You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Due for payment 2026-11-10] Replace ESLint with Oxlint in E/App #104516
E/App's ESLint CI is slow and wasteful when it has a cache miss or failure. The proposal reports that a cold run takes about 494 seconds on a 16-vCPU runner, while a warm run checks changed files in about 51 seconds. When the warm run fails, CI clears the cache and retries, so legitimate lint failures can take about 570 seconds to surface.
Proposed solution
Replace ESLint with Oxlint for E/App. Oxlint is Rust-based, supports many ESLint rules natively, and supports JavaScript plugins compatible with the ESLint plugin API, allowing us to retain existing and custom rules.
Benchmark
The side-by-side benchmark reports: on an 8-vCPU runner, Oxlint checks the whole repo in about 53 seconds. Over two weeks, Oxlint had a 55-second mean runtime versus ESLint's 121 seconds, with no cache retry or long failure tail. The proposal estimates monthly CI costs dropping from $788.56 to $173.63 (about $7,400 annual savings).
Merge Oxlint as a non-blocking job next to ESLint. ESLint stays the merge gate, and local linting doesn't change
While it's non-blocking, main updates the Oxlint baseline after every merge, so PRs only see Oxlint findings they add - we can analyze those, compare with eslint action results
If this merges before 2026-10-26, a small follow-up removes the temporary .npmrc release-age exception.
Watch it for a few days and compare its results with ESLint's
Make Oxlint blocking alongside ESLint for 2-3 days. Its baseline then works like ESLint's: a PR fixes or baselines its own findings, and main only tightens
Remove ESLint as the linter.
Rewrite old disable comments to the new rule ids (one workaround to remove)
turn on unused-disable rule reporting
Uninstalling the eslint package isn’t possible yet. oxlint runs some ESLint rules as JS plugins, eslint-config-expensify lists eslint as a regular dependency
Keep fixing Oxlint-only findings over time to shrink the baseline - we do it right now, but finally with some real guards
Drop workarounds as upstream fixes land. Keep updating oxc version, look for opportunities to remove some of workarounds, to keep config smooth
I will keep updating this comparison from time to time (results combined)
round 1: from oxlint merge to 2026-10-28 07:57 UTC
round 2 added runs from 07:57 to 12:44 UTC
round 3 added runs from 12:44 to 14:47 UTC
round 4 added runs from 14:47 UTC to 2026-10-29 07:50 UTC
Verdict so far
Oxlint never passed a run ESLint failed (0/407 comparable pairs). Apart from the known, accepted prefer-at[0] gap, every ESLint error was also reported by Oxlint. Oxlint was never less safe in this window.
Counts
ESLint
Oxlint
Runs
441
442
Completed
411
421
Cancelled / action_required
26 / 4
17 / 4
PASS
394
355
FAIL
17
66
Timing (completed runs)
ESLint
Oxlint
Runner
blacksmith 16 vCPU
blacksmith 8 vCPU, 4 shards
Job total
1414.5 min
750.6 min (46.9% faster)
Job avg / median / p90 / max
207 / 89 / 764 / 2128 s
107 / 96 / 155 / 309 s
Lint step total
1120.4 min
472.3 min (57.8% faster)
Lint step avg / median / p90 / max
164 / 56 / 717 / 2101 s
67 / 67 / 90 / 132 s
ESLint is fast on a warm cache (median 53 s) but 48 runs hit a cold cache (200 s to 35 min); 27 of them between 13:00 and 15:00 UTC on 2026-10-28, after a lint-rule change invalidated the cache key. Oxlint has no cache and stays flat.
Agreement
Category
Pairs
Both PASS
348
Both FAIL
17
Oxlint FAIL, ESLint PASS (stricter)
42
ESLint FAIL, Oxlint PASS (missed)
0 ✅
Not comparable (cancelled / action_required)
33
Both FAIL: same findings?
16 of 17 pairs: every ESLint finding (file + line + rule) was also reported by Oxlint. Column differences and rule-name prefixes (no-shadow vs @typescript-eslint/no-shadow) are treated as the same finding. 1 partial miss:
undelete-transactions-actor efb3354 (ESLint 40623100940 / Oxlint 40623100945): Oxlint reported 2 of 3. Missed rulesdir/prefer-at at tests/actions/TransactionTest.ts:2117:60 (call[0]). oxlint does not flag literal [0] indexing. Known, accepted gap from unicorn/prefer-at
no-unnecessary-type-assertion: tsgolint flags more assertions than typescript-eslint; oxlint.config.mts:728 notes tsgolint's TS7 inference disagrees with the repo's TS 6.0.2. Mostly test files (BankAccountUtilsTest, SearchPageTest, ReportAlternateTextUtilsTest, useActionAvatarTest, ...).
no-deprecated/originalMessage on writes: tsgolint also reports writes to deprecated properties (oxlint.config.mts:1214), ESLint does not.
import/no-cycle: enabled only in Oxlint (oxlint.config.mts:323); cycles via BillPayUtils and via Welcome, OnboardingFlow, App.ts.
no-unsafe-* (addUtilsToWindow.ts, focusComposerWithDelay/index.ts): true positives; typecheck failed on the same lines. ESLint missed them because its cache skipped files the PR did not change.
