Fix rails72 multibuildpack staging timeout by removing unused webpacker/sass-rails

[ivanovac]

Problem

Builds #19-28 fail with CF's hard 15-minute staging timeout. The root cause is not webpack compilation time (which only takes ~1m12s) — it is cumulative staging time exceeding 15 minutes:

Step	Time
Go binary compilation (supply)	~3 min
Python bootstrap + Node install	~25s
yarn install	~32s
bundle install (78 gems, includes sassc native C extension)	~9 min
Asset precompilation (webpack)	~1m12s
Total	~15-16 min → timeout

Previous PRs (#1178, #1179, #1180) addressed symptoms (test timeouts, app start timeout, manifest variable substitution) but not this root cause. PR #1181's fake manifest approach would save ~1m12s but the remaining ~14min still exceeds the limit.

Solution

The test only asserts two things:

Ruby version: ruby 3.x.x
Node version: vX.X.X

The controller just runs ruby -v and node -v. It does not use webpacker, sass-rails, webpack, SCSS, or the asset pipeline in any way. Remove those gems and everything they pull in.

Changes

• Gemfile: drop webpacker and sass-rails (only 2 gems declared now: rails + puma)
• Gemfile.lock: regenerated — 68 gems down from 78; sassc, ffi, webpacker, sprockets, nokogiri and ~10 others removed
• config/application.rb: drop sprockets/railtie require
• config/environments/production.rb: drop asset pipeline config blocks
• Deleted: package.json, yarn.lock, babel.config.js, postcss.config.js, app/javascript/, app/assets/, bin/webpack, bin/webpack-dev-server, bin/yarn
• .gitignore: drop node/yarn entries

Expected result

Step	Before	After
bundle install	~9 min (78 gems, sassc native)	~3-4 min (~68 gems, no native C)
Asset precompilation	~1m12s	skipped entirely (no sprockets/webpacker)
Total staging	~15-16 min (timeout)	~8-10 min ✅

Related PRs

• Fix multibuildpack integration test timeouts (builds #19-22 failures) #1178 (merged): Extended test timeout to 90s
• Add manifest.yml with increased timeout for Rails72 multibuildpack test fixture #1179 (merged): Added manifest.yml with 180s app start timeout
• Fix manifest.yml variable substitution error in Rails72 multibuildpack test #1180 (merged): Fixed manifest variable substitution error
• Skip asset precompilation in rails72 multibuildpack test to avoid CF staging timeout #1181 (open): Fake sprockets manifest — superseded by this PR (does not fix root cause)

[tnikolova82]

❌ PR Fails Integration Tests - Root Cause Identified

This PR breaks the nodejs multibuildpack test because it deletes entirely.

Test Failure Evidence

From the failed integration test run:

=== NAME  TestIntegration/integration/MultiBuildpack/when_supplied_with_nodejs/finds_the_supplied_dependency_in_the_runtime_container
    Timed out after 90.001s.
    Get "http://localhost:32781": dial tcp [::1]:32781: connect: connection refused

Why It Fails

The multibuildpack test uses nodejs-buildpack + ruby-buildpack together:

1. nodejs-buildpack detection: Looks for package.json to know it should install node
2. This PR deletes package.json → nodejs-buildpack doesn't install node
3. Without node: The app controller can't respond to node -v command
4. Result: App never starts, test times out

The Analysis Is Partially Correct

You're right that the root bottleneck is bundle install time (~9 min with 78 gems including sassc native extensions), not asset precompilation (~1m12s). Removing webpacker/sass-rails would help.

BUT you also removed the Node.js dependencies, which breaks the test's actual purpose.

What The Test Actually Needs

# The test ONLY checks these two things:
Eventually(deployment).Should(Serve(ContainSubstring("Ruby version: ruby 3.")))
Eventually(deployment).Should(Serve(MatchRegexp(`Node version: v\d+\.\d+\.\d+`)))

The app controller runs:

node_version = `node -v`  # ← NEEDS node to be installed
ruby_version = `ruby -v`

Correct Solution

Hybrid approach - keep what's needed, remove what's not:

File	Current	This PR	Should Be
package.json	Full webpacker deps	❌ Deleted	✅ Minimal: {"name":"app","private":true}
yarn.lock	Full lockfile	❌ Deleted	✅ Keep (or regenerate minimal)
Gemfile	Has webpacker/sass-rails	❌ Deleted	✅ Remove webpacker/sass-rails
Gemfile.lock	78 gems	❌ Deleted	✅ Regenerate (~68 gems)

Expected Staging Time

Phase	Current Master (PR #1181)	This PR (Broken)	Correct Fix
Go compilation	~3 min	~3 min	~3 min
Node/Python setup	~25s	❌ Skipped	~25s
yarn install	~32s	❌ Skipped	~5s (no deps)
bundle install	~9 min (78 gems, native)	❌ Fails	~3-4 min (68 gems, no native)
Asset precompile	✅ Skipped (PR #1181 fake manifest)	N/A	✅ Skipped
Total	~13-14 min (risky)	❌ Test fails	✅ ~8 min (safe)

Recommendation

Close this PR and create a new one that:

1. ✅ Keeps minimal package.json for nodejs-buildpack detection
2. ✅ Removes webpacker/sass-rails from Gemfile
3. ✅ Regenerates Gemfile.lock with fewer gems
4. ✅ Keeps PR Skip asset precompilation in rails72 multibuildpack test to avoid CF staging timeout #1181's fake sprockets manifest

This gives us the bundle install speedup you identified without breaking the nodejs multibuildpack detection.

Current Status

PR #1181 is already merged and includes the fake sprockets manifest that skips asset precompilation. It might be sufficient if build #28 passes. If build #28 shows staging is still too close to the 15min limit, then we should implement the hybrid fix above.