Files

T

Antigravity Agent 504fd5fb42 [G1-Sync] Manual knowledge update

2026-05-10 22:08:15 +09:00

6.9 KiB

Raw Blame History

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

title

Memory Leak Debugging in JavaScript

매 한 줄

"매 unintended retention — 매 GC 매 reach 가능한 reference chain 매 끊지 못해 매 heap 매 grows unbounded". JS 매 mark-and-sweep GC 자동이지만 매 closure/listener/global/timer 매 long-lived reference 매 object lifecycle 매 의도와 분리시키면 매 leak 발생, 매 Chrome DevTools Heap Snapshot 매 진단 standard.

매 핵심

매 leak sources (top 5)

Detached DOM nodes: 매 element removed from tree 매 JS reference 잔존.
Event listeners: 매 addEventListener 매 removeEventListener 없이 매 component unmount.
Timers: setInterval/setTimeout 매 cleanup 누락 매 closure 매 모두 retain.
Closures: outer scope variables 매 inner function 매 capture 후 매 long-lived.
Global accumulation: window/globalThis 매 cache/array 매 unbounded push.

매 detection tools

Chrome DevTools Memory: Heap snapshot, allocation timeline, allocation sampling.
performance.measureUserAgentSpecificMemory() (Chrome 89+): 매 cross-origin isolated context.
Node.js: --inspect + Chrome DevTools, heapdump module, --heap-prof flag.
WeakRef + FinalizationRegistry: 매 GC 관찰 (debugging only).

매 응용

SPA 매 route navigation 매 retain leak 진단.
Long-running dashboard 매 hour-scale leak 감시.
Node.js server 매 RSS growth 매 root cause.
React/Vue component lifecycle leak detection.

💻 패턴

Heap snapshot 3-snapshot technique

1. App 초기 load → Snapshot 1 (baseline)
2. Suspect action 수행 (modal open/close ×10) → Snapshot 2
3. 동일 action 재수행 → Snapshot 3
4. Snapshot 3 의 Comparison → Snapshot 1
5. "Allocated between snapshots 1 and 3" 의 still-alive objects = leak

Detached DOM 탐색 (DevTools Console)

// Heap snapshot Class filter:
//   "Detached HTMLDivElement"
//   "Detached HTMLElement"
// 매 instance 매 retainer chain 매 inspect — 매 root retainer 매 leak 출처

Event listener leak — fix pattern

// 매 BAD
class Widget {
  constructor() {
    window.addEventListener('resize', this.onResize.bind(this));
  }
  onResize() { /* ... */ }
  destroy() { /* listener still attached */ }
}

// 매 GOOD
class Widget {
  constructor() {
    this.onResize = this.onResize.bind(this);
    window.addEventListener('resize', this.onResize);
  }
  onResize() { /* ... */ }
  destroy() {
    window.removeEventListener('resize', this.onResize);
  }
}

AbortController 매 modern cleanup

class Component {
  constructor() {
    this.ac = new AbortController();
    const { signal } = this.ac;
    window.addEventListener('scroll', this.onScroll, { signal });
    window.addEventListener('resize', this.onResize, { signal });
    fetch('/api', { signal });
  }
  destroy() {
    this.ac.abort(); // 매 모든 listener + fetch 매 한 번에 cleanup
  }
}

Timer leak fix

// 매 BAD — closure captures large data
function startPolling(bigData) {
  setInterval(() => {
    console.log(bigData.length); // bigData retained forever
  }, 1000);
}

// 매 GOOD — explicit handle + cleanup
const handle = setInterval(poll, 1000);
function stop() { clearInterval(handle); }

WeakMap 매 cache without leak

// 매 BAD — Map 매 key 매 GC X
const cache = new Map();
function getMeta(node) {
  if (!cache.has(node)) cache.set(node, computeMeta(node));
  return cache.get(node); // node removed from DOM but still in cache
}

// 매 GOOD — WeakMap key 매 GC 가능
const cache = new WeakMap();
function getMeta(node) {
  if (!cache.has(node)) cache.set(node, computeMeta(node));
  return cache.get(node);
}

performance.measureUserAgentSpecificMemory

// crossOriginIsolated context (COOP+COEP headers) 필요
if (crossOriginIsolated && performance.measureUserAgentSpecificMemory) {
  const result = await performance.measureUserAgentSpecificMemory();
  console.log('bytes:', result.bytes);
  console.table(result.breakdown);
}

FinalizationRegistry 매 GC 관찰 (debug only)

const registry = new FinalizationRegistry((tag) => {
  console.log(`GC'd: ${tag}`);
});

class Tracked {
  constructor(name) {
    registry.register(this, name);
  }
}

new Tracked('widget-1'); // → "GC'd: widget-1" eventually (or never)

Node.js heap snapshot

node --inspect server.js
# 매 chrome://inspect → Memory → Take heap snapshot
# 또는 programmatic:

import { writeHeapSnapshot } from 'node:v8';
const path = writeHeapSnapshot(); // .heapsnapshot file
console.log(`Snapshot: ${path}`);

매 결정 기준

상황	Tool/Approach
Browser SPA growing memory	DevTools Heap Snapshot 3-snapshot
매 frame allocation hotspot	Allocation timeline (sampling)
Detached DOM 의심	Class filter "Detached " in snapshot
Node.js RSS growth	writeHeapSnapshot + Chrome DevTools
Continuous monitoring (production)	performance.measureUserAgentSpecificMemory
Event listener leak	AbortController 매 unified cleanup

기본값: 매 Heap Snapshot 3-snapshot diff — 매 retainer chain 매 따라 root 매 식별.

🔗 Graph

부모: JavaScript-Performance · Garbage-Collection
변형: Node-js-Memory-Profiling · V8-Heap-Analysis
응용: SPA-Performance · Long-Running-Apps
Adjacent: Chrome-DevTools · WeakMap · AbortController · FinalizationRegistry

🤖 LLM 활용

언제: 매 SPA/long-running app 의 메모리 증가, 매 unmount 후 referent 잔존, 매 production memory metrics 의 anomaly. 언제 X: 매 short-lived script (CLI tool), 매 GC pause 문제 (different — GC tuning territory).

❌ 안티패턴

delete keyword 의존: 매 reference 매 nullify 안 함 — 매 다른 reference 매 retain.
window.gc() 매 production: 매 only with --expose-gc flag, 매 hint 일 뿐.
Allocation timeline 매 production trace: 매 overhead 매 큼 — 매 staging 에서.
One-snapshot 진단: 매 baseline 없으면 매 noise 와 leak 매 구분 불가.
DevTools 매 incognito 가정: 매 extension 매 heap pollution — 매 incognito + 매 disabled extensions.

🧪 검증 / 중복

Verified (Chrome DevTools docs, V8 blog, Node.js v8 module).
신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — leak source taxonomy + DevTools workflow + AbortController/WeakMap patterns

6.9 KiB Raw Blame History Unescape Escape