[worker] Resolve runtime cache settings from job env and fallback on npm cache failures by AbbanMustafa · Pull Request #3855 · expo/eas-cli

AbbanMustafa · 2026-06-12T00:21:33Z

Why

We want to enable npm cache proxy usage per build job, but the worker was loading runtime settings before the job environment was available. That meant job-scoped environment variables like EAS_USE_NPM_CACHE=1 https://github.com/expo/universe/pull/27727 could not control cache URL resolution.

We also need cache proxy rollout to be safe: if the proxy is unavailable or misconfigured, builds should fall back to the normal npm registry behavior instead of failing during the post-prebuild install.

How

Move worker runtime settings loading into createBuildContext, where the job env is available.
Resolve runtime cache URLs from job env plus runner cache URL fallbacks.
Set NPM_CONFIG_REGISTRY only when the job has enabled npm cache usage.
Add post-prebuild npm install fallback:
- enabled only when EAS_USE_NPM_CACHE=1
- detects install failures caused by the configured npm cache registry
- reports warning-level Sentry events
- retries the same npm install after removing only NPM_CONFIG_REGISTRY, so npm falls back to project/user/default registry settings
Add Sentry reporting for non-fatal npm cache registry errors that npm logs while still exiting 0.

Test Plan

Added unit coverage for job-env runtime cache settings resolution.
Added unit coverage for post-prebuild npm cache fallback behavior:
- retries without the cache registry when the cache install fails
- reports non-fatal cache registry errors to Sentry when npm logs an error but exits 0
- does not retry when EAS_USE_NPM_CACHE is not enabled
Validated the failure path with a local build by setting EAS_USE_NPM_CACHE=1 from www. Cache private network could not resolve/connect to the internal cache host, which reproduced the proxy failure path and confirmed the build retried through the normal npm registry behavior, and logged to sentry

https://expoio.sentry.io/issues/7511993527/?environment=development&project=1837720&query=is%3Aunresolved&referrer=issue-stream&statsPeriod=14d

NOOP when env var is not set

…npm cache failures

github-actions · 2026-06-12T00:21:49Z

Subscribed to pull request

File Patterns	Mentions
*/	@douglowder

Generated by CodeMention

codecov · 2026-06-12T00:27:06Z

Codecov Report

❌ Patch coverage is 95.60440% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 58.48%. Comparing base (22fc4da) to head (c947944).

Files with missing lines	Patch %	Lines
...ages/build-tools/src/common/installDependencies.ts	96.37%	2 Missing ⚠️
packages/build-tools/src/runtimeSettings.ts	92.60%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3855      +/-   ##
==========================================
+ Coverage   58.42%   58.48%   +0.07%     
==========================================
  Files         917      919       +2     
  Lines       39994    40129     +135     
  Branches     8418     8445      +27     
==========================================
+ Hits        23363    23467     +104     
- Misses      16536    16566      +30     
- Partials       95       96       +1

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

…npm cache failures

AbbanMustafa · 2026-06-12T03:41:50Z

+  useFrozenLockfile: boolean;
+}): Promise<void> {
+  const npmCacheUrl = env.NPM_CONFIG_REGISTRY;
+  const npmCacheErrorTracker = createNpmCacheRegistryErrorTracker({ env, npmCacheUrl });


this captures cache failures that npm only prints to stdout/stderr, like audit endpoint ENOTFOUND, so we still log them to Sentry when npm exits 0 and skips the catch/fallback path

sjchmiela

needs some polish

sjchmiela · 2026-06-12T07:00:56Z

  const cocoapodsCacheUrl = RuntimeSettings.getCocoapodsCacheUrl();

  setEnv(env, 'NPM_CACHE_URL', npmCacheUrl);
+  setEnv(env, 'NPM_CONFIG_REGISTRY', npmCacheUrl);


gosh it took me so long to find docs proving this should work lol

does this replace npm config set registry?

