## Summary
- export precompiled modules helpers from `@expo/build-tools` and start preparation when flagged builds enable precompiled modules
- wait for precompiled module preparation before running `pod install` in both build-tools pod flows
- set worker env and config for precompiled module downloads, including proxy-aware download handling and coverage for the new behavior
## Testing
<img width="812" height="325" alt="Zrzut ekranu 2026-03-28 o 22 26 35" src="https://github.com/user-attachments/assets/9e4deb10-d233-4bb6-ac5d-48cef31c17ed" />
<img width="1032" height="548" alt="Zrzut ekranu 2026-03-28 o 22 27 27" src="https://github.com/user-attachments/assets/a6bbcb75-4845-4dce-9f31-5097ece800ae" />
## Summary
- add a dedicated `packages/eas-cli/src/sentry.ts` module that initializes `@sentry/node` for eas-cli
- capture command failures from `EasCommand`, including `EasCommandError`, and flush Sentry on exit
- propagate `EAS_CLI_SENTRY_DSN` into the worker child-process environment so worker-run eas-cli commands can report errors too
https://exponent-internal.slack.com/archives/C06EFBQK3B7/p1773432161848589?thread_ts=1773417343.042269&cid=C06EFBQK3B7
## Validation
- `yarn test src/commandUtils/__tests__/EasCommand-test.ts --runInBand`
- `yarn test:unit --runInBand src/__unit__/env.test.ts`
- `yarn build` in `packages/eas-cli`
- `node ./bin/run --version`
- `node ./bin/run --help`
- `node ./bin/run analytics`
- `EAS_CLI_SENTRY_DSN=https://[email protected]/1 node ./bin/run analytics`
## Manual Sentry Verification
- ran the compiled CLI against the test Sentry project `expo-stanley-test/test-local` with a real DSN
- forced an `EasCommandError` from `eas analytics` and confirmed the CLI failed locally as expected
- verified that Sentry ingested the event and created issue `TEST-LOCAL-1`
- confirmed the event in Sentry at `https://expo-stanley-test.sentry.io/issues/TEST-LOCAL-1`
## Summary
- standardize workspace clean/rebuild/prepack deletion steps on `rimraf`
- add `rimraf` where package-local scripts invoke it directly
- keep `worker` and root clean flows working from a clean install
Also adds Codex config for new workspace.
And the problem it solves is that sometimes `yarn clean` would fail if there were no `*.tsbuildinfo` files. Something weird about zsh in package.json
## Validation
- mise exec -- yarn clean
- mise exec -- yarn install --immutable
- mise exec -- yarn build
- mise exec -- yarn typecheck
- mise exec -- yarn clean (in packages/worker)
- mise exec -- yarn clean (in packages/logger)
Summary
- add the missing `ITSAppUsesNonExemptEncryption` boolean to `packages/worker/app.json` so the worker app can declare non-exempt encryption usage and avoid iOS config errors
Should fix system tests!
## Summary
- enable worker HTTP log streaming for staging using https://staging-logs.expo.dev/logs/
This is not tied to anything in UI yet. Just want to see the logs flowing.
## Summary
- stop mutating the original error instance in worker Sentry reporting
- attach the worker message to Sentry as an extra instead of rewriting the exception message — matches `universe/server-logging`
- keep the launcher-facing error payload unchanged
- fixes integration tests
## Testing
- yarn test:integration:watch stateSync
- yarn test:unit sentry
## Summary
- introduce `SystemError` as the explicit system-failure type
- replace in-repo `BuildError` usage with `ExpoError`, `UserError`, and `SystemError`
- keep unknown errors on the user-error side
## Changes
- remove the old `BuildError` class from the in-repo error flow
- make `SystemError` always publish `SERVER_ERROR` while keeping `trackingCode` for internal use
- update worker handling, analytics, and state sync to use the shared `ExpoError` contract
- avoid mutating the user-facing error object during Sentry reporting
## Testing
- `packages/eas-build-job`: `npm exec -- corepack yarn build && npm exec -- corepack yarn test src/__tests__/errors.test.ts --runInBand`
- `packages/build-tools`: `npm exec -- corepack yarn jest-unit src/buildErrors/__tests__/detectError.test.ts --runInBand`
- `packages/worker`: `npm exec -- corepack yarn test:integration src/__integration__/stateSync.test.ts`
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
We want to push logs to a logs service.
# How
Initially I wanted to set up a WebSocket connection between worker and the logs service to reduce the HTTP overhead, but I found having a persistent/reconnecting WebSocket connection is not trivial and HTTP should be ok.
Confirmed failure to upload is not logged in the GCS log.
