Rename the TDD tag system from tdd_bug/tdd_bug_<N> to tdd_issue/tdd_issue_<N>
across the entire codebase. The tdd_expected_fail tag is unchanged.
The TDD expected-failure workflow is not limited to bug fixes — it applies
equally to any issue type (features, tasks, refactors). The _bug suffix was
misleading and narrowed the perceived scope. The new _issue suffix accurately
reflects that the TDD tagging system applies to any Forgejo issue.
Changes span 92 files:
- features/environment.py: validate_tdd_tags(), should_invert_result(), and
apply_tdd_inversion() updated — regex, variables, error messages
- robot/tdd_expected_fail_listener.py: _validate_tdd_tags(), _should_invert_result(),
start_test(), end_test() updated consistently
- 33 Behave .feature files: all @tdd_bug/@tdd_bug_<N> tags renamed
- 29 Robot .robot files: all tdd_bug/tdd_bug_<N> tags renamed
- 3 Robot fixture files renamed (tdd_bug_alone, tdd_missing_tdd_bug,
tdd_expected_fail_missing_bug_n) with content and references updated
- Tag validation tests and helpers updated (function names, command dispatch
keys, output strings, fixture references)
- CONTRIBUTING.md: section renamed from 'TDD Bug Test Tags' to
'TDD Issue Test Tags', all tag references and examples updated
- noxfile.py: comment references updated
- Step definition files, mock helpers, and benchmark files: docstring
references updated
ISSUES CLOSED: #965
## Summary
Adds TDD bug-capture tests proving that `agents init --yes` fails to bypass the
migration approval prompt in a TTY environment (bug #783). These tests follow the
mandatory TDD bug-fix workflow defined in CONTRIBUTING.md §Bug Fix Workflow.
### Changes
- **Behave feature** (`features/tdd_init_yes_no_input.feature`): Two scenarios
tagged `@tdd_expected_fail @tdd_bug @tdd_bug_783` that invoke `agents init --yes`
in a fresh environment with the migration prompt simulated as declining (TTY with
no input). The tests verify the command exits successfully and does not display the
"Apply migrations now?" prompt.
- **Behave steps** (`features/steps/tdd_init_yes_no_input_steps.py`): Step definitions
that create an isolated temp directory, clear auto-apply and database-URL environment
variables, patch `sys.stdin` on the real `sys` module (correct mock target per project
convention), and replace `MigrationRunner._default_prompt_for_migration` with a
function that returns `False` (simulating a TTY prompt where the user declines).
Temp directory and env var cleanup is registered via `context.add_cleanup()`.
- **Robot Framework test** (`robot/tdd_init_yes_no_input.robot`): Two integration test
cases tagged `tdd_bug`, `tdd_bug_783`, `tdd_expected_fail` matching the Behave scenarios.
- **Robot helper** (`robot/helper_tdd_init_yes_no_input.py`): Helper script that exercises
the same code path using the same mock strategy as the Behave steps.
### Bug Reproduction Mechanism
The core challenge is that `CliRunner` replaces `sys.stdin` during `invoke()`, making
direct `isatty()` patching ineffective for controlling the prompt path. The mock strategy
addresses this with a two-pronged approach:
1. **`patch.object(sys, "stdin", mock_stdin)`** — Patches `sys.stdin` directly on the
real `sys` module (correct mock target per the project's established pattern in
`features/steps/migration_runner_steps.py`). Documents the intent of simulating a TTY.
2. **`patch.object(MigrationRunner, "_default_prompt_for_migration", ...)`** — Replaces
the prompt function with one that returns `False` (simulating a TTY user declining
migration). This is the mechanism that actually exercises the bug path, since
CliRunner's stdin replacement bypasses the `isatty()` check.
3. **`CLEVERAGENTS_DATABASE_URL` with non-template filename** — Sets the database URL
to a path whose filename does not match the `before_scenario` template-DB prefixes,
forcing the real Alembic migration path instead of the test fast-path.
With bug #783 present, `require_confirmation=True` is hardcoded in
`UnitOfWork._ensure_database_initialized()`, so the prompt fires, returns `False`,
causing `MigrationNotApprovedError` and a non-zero exit code. After the fix, `--yes`
should bypass the prompt entirely.
### Root Cause (for the bug fix developer)
`init_command()` in `cleveragents.cli.commands.project` receives the `--yes` flag but
only uses it to control output format. It does not forward `yes` to the migration runner.
The migration runner's `init_or_upgrade()` is called via
`unit_of_work._ensure_database_initialized()` with `require_confirmation=True` hardcoded.
### Scenario 2 Limitation
While the bug is present, Scenario 2's first assertion (exit code 0) fails and Behave
skips subsequent steps. The "contains Initialized" assertion is only evaluated after the
bug is fixed, providing distinct post-fix regression value.
### Quality Gate Results
- `nox -s lint` — ✅ passed
- `nox -s typecheck` — ✅ passed (0 errors)
- `nox -s unit_tests` — ✅ passed (387 features, 11121 scenarios, 0 failed)
- `nox -s integration_tests` — ✅ passed (1561 tests, 0 failed)
- `nox -s e2e_tests` — ✅ passed (16 tests, 0 failed)
- `nox -s coverage_report` — ✅ passed (97% coverage)
Closes#842
Reviewed-on: cleveragents/cleveragents-core#1049
Co-authored-by: Rui Hu <rui.hu@cleverthis.com>
Co-committed-by: Rui Hu <rui.hu@cleverthis.com>
freemo
hurui200320