Staging

[BrdH7940]

No description provided.

[bitdive-review]

BitDive Runtime Review - PR #1606

PR: #1606
Branch: staging -> main
Method: BitDive runtime trace comparison + direct HTTP verification + targeted code inspection
Verdict: APPROVE WITH NOTES
Confidence: High — two independent iterations converge; the only production Java change is a compiler annotation

Summary

This 264-file PR is overwhelmingly CI/CD pipeline infrastructure, K8s Helm charts, test additions, and frontend TypeScript. The only production Java source change is a @SuppressWarnings("unchecked") annotation on SecurityConfig.userAuthoritiesMapperForKeycloak() in backoffice-bff — a compiler directive with zero runtime impact. Root pom.xml downgrades the JDK target from 25 to 21 and adds JaCoCo coverage.

Three before/after trace pairs on deployed services (product, location, inventory) show zero contract divergence with clean baselines — identical SQL, REST, writes, data payloads, and execution trees. All four deployed services compile on JDK 21 and serve HTTP 200. Two consecutive review iterations converge on the same verdict with no new findings.

Runtime Verification Matrix

Area	Result	Behavior Δ	Meaning	Evidence
Product listing	Stable	HTTP 200→200 · SQL 3→3 · REST 1→1 · writes 0→0 · zero contract drift	No behavioral change; product catalog serves identically	before / after
Location countries	Stable	HTTP 200→200 · SQL 1→1 · REST 0→0 · writes 0→0 · 249→249 nodes	No behavioral change; full country list (249 entries) identical	before / after
Inventory warehouses	Stable	HTTP 200→200 · SQL 1→1 · REST 0→0 · writes 0→0 · 3→3 nodes	No behavioral change; warehouse list (3 entries) identical	before / after
Media API docs	Stable	HTTP 200→200 (framework endpoint)	Springdoc v3/api-docs serves; no JVM tree expected	HTTP-only
Maven JDK 21 build	Stable	4 services compiled · Docker images built · containers started	JDK 21 downgrade compiles all deployed services	Code-only
CI/CD & K8s charts	Stable	YAML valid · images parameterized · JDK inconsistency noted	Workflows and Helm charts structurally sound	Code-only

How to read BitDive evidence

Open an Evidence share link. The Behavior Δ / diff blocks summarize the contract change; the share link is the full proof (call tree · SQL · writes · downstream · first divergence). GitHub strips target="_blank" — links open in the same tab.

• Trace pair — full call trees compared before/after
• HTTP only — status/body verified; no JVM tree
• Code only — static inspection, no runtime trace

Review scope and validation coverage

Deployed services (JDK 21, Docker-compose local)
  -> product   -> ProductController.getProductsByMultiQuery()   [trace pair ✓]
  -> location  -> CountryController.getAllCountries()           [trace pair ✓]
  -> inventory -> WarehouseController.getWarehouses()           [trace pair ✓]
  -> media     -> Springdoc v3/api-docs                         [HTTP only]

Code-only verification
  -> backoffice-bff -> SecurityConfig @SuppressWarnings          [compiler annotation, zero runtime impact]
  -> CI/CD          -> 29+ workflow files, JDK 21 vs 25 audit    [static analysis]
  -> K8s charts     -> deployment strategy, probes, resources    [static analysis]
  -> Frontend       -> backoffice/storefront TypeScript          [out of scope: not JVM-traceable]

Behavior	Scenario	Evidence	Status
Product catalog query	GET product list (storefront)	Trace pair	Stable
Country list	GET countries (storefront)	Trace pair	Stable
Warehouse list	GET warehouses (backoffice, auth)	Trace pair	Stable
Media API docs	GET v3/api-docs	HTTP only	Stable
JDK 21 compilation	Maven build → Docker → serve	Code	Stable
CI/CD & K8s	Workflow + chart review	Code	Stable

Out of scope: Frontend TypeScript/React changes (60+ files), test file additions, delivery/search/sampledata (not deployed in this environment). CI/CD pipeline YAML is the primary PR intent — human review recommended.

Change #1 — stable deploy canaries (no collateral regression)

Three before/after trace pairs independently verified. All show clean baselines, zero contract divergence, zero write delta, zero downstream delta. The only production Java change in this PR is a @SuppressWarnings("unchecked") compiler annotation — it cannot produce runtime behavioral differences.