# Test Plan
Tested manually.
https://github.com/user-attachments/assets/f80f0482-73f5-4dfb-855f-2361ed04aa4f
## Summary
- remove the duplicated `userFacingMessage` and `userFacingErrorCode` fields from the build error flow
- keep current runtime behavior and unknown-error semantics
- preserve the current worker protocol shape
Stanley: I don't want to have separate `userFacingMessage` and `message` / `userFacingErrorCode` and `errorCode`. It looks like the reason we had them separate was to log "UNKNOWN_ERROR" to the user if we encounter an internal issue like "NPM_CORRUPTED_PACKAGE". This pull request inverts the logic — we're going to have a `trackingCode` so we keep logging "NPM_CORRUPTED_PACKAGE" to Sentry and the default `errorCode` is the one we log to the user.
## Changes
- make `BuildError.format()` use canonical `message` and `errorCode`
- make `build-tools` store internal-only classification in `trackingCode` instead of parallel public fields
- make worker derive `internalErrorCode` from `trackingCode ?? errorCode`
- delete the obsolete `userFacing...` fields from `BuildError`
## Testing
- `packages/eas-build-job`: `npm exec -- corepack yarn test src/__tests__/errors.test.ts --runInBand`
- `packages/build-tools`: `npm exec -- corepack yarn jest-unit src/buildErrors/__tests__/detectError.test.ts --runInBand`
- `packages/worker`: `npm exec -- corepack yarn test:integration src/__integration__/stateSync.test.ts`
## Why
Generic/custom workflow steps can throw `UserFacingError`, but worker-level error handling currently maps any non-`BuildError` to `UnknownBuildError` / `UnknownCustomBuildError`. That causes actionable step errors to be persisted and exposed as `UNKNOWN_ERROR`.
## How
- add a `toBuildError` conversion in worker service catch path
- preserve existing `BuildError` values
- convert `UserFacingError` into `BuildError` with the same code/message/docsUrl
- keep unknown fallbacks only for truly unstructured errors
- add integration coverage in `stateSync` to assert generic job `UserFacingError` is exposed in `externalBuildError` and `internalErrorCode`
We should make a bigger refactor to error handling, but it's more involved — https://github.com/expo/eas-cli/pull/3447.
## Test Plan
CI should pass.
```
PASS src/__integration__/stateSync.test.ts
State sync mechanism
query new worker
✓ should return 'new' state (210 ms)
query worker after dispatch
✓ should return 'in-progress' state (170 ms)
query worker after build
✓ should return 'success' state (2062 ms)
query worker after error
✓ should return 'error' state with artifacts from error (88 ms)
✓ should expose user-facing error from generic job (60 ms)
```
Co-authored-by: Copilot <[email protected]>
## Why
In https://github.com/expo/eas-cli/pull/3406 EAS workflow building the worker was **not** updated to Yarn v4. This breaks building worker.
I tried adding `defaults.tools.yarn: 4.12.0`, but this unfortunately [is **not** the way to use Yarn v4 on EAS as I've learned](https://docs.expo.dev/more/create-expo/#yarn-2-modern). The right way is using Corepack.
However, previously we explicitly did **not** use Corepack, because `pnpm --version` check the worker does in the beginning of its run was failing because it was run without `cwd`, so it was running within worker package which was configured with Yarn.
Since we want to use Yarn v4, we need to figure out a way to use Corepack.
## How
- Enabled Corepack.
- Set up `yarn` executable like the one we have in `universe`.
- Changed `pnpm --version` to **not** run inside worker (so it does not error).
- Removed `setup-node` for specific Node versions in favor of a Mise override.
## Test plan
- Corepack + Yarn resolution checks (local):
- `mise exec [email protected].4 -- bash -lc 'corepack enable; yarn --version'` => `4.12.0`
- same check from `scripts/` => `4.12.0`
- Worker runtime unit test (local):
- `mise exec [email protected].4 -- bash -lc 'corepack enable; yarn workspace @expo/worker test:unit -- runtimeEnvironment.test.ts'`
- GitHub workflow runs:
- success on Corepack worker workflow update: https://github.com/expo/eas-cli/actions/runs/22133707462
- latest run for full set of changes: https://github.com/expo/eas-cli/actions/runs/22133932635
## Why
- Repository maintenance: standardize on Yarn 4 managed via Mise.
- Prepare for introducing minimum package-age constraints in dependency policy.
## How
- Migrated repo tooling to Yarn 4 with Mise.
- Updated CI and scripts for Yarn 4 behavior:
- Lerna invoked via `yarn run -T lerna`.
- Coverage args forwarded correctly (`-- --coverage`).
- Removed unsupported Yarn 4 `--silent` usage.
- Added missing package-local script dependencies surfaced by Yarn 4 (`jest`, `oxlint`, `rimraf`).
