Pin nwsapi to 2.2.24 to fix jsdom selector recursion

[ravikiranvm]

Fixes OPS-4838.

Pins nwsapi (jsdom's CSS selector engine, transitive via jest-environment-jsdom → jsdom@26) to 2.2.24 through the root overrides block. Test-only dependency; no runtime code changes.

Additional Notes

Why

• The Jest 29 → 30 upgrade (Upgrade Jest from 29 to 30 #2491) regenerated the lockfile and re-resolved nwsapi from 2.2.24 to 2.2.27.
• nwsapi 2.2.25–2.2.27 implement :modal / :popover-open / :fullscreen by delegating to Element.matches. Under jsdom there is no native matcher, so the call re-enters nwsapi and recurses to a stack overflow that a try/catch swallows. One matches(':modal') costs ~150 ms.
• floating-ui's isTopLayer() calls matches(':modal') while positioning every Radix popper. Any test that opens a DropdownMenu, Popover, Select or Tooltip now takes 7–10 s and hits the 5 s Jest timeout.
• OSS is green only because no current test opens a popper. The internal repo's Test UI job has been red on main since 2026-09-10 for exactly this reason. Landing the pin here fixes both via the sync.

Upstream

• Regression in 2.2.25+: portal-rendered elements cause @testing-library/user-event to hang under JSDOM dperini/nwsapi#214 — maintainers confirmed the reentry bug and fixed it on master (715f652), not yet published to npm (latest is 2.2.27, master is 2.3.0-prerelease). Remove this override once a release containing that commit ships.
• Independently hit during the same Jest 30 upgrade: Pin nwsapi to 2.2.25 to avoid selector recursion in jsdom Automattic/vip-design-system#721.

Alternatives considered

• Raise testTimeout on affected tests: tests still take 7–10 s each, suites run for minutes, every future popper test pays the same cost. Rejected.
• Per-test userEvent.setup() instances: measured 8.5–10.2 s for the same click across four user-event configurations. The time is in floating-ui, not user-event. Rejected.
• Shim Element.prototype.matches in Jest setup to return false for the display-state pseudo-classes: works (59 ms), but monkey-patches a DOM prototype in every jsdom test and hides real :open/:closed state. The pin is smaller and sits at the root cause.

Safety of the rollback

• 2.2.24 is what every suite ran on until 2026-09-10. Both versions are MIT; THIRD_PARTY_LICENSES.txt is unaffected since the generator only covers production deps.
• What 2.2.25–2.2.27 added and we lose: the display-state pseudo-classes (the broken feature), :autofill aliases, a :has() sibling fix and :focus-visible/:focus-within fixes. Nothing in either repo passes these to querySelector/matches; the only hits are Tailwind class strings jsdom never evaluates.

Testing Checklist

• I tested the feature thoroughly, including edge cases
• I verified all affected areas still work as expected
• Automated tests were added/updated if necessary — none needed; existing Radix tests are the regression check
• Changes are backwards compatible with any existing data, otherwise a migration script is provided

What was run

• OSS: npm install resolves jest-environment-jsdom/node_modules/nwsapi@2.2.24; bare jsdom 26 matches(':modal') now throws unknown pseudo-class in 3 ms instead of taking 151 ms. nx test ui-components: 203 tests pass.
• Internal repo with 2.2.24 swapped in: campaign-card.test.tsx 85 s → 1.8 s; nx test ui-components 311 pass (14.6 s); nx test react-ui 1393 pass (20.1 s).

🤖 Generated with Claude Code

[linear]

OPS-4838

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud