fix: guard GoogleOneTapContainer against SSR window access

[brainbicycle]

The type of this PR is: Fix

Description

Removing the feature flag dropped the short-circuit that previously kept the window.location.pathname check from running during SSR, causing a "window is not defined" 500. Guard the enabled check with a typeof window test so it resolves to false on the server.

Assisted-By: Claude noreply@anthropic.com

[relativeci]

#9568 Bundle Size — 9.02MiB (~+0.01%).

9cd1d17(current) vs 914f877 main#9565(baseline)

Warning

Bundle contains 29 duplicate packages – View duplicate packages

Bundle metrics   1 change 1 regression

	Current #9568	Baseline #9565
Initial JS	2.94MiB(~+0.01%)	2.94MiB
Initial CSS	0B	0B
Cache Invalidation	97.48%	97.48%
Chunks	111	111
Assets	114	114
Modules	5704	5704
Duplicate Modules	691	691
Duplicate Code	5.05%	5.05%
Packages	248	248
Duplicate Packages	28	28

Bundle size by type   1 change 1 regression

	Current #9568	Baseline #9565
JS	8.8MiB (~+0.01%)	8.8MiB
Other	232.77KiB	232.77KiB

Bundle analysis report Branch brian/one-tap-cleanup-follow-up Project dashboard

---

^{Generated by RelativeCI Documentation Report issue}

[claude]

Summary

Adds a typeof window !== "undefined" short-circuit to the enabled computation in GoogleOneTapContainer, so the window.location.pathname check no longer runs during SSR. Fixes the "window is not defined" 500 introduced when the feature flag was removed.

Issues Found

None. The guard is placed first in the && chain, so window.location.pathname on line 43 is never reached on the server. The remaining window accesses (lines 47, 57–59, 65) all live inside useEffect, which never runs during SSR, so they need no guard.

Areas Reviewed

• SSR safety: Correct. enabled resolves to false on the server, the component returns null, and no browser global is touched at render time.
• Testing: src/Utils/__tests__/GoogleOneTapContainer.jest.tsx already covers the enabled/disabled paths and auth-path behavior. The SSR guard isn't directly exercised, but jsdom always defines window, so a unit test can't meaningfully assert it — CI/runtime is the real check here.

Looks good to merge.

[anandaroop]

Thanks!