tarball Size $GITHUB_STEP_SUMMARY Output — Sprint 165 Option C Operational Visibility (Seed #165-1)

Goal

• Recover from zero operational visibility after Sprint 165 Option C (buildx tarball + trivy --input) was established
• Unable to quantify cost of 8 image tarball upload/download per PR → resolved by adding $GITHUB_STEP_SUMMARY output step to 3 build jobs
• Establish baseline data for future seed #165-2 zstd compression comparison
• User selection A: seed #165-1 standalone recovery (safe incremental improvement), #165-2 deferred to Sprint 167+ after baseline established

Decisions

D0. Seed #165-1 Standalone Recovery — User Selection A (Safe Incremental Improvement)

• A adopted among Sprint 166 scope candidates A~D (user decision):
  • A (seed #165-1 standalone) / B (#165-1 + Sprint 164 #new-6 integrated) / C (#165-1 + #165-2 zstd measurement) / D (UAT visual verification first)
• Rationale: zstd comparison (#165-2) is meaningless without baseline data — measuring first then comparing next sprint is the correct order
• Sprint 165 Option C established (security gate) → Sprint 166 (visibility) → Sprint 167 (optimization comparison): step 2 of 3-sprint cycle

D1. tarball Size Only for First Introduction — cache hit/miss Info Deferred to Sprint 167+

• docker/build-push-action@v7 metadata output (containerimage.buildinfo) does not directly expose cache hit/miss
• Separate step using buildx du or imagetools inspect is needed — out of scope for this sprint
• stat -c %s byte size alone is sufficient to establish PR artifact upload/download cost baseline

D2. awk -v b="$SIZE_BYTES" Variable Passing Pattern — Shell Injection Prevention

• Adopted explicit awk -v key="$VAR" passing instead of relying on double-quote expand awk "BEGIN {... $SIZE_BYTES ...}"
• Double-quote pattern is only safe when SIZE_BYTES is an integer — awk syntax error possible if value format changes later
• awk -v is the standard safe pattern that clearly separates shell/awk boundaries

D3. stat -c %s GNU coreutils — ubuntu-latest Environment Guaranteed

• stat -c %s is GNU coreutils-specific (macOS uses stat -f %z)
• GitHub Actions ubuntu-latest guarantees GNU coreutils → platform difference irrelevant
• CI-only step, so macOS local execution mismatch has no cost

D4. Same Pattern for 3 Build Jobs (matrix vs hardcoding only difference) — DRY Naturally Applied

• build-services: uses ${{ matrix.service }} variable + if: steps.check.outputs.skip == 'false' guard
• build-frontend: /tmp/image-frontend.tar hardcoded (no skip guard — job itself has no skip guard)
• build-blog: /tmp/image-blog.tar hardcoded (no skip guard)

Implementation (1 PR, 34-Sprint Consecutive Branch Discipline)

Branch: feat/sprint-166-tarball-size-summary (new branch from main c0c48aa)

Phase A — ci.yml 3 Step Addition (d172eea)

Insert "Report tarball size" step immediately after Upload image tarball for Trivy scan in each of 3 build jobs (+41 lines):

• line 690 (build-services, matrix + skip guard), line 746 (build-frontend, hardcoded), line 824 (build-blog, hardcoded)

- name: Report tarball size
  run: |
    TARBALL=/tmp/image-frontend.tar   # build-services uses ${{ matrix.service }}, others hardcoded
    SIZE_BYTES=$(stat -c %s "$TARBALL")
    SIZE_MB=$(awk -v b="$SIZE_BYTES" 'BEGIN {printf "%.1f", b/1024/1024}')
    {
      echo "### 📦 frontend build artifact"
      echo "- tarball size: **${SIZE_MB} MB** (${SIZE_BYTES} bytes)"
      echo "- path: \`$TARBALL\`"
      echo "- retention: 1 day"
    } >> "$GITHUB_STEP_SUMMARY"

Phase B — ADR Record (This Commit)

• docs/adr/sprints/sprint-166.md (KR) + docs/adr-en/sprints/sprint-166.md (EN 1:1 mapping)
• docs/adr/README.md count 105→106, range 62165→62166 (lines 18/52/54)

Risk/Regression Blocking

Prediction 1: stat Platform Difference

stat -c %s (GNU) vs stat -f %z (macOS) difference → ubuntu-latest guarantees GNU coreutils. CI-only step, so local mismatch has no cost.

Prediction 2: $GITHUB_STEP_SUMMARY Size Limit 1 MiB/job

This step adds ~200 bytes per job → irrelevant to limit (1 MiB). CI standard env var, automatically injected in all steps.

Prediction 3: Step Execution Without tarball Generated

build-services: if: steps.check.outputs.skip == 'false' guard activates step only after build-push-action runs. frontend/blog: tarball only exists when build-push-action succeeds → always safe at step arrival point.

Verification

• Local: python3 yaml.safe_load PASS (YAML OK), node scripts/check-adr-en-coverage.mjs --strict 115/115 PASS, node scripts/check-doc-refs.mjs 0 broken refs
• CI (PR stage): build-services 6 matrix + build-frontend + build-blog job Summaries display ### 📦 {service} build artifact H3 + size MB (new UAT item 1)
• trivy-scan 8 matrix no regression (--input tarball matching unchanged)

Results

4 changed files:

• 1 modified: .github/workflows/ci.yml +41 lines (3 steps added: build-services/frontend/blog)
• 2 new: docs/adr/sprints/sprint-166.md (KR) + docs/adr-en/sprints/sprint-166.md (EN 1:1 mapping)
• 1 modified: docs/adr/README.md (lines 18/52/54 — count 105→106, range 62165→62166)

New Patterns

• Build job Summary visibility pattern — standard monitoring pattern: add $GITHUB_STEP_SUMMARY output step immediately after build artifact creation → upload. Applies uniformly to both matrix and hardcoded jobs. Sprint 165 Option C established → immediate visibility follow-up step → step 1 of operations-observation-based optimization cycle (seed #165-2)
• awk -v variable passing = shell safety standard pattern — explicit awk -v key="$VAR" passing instead of double-quote expand dependency. Clarifies shell/awk boundary in yaml run scripts. Established as standard pattern for future CI yaml script writing
• Baseline data acquisition first + comparison forward-fix split — seed #165-2 zstd comparison is meaningless without baseline. This sprint collects data → next sprint compares. Regression isolation via split + measurement value first pattern