[LukasMod]

Explanation of Change

Adds Oxlint as a second linter behind the existing npm run lint, non-blocking in CI. ESLint stays
the gate. Nothing about how you lint locally changes yet.

npm run lint                      # ESLint, unchanged, still the merge gate
npm run lint -- --linter=oxlint   # same pipeline, same seatbelt, Oxlint

Why

Oxlint lints the whole repo in about a tenth of ESLint's time, on half the runner. Failing runs
widen the gap: the ESLint job clears its cache and lints again from scratch to rule out stale
results, and Oxlint has no cache, so it never retries.

	ESLint	Oxlint
CI runner	16 vCPU	8 vCPU
CI step, clean, cold ESLint cache	494 s	53 s
CI step, clean, warm ESLint cache	51s - changed files only	53s - whole repo

With a warm cache the two cost about the same wall time, at half the runner for Oxlint. Cold or
failing, Oxlint is about 17 to 19 times cheaper in vCPU-seconds.

Side by side (2 weeks, Oct 6-20, 2026)

	ESLint	Oxlint
Runs (2 wk)	3541	7 (branch only)
Runner	16 vCPU	8 vCPU
Median step (warm)	63 s	53 s
Mean step (all)	121 s	55 s
Mean on failure	570 s (2nd cold pass)	n/a - no retry, no cache
p95	608 s	<=68 s (no tail)
vCPU-seconds / run (mean)	~1930	~440

How it fits

scripts/lint/index.ts runs a Linter -> Processor[] -> Formatter pipeline. Both linters implement
the same Linter port and share the seatbelt, the formatter and the exit codes. Each keeps its own
baseline: config/eslint/eslint.seatbelt.tsv and config/oxlint/oxlint.seatbelt.tsv. The Oxlint
job runs on every PR with continue-on-error. While it is a shadow job, the post-merge run on main
updates its baseline in both directions (SEATBELT_INCREASE=all): findings a merge adds are absorbed
and fixed ones are removed, and OSBotify commits the result as "Auto-update oxlint-seatbelt baseline".
A PR is then warned only about findings it adds. Runs started by OSBotify's own commit only tighten,
so the baseline settles instead of looping. Once Oxlint is the blocking linter, main goes back to
tighten-only, as ESLint's does.

Tool versions: oxlint 1.89.0, oxc-transform-react 0.155.0, oxlint-tsgolint 7.0.2003.

Workarounds

What each unusual part of the diff is for.

workaround	problem it solves
OxlintLinter runs N single-threaded shards plus one type-aware process	Oxlint runs all JS-plugin rules on one thread (oxc#26621), so extra cores only help as extra processes. Type-aware rules run once in oxlint-tsgolint, which builds a multi-GB TypeScript program. OxlintLinter derives one JSON config per leg from oxlint.config.mts at the start of a run and deletes both at the end. The sharded run reports the same findings as a single process
about 100 rules hosted as JS plugins under config/oxlint/plugins/	Oxlint has no native equivalent. The original ESLint rule source runs inside Oxlint's plugin runtime, so each rule behaves as it does under ESLint. Every hosted rule has a fixture, a replayed upstream test, or a probe in the POC PR's migration scripts (#98027)
eslintDirectives.mjs makes hosted rules honour their ESLint id in disable comments	Oxlint reserves the react, react-hooks, import and @typescript-eslint prefixes, so an existing eslint-disable-next-line react/jsx-no-bind would not suppress hosted/jsx-no-bind. It follows ESLint's placement rules too: eslint-disable/eslint-enable count only in block comments, and a directive on the same line as a report is compared by column
React Compiler rules go through a Rust bridge over oxc-transform-react, with reportDiagnostics: true	Oxlint's native react/* compiler rules drop every diagnostic in a component once one eslint-disable names react-hooks/exhaustive-deps or rules-of-hooks (oxc#26277). Hosting eslint-plugin-react-hooks as a JS plugin works but is slow: its JS compiler took 52 s of a 121 s whole-repo run, against 5.2 s for the Rust one. Without reportDiagnostics the binding returns only fatal diagnostics (oxc#26318), so non-blocking rules like set-state-in-effect looked clean. With it, set-state-in-effect reports the same locations as ESLint
the React Compiler gate (reactCompilerGate.mjs) wraps three hosted rules	jsx-no-constructed-context-values, no-inline-useOnyx-selector and the useCallback/useMemo advice from exhaustive-deps are false positives in files both React compilers memoize, and the fix they suggest is banned there. Oxlint has no processor, so the gate drops them inside the plugin, which also keeps them out of the editor. In the pipeline ReactCompilerFilter applies the same rule to either linter's output
the bridge sets eslintSuppressionRules: []	otherwise the compiler skips every function under an eslint-disable-next-line react-hooks/exhaustive-deps comment, and the repo has 230. Lint only: setting it in the build configs would change what ships
hosted/no-deprecated-self-closing-jsx stands in for the @deprecated check on <X />	tsgolint only treats an opening element as a call, so a self-closing tag never reaches its deprecation check (typescript-eslint works on ESTree, where both forms are JSXOpeningElement). The stand-in reports under the @typescript-eslint/no-deprecated id, so findings file onto the same per-symbol seatbelt rows, and it is enabled in the same scopes as the typed rule. Every MenuItem usage in the repo is self-closing. Delete the stand-in when tsgolint ships the fix; a fixture already fails on that signal
categories: {correctness: 'off'}	every rule is listed explicitly to mirror eslint.config.mjs, so Oxlint's default category cannot switch on rules ESLint does not run. An extends or a recommended category would also pull in the native react/* compiler rules, without a warning (oxc#26277)
shards take every Nth file, not a slice	the file list is path-sorted, so slices pile all the .tsx work into one shard
exec node_modules/.bin/oxlint, never npx oxlint	npx passes one string to sh -c, and a shard list exceeds Linux's per-argument cap
retry a dead leg once, serially	a leg can exit with no JSON or run out of memory. One retry costs less than failing the whole run
--fix runs a single process, no shards	a shard's fix shifts lines in a file the type-aware process then lints, so one finding came back twice at two columns. --fix is manual-only (not used by CI, hooks or lint-changed), so the slower run costs nothing day to day
--shards=auto|N, auto by default from cores and vm_stat on macOS	auto takes half the cores (measured fastest on 14 cores) and what memory allows. Node's free-memory reading counts free pages only and sizes the run down to one shard on an idle machine. CI passes --shards=4; 0 or anything invalid exits with an error
a codeless Oxlint diagnostic is a parse error unless it has no location	Oxlint reports parse errors and crashing JS plugins both without a rule code. A parse error is reported like ESLint does it (ruleID: null, Parsing error: ...) and the other findings survive; a crash, which has no location, fails the run
ruleNames.mjs maps Oxlint diagnostic codes back to ESLint rule ids	seatbelt rows and disable comments use ESLint ids: hosted(naming-convention) is @typescript-eslint/naming-convention, typescript(...) is @typescript-eslint/..., plus two renamed core rules. A bun test fails if the mapping drifts
import/resolver and import/extensions settings copied from ESLint	the hosted eslint-plugin-import rules read them; without them no-useless-path-segments and no-relative-packages cannot resolve .ts/.tsx targets and stay silent
a test checks every plugins array against Oxlint's schema	Oxlint ignores an unknown plugin name in an override without warning, and the rule that needed it reports nothing
the HeaderWithBackButton import ban has two group entries	ESLint matches with gitignore semantics, so a directory name covers everything below it. Oxlint matches the path exactly
no-unsafe-* and no-throw-literal off under scripts/**, .github/**	tsgolint is stricter than typescript-eslint on that family. Without the override it reports about 695 findings ESLint never did, all in loosely typed Actions glue
no-deprecated off on a list of 83 files	tsgolint flags deprecated write sites that typescript-eslint skips. We accept losing 5 read-site findings on those files
unicorn/prefer-at replaces rulesdir/prefer-at	partial cover: it catches x[x.length - N] and misses plain arr[0]. One of two rule downgrades we accepted. The other is boolean-conditional-rendering, which needs type information jsPlugins cannot see yet
package.json overrides pins vite-plus to ^0.1.15	npm 11 auto-resolves oxlint's optional vite-plus: "*" peer, and the latest vite-plus pulls a vitest@5 chain that requires @types/node >= 22 while the repo pins 20. Without the pin, installs and every future oxlint bump fail with ERESOLVE. Nothing imports vite-plus. TEMPORARY until @types/node reaches 22+
oxlint scopes in .npmrc's min-release-age-exclude	min-release-age=7 would block the current oxlint and oxc-transform-react releases. The native binding scopes are listed too, so npm does not keep a stale binding without telling anyone. TEMPORARY: both releases clear the hold on 2026-10-26 after 17:24 UTC, and a follow-up removes the four lines if this merges first

Parity, whole repo

Both linters run through the same pipeline: ESLint 3668 findings, Oxlint 5038.
After the latest merge from main the baselines hold:

• eslint: 3668 findings in 1966 rows
• oxlint: 5038 findings in 2597 rows

Read by direction, at the seatbelt's granularity (file, rule, count):

• 1912 (file, rule) pairs are identical, 3523 findings.
• Oxlint adds 1398 findings ESLint does not report. These only make the gate stricter.
• Oxlint hides 28 that ESLint reports:
  • 23 react-hooks/refs across 10 files. Most are duplicate collapsing: the compiler emits the same
    diagnostic several times at one location, and the bridge keeps fewer copies. The bridge misses 3
    locations (__mocks__/react-native-safe-area-context.tsx x2, AboutPage.tsx), and only AboutPage.tsx
    ends up with no refs finding at all.
  • 5 no-deprecated read sites (getReportTransactions, 4 files under src/libs/actions/) inside the
    83-file override above, an accepted cost.

Where the 1398 extra findings come from:

• Two type-aware rules ESLint enables but cannot run here: no-unnecessary-type-assertion (805)
  and import/no-cycle (153).
• tsgolint is stricter: no-deprecated write sites (+251), no-unsafe-type-assertion (+60),
  no-unsafe-return (+8). Kept and grandfathered.
• The bridge over-reports preserve-manual-memoization (+43).
• react-hooks/purity (+28), all Date called during render. ESLint's compiler build misses these real
  impure reads; we keep them to clean up over time.
• About 50 more across shared rules, one to four each, with import/no-named-as-default
  the outlier at 13. unicorn/prefer-at and no-unsafe-optional-chaining add 4 each with no ESLint
  counterpart.

rulesdir/boolean-conditional-rendering is the one rule with no Oxlint counterpart that we lose. no-invalid-this is
not run either, because Oxlint's plugin bridge throws on the API it calls.

Plan

This PR bumps oxc-transform-react, the React Compiler the web build and Jest use, to 0.155.0 (the bridge needs
reportDiagnostics, added in 0.152.0 by oxc#26624). The .npmrc exclusion is only needed until oxlint 1.89.0 and
oxc-transform-react 0.155.0 are past npm's 7-day hold, on 2026-10-26 after 17:24 UTC. If this merges before
then, a follow-up PR removes the four min-release-age-exclude lines.

1. Land it as non-blocking, keep its own baseline (absorbing on main, see above), watch it for a day, and compare the results between ESLint and Oxlint.
2. Make Oxlint blocking next to ESLint for 2-3 days: drop SEATBELT_INCREASE, continue-on-error and the report step from oxlint.yml, add oxlint to confirmPassingBuild in preDeploy.yml, make oxlint the default in scripts/lint/index.ts, and review the absorbed baseline first.
3. Remove ESLint, with no going back. Then rewrite disable comments that name hosted rules by their ESLint id to the hosted/* id, delete eslintDirectives.mjs, and turn on Oxlint's unused-disable reporting.
4. Fix Oxlint-only findings over time to shrink the baseline.

Workarounds to delete when upstream moves: the bridge when oxc#26277 is fixed, sharding when oxc#26621 is, the self-closing JSX stand-in when tsgolint ships its fix, the vite-plus pin when @types/node reaches 22.

POC PR with migration scripts and tests: #98027

Fixed Issues

$ #104516
PROPOSAL:

Tests

• Verify that no errors appear in the JS console

Offline tests

QA Steps

• Verify that no errors appear in the JS console

PR Author Checklist

• I linked the correct issue in the ### Fixed Issues section above
• I wrote clear testing steps that cover the changes made in this PR
  • I added steps for local testing in the Tests section
  • I added steps for the expected offline behavior in the Offline steps section
  • I added steps for Staging and/or Production testing in the QA steps section
  • I added steps to cover failure scenarios (i.e. verify an input displays the correct error message if the entered data is not correct)
  • I turned off my network connection and tested it while offline to ensure it matches the expected behavior (i.e. verify the default avatar icon is displayed if app is offline)
  • I tested this PR with a High Traffic account against the staging or production API to ensure there are no regressions (e.g. long loading states that impact usability).
• I included screenshots or videos for tests on all platforms
• I ran the tests on all platforms & verified they passed on:
  • Android: Native
  • Android: mWeb Chrome
  • iOS: Native
  • iOS: mWeb Safari
  • MacOS: Chrome / Safari
• I verified there are no console errors (if there's a console error not related to the PR, report it or open an issue for it to be fixed)
• I verified there are no new alerts related to the canBeMissing param for useOnyx
• I followed proper code patterns (see Reviewing the code)
  • I verified that any callback methods that were added or modified are named for what the method does and never what callback they handle (i.e. toggleReport and not onIconClick)
  • I verified that comments were added to code that is not self explanatory
  • I verified that any new or modified comments were clear, correct English, and explained "why" the code was doing something instead of only explaining "what" the code was doing.
  • I verified any copy / text shown in the product is localized by adding it to src/languages/* files and using the translation method
    • If any non-english text was added/modified, I used JaimeGPT to get English > Spanish translation. I then posted it in #expensify-open-source and it was approved by an internal Expensify engineer. Link to Slack message:
  • I verified all numbers, amounts, dates and phone numbers shown in the product are using the localization methods
  • I verified any copy / text that was added to the app is grammatically correct in English. It adheres to proper capitalization guidelines (note: only the first word of header/labels should be capitalized), and is either coming verbatim from figma or has been approved by marketing (in order to get marketing approval, ask the Bug Zero team member to add the Waiting for copy label to the issue)
  • I verified proper file naming conventions were followed for any new files or renamed files. All non-platform specific files are named after what they export and are not named "index.js". All platform-specific files are named for the platform the code supports as outlined in the README.
  • I verified the JSDocs style guidelines (in STYLE.md) were followed
• If a new code pattern is added I verified it was agreed to be used by multiple Expensify engineers
• I followed the guidelines as stated in the Review Guidelines
• I tested other components that can be impacted by my changes (i.e. if the PR modifies a shared library or component like Avatar, I verified the components using Avatar are working as expected)
• I verified all code is DRY (the PR doesn't include any logic written more than once, with the exception of tests)
• I verified any variables that can be defined as constants (ie. in CONST.ts or at the top of the file that uses the constant) are defined as such
• I verified that if a function's arguments changed that all usages have also been updated correctly
• If any new file was added I verified that:
  • The file has a description of what it does and/or why is needed at the top of the file if the code is not self explanatory
• If a new CSS style is added I verified that:
  • A similar style doesn't already exist
  • The style can't be created with an existing StyleUtils function (i.e. StyleUtils.getBackgroundAndBorderStyle(theme.componentBG))
• If new assets were added or existing ones were modified, I verified that:
  • The assets are optimized and compressed (for SVG files, run npm run compress-svg)
  • The assets load correctly across all supported platforms.
• If the PR modifies code that runs when editing or sending messages, I tested and verified there is no unexpected behavior for all supported markdown - URLs, single line code, code blocks, quotes, headings, bold, strikethrough, and italic.
• If the PR modifies a generic component, I tested and verified that those changes do not break usages of that component in the rest of the App (i.e. if a shared library or component like Avatar is modified, I verified that Avatar is working as expected in all cases)
• If the PR modifies a component related to any of the existing Storybook stories, I tested and verified all stories for that component are still working as expected.
• If the PR modifies a component or page that can be accessed by a direct deeplink, I verified that the code functions as expected when the deeplink is used - from a logged in and logged out account.
• If the PR modifies the UI (e.g. new buttons, new UI components, changing the padding/spacing/sizing, moving components, etc) or modifies the form input styles:
  • I verified that all the inputs inside a form are aligned with each other.
  • I added Design label and/or tagged @Expensify/design so the design team can review the changes.
• If a new page is added, I verified it's using the ScrollView component to make it scrollable when more elements are added to the page.
• I added unit tests for any new feature or bug fix in this PR to help automatically prevent regressions in this user flow.
• If the main branch was merged into this PR after a review, I tested again and verified the outcome was still expected according to the Test steps.

Screenshots/Videos

Android: Native

Android: mWeb Chrome

iOS: Native

iOS: mWeb Safari

MacOS: Chrome / Safari

[github-actions]

⚠️ This PR is possibly changing native code and/or updating libraries, it may cause problems with HybridApp. Please check if any patch updates are required in the HybridApp repo and run an AdHoc build to verify that HybridApp will not break. Ask Contributor Plus for help if you are not sure how to handle this. ⚠️

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
see 12 files with indirect coverage changes

[LukasMod]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 82f9a6d6fd

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[LukasMod]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 568fc81eee

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[LukasMod]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Keep it up!

Reviewed commit: daf3ac1f60

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[melvin-bot]

Hey, I noticed you changed src/languages/en.ts in a PR from a fork. For security reasons, translations are not generated automatically for PRs from forks.

