fix(test): Windows Unicode encoding in integration tests

[danielmeppiel]

Problem

After fixing the teardown issue (PR #295), the Windows release pipeline progressed to the next test suite (guardrailing hero scenario) and hit a new failure:

UnicodeEncodeError: 'charmap' codec can't encode character '\u2713' in position 0

Windows uses cp1252 encoding for stdout by default, which cannot encode Unicode symbols like ✓ and ✨. Additionally, subprocess.run(text=True) defaults to cp1252 for decoding APM output, which contains Rich-formatted Unicode.

Fix

1. Replaced Unicode symbols (✓, ✨) with ASCII equivalents ([OK], etc.) in test print statements
2. Added encoding='utf-8' + errors='replace' to all subprocess calls that capture APM output — ensures Rich's Unicode output can be decoded on Windows

Files changed

• tests/integration/test_guardrailing_hero_e2e.py — print statements + subprocess encoding
• tests/integration/test_mcp_registry_e2e.py — print statements + subprocess encoding (preemptive — next suite after guardrailing)
• tests/integration/test_ado_e2e.py — subprocess encoding (preemptive)

Context

• Pipeline run 23087068353 —
• Partially addresses [FEATURE] Clean up tearDown patterns in unit tests (gc.collect + sleep) #185

[Copilot]

Pull request overview

Fixes Windows integration test failures caused by non-UTF-8 console/codepage defaults by making test output ASCII-safe and forcing UTF-8 decoding for captured subprocess output.

Changes:

• Replace Unicode status symbols in E2E test print() statements with ASCII equivalents.
• Add encoding='utf-8' and errors='replace' to subprocess.run(..., text=True) calls that capture APM output.
• Apply the same subprocess decoding hardening to additional E2E suites (preemptive).

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
tests/integration/test_guardrailing_hero_e2e.py	Forces UTF-8 decoding for subprocess output and replaces Unicode symbols in status prints.
tests/integration/test_mcp_registry_e2e.py	Forces UTF-8 decoding for subprocess output and replaces most Unicode status prints with ASCII markers.
tests/integration/test_ado_e2e.py	Forces UTF-8 decoding for captured subprocess output.

Comments suppressed due to low confidence (1)

tests/integration/test_mcp_registry_e2e.py:417

• There is still a non-ASCII warning symbol printed in this flow ("⚠ Copilot configuration not created..."). On Windows with a non-UTF-8 console encoding this can still raise UnicodeEncodeError, defeating the goal of replacing unicode symbols in test output. Replace it with an ASCII-only marker (e.g., "[WARN]") to keep this suite Windows-safe.

                    print(f"[OK] Copilot configuration tested with {len(mcp_servers)} servers")
                    
                except json.JSONDecodeError as e:
                    pytest.fail(f"Invalid JSON in Copilot config: {e}\nContent: {config_content}")
            else:

[Copilot]

Pull request overview

This PR hardens Windows integration/E2E tests against Unicode encoding/decoding issues by switching test log markers to ASCII and forcing UTF-8 decoding for subprocess output that captures APM CLI output.

Changes:

• Replace Unicode markers in test print() statements (e.g., ✓, ✨) with ASCII equivalents (e.g., [OK]).
• Add encoding='utf-8' and errors='replace' to subprocess.run(...) / Popen(...) calls that capture text output.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
tests/integration/test_guardrailing_hero_e2e.py	Uses UTF-8 decoding for subprocess output and replaces Unicode log markers with ASCII.
tests/integration/test_mcp_registry_e2e.py	Uses UTF-8 decoding for subprocess output and replaces Unicode log markers with ASCII.
tests/integration/test_ado_e2e.py	Uses UTF-8 decoding for captured subprocess output (preemptive Windows hardening).

Comments suppressed due to low confidence (1)

tests/integration/test_mcp_registry_e2e.py:154

• These logs still print result.stdout directly. apm mcp search/show uses Rich Table(...) output (box-drawing characters) and other commands emit Unicode symbols (e.g., "✓" in install output). On Windows with a cp1252 stdout, printing that captured text can still raise UnicodeEncodeError even after replacing the leading test marker. Consider sanitizing before printing (encode to the current stdout encoding with errors='replace'/backslashreplace), or avoid printing captured stdout in the test output.

        print(f"[OK] MCP search found GitHub servers:\n{result.stdout[:500]}...")