OCPBUGS-105789: Recover expired Playwright sessions via auto re-login - #16954
OCPBUGS-105789: Recover expired Playwright sessions via auto re-login#16954rhamilto wants to merge 7 commits into
Conversation
Addresses multiple sources of flakiness in the web terminal Playwright e2e tests: - warmupSPA now detects the login page and re-authenticates instead of retrying until timeout when the auth session has expired - waitForTerminalIconVisible replaced manual retry loop with toPass() to prevent "page closed" errors when the test timeout fires mid-reload - waitForTerminalWindow default timeout increased from 60s to 120s for slow DevWorkspace provisioning - navigateToWebTerminalConfig uses longer timeouts for the actions menu and Web Terminal tab which load dynamically - All web terminal tests marked test.slow() to triple the default timeout - Extracted getBaseURL, getAdminCredentials, getDeveloperCredentials into login-helper.ts to eliminate duplication across auth setup files and warmupSPA Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The Playwright e2e suite authenticates once in the setup projects and freezes the session into a storageState snapshot that every test reuses. Nothing re-authenticates, so when the OpenShift OAuth token expires during a long run, navigations silently redirect to the login page and tests hang waiting for elements that never appear. Add a self-healing auth fixture that overrides the built-in page fixture: a main-frame navigation listener detects when a navigation lands on the login page and transparently re-runs performLogin for the active persona (resolved from the project storageState path), then refreshes the stored session so later tests reuse the fresh state. Re-entrancy is guarded so the login flow's own navigations don't recurse. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Pipeline controller notification For optional jobs, comment This repository is configured in: LGTM mode |
openshift-ci-robot
added
jira/valid-reference
|
@rhamilto: This pull request references Jira Issue OCPBUGS-105789, which is valid. The bug has been moved to the POST state. 3 validation(s) were run on this bug
The bug has been updated to refer to the pull request using the external bug tracker. DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
|
/hold for #16953 to merge first |
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
WalkthroughThe PR centralizes Playwright authentication, adds session recovery after login redirects, integrates recovery into page navigation, increases Web Terminal wait and retry limits, and marks affected tests as slow. ChangesE2E reliability
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Test as Playwright test
participant Fixture as page fixture
participant Recovery as session recovery
participant Login as login helper
participant Page as Playwright Page
Test->>Fixture: navigate with page.goto
Fixture->>Recovery: record URL and check expiration
Recovery->>Page: detect login redirect
Recovery->>Login: resolve credentials
Login-->>Recovery: return credentials
Recovery->>Page: performLogin and save storage state
Recovery->>Page: restore intended URL
Fixture-->>Test: return navigation response
Suggested labels: Suggested reviewers: 🚥 Pre-merge checks | ✅ 14 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (14 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
openshift-ci
Bot
added
the
do-not-merge/hold
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: rhamilto The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
openshift-ci
Bot
added
the
approved
|
@rhamilto: This pull request references Jira Issue OCPBUGS-105789, which is valid. 3 validation(s) were run on this bug
DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@frontend/e2e/fixtures/auth-fixture.ts`:
- Around line 81-85: Update attachSessionRecovery and the navigation flow used
by performLogin so session recovery is awaited rather than started
fire-and-forget. Wrap navigation to preserve the original target, recover an
expired session, and retry that target before allowing subsequent test actions
to continue; add a regression test covering an expired-session redirect.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: c0fb6ea2-b79d-45c4-bf5a-1aff73050158
📒 Files selected for processing (11)
frontend/e2e/fixtures/auth-fixture.tsfrontend/e2e/fixtures/index.tsfrontend/e2e/pages/web-terminal-config-page.tsfrontend/e2e/pages/web-terminal-page.tsfrontend/e2e/setup/admin-auth.setup.tsfrontend/e2e/setup/developer-auth.setup.tsfrontend/e2e/setup/login-helper.tsfrontend/e2e/tests/webterminal/developer/web-terminal-basic.spec.tsfrontend/e2e/tests/webterminal/developer/web-terminal-devuser.spec.tsfrontend/e2e/tests/webterminal/web-terminal-admin.spec.tsfrontend/e2e/tests/webterminal/web-terminal-config.spec.ts
- Restore the intended route after re-login so deep-link tests resume where they were headed instead of on the console home page (CodeRabbit). - Detect the login page with a bounded wait on auth-URL navigations so a still-rendering OAuth login form isn't missed, while keeping normal navigations on an instantaneous check to avoid per-navigation latency (CodeRabbit + timing race). - Claim the re-login slot synchronously to close a TOCTOU window where two concurrent navigation events could each launch a login. - Hoist the Frame type import and clarify the proactive-check comment. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
/pipeline required |
|
Scheduling tests matching the |
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@frontend/e2e/fixtures/auth-fixture.ts`:
- Around line 31-32: Update isAuthUrl to parse the URL and evaluate only its
normalized pathname, avoiding matches from query strings or unrelated URL text.
