# Dora QA targets

#

# Local-first quality gates. Same scripts run in CI.

# See docs/plan-agentic-qa-strategy.md for the full strategy.

#

# The ladder (increasing thoroughness, increasing runtime):

#

# make qa-fast ~1 min pre-commit sanity

# make qa-full ~5-10 min pre-push

# make qa-deep ~15 min target Tier 1 gate, stronger than today's CI

# (adds coverage, adversarial, mutants, semver -

# kept laptop-only, see strategy doc §5)

# make qa-nightly ~3-4 hours Full parity with .github/workflows/nightly.yml

# (qa-deep + proptest@1000 + miri + example-smoke

# + ci-nightly-jobs). After the #1716 rebalance,

# nightly.yml has 27 test jobs (re-counted

# in #2999): example-smoke

# covers 4 (smoke-suite/log-sinks/service-action/

# streaming); scripts/qa/ci-nightly-jobs.sh drives

# 20 more with platform-aware dispatch

# (record-replay, cluster-smoke, cluster-e2e [Linux],

# cluster-record-replay [Linux],

# topic-and-top, cpu-affinity [Linux], redb-backend,

# daemon-reconnect [Linux], state-reconstruction,

# multi-daemon-late-subscriber [Linux],

# test-cross-platform, examples, cli-tests,

# bench-example, cross-check, ros2-bridge [Linux+ROS2],

# ros2-zenoh-humble, ros2-zenoh-kilted,

# msrv, kani-proofs [skipped if Kani absent]).

# Green local run on platform X predicts

# green CI nightly for platform X's jobs.

# make qa-release-gate Tier 3 automatable parts (deep + semver;

# audit/dogfood are human)

# make qa-mutation-audit ~10-18 hrs full-repo cargo-mutants; deliberate

# test-quality audit, NOT every nightly

#

# Orthogonal to the ladder:

#

# make qa-examples ~15-20 min run all smoke-eligible example

# dataflows end-to-end (wraps

# scripts/smoke-all.sh). The script

# skips examples that need CUDA, ROS2,

# webcam, multi-machine deploy, C/C++

# toolchains, or interactive CLI --

# run `scripts/smoke-all.sh -h` to see

# the SKIP list. Not part of qa-*

# ladder by design: the ladder

# excludes dora-examples tests to

# keep per-commit budgets tight.

#

# make qa-test-python ~1 min warm unit tests of the three PyO3 crates,

# which every other qa-* target

# excludes. Needs a Python >= 3.11

# with a shared libpython; run it

# after touching apis/python/* or

# libraries/extensions/ros2-bridge/

# python. CI runs the same command in

# ci.yml's contract-tests job.

#

# `make qa-tier1` is a back-compat alias for `make qa-deep`.

.PHONY: qa qa-fast qa-full qa-deep qa-tier1 qa-nightly qa-release-gate qa-mutation-audit \

qa-examples qa-cluster-e2e qa-cluster-record-replay qa-docker-slim \

ros2-zenoh-humble ros2-zenoh-kilted \

qa-fmt qa-audit qa-unwrap qa-secret-files qa-publish-graph \

qa-package-includes qa-lockfile \

qa-ci-reporting qa-ci-reporting-selftest qa-clippy qa-test qa-test-python \

qa-test-python-node qa-coverage qa-mutants qa-semver qa-breaking qa-breaking-update \

qa-adversarial qa-kani qa-pgo qa-install qa-pgo-install qa-kani-install \

qa-verify-release

qa: qa-fast

qa-fast:

@scripts/qa/all.sh --fast

qa-full:

@scripts/qa/all.sh --full

qa-deep:

@scripts/qa/all.sh --deep

# Back-compat alias. Prefer `make qa-deep` in new docs/scripts.

qa-tier1: qa-deep

qa-nightly:

@scripts/qa/all.sh --nightly

ros2-zenoh-humble:

@scripts/ros2-zenoh-interop.sh humble all

ros2-zenoh-kilted:

@scripts/ros2-zenoh-interop.sh kilted all

qa-release-gate:

@scripts/qa/all.sh --release-gate

qa-mutation-audit:

