Honor $BUNDLE_JOBS for bundle install

README.md

@@ -204,6 +204,9 @@ They should be set in the `env` at the job level as shown in the [example](#matr
 To find the correct the environment variable name, see the [Bundler docs](https://bundler.io/man/bundle-config.1.html) or look at `.bundle/config` after running `bundle config --local KEY VALUE` locally.
 You might need to `"`-quote the environment variable name in YAML if it has unusual characters like `/`.
 
+By default, `bundle install` is run with `--jobs` set to the number of available processors (up to 8).
+This can be changed by setting `BUNDLE_JOBS`.
+
 To perform caching, this action will use `bundle config --local path $PWD/vendor/bundle`.  
 Therefore, the Bundler `path` should not be changed in your workflow for the cache to work (no `bundle config path`).
 

bundler.js

@@ -214,10 +214,15 @@ export async function bundleInstall(gemfile, lockFile, platform, engine, rubyVer
     console.log(`Found cache for key: ${cachedKey}`)
   }
 
-  // Number of jobs should scale with runner, up to a point
-  const jobs = Math.min(os.availableParallelism(), 8)
+  const installArgs = ['install']
+  // Honor $BUNDLE_JOBS if set, as passing --jobs would override it
+  if (!process.env['BUNDLE_JOBS']) {
+    // Number of jobs should scale with runner, up to a point
+    const jobs = Math.min(os.availableParallelism(), 8)
+    installArgs.push('--jobs', `${jobs}`)
+  }
   // Always run 'bundle install' to list the gems
-  await exec.exec('bundle', ['install', '--jobs', `${jobs}`])
+  await exec.exec('bundle', installArgs)
 
   // @actions/cache only allows to save for non-existing keys
   // Also, skip saving cache for merge_group event