Anchor the authentication-path regex to the complete pathname shape with ^ and
$, while preserving recognition of OAuth, OAuth2, login, and auth routes.
- Around line 60-68: Update the authentication recovery flow around
saveStorageState to serialize storage-state writes by statePath. Use a shared
lock keyed by each statePath so concurrent workers targeting kubeadmin.json or
developer.json cannot write simultaneously, and ensure the lock is released on
every success and failure path; preserve the existing re-login coordination
behavior.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: b870fd4e-cb46-41f2-92bd-70e5cc86b2e9
📒 Files selected for processing (3)
frontend/e2e/fixtures/auth-fixture.tsfrontend/e2e/fixtures/index.tsfrontend/e2e/setup/login-helper.ts
🚧 Files skipped from review as they are similar to previous changes (2)
- frontend/e2e/fixtures/index.ts
- frontend/e2e/setup/login-helper.ts
| function isAuthUrl(url: string): boolean { | ||
| return /\/oauth\/|\/oauth2\/|\/login(\/|$|\?)|\/auth\//.test(url); |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Anchor authentication URL matching to the pathname.
The current expression matches /oauth/ and /auth/ anywhere in the full URL. A normal application route or query value can then be classified as an auth URL. This prevents lastAppUrl updates and can start unnecessary recovery.
Parse the URL, normalize its pathname, and match the complete auth-path shape.
Proposed fix
function isAuthUrl(url: string): boolean {
- return /\/oauth\/|\/oauth2\/|\/login(\/|$|\?)|\/auth\//.test(url);
+ const pathname = new URL(url).pathname.normalize('NFC');
+ return /^\/(?:oauth2?|login|auth)(?:\/.*)?$/.test(pathname);
}As per path instructions, “Normalize Unicode and anchor regexes (^$); watch for ReDoS.”
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| function isAuthUrl(url: string): boolean { | |
| return /\/oauth\/|\/oauth2\/|\/login(\/|$|\?)|\/auth\//.test(url); | |
| function isAuthUrl(url: string): boolean { | |
| const pathname = new URL(url).pathname.normalize('NFC'); | |
| return /^\/(?:oauth2?|login|auth)(?:\/.*)?$/.test(pathname); | |
| } |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@frontend/e2e/fixtures/auth-fixture.ts` around lines 31 - 32, Update isAuthUrl
to parse the URL and evaluate only its normalized pathname, avoiding matches
from query strings or unrelated URL text. Anchor the authentication-path regex
to the complete pathname shape with ^ and $, while preserving recognition of
OAuth, OAuth2, login, and auth routes.