Rules only in Oxlint config: react-hooks/purity, unicorn/prefer-at.
Resolving PRs: #101292 merged 2026-10-27, approved by abzokhattab, roryabraham. #105376 merged 2026-11-02, approved by huult, roryabraham. Both by LukasMod. No revert.
roster.md lists abzokhattab as a Backend Contributor and as a C+.
Activity
[-]Replace ESLint with Oxlint in E/App CI[/-][+]Replace ESLint with Oxlint in E/App[/+]roryabraham commentedon Oct 20, 2026
@LukasMod has a draft PR here: #101292
LukasMod commentedon Oct 22, 2026
I marked it as ready for review
How I see the plan:
mainupdates the Oxlint baseline after every merge, so PRs only see Oxlint findings they add - we can analyze those, compare with eslint action resultsmainonly tightensLukasMod commentedon Oct 26, 2026
PR updated
LukasMod commentedon Oct 28, 2026
I will keep updating this comparison from time to time (results combined)
Verdict so far
Oxlint never passed a run ESLint failed (0/407 comparable pairs). Apart from the known, accepted
prefer-at[0]gap, every ESLint error was also reported by Oxlint. Oxlint was never less safe in this window.Counts
Timing (completed runs)
ESLint is fast on a warm cache (median 53 s) but 48 runs hit a cold cache (200 s to 35 min); 27 of them between 13:00 and 15:00 UTC on 2026-10-28, after a lint-rule change invalidated the cache key. Oxlint has no cache and stays flat.
Agreement
Both FAIL: same findings?
16 of 17 pairs: every ESLint finding (file + line + rule) was also reported by Oxlint. Column differences and rule-name prefixes (
no-shadowvs@typescript-eslint/no-shadow) are treated as the same finding. 1 partial miss:rulesdir/prefer-atattests/actions/TransactionTest.ts:2117:60(call[0]). oxlint does not flag literal[0]indexing. Known, accepted gap from unicorn/prefer-atOxlint stricter (42 pairs), root causes
no-unnecessary-type-assertionno-deprecated/originalMessageimport/no-cycleno-unsafe-*react-hooks/purity,unicorn/prefer-at,no-deprecated/MenuItemWithTopDescriptionno-unnecessary-type-assertion: tsgolint flags more assertions than typescript-eslint;oxlint.config.mts:728notes tsgolint's TS7 inference disagrees with the repo's TS 6.0.2. Mostly test files (BankAccountUtilsTest, SearchPageTest, ReportAlternateTextUtilsTest, useActionAvatarTest, ...).no-deprecated/originalMessageon writes: tsgolint also reports writes to deprecated properties (oxlint.config.mts:1214), ESLint does not.import/no-cycle: enabled only in Oxlint (oxlint.config.mts:323); cycles via BillPayUtils and via Welcome, OnboardingFlow, App.ts.no-unsafe-*(addUtilsToWindow.ts,focusComposerWithDelay/index.ts): true positives;typecheckfailed on the same lines. ESLint missed them because its cache skipped files the PR did not change.react-hooks/purity,unicorn/prefer-at.19 remaining items
melvin-bot commentedon Nov 9, 2026
Triggered auto assignment to @mallenexpensify (
Awaiting Payment)melvin-bot commentedon Nov 9, 2026
Payment Summary
Resolving PRs:
[NoQA] Oxlint linting #101292
[No QA] chore: make oxlint the blocking linter and remove ESLint #105376
Reviewer: @huult owed $175 via NewDot
Contributor: @LukasMod does not require payment (Contractor)
BugZero Checklist (@mallenexpensify)
melvin-bot commentedon Nov 12, 2026
@mallenexpensify Whoops! This issue is 2 days overdue. Let's get this updated quick!
mallenexpensify commentedon Nov 13, 2026
Claude posting on behalf of Matt:
Your call. Job price: $175 (standard, created 2026-10-20, no Help Wanted).
No deploy blockers or regressions on the issue or the PRs.
The C+ checklist says there are no test steps, because no user-facing behavior changed.
Next: decide if abzokhattab is paid, then pay. No Caps+Y.
Evidence
PAYPREP v1mallenexpensify commentedon Nov 13, 2026
Payment Summary
Contributor: @abzokhattab due $100 via NewDot
Contributor+: @huult due $100 via NewDot
This is an edge case, I did 50% for each cuz they were
[NO QA]and one didn't have test vids (other was a single screenshot)Will leave open for feedback then close once I end up here agin.
huult commentedon Nov 14, 2026
@mallenexpensify Agreed, thanks!
JmillsExpensify commentedon Nov 14, 2026
$100 approved for @huult