Skip to content

1.2.0 — 2026-09-24

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 24 Sep 14:25
v1.2.0
97fbff4
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Release Notes

Field trials on six stacks (Python, Rust, TypeScript on Bun, Go, Java on Maven, C++ with
GoogleTest) found cases where rqtk reported a requirement verified when it wasn't. This release
fixes them; some checks are stricter as a result.

Upgrade notes

  • coverage --strict requires every requirement to be Verified. It used to pass Planned
    and In Progress requirements. Add --allow planned (and --allow in-progress) to accept
    requirements written ahead of their implementation.
  • RQ030 is an error: a verifies tag with no test declaration below it can never be
    verified. Only comments and attributes may sit between a tag and its test now.
  • verify exits 1 when the results match no linked test, or a linked test matches several
    distinct tests.
  • Suspect lasts longer. A requirement stays Suspect after it changes, even when the same
    unchanged tests pass again, and after a parent requirement or need changes; see Review debt
    below. Run your tests and rqtk verify once after upgrading and commit
    .rqtk/evidence.toml, so the evidence records test source hashes and upstream hashes.
  • Evidence written by 1.2 can't be read by 1.1 (it has new fields). Upgrade everyone who
    runs rqtk verify together. From 1.2 on, unknown fields in the evidence file are ignored.
  • Rust: #[requirements_docs] moved to the requirements-docs feature;
    features = ["macros"] now brings only #[verifies].

Fixed: verdicts

  • Test declarations in Java, Kotlin, C#, C and C++ (including GoogleTest TEST/TEST_F/
    TEST_P and Catch2 TEST_CASE) are recognised; before, nothing in those languages could be
    verified.
  • A tag binds only to the test directly below it. It used to bind to the next test it
    recognised within 12 lines, so it.each(…) gave its activity to an unrelated test.
  • JS/TS .only, .skip, .concurrent, .each(…) and describe groups; Go methods; Python
    async def; Kotlin backtick names.
  • Results are matched by the file and line the runner reports where available, then by what
    the link's path says about the classname. A name that still matches several distinct tests
    is reported as ambiguous instead of all of them deciding the outcome.
  • CTest and GoogleTest disabled tests (status="disabled", status="notrun") count as
    skipped, not passed.
  • verify no longer says "Evidence is up to date" when nothing matched.

Added

  • Review debt. Evidence records a hash of each activity's test source and of everything
    upstream of the requirement. A requirement re-verified by unchanged tests after it changed,
    or whose parent or need changed, is Suspect until its tests change or someone records
    rqtk review <ID> [--note …]. coverage, context and report say why a requirement is
    Suspect (suspect_reasons in JSON).
  • Table-driven and parameterised tests: // rqtk: verifies VA-… case "name" on a table
    row, or verifies("VA-…", case = "…"), links one case: Go subtests, pytest parameters,
    rows of it.each([...]), and GoogleTest parameter values.
  • JUnit @DisplayName / @ParameterizedTest(name = …) and xUnit DisplayName are read from
    the annotations, so tests match under Gradle, which reports display names.
  • Evidence names tests as <source file>::<name>, the same whichever runner reported them
    (Bun and vitest, CTest and GoogleTest).
  • verify counts linked tests that got no result, and warns when it records evidence with no
    commit to record it against.
  • init --hook installs the pre-commit hook.
  • The report header shows uncommitted requirement changes and the latest baseline.
  • rqtk add writes complete requirements: --parent, --satisfies, --criteria,
    --activity (IDs VA-<CATEGORY>-<NUMBER>-<NN>), --priority, --method, --level,
    --phase. It refuses a category that requires a parent when none is given.
  • rqtk add-activity <ID> --name …, and add-need --rationale.
  • coverage --strict --allow planned|in-progress.
  • rqtk report opens with what needs attention, shows the tests and commit (or date and files)
    behind every activity, and adds a stakeholders table and a need → requirement → activity →
    evidence matrix.
  • impact names the tests to run for each activity and separates activities whose evidence
    already covers the change (tests, done).
  • Lint rule RQ031: an activity's evidence file doesn't exist.
  • activity_states in verification JSON: activity states as objects.
  • init names the project after Cargo.toml, pyproject.toml, package.json or go.mod, and
    warns outside a git repository.

Changed

  • init creates the example stakeholder and need only with --example.
  • New requirements get priority Medium (or the middle configured level), not the first level.
  • New requirement and need files carry no content_hash, so editing them leaves no stale-hash
    warning. rehash refreshes stored hashes only; rehash --all stamps every file.
  • validation.require_parent_for_categories is the documented name of the parent rule;
    require_parent_for_levels still works.
  • identification.id_pattern is optional: without it, IDs are checked against
    <prefix>-<CATEGORY>-<NUMBER> for the configured categories, so adding a category is one
    table. init no longer writes it.
  • Everything except history (impact, diff, log, baseline) works outside a git
    repository; lint used to fail there.
  • New stakeholder and need IDs follow zero_padding (STK-0001), or the width a project's
    existing IDs already use.
  • Requirements in top-level categories default to verification level System.
  • trace says when a requirement is top-level or a leaf instead of printing empty lists, and
    shows titles.
  • rqtk = { default-features = false, features = ["macros"] } builds 16 crates instead of about
    270: the macros look activities up with a TOML parser, and the rqtk crate's dependencies
    hang off its lib, cli and macros features.
  • scan shows what each link is attached to and counts links not attached to a test.

Fixed: other

  • A closed stdout (rqtk … | head) ends the command quietly instead of panicking.
  • The workspace builds on macOS with a plain cargo build (the Python extension links).
  • skills install says "Added" when it adds its block to an existing AGENTS.md.

Install rqtk 1.2.0

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/wdoppenberg/rqtk/releases/download/v1.2.0/rqtk-installer.sh | sh

Install prebuilt binaries via powershell script

powershell -ExecutionPolicy Bypass -c "irm https://github.com/wdoppenberg/rqtk/releases/download/v1.2.0/rqtk-installer.ps1 | iex"

Download rqtk 1.2.0

File Platform Checksum
rqtk-aarch64-apple-darwin.tar.xz Apple Silicon macOS checksum
rqtk-x86_64-apple-darwin.tar.xz Intel macOS checksum
rqtk-x86_64-pc-windows-msvc.zip x64 Windows checksum
rqtk-aarch64-unknown-linux-gnu.tar.xz ARM64 Linux checksum
rqtk-x86_64-unknown-linux-gnu.tar.xz x64 Linux checksum
Assets 18
Loading