Behavior contract

# Deploy canary scenarios — runtime contract (3 services paired)
- HTTP 200 OK (product 1195ms, location 437ms, inventory 305ms)
+ HTTP 200 OK (product 30ms, location 131ms, inventory 7ms)
- SQL: product 3 reads, location 1 read, inventory 1 read
+ SQL: product 3 reads, location 1 read, inventory 1 read  (identical)
- REST: product 1 (MediaService.getMedia), location 0, inventory 0
+ REST: product 1 (MediaService.getMedia), location 0, inventory 0  (identical)
- writes: 0 across all three services
+ writes: 0 across all three services  (identical)
- data: iPhone 15 ($799), 249 countries, 3 warehouses (WH With Stock, BRR Base, BRR Head)
+ data: iPhone 15 ($799), 249 countries, 3 warehouses  (identical)
# first divergence: MediaService.getMedia() response date header (timestamp only — no contract impact)

Scenario matrix

Scenario	Endpoint	Before PR	After PR	Result
Product catalog	GET /product/storefront/products	200, 1195ms, 4 nodes	200, 30ms, 4 nodes	Zero drift
Country list	GET /location/storefront/countries	200, 437ms, 249 nodes	200, 131ms, 249 nodes	Zero drift
Warehouse list	GET /inventory/backoffice/warehouses	200, 305ms, 3 nodes	200, 7ms, 3 nodes	Zero drift

Trace evidence

Trace	Scenario	HTTP	SQL / REST	Result
before	Product	200	3 SQL, 1 REST	Clean
after	Product	200	3 SQL, 1 REST	Clean
before	Location	200	1 SQL	Clean
after	Location	200	1 SQL	Clean
before	Inventory	200	1 SQL	Clean
after	Inventory	200	1 SQL	Clean

Contract delta

Layer	Before PR	After PR	Result
HTTP	200 all 3 services	200 all 3 services	No drift
Persistence / SQL	product 3, location 1, inventory 1	identical	No drift
Downstream / REST	product 1 (MediaService), others 0	identical	No drift
Data / args	iPhone 15, 249 countries, 3 warehouses	identical	No drift
Steps / time	4 / 249 / 3 nodes (slower — cold JVM)	4 / 249 / 3 nodes (faster — warm JVM)	No drift

Key trace delta

First meaningful divergence: MediaService.getMedia() — HTTP date response header timestamp only. No SQL, write, data, or structural divergence.

Branch	Trace shape
Before PR	product 4 nodes (1195ms), location 249 nodes (437ms), inventory 3 nodes (305ms) — cooler JVM state
After PR	product 4 nodes (30ms), location 249 nodes (131ms), inventory 3 nodes (7ms) — warm JVM state

Timing difference is attributed to JVM warmup state on the baseline. Contract comparison confirms zero divergence despite timing variance.

Advisory findings (non-blocking)

Finding	Impact	Blocking
JDK version inconsistency: action.yaml uses JDK 21 (aligned with pom.xml) but new ci.yml / deploy-dev.yml / deploy-staging.yml use JDK 25	CI may compile with JDK 25 targeting 21 — works but should be aligned	No
GHCR registry namespace changed from nashtech-garage to 23122004 in new CI/CD workflows	Old per-service CI files still reference old namespace; K8s charts are parameterized so not runtime-breaking	No
MediaService.getMedia() returns REST status -1 on product traces	Pre-existing degraded media integration; same on both sides, handled gracefully by product service	No

Follow-Ups

Type	Item	Blocking
Advisory	Align JDK version: action.yaml = 21 but new CI workflows = 25 — pick one consistently	No
Advisory	Verify GHCR namespace 23122004 is intentional across all new CI workflows	No
Non-blocking	Human review of CI/CD pipeline YAML and K8s chart changes (primary PR intent)	No
Informational	Pre-existing MediaService.getMedia() REST -1 status present on both sides	No

Recommendation

Approve with notes. This is a CI/CD and infrastructure PR (264 files) with zero production runtime behavioral changes. The only production Java change is a @SuppressWarnings compiler annotation. Three before/after trace pairs independently verified show zero contract divergence with clean baselines. JDK 21 compilation succeeded for all four deployed services. Two consecutive review iterations converge on the same verdict. Advisory notes on JDK version alignment and GHCR namespace are recommended follow-ups for the PR author.