@scripts/qa/all.sh --mutation-audit

# Run all smoke-eligible example dataflows end-to-end via

# scripts/smoke-all.sh. Skips examples requiring CUDA, ROS2, webcam,

# multi-machine deploy, C/C++ toolchains, or interactive CLI.

# Budget ~15-20 min. Pass flags through, e.g.

# make qa-examples ARGS="--rust-only"

# make qa-examples ARGS="-v" # stream dora output live

qa-examples:

@scripts/smoke-all.sh $(ARGS)

# Build the `dora-slim` container image and run a dataflow inside it, the

# same two steps `.github/workflows/docker-image.yml` gates on. Needs a working

# Docker daemon; deliberately NOT in the qa-fast/full/deep ladder, since it

# costs minutes and a daemon the ladder does not assume. Run it when you touch

# `docker/`.

qa-docker-slim:

@docker build docker/slim -t dora-slim:local

@docker run --rm -v "$$PWD/docker/slim/smoke.sh:/smoke.sh:ro" dora-slim:local bash /smoke.sh

# Real-sshd end-to-end test of `dora cluster up/status/down`. Linux-only.

# Hard-fails if openssh-server is not installed (install via

# `sudo apt-get install -y openssh-server`). Subset of qa-nightly; this

# target is here so developers can run just this job locally without

# invoking the full nightly suite.

qa-cluster-e2e:

@scripts/qa/ci-nightly-jobs.sh cluster-e2e

# Stitches cluster-e2e + record-replay (#2013): record a dataflow spread

# across a 3-daemon SSH cluster, replay it locally, and validate the replayed

# state against the seed(42) baseline. Linux-only; same openssh-server

# requirement as qa-cluster-e2e. Subset of qa-nightly, broken out for local

# runs.

qa-cluster-record-replay:

@scripts/qa/ci-nightly-jobs.sh cluster-record-replay

# Individual gates

qa-fmt:

@cargo fmt --all -- --check

qa-audit:

@scripts/qa/audit.sh

qa-unwrap:

@scripts/qa/unwrap-budget.sh

# Credential-file gate (#2194): fail if git tracks a file whose name says

# it holds a secret. .gitignore did not stop `.adora-token` reaching the

# public repo, because it only covers files git is not already tracking.

qa-secret-files:

@scripts/qa/secret-files.sh

# Manifest-only crates.io publish-graph gate (#3304): no published crate

# may depend on a `publish = false` one, and release.yml's ordered list

# must name every dependency before its dependents.

qa-publish-graph:

@scripts/qa/publish-graph.sh

# Build-time file inclusion gate (#3400): every `include_str!` /

# `include_bytes!` target of a publishable crate must be a git-tracked

# file inside that crate, or the published `.crate` will not have it.

qa-package-includes:

@scripts/qa/package-includes.sh

# Cargo.lock freshness gate (#3512): the committed lock must already

# satisfy every workspace manifest. Plain cargo commands rewrite the lock

# in place, so a stale entry is green locally and only surfaces 3-4 hours

# later in the nightly's --locked builds (`msrv`, cluster `cargo install`).

qa-lockfile:

@scripts/qa/lockfile.sh

# Structural check on nightly.yml failure reporting. Parses the workflow

# only -- runs no nightly job, takes well under a second.

qa-ci-reporting:

@scripts/qa/ci-nightly-reporting.sh

# The gate's own tests: feed known-broken workflows through the real script

# and assert it goes red. A structural check that only ever prints OK is

# indistinguishable from one that stopped parsing.

qa-ci-reporting-selftest:

@python3 scripts/qa/tests/test_ci_nightly_reporting.py

qa-clippy:

@cargo clippy --all --all-targets \

--exclude dora-node-api-python \

--exclude dora-operator-api-python \

--exclude dora-ros2-bridge-python \

-- -D warnings

qa-test:

@cargo test --all \

--exclude dora-node-api-python \

--exclude dora-operator-api-python \

--exclude dora-ros2-bridge-python \

--exclude dora-runtime-python \

--exclude dora-cli-api-python \

--exclude dora-examples

# The unit tests of the PyO3 crates `qa-test` excludes. Kept a separate

# target, not folded into `qa-test`: cargo builds these crates without

# `pyo3/extension-module` (unlike the maturin wheel), so the test binaries

# link libpython directly and need an interpreter >= 3.11 with a shared

# library — a machine set up only for Rust work would start failing the

# everyday gate. CI runs this same target in ci.yml's `contract-tests` job,

# which sets Python up explicitly.

#

# `dora-runtime-python` is the operator runtime's Python backend. Its tests

# cover the cross-language arms of the runtime split — above all that a

# shared-library operator reaching the Python runtime is *delegated* to the

# shared-lib backend rather than rejected, which is what keeps native

# operators working under an embedded-Python daemon.

qa-test-python:

