perf(pm): skip lifecycle reads when scripts are ignored

[elrrrrrrr]

Summary

• skip package lifecycle-script reads during rebuild when install runs with --ignore-scripts
• keep binary linking behavior unchanged
• add coverage that ignore mode does not populate lifecycle scripts

Hypothesis

p3/p4 benchmark runs use utoo install --ignore-scripts. Rebuild currently still reads package.json for packages marked with hasInstallScript, even though those scripts cannot run. Avoiding those reads should reduce warm-link filesystem work and async scheduling overhead.

Validation

• cargo fmt
• cargo test -p utoo-pm service::package::tests::test_collect_packages_from_lock_with_scripts
• cargo clippy -p utoo-pm --all-targets -- -D warnings --no-deps

Benchmark Plan

• GHA label-triggered pm-bench-phases, Linux/npmjs only
• Compare p3/p4 wall and vCtx/iCtx against perf(pm): add resolver demand bfs scheduler #2948 baseline and perf(pm): run install clone workers on rayon #2965

[gemini-code-assist]

Code Review

This pull request refines the logic for reading lifecycle scripts in the PackageService by ensuring they are only loaded when the script policy is set to Run and scripts are actually present. Additionally, it includes a new test assertion to verify that ignored scripts are correctly excluded from being read. I have no feedback to provide.

[github-actions]

📊 pm-bench-phases · 16146fb · linux (ubuntu-latest)

Workflow run — ant-design

PMs: utoo (this branch) · utoo-npm (latest published) · bun (latest)

npmjs.org

p0_full_cold

PM	wall	±σ	user	sys	RSS	pgMinor
bun	9.62s	0.35s	10.50s	10.08s	658M	303.8K
utoo-next	8.09s	0.31s	10.61s	12.19s	990M	128.5K
utoo-npm	7.93s	0.32s	10.77s	12.16s	983M	124.5K
utoo	7.91s	0.19s	11.48s	12.20s	968M	150.3K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	19.0K	19.2K	1.20G	7M	1.90G	1.77G	1M
utoo-next	128.6K	92.0K	1.17G	5M	1.73G	1.72G	2M
utoo-npm	127.5K	93.3K	1.17G	5M	1.73G	1.72G	2M
utoo	119.0K	92.2K	1.17G	6M	1.72G	1.72G	2M

p1_resolve

PM	wall	±σ	user	sys	RSS	pgMinor
bun	2.29s	0.05s	4.03s	1.20s	490M	191.7K
utoo-next	3.26s	0.05s	5.42s	2.32s	613M	79.9K
utoo-npm	3.33s	0.02s	5.42s	2.33s	610M	75.8K
utoo	2.52s	0.10s	6.08s	1.75s	663M	122.5K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	11.7K	4.1K	204M	3M	108M	-	1M
utoo-next	83.1K	91.4K	201M	3M	7M	3M	2M
utoo-npm	82.5K	96.4K	201M	3M	7M	3M	2M
utoo	16.8K	22.3K	204M	3M	7M	3M	2M

p3_cold_install

PM	wall	±σ	user	sys	RSS	pgMinor
bun	6.69s	0.23s	6.42s	9.82s	590M	202.7K
utoo-next	6.85s	1.17s	5.16s	10.56s	431M	58.1K
utoo-npm	6.03s	0.20s	5.25s	10.45s	496M	63.1K
utoo	6.74s	1.50s	5.25s	10.80s	495M	61.8K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	6.4K	8.3K	1.00G	4M	1.77G	1.77G	1M
utoo-next	108.8K	51.4K	999M	3M	1.72G	1.72G	2M
utoo-npm	94.9K	46.8K	999M	2M	1.72G	1.72G	2M
utoo	121.6K	79.0K	999M	3M	1.72G	1.72G	2M

p4_warm_link

PM	wall	±σ	user	sys	RSS	pgMinor
bun	3.42s	0.03s	0.20s	2.43s	135M	32.2K
utoo-next	2.52s	0.24s	0.50s	3.82s	80M	18.3K
utoo-npm	2.27s	0.09s	0.50s	3.78s	80M	18.2K
utoo	2.40s	0.08s	0.53s	3.86s	63M	14.3K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	276	23	5M	49K	1.93G	1.76G	1M
utoo-next	40.1K	18.3K	324K	15K	1.72G	1.72G	2M
utoo-npm	39.6K	17.5K	317K	12K	1.72G	1.72G	2M
utoo	52.8K	24.8K	317K	9K	1.73G	1.72G	2M

npmmirror.com: no output captured.

[elrrrrrrr]

GHA run 1 read

Run: https://github.com/utooland/utoo/actions/runs/26016519226

phase	utoo wall	utoo ctx	same-run utoo-next	same-run utoo-npm	same-run bun	read
p3_cold_install	6.74s ±1.50	121.6K / 79.0K	6.85s, 108.8K / 51.4K	6.03s, 94.9K / 46.8K	6.69s, 6.4K / 8.3K	no ctx win; iCtx regresses materially
p4_warm_link	2.40s ±0.08	52.8K / 24.8K	2.52s, 40.1K / 18.3K	2.27s, 39.6K / 17.5K	3.42s, 276 / 23	not positive; p4 ctx is worse than same-run next/npm

Conclusion: reject this single-factor experiment. Skipping lifecycle package.json reads under --ignore-scripts is correct-looking but not a measurable p3/p4 lever in this benchmark. Do not fold unless a later combined change proves it is needed for correctness rather than perf.

[elrrrrrrr]

Closing this PM performance experiment after the investigation phase. The benchmark data and conclusions are preserved in the PR body/comments; we will split the validated pieces into smaller reviewable PRs for the formal ship path.