Source: Path instructions
| // Claim the re-login slot synchronously — before any await — so concurrent | ||
| // navigation events can't both pass the check above and launch duplicate | ||
| // logins. The deferred is published now and settled once we know whether a | ||
| // recovery is actually needed; if not, we release the slot immediately. | ||
| let release!: () => void; | ||
| const claim = new Promise<void>((resolve) => { | ||
| release = resolve; | ||
| }); | ||
| reloginInProgress.set(page, claim); |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- storage-state configuration ---'
fd -i 'playwright.config.*|package.json' frontend -x sh -c \
'printf "\n--- %s ---\n" "$1"; rg -n -C 3 "storageState|workers|fullyParallel" "$1" || true' sh {}
printf '%s\n' '--- storage-state persistence implementation ---'
rg -n -C 5 --glob '*.ts' \
'function saveStorageState|const saveStorageState|saveStorageState\(|storageState\(\{.*path' \
frontend/e2eRepository: openshift/console
Length of output: 6073
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- playwright projects and storage-state paths ---'
sed -n '1,190p' frontend/playwright.config.ts
printf '%s\n' '--- auth fixture and fixture scopes ---'
sed -n '1,150p' frontend/e2e/fixtures/auth-fixture.ts
sed -n '1,180p' frontend/e2e/fixtures/index.ts
printf '%s\n' '--- storage-state path definitions and save helper ---'
rg -n -C 8 --glob '*.ts' \
'adminStorageState|developerStorageState|STORAGE_STATE_DIR|storageState\(\{ path|saveStorageState' frontend/e2e frontend/playwright.config.tsRepository: openshift/console
Length of output: 22794
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- Playwright version ---'
rg -n -C 2 '"`@playwright/test`"|"playwright"' frontend/package.json package.json frontend/yarn.lock yarn.lock pnpm-lock.yaml package-lock.json 2>/dev/null || true
printf '%s\n' '--- repository-level synchronization around auth-state writes ---'
rg -n -C 4 --glob '!node_modules/**' \
'kubeadmin\.json|developer\.json|storageState\(\{ *path|saveStorageState|lockfile|proper-lockfile|mkdir.*lock|writeFile.*storage' .Repository: openshift/console
Length of output: 16321
🌐 Web query:
Playwright storageState({ path }) implementation atomic file write behavior
💡 Result:
The Playwright storageState({ path }) method does not implement atomic file write behavior [1][2]. When this method is called, Playwright serializes the browser context's state (cookies, local storage, and optionally IndexedDB) and writes it directly to the specified path using standard file system operations [1]. Internally, Playwright uses Node.js's fs.promises.writeFile to save the state as a JSON file [1][3][4]. This process creates or overwrites the file directly [1][2]. Because it does not use an atomic pattern—such as writing to a temporary file and then renaming it—there is a possibility that a process reading the file simultaneously could encounter a partially written or corrupted state if the write operation is interrupted [1][2]. Additionally, the method does not provide built-in protections against race conditions or concurrent writes to the same path [2]. If multiple test workers or processes attempt to call storageState with the same output path simultaneously, their writes may interleave or cause conflicts [2]. To mitigate these risks in environments where concurrency is a concern, it is recommended to: 1. Use unique paths for different workers or test runs (e.g., by utilizing test.info().outputPath()) [2]. 2. Avoid sharing the same state file across concurrent processes that might write to it [2]. 3. Treat the file as an immutable artifact once it has been generated by a setup process, ensuring that subsequent test workers only perform read operations [2].
Citations:
- 1: https://leeroopedia.com/index.php/Implementation:Microsoft_Playwright_StorageState
- 2: https://mergify.com/blog/playwright-storagestate-shared-auth-leakage
- 3: https://github.com/microsoft/playwright/blob/5790370e/packages/utils/serializedFS.ts
- 4: https://github.com/microsoft/playwright/blob/c0cc9802/packages/playwright-core/src/server/utils/fileUtils.ts
Serialize storage-state writes by statePath.
WORKERS can run multiple workers in one project, but all admin workers use kubeadmin.json and all developer workers use developer.json. saveStorageState writes directly without a lock or atomic replacement. Concurrent recovery can corrupt or overwrite the shared state. Use a lock keyed by statePath or assign each worker a separate state file.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@frontend/e2e/fixtures/auth-fixture.ts` around lines 60 - 68, Update the
authentication recovery flow around saveStorageState to serialize storage-state
writes by statePath. Use a shared lock keyed by each statePath so concurrent
workers targeting kubeadmin.json or developer.json cannot write simultaneously,
and ensure the lock is released on every success and failure path; preserve the
existing re-login coordination behavior.
The prior review pass introduced a bounded-wait path in isOnLoginPage that
called locator.waitFor({ timeout: timeoutMs }). Playwright treats timeout: 0
as "wait forever" rather than "check instantaneously", so the hot-path
callers that pass 0 would have blocked every navigation indefinitely,
deadlocking the whole suite.
Guard the zero case with a non-waiting isVisible() check and only fall back
to waitFor when a bounded window is explicitly requested.