- Increased `EasCommand` coverage test timeout from 30s to 60s. (Was consistently failing in CI.)
## Test Plan
- CI only: `Run tests` matrix (including Node 20 coverage) and other touched workflows.
- Release behavior will be verified during the normal release run.
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
For the new logs experience we need to identify every log with a unique ID.
Also, in the new logs experience we're going to be handling logs in plaintext, uncompressed, so size starts to matter.
# How
Decorating every log with a unique ID is surprisingly difficult with bunyan. It only runs serializers on the object passed in, I tried adding a `log_id` serializer `serializers: { log_id: () => uuidv7() }` and a non-empty value to build logger, but it was getting serialized in the moment of logger creation.
Instead, I'm adding `log_id` in the `Transform` that also masks secrets. The overall architecture looks like this now:
<img width="1742" height="1126" alt="Bez tytułu-2026-02-04-1436-3" src="https://github.com/user-attachments/assets/956e1b58-75ca-41c5-92ba-8229dd981ff6" />
Also removes `service: 'worker'` and `hostname` from logs. `service` was only being used [here](https://github.com/expo/universe/blob/3246198a3f913bf6fd0fd82e8e3df402cb59e82e/server/website/common/logs-utils.ts#L218) which I don't think is a thing right now.
# Test Plan
Added unit test that ensures `Transform` adds `logId`.
<img width="1257" height="1102" alt="Zrzut ekranu 2026-02-12 o 16 02 32" src="https://github.com/user-attachments/assets/4cfbbf43-744f-4af0-9faf-de0cab54f178" />
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
We're moving to Oxc in universe too. The toolset is faster than ESLint / Prettier.
# How
Asked Codex to migrate the repository. I don't think all rules got migrated, unfortunately, but also the lint are in a bit weird state (different rules in `eas-cli` vs the rest of the repo). I hope we can unify the rules soon.
# Test Plan
CI should pass.
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
Release failed because in `prepack` in some packages we delete `dist`, but don't delete `.tsbuildinfo` file.
# How
Made Codex fix all the package scripts. I _think_ we also could skip `yarn build` before publishing.
# Test Plan
Try to release and see if it works.
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
I think we migrated from Volta to Mise in… most if not all places?
# How
I think I let Codex do all the work, nothing too difficult. Grepped repo for "volta" leftover afterwards.
One thing I'm not sure of is whether worker is going to work. I remember trying to remove Volta in `turtle-v2` and it wasn't working for some reason? We'll see.
# Test Plan
Compared these changes with https://github.com/expo/universe/commit/acca8169746c5cb4d708add083f77b927341ae4a.
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
We moved more packages into this monorepo and the existing test setup didn’t scale well:
- the root Jest config only covered a subset of packages,
- PRs ran heavier “all tests” workflows,
- worker builds ran on PRs even though typecheck + tests were enough.
This PR aligns testing with the dependency graph, reduces unnecessary CI work, and standardizes Jest config handling.
# How
- Switched root testing to Lerna and updated docs.
- CI now runs affected+dependent tests on PRs and all packages on main.
- Coverage runs via Lerna with per-package reports (no merge) and is kept for PRs.
- Worker builds are skipped on PRs; worker workflow triggers include dependency paths on main.
- All Jest configs are now TypeScript with a shared base (`jest/jest.shared.config.ts`) and use `@jest/types` to avoid TS config parsing errors.
- CI stays unit-only (no integration tests).
# Test Plan
Not run locally. CI covers:
- `yarn lerna run test` (affected on PRs, all on main)
- `yarn lerna run test -- --coverage`
`package.sh` currently does not copy workspace packages into the temporary build folder which makes worker pull packages from NPM.
Instead, we want to use local versions from workspace.
Added `copy_package` calls for all dependencies of `@expo/worker`.
It was done in a similar way in `turtle-v2` — https://github.com/expo/turtle-v2/blob/245eaf23492f2dcd9204a4606bb95c109208d29e/src/services/worker/package.sh#L46-L47.
Ran `./package.sh` and confirmed `@expo/eas-build-job` and other packages are symlinked.
## Consolidate eas-build packages into eas-cli monorepo
This PR completes the migration of packages from the `eas-build` repository into `eas-cli`, consolidating all EAS Build tooling into a single monorepo.
