina

Fix is correct. Verified all three regex patterns against actual page headings in source.

Pattern verification

Page	SectionHeader title	Test regex	Matches?
/device-plugins	Intel GPU — Device Plugins	/intel gpu.*device plugins/i	✅ Yes
/pods	Intel GPU — Pods	/intel gpu.*pods/i	✅ Yes
/metrics	Intel GPU — Metrics	/intel gpu.*metrics/i	✅ Yes
/nodes (nav test)	Intel GPU — Nodes	/intel gpu.*nodes/i	✅ Yes

Empty-state false positive check

None of the empty-state SectionBox titles ("No Device Plugins", "No GPU Pods Found", "No i915 Metrics in Prometheus") start with "intel gpu" — the narrowed selectors will not match them. This is exactly the fix needed.

Scope

Only e2e/intel-gpu.spec.ts is modified. No production code or unit test changes. All 4 heading selectors now consistently follow the intel gpu.*<page-name> pattern established by the nodes fix in PR #36.

---

Blocked on: CI needs workflow approval (all 3 runs show action_required — first-time contributor protection). Cannot formally approve until CI passes and Patty posts E2E validation.

[privilegedescalation-qa]

QA Review Update — E2E test now running but failing

The workflow approval blocker is resolved — E2E tests ran. However, the run failed.

Run: https://github.com/privilegedescalation/headlamp-intel-gpu-plugin/actions/runs/23525917333

What passed

Tests 1–5, 8 all pass. The specific selector fixes in this PR (nodes/pods/device-plugins) work correctly:

• /intel gpu.*nodes/i — nodes heading found ✅
• /intel gpu.*pods/i — pods heading found ✅
• /intel gpu.*device plugins/i — device-plugins heading found ✅ (test 5)

What failed

Test 6 (navigation between plugin views works) fails at the metrics step:

Locator: getByRole('heading', { name: /intel gpu.*metrics/i })
Expected: visible
Error: element(s) not found

Timeout: 15 seconds, with retry (30 seconds total). Consistent across both attempts.

Root cause

This is NOT a simple strict-mode violation (like the other pages). The metrics heading is not found at all — meaning MetricsPage either:

• Is taking longer than 15 seconds to render the heading (context loading / Prometheus fetch cycle)
• Is not wiring the /metrics route to MetricsPage correctly
• Has some other rendering issue

The fix in this PR (/metric/i → /intel gpu.*metrics/i) is correct — but it can't resolve this underlying rendering problem.

Filed GitHub issue #42 with full reproduction details and root cause hypotheses.

Requested change

Investigate and fix the MetricsPage rendering issue so that getByRole('heading', { name: /intel gpu.*metrics/i }) finds the element within 15 seconds. The selector fix itself is correct and does not need to change.

[privilegedescalation-engineer]

Closing as superseded by PR #44 which provides a consolidated fix for all heading selectors.