test

name	test
description	Run TML tests using the MCP test tool. Use when the user says "test", "roda os testes", "run tests", or wants to verify changes. Supports running specific test files, filtered tests, specific suites, or the full suite.
user-invocable	true
argument-hint	optional suite name, file path, filter, no-cache, fast

Delegation: For large test runs or test analysis, use the Agent tool to dispatch a tester agent with model: sonnet. For quick single-file tests, execute directly.

Test Workflow

Use the mcp__tml__test MCP tool for ALL test operations. NEVER use Bash to run tml.exe test.

Determine What to Test

Based on $ARGUMENTS:

• No arguments: Run full test suite with verbose: true
• A suite name (e.g., core/str, std/json, compiler/compiler): Run that suite with suite parameter
• A file path (e.g., lib/std/tests/net/ip.test.tml): Run that specific file with path parameter
• A directory (e.g., lib/std/tests/net/): Run all tests in that directory with path parameter
• A filter string (e.g., collections): Run matching tests with filter parameter
• "no-cache" or "fresh": Add no_cache: true to force recompilation
• "fast": Add fail_fast: true to stop on first failure

Suite Names

Suite names follow the pattern <lib>/<module>:

• core/str, core/fmt, core/num, core/iter, core/ops, core/alloc, ...
• std/json, std/collections, std/crypto, std/io, std/sync, std/net, ...
• compiler/compiler, compiler/runtime, compiler/borrow, ...

Use --list-suites via CLI to see all available suites.

Execution

Call mcp__tml__test with appropriate parameters:

• suite: Suite group name (e.g., "core/str", "std/json") — preferred for module-level testing
• path: Specific file or directory
• filter: Test name filter (file path substring)
• verbose: Always true for useful output
• no_cache: Force full recompilation when requested
• fail_fast: Stop on first failure when requested
• structured: Use for machine-readable results (total/passed/failed/failures)

IMPORTANT: Prefer Suite Over Full Suite

When verifying changes to a specific module, ALWAYS use suite parameter instead of running the full test suite. This is dramatically faster:

• suite="core/str" → ~3s (241 tests)
• suite="std/json" → ~4s (153 tests)
• Full suite → minutes (8496 tests)

Only run the full suite when explicitly asked or when changes are cross-cutting.

Report Results

After tests complete:

• Report total/passed/failed counts
• If tests failed, show which specific tests failed and why
• If all passed, confirm success concisely