Skip to content

fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/1: Give whole-tree guard tests the loaded time limit and skip unparsed modules in the CI-runnability guard - #982

Merged
bwmatson merged 1 commit into
mainfrom
spec/fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/1
Nov 16, 2026
Merged

bwmatson merged 1 commit into
mainfrom
spec/fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/1

Conversation

@bwmatson

Copy link
Copy Markdown
Owner

Start here

Already verified by the pipeline

  • unit-test result: passed
  • Review took 1 review round

Review tier: full — the planned estimate is full

Why

The test runs the CI-runnability guard over the whole tests tree: check_tests() in tests/ci_runnable.py calls
check_modules (tests/wide_scans.py), which takes every test_*.py under tests/ (root.rglob), reads it and
parses it with ast.parse. For every test that carries the integration selector, _violators then follows the
test's same-module fixtures and helpers (_reached) and calls AgentStart.of_function(function, tree) for each

… The full reason is in openspec/changes/fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/proposal.md.

What this pull request does

Done when every test listed below carries the loaded limit, the CI-runnability guard parses only modules whose text contains integration, the guard's logic tests and its baseline are unchanged, and the changed tests pass under the repeat check.

Makes the whole-repository guard scans in the test suite finish within their time limit under load, by giving them the loaded time limit and having the CI-runnability guard skip modules that cannot hold an integration-marked test.

  • The CI-runnability guard (tests/ci_runnable.py) no longer parses a test module whose text does not contain integration, since such a module cannot carry the selector the guard checks; its docstring says so.
  • test_no_integration_test_of_the_kit_needs_more_than_ci_has runs with the loaded time limit (LOADED_TEST_TIME_LIMIT) instead of the default.
  • The other whole-tree scans get the same loaded limit: the wide-scans selector check, the live-tag drift comparison against the old glob, the secret-leak model scan and the no-thread-pool source drift test.
  • The guard's logic tests and its baseline are unchanged.

Based on main, with nothing unmerged beneath it — ready to merge whenever the checks look right.

Unit fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/1 of change fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38, task group(s) 1.
Spec: openspec/changes/fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/ in the planning repo.

Assumptions

This unit assumes nothing unmerged: everything it builds on is already in main.

Verification

This is a unit-test unit: everything it needs is fakes, fixtures or throwaway containers, so GitHub Actions runs the whole of it. No e2e-test run was required.

How this was built

The commit order is checked mechanically; lint, formatting and types pass at the tip.


Opened by the spec-driven pipeline. The build agent never merges. The pipeline merges only
a unit review classified auto and policy allowed; a person merges any other, after checking
the stack order above.

…_of_the_kit_needs_more_than_ci_has` deterministic

The tests come first and fail until the implementation.

The test runs the CI-runnability guard over the whole tests tree: `check_tests()` in `tests/ci_runnable.py` calls `check_modules` (`tests/wide_scans.py`), which takes every `test_*.py` under `tests/`…

Done when every test listed below carries the loaded limit, the CI-runnability guard parses only modules whose text contains `integration`, the guard's logic tests and its baseline are unchanged, and…

Unit: fix-test-tests-test-ci-runnable-guard-py-test-no-integration-dcc32e38/1
@bwmatson bwmatson added change-fix-test-tests-test-ci-runnable-guard-py-te The change this pull request belongs to checking In review, but its checks are still running review-full Review asks for a full read from a person in-review Built and waiting for human review and removed checking In review, but its checks are still running labels Nov 16, 2026
@bwmatson
bwmatson merged commit a43bcb7 into main Nov 16, 2026
5 checks passed
@bwmatson bwmatson added settling Merged; waiting for its checks to complete and removed in-review Built and waiting for human review settling Merged; waiting for its checks to complete labels Nov 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

change-fix-test-tests-test-ci-runnable-guard-py-te The change this pull request belongs to review-full Review asks for a full read from a person

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant