Testing Without Tautologies

Testing Without Tautologies

Core Idea

Tests should fail when protected behavior breaks. A passing test helps only if it can catch a real problem.

Before writing or changing a test, ask: "What production change should make this test fail?" If you cannot answer, redesign the test.

Quality Gate

Before writing the test body, answer these:

• Who uses this? Prefer public APIs, HTTP responses, SSE events, persisted rows, parsed session output, rendered UI, or CLI output. Avoid private state.
• What example proves it? Use concrete inputs and literal expected outputs. For parsers, hand-write the fixture JSONL and the expected sessions/messages; do not run the parser to generate expectations.
• What break would this catch? Name the wrong branch, missing side effect, wrong argument, boundary case, or contract violation.
• Do we own it? Test our choices at framework, database, and filesystem boundaries. Do not re-test documented dependency mechanics.
• Can you state it? Given this setup, when the user/system does X, then Y observable behavior changes. If Y is not assertable, the test is not ready.

Required Checks

Apply these checks to every new or modified test:

1. Assert observable effects

   • Check returned values, rows in the DB, HTTP status and body, emitted SSE events, parsed message content, rendered DOM, or errors.
   • A no-assertion test is acceptable only when the failure mode is the subject, such as "this config rejects invalid input." Prefer explicit assertions anyway.
   • Use require.X for checks that must abort the test (setup errors, nil receivers, length before indexing) and assert.X for independent checks. Never if got != want { t.Fatalf(...) } in new tests.

2. Make fakes and mocks specific

   • Verify arguments, call counts, order, and branches when they are part of the contract.
   • Do not let a fake accept any input when the code must pass one value.

3. Separate branch doubles

   • Do not reuse one fake handler for success, error, incomplete, or malformed paths.
   • Give each branch its own fixture or spy so the wrong branch cannot satisfy the expectation.

4. Do not mock the subject

   • Mock dependencies, boundaries, and slow or nondeterministic collaborators. Prefer real in-process collaborators: testDB(t) for SQLite, httptest for handlers, t.TempDir() fixture files for parsers and sync.
   • Do not replace the parser, handler, query, sync step, or component under test.
   • Do not fabricate impossible state to provoke an error (for example, dropping a column so a query fails). Return a typed error from a seam and unit-test the error-to-message mapping directly.

5. Investigate failures before changing expectations

   • Do not flip expected values just to make a failing test pass.
   • First decide whether the production change is intended. Then update the test to describe the new contract.

6. Avoid mirror assertions

   • Do not compute expected values with the same logic under test. Do not build the expected FTS query with the query builder, or the expected timestamp with the same timeutil call the code makes.
   • Use literals, hand-checked fixtures, small examples, or invariant/property assertions.
   • Keep test logic simple enough to review by inspection. Table-driven tests with literal want fields are the preferred shape.

7. Do not test upstream functionality

   • Do not prove that net/http routing, mattn/go-sqlite3, FTS5 tokenizer mechanics, fsnotify, pgx, Svelte reactivity, or Vite work as documented.
   • Test your boundary contract instead: route registration, the SQL your code emits, value handoff from HTTP params to DB queries, migration effects, SSE payload shape, and error responses.
   • For surprising upstream behavior (an FTS5 quoting rule, a WAL edge case, a watcher event ordering), write a narrow characterization test around your integration point and name the upstream assumption in the test name or a comment.

8. Avoid blindingly obvious current-code assertions

   • Do not test that the implementation is written the way it is written now.
   • Skip tests for plain constructor assignment, getters, trivial forwarding, constants, and data-only structs (for example, that an AgentDef entry has the fields it was declared with).
   • Test them only when they validate, normalize, default, derive, copy, enforce permissions, handle errors, cause side effects, or protect compatibility.
   • Prefer the first consumer-visible result that depends on the fields: a discovered session, an API response, a rendered row.

9. Exercise shell scripts, do not read them

   • Shell script tests must run the script against controlled inputs and assert outputs, side effects, or exit codes.
   • Never read a script's source and assert that it contains a specific line, flag, or snippet.

10. Never write negative-existence tests

    • When removing or refactoring code, never add a test asserting that a function, method, file, import, or symbol does not exist — grepping the source tree, reflecting over types, or asserting a compile failure.
    • Deletion is proven by the deletion itself: the code is gone, the build compiles, and the behavior tests for the replacement pass. A "still-deleted" test protects nothing, breaks legitimate future reuse of the name, and outlives the migration it policed.
    • If the concern is that the old path could silently come back into use, test the new path's observable contract instead — the behavior that would break if someone rewired it.
    • Removing such guard tests is part of finishing a cleanup, not a test coverage regression.

Backend Parity

When a behavior is required to match across SQLite and PostgreSQL, protect it on both sides: the SQLite test in internal/db and the PG test in internal/postgres (behind the pgtest tag) should assert the same observable contract — same filtering, ordering, aggregation, and edge cases. A parity bug that only one backend's test can catch is a parity bug the suite will miss.

Test Level

Use the narrowest test that can catch the break:

• Parser, sync, db, config, and server logic: Go unit tests with testDB(t), httptest, and t.TempDir() fixtures. Run with make test (needs CGO_ENABLED=1 and -tags "fts5").
• PostgreSQL behavior: integration tests with the pgtest tag via make test-postgres.
• Frontend logic and components: colocated *.test.ts files.
• User-visible workflows across the HTTP/SQLite/SPA stack: Playwright specs in frontend/e2e/, seeded via cmd/testfixture.

Keep e2e tests non-tautological: assert the workflow result, stored state, rendered UI, or API contract — not just that the server started or the page did not crash.

Mutation Check

Before finishing, mentally mutate the production code. At least one relevant test should fail for each realistic mutation.

• Wrong constant or argument.
• Wrong branch handler.
• Missing state change (row not written, session not resynced).
• Empty/default return.
• Missing side effect (no SSE event, no FTS row).
• Broken fake at a boundary your code should notice.
• Renamed or rearranged private fields with behavior preserved.
• Missing validation for zero, empty, nil, unauthorized, or malformed input.

If none fail, the test is probably tautological.

Red Flags

• Reuses the same setup/assertion object, guaranteeing equality.
• Can fail only through panic, error, missing selector, or server crash.
• Still matters if only the framework/library remains.
• Translates a constructor, getter, setter, mapper, or wrapper line by line.
• Exists for coverage without checking side effects, boundaries, or outcomes.
• Hides expected values behind loops, formatters, builders, or helpers.
• Greps source files (Go or shell) for implementation strings instead of observing behavior.
• Asserts that a removed function, file, or symbol stays removed.