Add e2e/tests/console/app/session-recovery.spec.ts, which pins the login-page
detection primitive the recovery flow depends on. It renders markup via
data: URLs so no cluster is required, and it caught the timeout:0 deadlock.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
/pipeline required |
|
Scheduling tests matching the |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@frontend/e2e/fixtures/auth-fixture.ts`:
- Around line 108-112: Update the navigation wrapper around
recoverSessionIfExpired so recovery-initiated navigations are marked before
invoking the original page.goto, allowing the wrapper to bypass recovery for
those calls and avoid waiting on its own claim. Apply the same marker/skip
behavior to navigations within performLogin, and add a regression test covering
route restoration through the wrapped page fixture when intendedUrl is set.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 6efc1919-1349-4967-96bf-e12c4a4f23d1
📒 Files selected for processing (4)
frontend/e2e/fixtures/auth-fixture.tsfrontend/e2e/fixtures/index.tsfrontend/e2e/setup/login-helper.tsfrontend/e2e/tests/console/app/session-recovery.spec.ts
🚧 Files skipped from review as they are similar to previous changes (2)
- frontend/e2e/fixtures/index.ts
- frontend/e2e/setup/login-helper.ts
Address two issues from the pre-push review: - The page.goto override drove recovery, but recovery itself navigates through that same overridden goto (performLogin's initial goto and the route-restoration goto). Re-entering recovery there would await the very claim promise it is nested inside, deadlocking the override. Add an isRecoveryInProgress guard so recovery-owned navigations pass straight through to the original goto. - The shared recovery claim resolved even when performLogin threw, so a caller that joined an in-flight recovery (the reloginInProgress branch) saw a false success. Reject the claim on failure so every awaiter observes the same outcome, with a no-op catch on the stored copy to avoid an unhandled rejection when nobody joins. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
/pipeline required |
|
Scheduling tests matching the |
openshift-ci
Bot
added
the
tide/merge-method-squash
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@frontend/e2e/fixtures/auth-fixture.ts`:
- Around line 75-86: Update the reloginInProgress map and shared claim in
recoverSessionIfExpired to use Promise<boolean>, return the existing claim
directly to joiners, resolve false when the login page is absent or credentials
are unavailable, and resolve true only after successful recovery.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 95f85d2f-bd3e-4e0e-8124-8ca14e446085
📒 Files selected for processing (2)
frontend/e2e/fixtures/auth-fixture.tsfrontend/e2e/fixtures/index.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- frontend/e2e/fixtures/index.ts
| // The claim resolves on success and rejects on failure so joiners (the | ||
| // `existing` branch above) observe the same outcome instead of a false success. | ||
| let release!: () => void; | ||
| let fail!: (error: unknown) => void; | ||
| const claim = new Promise<void>((resolve, reject) => { | ||
| release = resolve; | ||
| fail = reject; | ||
| }); | ||
| // A rejected claim that nobody awaits is an unhandled rejection; attach a | ||
| // no-op catch to the stored copy so only explicit awaiters see the error. | ||
| claim.catch(() => {}); | ||
| reloginInProgress.set(page, claim); |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Propagate the actual recovery result to joiners.
The shared claim is Promise<void>, but recoverSessionIfExpired returns Promise<boolean>. A caller that joins at Lines 64-67 always receives true. The owner resolves the claim without recovery when the login page is not detected or credentials are unavailable.
Use Promise<boolean> for the claim. Resolve false on both no-op paths and true after successful recovery. Return the existing claim to joiners.
Proposed fix
- let release!: () => void;
+ let release!: (recovered: boolean) => void;
...
- const claim = new Promise<void>((resolve, reject) => {
+ const claim = new Promise<boolean>((resolve, reject) => {
...
if (existing) {
- await existing;
- return true;
+ return existing;
}
...
- release();
+ release(false);
...
- release();
+ release(false);
...
- release();
+ release(true);Also update the reloginInProgress map value type to Promise<boolean>.
Also applies to: 117-118
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@frontend/e2e/fixtures/auth-fixture.ts` around lines 75 - 86, Update the
reloginInProgress map and shared claim in recoverSessionIfExpired to use
Promise<boolean>, return the existing claim directly to joiners, resolve false
when the login page is absent or credentials are unavailable, and resolve true
only after successful recovery.
