Files

T

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

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

6.7 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

브라우저 메모리 누수 탐지(Browser Memory Leak Detection)

매 한 줄

"매 SPA 가 시간이 지날수록 느려진다면 90% memory leak". JS 의 GC 는 매 reachable object 만 살리지만, detached DOM, forgotten timer, closure capture, global accumulation 이 매 unintended retention 을 만든다. 매 Chrome DevTools Memory panel + Performance panel 의 heap snapshot diff + allocation timeline 이 매 standard tool.

매 핵심

매 4 가지 흔한 leak pattern

Detached DOM: 매 element 를 removeChild 했지만 JS reference 가 살아있음 → DOM tree + element subtree 통째로 retained.
Forgotten timer / listener: setInterval / addEventListener 의 callback closure 가 매 component scope 를 hold.
Closure over big data: outer function 의 large array 를 inner closure 가 무의식적으로 capture.
Global accumulation: window.cache[key] = ... 식의 무한 적재.

매 detection workflow (3 snapshot technique)

App load → snapshot A.
매 leak-suspect action 100 회 (open/close modal …) → snapshot B.
매 동일 action 100 회 더 → snapshot C.
Comparison: B → C 사이 Delta 가 0 이어야 정상. 매 keep growing → leak.

매 응용

SPA debugging (React/Vue route transition leak).
Memory regression CI (puppeteer + heap snapshot).
Production monitoring (performance.memory).

💻 패턴

Detect detached DOM (DevTools console)

// 매 Memory panel → "Detached" filter 로 search
// 또는 console 에서:
function findDetached() {
  const all = $$('*'); // visible
  // queryObjects(HTMLElement) — DevTools console 전용
  console.log('visible:', all.length);
}
queryObjects(HTMLElement); // 매 DevTools console only

Cleanup pattern (idiomatic)

// React useEffect cleanup
useEffect(() => {
  const id = setInterval(tick, 1000);
  const onResize = () => recalc();
  window.addEventListener('resize', onResize);
  return () => {
    clearInterval(id);
    window.removeEventListener('resize', onResize);
  };
}, []);

WeakMap / WeakRef for caches

// 매 strong ref → leak. WeakMap → key GC 되면 entry 자동 사라짐
const meta = new WeakMap();
function attach(node, info) { meta.set(node, info); }
// node 가 detached 되고 reference 끊기면 entry GC

// WeakRef + FinalizationRegistry (2021+)
const reg = new FinalizationRegistry(id => console.log('GC:', id));
const w = new WeakRef(obj);
reg.register(obj, 'my-resource');

performance.memory monitoring (Chrome only)

function logMemory() {
  const m = performance.memory;
  console.log({
    used:   (m.usedJSHeapSize  / 1048576).toFixed(1) + 'MB',
    total:  (m.totalJSHeapSize / 1048576).toFixed(1) + 'MB',
    limit:  (m.jsHeapSizeLimit / 1048576).toFixed(1) + 'MB',
  });
}
setInterval(logMemory, 5000);

3-snapshot diff (puppeteer CI)

import puppeteer from 'puppeteer';
const browser = await puppeteer.launch();
const page = await browser.newPage();
await page.goto('http://localhost:3000');

const session = await page.target().createCDPSession();
async function snapshot() {
  await session.send('HeapProfiler.collectGarbage');
  const m = await page.evaluate(() => performance.memory.usedJSHeapSize);
  return m;
}

const a = await snapshot();
for (let i = 0; i < 100; i++) await page.click('#open-modal'), await page.click('#close-modal');
const b = await snapshot();
for (let i = 0; i < 100; i++) await page.click('#open-modal'), await page.click('#close-modal');
const c = await snapshot();

const growth = c - b;
if (growth > 1_000_000) throw new Error(`Memory leak: +${growth} bytes between B→C`);

AbortController for fetch / event cleanup

const ac = new AbortController();
window.addEventListener('scroll', onScroll, { signal: ac.signal });
fetch(url, { signal: ac.signal });
// component unmount:
ac.abort(); // 매 listener + fetch 동시 cleanup

Common React leak — stale closure capture

// 매 BAD — bigData 를 closure 가 잡음, unmount 후에도 retain
useEffect(() => {
  const bigData = computeMassive();
  const id = setTimeout(() => console.log(bigData[0]), 60_000);
  // cleanup 없음 → leak
}, []);

// 매 GOOD
useEffect(() => {
  const bigData = computeMassive();
  const id = setTimeout(() => console.log(bigData[0]), 60_000);
  return () => clearTimeout(id);
}, []);

매 결정 기준

상황	Tool
ad-hoc 탐색	DevTools Memory → Heap snapshot
allocation 출처 추적	Performance → record + Memory checkbox
detached DOM	Memory → "Detached" filter
CI regression	puppeteer + `HeapProfiler.collectGarbage` + `performance.memory`
production monitoring	`performance.memory` polling → telemetry

기본값: 매 3-snapshot heap diff (A=load, B=after-100-actions, C=after-200-actions). B→C 0 growth 가 합격선.

🔗 Graph

부모: Browser_Performance · Memory_Management
변형: Heap_Snapshot · Allocation_Profiling · Detached_DOM
응용: React_Memory_Leak · SPA_Performance
Adjacent: WeakMap · WeakRef · FinalizationRegistry · AbortController · 렌더링 최적화 개념 설명 자료

🤖 LLM 활용

언제: SPA 가 time-on-page 로 점점 느려질 때, mobile OOM, infinite scroll page 의 retention. 언제 X: server-side memory leak (그건 Node heap profiler 영역).

❌ 안티패턴

GC 강제 호출 시도: 매 gc() 는 standard 아님. DevTools "Collect garbage" 버튼이 유일.
performance.memory 만 보고 판단: 매 Chromium 전용, Firefox/Safari 없음. heap snapshot 이 ground truth.
removeChild 만 하고 reference cleanup 안 함: 매 detached DOM leak 의 source.
setInterval 매 component mount 마다 + cleanup 없음: 매 SPA killer.

🧪 검증 / 중복

Verified (Chrome DevTools docs, web.dev memory profiling, V8 blog).
신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — leak patterns + 3-snapshot diff + WeakRef/AbortController patterns

6.7 KiB Raw Blame History