Skip to content

testing.md

Latest commit

 

History

History
88 lines (65 loc) · 3.54 KB

testing.md

File metadata and controls

88 lines (65 loc) · 3.54 KB
Raw

Testing

Philosophy

Real tests, no mocks. We test with real SQLite databases, real file systems, real sockets, and a real @parcel/watcher. The only things not tested are actual Claude API calls and GUI frontend behavior.

Test Tiers

Unit Tests (test/unit/)

Fast, pure logic. Run on all CI targets.

File What it tests
jsonl-parser.test.ts All event types, malformed input, partial lines, mixed line endings, incremental parsing
checkpoint.test.ts File identity hashing, checkpoint decisions (skip/incremental/full/deleted)
claude-paths.test.ts JSONL path parsing, project slug extraction
session-discovery.test.ts Slug algorithm, path round-tripping
schema.test.ts Migration runner, table creation, constraints (unique, cascade)
protocol.test.ts JSON-RPC parsing, response/error/notification creation, subscription manager

Integration Tests (test/integration/)

Real I/O. Run on Linux x64, macOS ARM64, Windows x64.

File What it tests
indexer.test.ts JSONL → SQLite pipeline, incremental indexing, malformed input handling
watcher.test.ts @parcel/watcher detecting file changes, debounced processing
ipc.test.ts Real net.Server + net.Socket, JSON-RPC request/response
search.test.ts FTS5 queries, filtered search by project/role, ranking
checkpoint.test.ts Truncation, replacement, partial lines, deletion, idempotent re-index
idempotency.test.ts Multiple fullScans produce identical state, no duplicate rows

E2E Tests (test/e2e/)

Spawn real daemon processes. Run on Linux x64, macOS ARM64.

File What it tests
daemon-lifecycle.test.ts Start daemon, send requests, write JSONL, query, shutdown
watch-index-query.test.ts Write JSONL incrementally, verify watcher + indexer + search pipeline

Fixtures

JSONL fixtures in test/fixtures/conversations/:

File Purpose
simple-chat.jsonl Basic user/assistant exchange
multi-turn-with-tools.jsonl Tool use blocks (Read, Edit, Write), tool results, thinking
thinking-blocks.jsonl Extended thinking content
malformed-lines.jsonl Invalid JSON mixed with valid lines
mixed-line-endings.jsonl CRLF line endings
empty.jsonl Empty file

Running Tests

npm test                  # All tests
npm run test:unit         # Unit only
npm run test:integration  # Integration only
npm run test:e2e          # E2E only
npm run test:watch        # Unit tests in watch mode

CI Matrix

Target Unit Integration E2E
Linux x64 yes yes yes
Linux ARM64 yes — —
macOS ARM64 yes yes yes
Windows x64 yes yes —

Cross-Platform Test Helpers

  • All tests use TESSYN_CLAUDE_DIR override pointing to temp directories (fs.mkdtempSync)
  • Socket paths: /tmp/tessyn-test-<pid>-<timestamp>.sock on Unix, \\.\pipe\tessyn-test-<pid>-<timestamp> on Windows
  • macOS: @parcel/watcher returns paths with /private/var/ prefix (symlink resolution) — the resolveRealPath() helper handles this
  • Watcher tests poll for results with timeouts rather than fixed waits (FSEvents latency varies)
  • JSONL fixtures include both \n and \r\n line endings

What We Don't Test

  • Claude API calls (we test JSONL parsing and claude CLI availability, not the LLM itself)
  • GUI frontend behavior (frontend's responsibility)
  • Performance at scale (premature — revisit with real usage data)