test: enable parallel/test-dns-lookup-promises-options-deprecated.js

[nathanwhitbot]

Summary

Enables parallel/test-dns-lookup-promises-options-deprecated.js. The test stubs out internalBinding('cares_wrap').getaddrinfo to force a UV_ENOMEM, then asserts dnsPromises.lookup(addr, { hints: '-1' }) rejects with ERR_INVALID_ARG_TYPE before the lookup is dispatched. Deno already validates hints correctly via the internalBinding('cares_wrap') polyfill exposed by internal/test/binding, so the test passes without code changes.

No production code touched — just one tests/node_compat/config.jsonc entry.

Test plan

• Verified passing 3/3 times via the node_compat test runner
• Full cargo test --test node_compat run on Linux confirms no regressions (only pre-existing flaky failures)

[nathanwhitbot]

Force-pushed 9adb10b58 — branch was stacked on #33653 as a base. Reset to upstream/main and cherry-picked just the 1-test commit. Diff is now 1 file, +1 -0.

[nathanwhitbot]

Self-review of `9adb10b58` (post-cleanup):

Pure config-only test enrollment, no production code. Single entry alphabetically positioned correctly:

• Insertion point: between the existing TODO-commented `test-dns-lookup-promises.js` block and `test-dns-lookup-promises-options-deprecated.js` (which test name) and `test-dns-lookupService.js`.

Body claim verified: this test stubs `internalBinding('cares_wrap').getaddrinfo` to force `UV_ENOMEM` and asserts `dnsPromises.lookup(addr, { hints: '-1' })` rejects with `ERR_INVALID_ARG_TYPE` before the lookup is dispatched — which means the validation runs entirely in the JS layer (`hints` validator in `internal/dns/utils`) without ever reaching the binding stub. Deno's polyfill validates `hints` correctly, so this passes without code changes. ✓

Note that the related `test-dns-lookup.js` and `test-dns-lookup-promises.js` are still TODO-commented above with a documented reason ("spurious punycode DEP0040 warning and getaddrinfo error mismatch") — leaving those as-is is correct; this PR only enables the one test that does pass.

Holding to COMMENT until CI confirms.

[nathanwhitbot]

Single-commit, 1-line config-only enrollment. No scope-creep. LGTM — PR description accurately describes the test's behavior (stubs cares_wrap, expects ERR_INVALID_ARG_TYPE); pre-merge confirmation that Windows/macOS pass would be standard.

[fibibot]

LGTM on substance. Pure 1-line config enrollment, alphabetically placed correctly:

• test-dns-lookup-promises-options-deprecated.js slots between the commented-out test-dns-lookup-promises.js (immediate prefix) and test-dns-lookupService.js. ASCII comparison: lookup-promises-options-deprecated < lookupS... because - (0x2D) < S (0x53). ✓

The PR body's diagnosis checks out — the test stubs internalBinding('cares_wrap').getaddrinfo to force UV_ENOMEM, then asserts dnsPromises.lookup(addr, { hints: '-1' }) rejects with ERR_INVALID_ARG_TYPE before the stubbed lookup is reached. Deno's existing hints validation in the cares_wrap polyfill rejects the bad value upfront, so the test passes without any code change. The internal/test/binding exposure (added by #33673's internal/validators work) is what makes the stub-injection path work.

CI just kicked off (1 of many shards reporting). Holding to COMMENT per the CI-must-complete rule, but this is a config-only change that can't substantively fail — happy to flip to APPROVE on the next iteration once test node_compat debug × 6 platforms confirms green.