If you want to automatically generate translations for other locales, an Expensify employee will have to:

1. Look at the code and make sure there are no malicious changes.
2. Run the Generate static translations GitHub workflow. If you have write access and the K2 extension, you can simply click: [this button]

Alternatively, if you are an external contributor, you can run the translation script locally with your own OpenAI API key. To learn more, try running:

npx bun ./scripts/generateTranslations.ts --help

Typically, you'd want to translate only what you changed by running npx bun ./scripts/generateTranslations.ts --compare-ref main

[LukasMod]

@MelvinBot review

[MelvinBot]

Review

I found no blocking bugs. The default npm run lint (ESLint) path behaves the same, a crashed Oxlint leg is never treated as clean, and the Oxlint job can't block deploys. Before merge, look at the first two items below. The rest are small parity gaps.

Worth fixing or calling out

1. This PR also changes the production web compiler. package.json:352 bumps oxc-transform-react from 0.148.0 to 0.155.0. That package is the React Compiler for the shipped web bundle (config/rsbuild/loaders/oxc-react-compiler-loader.mjs), the compliance check, and Jest. Version 0.155.0 is two days old and skips the 7-day hold through .npmrc:16-17. That exclusion also covers every future version. Consider splitting the bump into its own PR, or mention it in the description and test the web build.
2. A parse error fails the whole run with a misleading message. OxlintLinter.ts:314-322 treats any diagnostic with no code as "a JS plugin threw" and exits 2. If oxc parser errors also have no code, one file with a syntax error drops every other finding. ESLint reports that case as a ruleId: null error and keeps going. I couldn't run Oxlint to check what parser errors look like.

Minor parity gaps (each verified in code)

• Import resolver settings are missing. oxlint.config.mts:120-123 has no import/resolver or import/extensions, but ESLint sets both for TypeScript. If Oxlint passes settings to JS plugins, hosted/no-useless-path-segments and hosted/no-relative-packages can't resolve .ts/.tsx targets and stay silent.
• TypeScript option mismatches. prefer-promise-reject-errors keeps allowEmptyReject: true in TS files, but ESLint's TS config uses plain 'error'. prefer-const keeps ignoreReadBeforeAssign: true, but ESLint's TS config doesn't.
• The OnyxUtils import ban also applies to JS files. The ban at oxlint.config.mts:31-38 reaches js/jsx/mjs/cjs files. ESLint applies it only to ts/tsx files, through @typescript-eslint/no-restricted-imports.
• Compiler diagnostics without a label land at 1:0. reactCompilerRust.mjs:200-206 falls back to offset 0. The ESLint plugin skips diagnostics that have no location, and no next-line disable can reach a finding at 1:0.
• .mts/.cts files are compiled as JSX. getLang in reactCompilerRust.mjs:66-74 does this, so TypeScript syntax in those files fails to parse and they get no compiler diagnostics.
• --fix can count native findings twice. Under --fix the legs run one after another (OxlintLinter.ts:389-398). A type-aware autofix such as dot-notation moves the columns of findings the shards already reported. The dedupe by message key then misses, and the seatbelt can show false new errors.
• Shard count 0 means different things. OXLINT_SHARDS=0 gives one shard, but --shards=0 means auto.
• eslintDirectives.mjs edge cases. It treats // eslint-disable foo in a line comment as a block disable, while ESLint ignores that form. It also ignores where a directive sits on a line. Nothing in the repo hits these today.
• Unused disable comments are no longer flagged. ESLint sets reportUnusedDisableDirectives: 'error', and the Oxlint run has no equivalent.

Checked and looks correct

• continue-on-error is on the lint step, and preDeploy.yml keeps oxlint out of confirmPassingBuild's needs, so deploys don't depend on it.
• Baseline tightening on main uses the same token setup as lint.yml. A crashed run (exit code above 1) never reaches the seatbelt. If both workflows push at once, the second push is rejected and retried on the next run.
• Shards split files with no overlap. Fatal legs are retried once and then fail the run. Temp configs are removed in finally and gitignored.
• The ESLint seatbelt keys and processors are unchanged. The only visible change is that the "removed N baseline row(s)" message now goes to stderr.
• The lockfile only adds oxlint, tsgolint and oxc packages, all as dev dependencies. The vite-plus override affects an optional peer that isn't installed.

---

view run

[LukasMod]

Working on Melvin findings

[github-actions]

This PR adds a new Onyx.connectWithoutView call, so I've requested a review from the Onyx performance reviewers (@tgolen, @mountiny, @luacmartins, @chuckdries) — a review from any one of them is enough. Please add a link in your PR description to the Slack discussion where the @frontend-performance team approved using connectWithoutView here.

[LukasMod]

I marked it as a draft until all feedback is not resolved so actions will not be triggered

[LukasMod]

@roryabraham Thanks for the review, I think everything is covered now. I updated config regarding latest changes on main