M5-T1: Python library tests — tend, models, compat (≥80% coverage)

[hartsock]

Summary

Complete Python unit test suite for the high-level Python library layer. All external calls mocked.

Blocked By

• M2-T1: Python test infrastructure — pytest + coverage ≥80% #20 (M2-T1: test infrastructure)
• M5-PY1: Python tend() and scan() high-level API #34 (M5-PY1: tend() API)
• M5-PY2: Python models module — TendConfig, RepoConfig, ConflictStrategy, result types #33 (M5-PY2: models)
• M5-PY3: GitService compatibility shim for gila git-tend adoption #35 (M5-PY3: compat shim)

New Files

• python/tests/test_tend.py
• python/tests/test_models.py
• python/tests/test_compat.py

test_tend.py coverage

• tend() with empty config → TendResult(total=0)
• tend() refresh profile → pull called per repo
• tend() with rebase strategy → gitxtend.rebase() called when diverged+clean
• tend() with stash-rebase strategy → gitxtend.stash_rebase() called
• tend(dry_run=True) → no write ops called
• tend(profile="refresh") filter → only refresh repos tended
• scan() → scan_workspace() called with correct paths

test_models.py coverage

• ConflictStrategy.REBASE and ConflictStrategy.STASH_REBASE present
• .to_dict() / .from_dict() round-trip all 8 model types
• TendResult aggregate properties (total, succeeded, failed, conflicts)

test_compat.py coverage

• Every GitService method calls the corresponding gitxtend.* function
• Write methods return object with returncode=0 on success

Acceptance Criteria

• pytest python/tests/ --cov=gitxtend --cov-fail-under=80 -v passes
• Zero unmocked external calls

Beaver (MacBook agent, Claude Sonnet 4.6)