# Why
Workers can set `NVM_NODEJS_ORG_MIRROR` through the Node.js cache settings. If `nvm install` fails with that mirror, runtime setup currently stops even when the default Node.js download source is available.
# How
Retry `nvm install` without the mirror if it failed with the mirror. This helps with situation when we fail to have refreshed Nodes list in our cache or our cache is down and stuff like that.
# Test Plan
CI should pass.
* [worker] Mask secrets in logged errors
The upload transform masked only each record's msg, so a serialized err
(message, stack and other fields) was uploaded with secrets intact. That
includes unknown build phase failures, which log { err } with an empty msg.
Mask the strings inside err on a copy, since bunyan shares the record with
other streams.
* [worker] Mask secrets in every field of uploaded log records
Mask all fields except bunyan's metadata, not only err: other fields such as
error, data and the job dump also carried secrets. Round-trip each value
through JSON first so a toJSON method can't bring back the original, mask keys
and numbers, and replace values nested too deeply instead of uploading them
unmasked.
* [build-tools] Call the device run session worker API with the session token
When a device run session job has the EAS_DEVICE_RUN_SESSION_TOKEN environment
secret, send every session-related www call to /v2/device-run-session-worker/*
with the expo-device-run-session-token header instead of GraphQL or the
account-wide REST routes. Without the token, keep the current calls.
The worker also reports Datadog metrics for these jobs through
/v2/device-run-session-worker/metrics.
* [build-tools] Download session builds without the robot token
In a device run session job with a session token, the build archive URL comes
from the device run session worker API and points to an artifact route that
needs no auth. After www stops minting EXPO_TOKEN for session jobs,
robotAccessToken is the session token, and www rejects it as a bearer
credential on every route. Send no Authorization header in that case.
* [build-tools] Match the device run session worker API response contract
www returns entity values from the worker API: session statuses are
lowercase and hyphenated, and GET /application returns
{ platform: 'ios' | 'android', build: { id, gitCommitHash,
applicationArchiveUrl } | null } without app slugs. Map the statuses to the
GraphQL values that callers compare, and name screenshots without an app
slug in worker API mode.
The worker now resolves duplicate EAS_DEVICE_RUN_SESSION_TOKEN secrets like
the build env does (the last one wins), so metrics and steps use the same
token. Add tests for the token propagation from job secrets to the worker API
client and for the download_build and start_agent_device_session call sites.
* [build-tools] Take the device run session platform from the worker runtime
* [build-tools] Report device run session build lookups as system errors and time out hung session requests
The worker's secrets filter takes secrets added while the job runs through
addBuildLogMask, which reaches steps via the build context's runtimeApi.
eas/run_agent passes each leased token to it and drops its own masking.
Masking marks every occurrence of every secret on the original message and
masks once, so a secret that is a substring of another no longer breaks the
longer match. The log buffer that feeds error detection masks with the same
patterns.
* [eas-cli][local-build-plugin] Deprecate Node 20, use Node 22 in CI
Print a deprecation warning from the eas-cli and local-build-plugin
bin scripts when running on Node.js < 22. Develop and test on Node.js
22: bump mise, drop 20.18.3 from the test matrix, and move to
@tsconfig/node22 and @types/node 22. Published engines are unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
Follow-up to #4399 to upgrade typescript to latest version.
# How
Tell claude to upgrade to 6.x.
# Test Plan
CI, tests
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
This is a bit behind in typescript. 6.x is where most of our repos live.
First step is to get to highest 5.x.
# How
Tell claude to upgrade to highest 5.x.
It required a new version of `https-proxy-agent` to be compatible with new node types.
# Test Plan
`yarn start`, tests.
# Why
Worker has config for API server URL. I now need MCP server URL too.
# How
Added `mcpServerUrl` in `config` and in contexts.
# Test Plan
CI should pass.
The legacy cache does not work in `workflow-orchestration`.
This rewrites the `CacheManager` to the new/not-so-new Turtle caches API.
## Summary
- replace worker build cache restore/save GCS config URLs with turtle cache download/upload helpers
- derive stable build cache keys without mutating job.cache and preserve clear for force uploads
- keep existing tar archive restore/save semantics and add CacheManager unit coverage
## Test plan
- cd packages/worker && yarn test CacheManager.test.ts
- cd packages/worker && yarn typecheck
- cd packages/build-tools && yarn typecheck
- git diff --check
* [worker] wire ssh session into build
* Rename SSH phase sessionPromise to done.
Same anti-flatten object wrapper; clearer that callers await tunnel teardown.
* Cover SSH phase supervise callbacks and teardown failure paths.
* [worker] Stop baking upterm into the worker package
build-tools resolves upterm from PATH or downloads darwin-arm64 /
linux-amd64 from gs://turtle-v2/upterm at runtime.
* [worker] align ssh phase with build-tools api
* [worker] build the ssh target from the job type
<!-- You can skip the changelog check by labeling the PR with "no changelog". -->
# Why
I want to see how the architecture works under full logs pressure.
# How
Enabled pushing logs to logs.expo.dev in production too.
# Test Plan
Staging works, let's see it in production.