Yes! they mention it here https://docs.npmjs.com/cli/v11/using-npm/config#environment-variables

Any environment variables that start with npm_config_ will be interpreted as a configuration parameter. For example, putting npm_config_foo=bar in your environment will set the foo configuration parameter to bar.

example

> echo "registry=https://from-npmrc.example" > /tmp/.npmrc > npm config get registry --userconfig=/tmp/.npmrc https://from-npmrc.example > export NPM_CONFIG_REGISTRY=https://from-env.example > npm config get registry --userconfig=/tmp/.npmrc https://from-env.example

should we remove call to npm config set registry then?

I think we should! posted validation below

sjchmiela · 2026-06-12T07:12:42Z

  const childLogger = buildLogger.child({ buildId });
+  await RuntimeSettings.loadAsync({
+    environment: config.env,
+    logger: childLogger,


so actually about this logger -- i don't think we should use the build-facing logger. i don't think we need users to see "Failed to fetch worker runtime settings, using defaults", especially if it's in an "(unnamed build phase)"

i kept this logger from the start so "if we ever were to inspect logs of the service" we would have a record of this failure. maybe it was a bad idea and we should just rely only on sentry and env vars printing as distinction?

sjchmiela · 2026-06-12T17:41:23Z

  const cocoapodsCacheUrl = RuntimeSettings.getCocoapodsCacheUrl();

  setEnv(env, 'NPM_CACHE_URL', npmCacheUrl);
+  setEnv(env, 'NPM_CONFIG_REGISTRY', npmCacheUrl);


should we remove call to npm config set registry then?

sjchmiela · 2026-06-12T17:43:33Z

+    const enabled =
+      envOverride === '1' ||
+      (envOverride !== '0' && runtimeSettings.caches?.[process.platform]?.npm);


i think if runtime settings is false we should NOT enable a given cache. runtime settings being false is for cache maintenance when we know it's going to not be available so even if a user has EAS_USE_NPM_CACHE we should not enable the cache for them

thank you yes, adding back to allow for killswitch

github-actions · 2026-06-12T18:04:15Z

⏩ The changelog entry check has been skipped since the "no changelog" label is present.

AbbanMustafa · 2026-06-12T18:13:52Z

Validated with updated changes that unset EAS_USE_NPM_CACHE results in a no op

and with env var set, install through cache is attempted and failure is reported to sentry

https://expoio.sentry.io/issues/7545404457/events/latest/?environment=development&project=1837720&query=is%3Aunresolved&referrer=latest-event&statsPeriod=1h

[worker] Resolve runtime cache settings from job env and fallback on …

f575e83

…npm cache failures

AbbanMustafa added the no changelog PR that doesn't require a changelog entry label Jun 12, 2026

[worker] Resolve runtime cache settings from job env and fallback on …

1e7c270

…npm cache failures

AbbanMustafa commented Jun 12, 2026

View reviewed changes

AbbanMustafa requested a review from sjchmiela June 12, 2026 04:23

sjchmiela requested changes Jun 12, 2026

View reviewed changes

AbbanMustafa added 7 commits June 12, 2026 11:23

merge conflict

8f8934f

no longer url fallback

f221991

use env over runtime stngs

ad5335e

report log to sentry only

4599f13

proper subclass errs

ad3de1e

proper subclass errs

a593246

proper subclass errs

3909e80

sjchmiela reviewed Jun 12, 2026

View reviewed changes

runtime setting override, rem set registry cmd

c947944

AbbanMustafa requested a review from sjchmiela June 12, 2026 18:33

Conversation

AbbanMustafa commented Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why

How

Test Plan

Uh oh!

github-actions Bot commented Jun 12, 2026

Uh oh!

codecov Bot commented Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sjchmiela left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Jun 12, 2026

Uh oh!

AbbanMustafa commented Jun 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

AbbanMustafa commented Jun 12, 2026 •

edited

Loading

codecov Bot commented Jun 12, 2026 •

edited

Loading