E2E Test Report – pdf-engine

Report ID: TEST-20260409-PDF Date: 2026-04-09 Service: pdf-engine (ods-platform) Operator: ADLC Pipeline (automated) + manual validation Session: First E2E after 9-day outage (BUG-003 recovery)

1. Executive Summary

pdf-engine is back online after the BUG-003 outage (2026-03-31 to 2026-04-09). Core functionality is healthy: template CRUD, job creation, authentication, and multi-tenant isolation all pass. One new bug discovered: BUG-004 – duplicate template names are accepted (HTTP 201) when the spec requires HTTP 409 Conflict. One test was blocked due to a test script crash unrelated to the service.

Critical Finding

BUG-004: Duplicate template name returns 201 instead of 409. The database unique index uq_templates_tenant_name_active exists but the Rust handler does not catch the PostgreSQL error code 23505 (unique_violation). This is a data integrity issue that allows ambiguous template names per tenant.

2. Test Environment

3. Test Results

Result Breakdown

• Templates CRUD: 5/6 PASS (1 FAIL is the duplicate name constraint bug)
• Jobs CRUD: 2/3 PASS (1 BLOCKED is test infrastructure, not service)
• Auth: 1/1 PASS
• Multi-tenancy: 1/1 PASS
• Health: 1/1 PASS

4. BUG-004: Duplicate Template Name Accepted

Summary

Description

When creating a template with a name that already exists for the same tenant, the API returns HTTP 201 (Created) and inserts a duplicate row. The spec requires HTTP 409 (Conflict).

Reproduction Steps

1. Authenticate with a valid OID JWT for tenant A.
2. Create a template: POST /v1/templates with {"name": "E2E-Final-113903", "content_type": "html", "body_html": "<p>original</p>"} – returns 201.
3. Create the same template again: POST /v1/templates with {"name": "E2E-Final-113903", "content_type": "html", "body_html": "<p>dup</p>"} – returns 201 (expected: 409).
4. Two templates now exist with the same name for the same tenant.

Expected vs Actual

Root Cause

The database unique index uq_templates_tenant_name_active ON pdf.templates (tenant_id, name) WHERE deleted_at IS NULL exists in the schema. However, the create_template handler in src/api/templates.rs does not match on sqlx::Error::Database with error code 23505 (unique_violation). The constraint error is either swallowed by a generic error handler or mapped to HTTP 500, which then gets caught as a different path.

Impact

• Multiple templates with the same name per tenant can exist
• Template selection by name becomes ambiguous
• PDF generation workflows that reference templates by name may produce inconsistent results
• Data cleanup required for any duplicates created in the interim

Recommended Fix

In the create_template handler, catch the PostgreSQL unique violation:

Err(sqlx::Error::Database(e)) if e.code() == Some(Cow::Borrowed("23505")) => {
    HttpResponse::Conflict().json(json!({
        "error": "conflict",
        "message": "Template with this name already exists"
    }))
}

Estimated effort: ~10 lines across src/api/templates.rs and src/repository/template_repo.rs.

Historical Context

This bug was previously identified as BUG-002 in the automated E2E suite run on 2026-03-26 (scenarios TPL-VAL-09 and TPL-VAL-10 failed with the same symptom: status 201 instead of 409). It was not prioritized during the BUG-003 outage recovery. The 2026-04-09 manual E2E session confirms the bug persists and has been reclassified as BUG-004 at HIGH severity.

5. Recovery Context: BUG-003 Outage

Timeline

BUG-003 Resolution Summary

• Root cause: sqlx connection string parameter ?search_path=pdf is silently ignored. The migrator defaulted to the public schema and panicked when migration 007 attempted to create a table that already existed.
• Fix: Added after_connect callback in the connection pool configuration to run SET search_path = pdf on every new connection.
• Verification: Migration 007 now applies correctly. All template and job tables exist in the pdf schema. Database has 0 duplicate entries.

Lessons Learned

1. PostgreSQL search_path must be set explicitly via after_connect, not via URL parameter, when using sqlx with schema isolation.
2. Container healthcheck requires curl in the Docker image – missing curl caused Coolify healthcheck failures and masked the actual migration error.
3. The is_preview env var must be set correctly for Coolify deployments to avoid incorrect build behavior.

6. Comparison with Previous Sessions

Key improvements since 2026-03-26: - Auth upgraded from HS256 to RS256 (aligned with OID) - BUG-003 (migration panic) resolved - CORS configuration fixed (was returning 400 on OPTIONS) - Container healthcheck operational

Remaining from 2026-03-26: - BUG-004 (duplicate name) still open - PDF manipulation endpoints (split, merge, rotate) still not implemented (P1+ scope) - FK violation error message not sanitized to spec wording

7. Blocked Test: TC-PDF-010

8. Recommendations

Immediate (before next E2E session)

1. Fix BUG-004 – Add 23505 error code handling in create_template handler. Estimated 10 lines. This is a data integrity issue and should be resolved before any production promotion.

2. Fix test script – TC-PDF-010 blocked due to variable propagation failure. Not a service bug but must be fixed for complete test coverage.

Short-term

3. Rerun full 69-scenario automated suite – The last automated run was 2026-03-26 with 68.1% pass rate. Many failures were related to BUG-003 (now fixed) and test harness issues. A rerun will establish the true current baseline.

4. Audit for duplicate templates – Query pdf.templates for any rows with duplicate (tenant_id, name) pairs created during the period when BUG-004 was active. Clean up if found.

5. Enable Coolify healthcheck – DevOps review noted health_check_enabled=False. Must be corrected before production deployment.

Medium-term

6. Implement PDF manipulation endpoints – Split, merge, rotate APIs are specified but not yet implemented (P1 scope). These accounted for most failures in the 2026-03-26 automated suite.

7. Set container memory limits – DevOps review noted no memory limits configured in Coolify.

9. Evidence Locations

Report generated 2026-04-09 by documentarian agent. Next action: fix BUG-004, then rerun full automated E2E suite.