test(0.2): adversarial filesystem suite + 0.1.x schema migration fixture (Tracks 9.9/9.11)

[pmclSF]

Summary

Two Track 9 (engineering hardening) deliverables. Both are pure
test additions — no behavior changes — and both raise adopter-
trust posture against scenarios the existing suite didn't
exercise. Both are explicitly post-0.2.0-blocking per the plan
but are safe to land now since they're additive.

• Track 9.9 — Adversarial filesystem suite. 8 new tests
  exercising the analyzer against deliberately weird filesystem
  inputs: binary file with .ts extension, oversize source, UTF-8
  BOM, NUL bytes, empty test file, nested .git directories, deep
  nesting, long filename.
• Track 9.11 — Schema migration fixture. Hand-crafted 0.1.x-
  shape JSON fixture + 3 tests that load + migrate it through the
  current code, asserting data preservation and round-trip
  idempotency.

What changed

New files (additive only):

• internal/analysis/adversarial_fs_test.go — 8 adversarial FS
  tests (one skipped on Windows due to path-length)
• internal/models/testdata/snapshot_v0_1_x_legacy.json —
  realistic 0.1.x snapshot fixture
• internal/models/migrate_fixture_test.go — 3 fixture-driven
  migration tests

What's NOT covered (with rationale)

• Symlink loops — behavior diverges across platforms (Linux
  follows; macOS errors; Windows may not have symlinks). Add
  per-platform tests when a real adopter hits one.
• Permission-denied directories — hard to set up portably
  without root. Manual smoke confirms the walker silently skips.

These exclusions are documented in the test file's package
comment so future contributors know the boundary.

Test plan

• go build ./... clean
• go test ./internal/analysis/... — 8 new adversarial-FS
  tests pass + existing tests unaffected
• go test ./internal/models/... — 3 new fixture tests +
  existing migrate tests all green
• go test ./... — full suite green
• make docs-verify passes

Plan tracker

Closes Track 9.9 + 9.11. Track 9 remaining: 9.1-9.6, 9.7
(truth-verify), 9.8 (docs-linkcheck), 9.10 (memory bench), 9.12
(field tier docs). Per the plan, all of Track 9 is explicitly
post-0.2.0-blocking; this PR is safe to land in either the 0.2.0
or 0.2.x window.

🤖 Generated with Claude Code

[github-actions]

Terrain AI Risk Review

Metric	Value
AI surfaces	13
Eval scenarios	16
Impacted scenarios	0
Uncovered surfaces	13

Decision: PASS — AI surfaces are covered.

[github-actions]

[PASS] Terrain — Safe to merge

All changed code is well protected by existing tests.

Metric	Value
Changed files	3 (0 source · 2 test)
Tests selected	2 of 774 (0% of suite)

---

Pre-existing issues (2)

• internal/analysis/adversarial_fs_test.go [LOW] — [staticSkippedTest] 1 of 8 tests statically skipped (12%) in internal/analysis/adversarial_fs_test.go.
• internal/analysis/adversarial_fs_test.go [LOW] — [fixtureFragilityHotspot] Fixture 'mustWriteAdversarial' is used by 8 tests across 1 files. A single change cascades widely.

---

Recommended tests

2 test(s) with exact coverage of 0 impacted unit(s).

Test	Confidence	Why
internal/analysis/adversarial_fs_test.go	exact	test file directly changed
internal/models/migrate_fixture_test.go	exact	test file directly changed

---

Limitations

• No coverage artifacts provided; protection gaps reflect missing data, not measured absence. Provide --coverage to improve accuracy.
• Mixed test cultures reduce cross-framework optimization confidence. Consider standardizing on fewer frameworks.

_{Generated by Terrain · terrain pr --json for machine-readable output}

Targeted Test Results

Terrain selected 2 test(s) instead of the full suite.

• Go tests: passed