### Background
Previously, the EAS Build libraries lived in a separate `expo/eas-build` repository. This made development and releases more complex, requiring coordination between two repos. By consolidating into `eas-cli`, we can:
- Develop and test changes across CLI and build tools in a single PR
- Simplify the release process
- Reduce dependency version drift between packages
### Migrated Packages
The following packages have been moved from `eas-build`:
| Package | npm Name | License |
|---------|----------|---------|
| `build-tools` | `@expo/build-tools` | BUSL-1.1 |
| `eas-build-job` | `@expo/eas-build-job` | MIT |
| `steps` | `@expo/steps` | BUSL-1.1 |
| `logger` | `@expo/logger` | BUSL-1.1 |
| `downloader` | `@expo/downloader` | BUSL-1.1 |
| `turtle-spawn` | `@expo/turtle-spawn` | BUSL-1.1 |
| `template-file` | `@expo/template-file` | BUSL-1.1 |
| `local-build-plugin` | `eas-cli-local-build-plugin` | BUSL-1.1 |
| `create-eas-build-function` | `create-eas-build-function` | MIT |
### Changes
**Package Configuration:**
- Updated `repository` and `bugs` URLs in all migrated package.json files to point to `expo/eas-cli`
- Updated README repository links
- Aligned dependency versions across `eas-cli`, `eas-json`, and `worker` to use workspace packages
**Documentation:**
- Expanded `CLAUDE.md` with comprehensive documentation covering all packages, architectural patterns, and development workflows
- Added `AGENTS.md` for AI agent guidance
**Licensing:**
- Added `LICENSE-BUSL` for BUSL-1.1 licensed packages
- Added `LICENSE-eas-build-job` for MIT licensed packages
**Build Infrastructure:**
- Added `.gitignore` entries for build artifacts: `dist`, `*.tsbuildinfo` (needed for `tsconfig.json` and `tsconfig.build.json`), vim swap files
- Added `bin/eas-cli-local-build-plugin` symlink for local build testing
**Dependency Cleanup:**
- Removed stale npm registry entries from `yarn.lock` - workspace packages now resolve locally
### Testing
- [x] `yarn install` completes successfully
- [x] `yarn build` builds all packages
- [x] `yarn test` passes
- [x] Worker builds and runs correctly against workspace packages
### Follow-up
After this PR is merged:
- Archive or deprecate the `expo/eas-build` repository
- Update any CI/CD pipelines that referenced the old repo
- Update internal documentation pointing to eas-build
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
Noticed we're not running tests on Node 18 anymore (since https://github.com/expo/eas-cli/commit/6bb09c4e99950a6b46b0b908a93e8fcd9b3002aa). I think we probably want to officially drop support for Node 18 too.
# How
Looked at https://github.com/expo/eas-cli/commit/eb4c5e073d02f38d7f90fbd6a1b2b5491dc40464, pinned Node to a newer version, updated types, adjusted `engines`.
# Test Plan
CI should pass.
<!-- If this PR requires a changelog entry, add it by commenting the PR with the command `/changelog-entry [breaking-change|new-feature|bug-fix|chore] [message]`. -->
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
> Apple's altool in Xcode 26.0 has a bug where it incorrectly selects the wrong app when multiple apps share a common bundle ID prefix (e.g., `com.example`, `com.example.app1`, `com.example.app2`). This causes builds to be uploaded to the wrong app on App Store Connect.
> See: https://github.com/fastlane/fastlane/issues/29698https://github.com/expo/universe/pull/23310 + https://exponent-internal.slack.com/archives/C06EFBQK3B7/p1765365928911459
# How
If `xcodebuild -version` matches "Xcode 26.0", we'll add `DELIVER_ALTOOL_ADDITIONAL_UPLOAD_PARAMETERS='--use-old-altool'`. This helps if the code does not re-override `$DELIVER_ALTOOL_ADDITIONAL_UPLOAD_PARAMETERS` further, in which case it should do `DELIVER_ALTOOL_ADDITIONAL_UPLOAD_PARAMETERS="overrides $DELIVER_ALTOOL_ADDITIONAL_UPLOAD_PARAMETERS"` or sth like that.
I hope this fixes Launch.
# Test Plan
Tested manually on my computer
```
> spawnSync("xcodebuild", ["-version"]).stdout.includes("Xcode")
true
> spawnSync("xcodebuild", ["-version"]).stdout.includes("Xcode 26.0")
false
(below did not throw)
> spawnSync("abc")
{
error: Error: spawnSync abc ENOENT
at Object.spawnSync (node:internal/child_process:1123:20)
at spawnSync (node:child_process:877:24)
at REPL12:1:1
at ContextifyScript.runInThisContext (node:vm:137:12)
at REPLServer.defaultEval (node:repl:598:22)
at bound (node:domain:433:15)
at REPLServer.runBound [as eval] (node:domain:444:12)
at REPLServer.onLine (node:repl:927:10)
at REPLServer.emit (node:events:536:35)
at REPLServer.emit (node:domain:489:12) {
errno: -2,
code: 'ENOENT',
syscall: 'spawnSync abc',
path: 'abc',
spawnargs: []
},
status: null,
signal: null,
output: null,
pid: 0,
stdout: null,
stderr: null
}
```