…links Address the second-round pre-push review: - awaitSessionRecovery no longer swallows recovery rejections. A re-login failure now propagates through the page.goto override so the active test fails with the original error instead of silently proceeding on the login page. - Record the intended destination in the page.goto override before navigating. When a deep link immediately redirects to the login page, the framenavigated handler may only ever observe the auth URL, so relying on it alone would restore a stale route; capturing the target up front makes recovery return to where the test was headed. Add a regression test covering deep-link recording: an app URL is remembered, and a subsequent auth-URL or about:blank navigation does not overwrite it. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
saveStorageState wrote directly to the destination file. Playwright workers are separate processes, so two workers recovering the same persona could interleave writes to the same storageState file, leaving another worker loading that file to observe half-written, invalid JSON. Write to a per-process temp file and rename it into place. A same-directory rename is atomic, so a concurrent reader always sees a complete file. Clean up the temp file if the write fails. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
/pipeline required |
|
Scheduling tests matching the |
|
@rhamilto: The following tests failed, say
Full PR test history. Your PR dashboard. DetailsInstructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
|
This may be overkill in light of #16911 |
2 participants
Analysis / Root cause:
The Playwright e2e suite authenticates exactly once, in the setup projects (
admin-auth.setup.ts,developer-auth.setup.ts), and freezes the session into astorageStatesnapshot. Every test project loads that static snapshot (playwright.config.ts) and nothing ever re-authenticates —performLoginis only referenced by the two setup files.When the OpenShift OAuth token expires during a long run, subsequent
page.goto()calls silently redirect to the login page, and tests hang waiting for elements (e.g.user-dropdown-toggle,#page-sidebar) that never appear, failing with generictoBeVisible/test-timeout errors. This showed up as widespread login-page hangs in CI.Jira: https://redhat.atlassian.net/browse/OCPBUGS-105789
Solution description:
Add a self-healing auth fixture that overrides Playwright's built-in
pagefixture:e2e/fixtures/auth-fixture.ts(new) — attaches a main-frameframenavigatedlistener that detects when a navigation lands on the login page and transparently re-runsperformLoginfor the active persona, then re-savesstorageStateso later tests reuse the fresh session. After re-login it restores the route the test was navigating to, so deep-link tests resume where they intended rather than on the console home page. Concurrency is handled by claiming the re-login slot synchronously (closing a TOCTOU window where two navigation events could each start a login), and recovery-owned navigations bypass thepage.gotooverride so they don't deadlock on their own in-progress claim. A failed re-login rejects the shared claim so any joining caller observes the failure instead of a false success. It logs a warning when recovery fires so future occurrences are visible in CI logs.e2e/setup/login-helper.ts— addsisOnLoginPage()andresolveCredentialsForStorageState()(dispatches to the existinggetAdminCredentials()/getDeveloperCredentials()helpers based on the project's storage-state filename).isOnLoginPage()supports an instantaneous check on the hot path and a bounded wait for still-rendering OAuth redirects; it uses a non-waitingisVisible()for the zero-timeout case because Playwright treatswaitFor({ timeout: 0 })as "wait forever".e2e/fixtures/index.ts— wires recovery into the sharedpagefixture and wrapspage.gotoso callers await any recovery the navigation itself triggered. Since every spec importstestfrom here, coverage is suite-wide with no per-test changes.e2e/tests/console/app/session-recovery.spec.ts(new) — regression test pinning the login-page detection primitive the recovery flow depends on. It renders markup viadata:URLs so no cluster is required.Auth-disabled clusters and unset developer credentials cause recovery to no-op safely.
Screenshots / screen recording:
Test setup:
Requires a cluster whose OAuth access-token lifetime is shorter than the full e2e run so the session expires mid-run; the recovery path then re-authenticates transparently. The
session-recovery.spec.tsdetection tests run without a cluster.Test cases:
session-recovery.spec.ts: login-page detection bydata-test-id, by username/password inputs, non-detection of an authenticated app page, bounded wait for a slow-rendering login form, andawaitSessionRecoveryno-op when idle.Browser conformance:
Additional info:
Test-infrastructure only — no product/runtime code changes.
Reviewers and assignees:
🤖 Generated with Claude Code
Summary by CodeRabbit
Bug Fixes
Tests