Files

T

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

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

6.6 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

Software Maintenance

매 한 줄

"매 software maintenance 의 corrective + adaptive + perfective + preventive 의 4 categories (ISO/IEC 14764)". 매 lifecycle cost 의 60-80% 의 maintenance — 매 dev 의 minority 의 only feature 의 build. 매 2026 의 game-changer 의 AI-aided refactor / debug (Claude Opus 4.7, Cursor, Sourcegraph Cody) + automated dep update (Renovate / Dependabot).

매 핵심

매 4 categories (ISO/IEC 14764)

Corrective: 매 bug fix — 매 reactive.
Adaptive: 매 환경 변화 (OS, runtime, API breaking change).
Perfective: 매 performance / readability / structure 의 improve.
Preventive: 매 future fault 의 anticipate (refactor, test, doc).

매 Maintenance 의 challenges

Knowledge erosion: 매 original author 의 leave → 매 dark code.
Dep rot: 매 transitive CVE / EOL.
Code rot: 매 entropy 의 increase.
Test fragility: 매 flaky 의 trust 의 erode.

매 Modern (2026) techniques

AI refactor / explain: Claude / Cursor 의 large diff 의 understand.
Automated dep: Renovate, Dependabot — 매 PR 의 auto + test 의 auto-merge.
Continuous profiling: pprof / Pyroscope / Parca 의 perf regression.
Code archaeology: git-blame + AI 의 historical context.
Automated migration: codemod (jscodeshift, Bowler), Ruff fix, OpenRewrite (Java).

매 응용

CVE patching pipeline.
Major version migration (Python 3.11 → 3.13, React 18 → 19).
Dead-code / dep prune.
Test backfill (untested legacy).

💻 패턴

Renovate config (auto dep update)

{
  "$schema": "https://docs.renovatebot.com/renovate-schema.json",
  "extends": ["config:recommended", ":dependencyDashboard"],
  "schedule": ["before 6am on Monday"],
  "automerge": true,
  "automergeType": "pr",
  "packageRules": [
    { "matchUpdateTypes": ["major"], "automerge": false },
    { "matchPackagePatterns": ["^@types/"], "automerge": true, "groupName": "types" }
  ],
  "vulnerabilityAlerts": { "labels": ["security"], "automerge": false }
}

AI-assisted refactor (Claude Code)

# 매 large legacy file 의 refactor — 매 small steps + tests
claude refactor src/legacy.py \
  --strategy "extract pure functions, add type hints, preserve behavior" \
  --test "pytest tests/test_legacy.py"

Codemod (jscodeshift, callback → async)

// transform.js
module.exports = function(file, api) {
  const j = api.jscodeshift;
  return j(file.source)
    .find(j.CallExpression, { callee: { name: "fs.readFile" } })
    .replaceWith(p => {
      const [path, cb] = p.node.arguments;
      return j.awaitExpression(
        j.callExpression(j.memberExpression(j.identifier("fs"), j.identifier("promises.readFile")), [path])
      );
    })
    .toSource();
};
// jscodeshift -t transform.js src/

OpenRewrite recipe (Java migration)

# rewrite.yml
type: specs.openrewrite.org/v1beta/recipe
name: com.acme.UpgradeJava21
recipeList:
  - org.openrewrite.java.migrate.UpgradeToJava21
  - org.openrewrite.java.migrate.UpgradeBuildToJava21
  - org.openrewrite.java.format.AutoFormat

Continuous profiling (Pyroscope)

import pyroscope
pyroscope.configure(
    application_name="payment-svc",
    server_address="http://pyroscope:4040",
    tags={"env": "prod", "version": "1.4.2"},
)

# 매 deploy 의 pre / post flame graph 의 diff 의 perf regression 의 catch

CVE patching pipeline

# .github/workflows/cve.yml
name: cve-patch
on: { schedule: [{ cron: "0 6 * * 1" }] }
jobs:
  scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: trivy fs --severity HIGH,CRITICAL --exit-code 1 .
      - run: |
          gh issue create --title "CVE scan failures $(date -I)" \
            --body-file trivy-report.txt \
            --label security

Dead-code detection (knip / unimported)

# 매 JS/TS 의 dead module / export
npx knip
# 매 Python
vulture src/ --min-confidence 80
# 매 Java
jdeps --print-module-deps build/libs/*.jar

Mutation testing (perfective)

# 매 Python: mutmut, 매 JS: stryker
# 매 test 의 actually catch 의 measure
mutmut run --paths-to-mutate=src/
mutmut html
# 매 score < 70% 의 의 test gap.

AI-explain (legacy code archaeology)

import anthropic
client = anthropic.Anthropic()

def explain(file_path: str):
    src = open(file_path).read()
    blame = subprocess.check_output(["git", "log", "-p", file_path]).decode()[:30000]
    msg = client.messages.create(
        model="claude-opus-4-7",
        max_tokens=2000,
        system="You are a senior engineer explaining legacy code. Cover: purpose, key invariants, hazards, refactor priorities.",
        messages=[{"role": "user", "content": f"FILE:\n{src}\n\nGIT LOG:\n{blame}"}],
    )
    return msg.content[0].text

매 결정 기준

상황	Action
Active CVE	Renovate auto-PR + emergency merge
Major version	OpenRewrite / codemod + staged rollout
Slow function	Continuous profiling + targeted fix
Untested legacy	Characterization tests first, refactor after
Dark code	AI-explain + pair with original author log

기본값: Renovate + Trivy + Pyroscope + AI-assist as standard stack.

🔗 Graph

부모: SDLC · Software-Engineering
변형: Refactoring · Technical-Debt · Code-Migration
응용: CVE-Patching · Continuous-Profiling · Codemod
Adjacent: Renovate · OpenRewrite · AI-Augmented-Review

🤖 LLM 활용

언제: legacy explain, refactor proposal, doc generate, codemod 의 draft. 언제 X: prod hot-fix without test (always test first), security patch 의 unattended merge.

❌ 안티패턴

Big-bang refactor: 매 small steps + green test 의 always.
No characterization test: 매 refactor 의 silent break.
Manual dep update: 매 Renovate / Dependabot 의 always.
Hot-fix to main without rollback: 매 PR + test + tag 의 keep.

🧪 검증 / 중복

Verified (ISO/IEC 14764, Renovate, OpenRewrite, Pyroscope docs).
신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — full categories + AI-aided modern stack

6.6 KiB Raw Blame History