RDFPFN792026: trace hydration branches through rendered values

[aidenybai]

Why

ReactBench exposed browser-global hydration branches that the rule missed when the predicate flowed through mutable aliases, local helpers, memoized objects, hook state, or compiled createElement output.

What

• trace browser-global predicates through render-phase helper and state flows
• recognize rendered local factory results without treating lookalike factories as React provenance
• require a same-file rendered consumer before reporting a browser-derived custom-hook initializer
• keep mount-gated probes and equivalent initial output quiet
• add ReactBench regression and fuzz coverage

ReactBench evidence

• live-doc contract: 42 source-provable units
• exact replay: 42/42 caught
• excluded: six initially-false mount gates and one cross-file-only hook consumer (ivEkXud)
• parse errors: 0
• reconstructed-source mismatches: 0

The ivEkXud hook initializer is intentionally not reported from its standalone source file. Its rendered consumer appears only in a separate patched file, so this file-local rule cannot prove the hydration mismatch.

Test plan

• nr test in packages/oxlint-plugin-react-doctor: 958 files, 24,819 passed, 201 skipped
• FUZZ_RULE=no-hydration-branch-on-browser-global FUZZ_ITERATIONS=500 FUZZ_SEED=792026 nr fuzz
• nr typecheck in the plugin and fuzz packages
• nr lint
• nr format:check
• nr smoke:json-report

The root test/typecheck orchestration was also attempted. Unrelated workers exhausted the shared runner after the affected packages had passed; the isolated affected-package gates are green.

Local RDE could not start because AXIOM_DATASET is not configured.

---

Note

Medium Risk
Large, intricate changes to a correctness lint rule’s static analysis; behavior shifts on many edge cases, though scope is limited to the plugin and tests.

Overview
Extends no-hydration-branch-on-browser-global so browser-global predicates are tracked through render-time helpers, mutable aliases, useState/useMemo flows, custom-hook return paths, and compiled createElement output—not only direct JSX conditionals.

Predicate resolution now follows let/var writes guarded by browser checks (including nested helper invocations), useMemo object properties, lazy useState initializers, and equality/bitwise forms that still differ by runtime; reflexive or duplicated-predicate comparisons are suppressed.

Reporting gates require a same-file rendered consumer for hook-returned browser state, treat locally invoked render helpers as in-output, and distinguish real React createElement from lookalike factories via explicit runtime require/import provenance. Equivalence checks use binding-aware structural comparison so shadowed branch-local JSX and mount-gated or suppressHydrationWarning consumers stay quiet when output is effectively stable.

Adds a large ReactBench regression suite and fuzz corpus cases for true positives and false-positive retractions.

^{Reviewed by Cursor Bugbot for commit 9cc9aeb. Bugbot is set up for automated code reviews on this repo. Configure here.}

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/eslint-plugin-react-doctor@1488

npm i https://pkg.pr.new/oxlint-plugin-react-doctor@1488

npm i https://pkg.pr.new/react-doctor@1488

commit: 9cc9aeb

[github-actions]

Interactive terminal E2E

Recorded from the built CLI at 9cc9aeb in a real terminal. The fixture holds Git busy for three seconds, so Scanning... must appear immediately after project selection, then exercises the compact interactive report.

Download the GIF and MP4 artifact

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit f5a3f62. Configure here.}