@cargo test --lib \

-p dora-cli-api-python \

-p dora-node-api-python \

-p dora-operator-api-python \

-p dora-ros2-bridge-python \

-p dora-runtime-python

# Python-level unit tests of the `dora` package under apis/python/node/tests/

# (the `dora` compat shim, the `dora.builder` API, and `dora.testing.MockNode`).

# These exercise the *installed* package, so they need `dora-rs` importable —

# run inside a venv that has `uv pip install -e apis/python/node` (plus pytest).

# CI runs this in ci.yml's `contract-tests` job, which sets that venv up (#3305).

qa-test-python-node:

@python -m pytest apis/python/node/tests/

qa-coverage:

@scripts/qa/coverage.sh

qa-mutants:

@scripts/qa/mutants.sh

qa-semver:

@scripts/qa/semver.sh

# The dora 1.x compatibility gate: every surface the 1.0 guarantee freezes,

# checked against the last released tag (docs/api-rust.md, "Stability scope

# at 1.0"). `ARGS="--fast"` drops to the no-compile checks (seconds) -- that

# subset is what PR CI runs and what `make qa-fast` includes; the default

# adds cargo-semver-checks and the snapshot-freshness rebuild.

#

# make qa-breaking

# make qa-breaking ARGS="--fast"

# make qa-breaking ARGS="--baseline v1.0.0"

qa-breaking:

@scripts/qa/breaking-changes.sh $(ARGS)

# Re-record the generated inputs the gate diffs: the `dora` command snapshot

# and the JSON schemas. Run this when a PR *adds* to either surface, and

# commit the result -- that diff is how the addition gets reviewed.

qa-breaking-update:

@scripts/qa/breaking-changes.sh --update

# Check that a published release is complete: every crate on release.yml's

# publish list is on crates.io, both wheels are on PyPI with the full

# platform matrix and an sdist, and the GitHub Release carries the CLI

# binaries plus the 8 C/C++ archives. release.yml runs this as its last job;

# run it by hand to audit an older tag. Pass VERSION=<x.y.z>.

#

# This reads back a *published* release. For the static publish-graph rules

# (list vs. manifests, ordering), see `make qa-publish-graph` — that one runs

# in PR CI and needs no network.

qa-verify-release:

@test -n "$(VERSION)" || { echo "usage: make qa-verify-release VERSION=1.0.0-rc.5"; exit 2; }

@python3 scripts/release/verify-release.py --version "$(VERSION)" $(ARGS)

# Adversarial LLM review of current diff (requires codex or claude CLI)

qa-adversarial:

@scripts/qa/adversarial.sh

# Formal verification: run the #[kani::proof] harnesses (cfg(kani)-gated,

# zero impact on normal builds). A passing proof covers every input in the

# harness state space, not just sampled values. Budget ~5-10 min cold,

# <1 min warm. Run when touching files that contain proofs (see

# docs/formal-verification.md) and in pre-release gating; not part of the

# per-commit qa-fast loop. Nightly CI runs the same proofs (kani-proofs

# job). Install via `make qa-kani-install`.

qa-kani:

@scripts/qa/kani.sh

# Profile-Guided Optimization measurement. Build dora-cli with cargo-pgo

# (instrument -> train on examples/benchmark -> optimize), then run the

# benchmark twice (baseline vs PGO) and print a side-by-side comparison.

# Budget ~30-40 min wall time, ~5-10 GB extra target/ artifacts. PGO results

# don't transfer across OS/arch — run this on your target platform to find

# out if PGO helps. See dora-rs/dora#331 for the methodology and the

# macOS-arm64 baseline data point (+25% throughput geomean).

qa-pgo:

@scripts/qa/pgo.sh

# One-shot tool installation

qa-install:

cargo install cargo-audit cargo-deny cargo-llvm-cov cargo-mutants cargo-semver-checks

rustup component add llvm-tools-preview

qa-pgo-install:

cargo install cargo-pgo

rustup component add llvm-tools-preview

qa-kani-install:

cargo install kani-verifier --locked

cargo kani setup