Skip to content
Parity 0.2.0
Menu

Proof#

This page records what has been run for Parity and what it showed. Every row traces to a source: MAINTAINING.md for the CI and the checklist, the CHANGELOG for the fixes, and the README for what is still owed, so a claim here can be checked instead of believed.

Last run#

Check What it shows Where it is recorded
go test ./..., go vet ./..., gofmt -l cmd internal tools The unit tests pass, vet is clean, and the formatting is clean. The terminal voice package internal/term is covered at 92% and race-clean git.wayne.edu CI, every push (MAINTAINING, "CI"); the coverage and race-clean line is in the CHANGELOG (Unreleased, "Tests")
govulncheck No known vulnerability in the code. It flagged GO-2026-6859, and CI and release builds moved to Go 1.26.11 to fix it git.wayne.edu CI, every push (MAINTAINING, "CI"); the move to Go 1.26.11 is in the CHANGELOG (Unreleased, "Security")
Secret detection (Preflight 0.9.0) The repository holds no secrets git.wayne.edu CI, every push (MAINTAINING, "CI"); the Preflight version is in the CHANGELOG (Unreleased)
Cross-compiled binaries, all six targets All six binaries build; the artifacts are kept a week git.wayne.edu CI, every push (MAINTAINING, "CI")
The site gate The docs and agent-ready files are current, and every docs link resolves git.wayne.edu CI (MAINTAINING, "The site")
test/install/install-test.sh The installers work against the packaged downloads tree git.wayne.edu CI, on every tag (MAINTAINING, "CI")
install-test.ps1 on Windows PowerShell 5.1 and 7, and install-test.sh on macOS arm64 The Windows and macOS installers work on their own operating systems GitHub CI, on tags (MAINTAINING, "CI")
linux-up-ci parity up, parity status and parity test on testdata/hello run end to end on a Linux Docker daemon, against the CI fixture image (not Citadel's real image) GitHub CI, ubuntu-latest (MAINTAINING, "CI")

Real machines#

The release checklist's real-machine steps (MAINTAINING, "Cutting a release", steps 3 to 5), from its "Checklist runs" table, newest first:

Date OS Engine Mode Parity Result
2026-11-19 Windows 11 x64 Docker Desktop 29.9.2, Compose 5.6.1 mount and bake 0a01fef Passed: mount: check ready (engine.memory, mount-speed), up, test 6 of 6, again at 30, 60, 90 and 120 seconds, status matches Citadel, down --volumes. bake: check ready (engine.memory), up, test 6 of 6, again at 30, 60, 90 and 120 seconds, status matches Citadel, down --volumes. The Podman leg was not run (Podman's machine was not started); the 2026-11-10 Podman row stands.
2026-11-11 Windows 11 x64 Docker Desktop 29.9.2, Compose 5.6.1 mount 356c7cc (modified) Passed: check ready, up, test 6 of 6 at once and again at 30, 60, 90 and 120 seconds, status matches Citadel, down --volumes. The first tools/checklist run, from its own branch
2026-11-11 Windows 11 x64 Docker Desktop 29.9.2, Compose 5.6.1 bake 356c7cc (modified) Passed: check ready, up, test 6 of 6 at once and again at 30, 60, 90 and 120 seconds, status matches Citadel, down --volumes. The mount leg of the same run lost every container at the 120-second test when something outside Parity killed all the engine's containers at once; the row above is its rerun
2026-11-10 Windows 11 x64 Podman 5.8.9, rootless machine on WSL 2, podman compose running Docker Desktop's Compose 5.6.1 mount and bake 64bf3e8 Passed: check ready, image get over git LFS (379 MB, sha256 and image id match), up, test 6 of 6 at once and again at 30, 60, 90 and 120 seconds, status matches Citadel, down --volumes. In about 30 more runs test failed one check twice: once when Podman's SSH connection to its machine dropped (exited 125), which parity test now tries once more; the other run's detail was not kept
2026-11-10 Windows 11 x64 Docker Desktop 29.9.2, Compose 5.6.1 mount 64bf3e8 Passed: check ready (the mount-speed warning, as expected), up, test 6 of 6 at once and again at 30, 60 and 90 seconds, status matches Citadel, down
Apple Silicon Podman (Rosetta on, then off) and Docker Desktop pending
Linux Docker, in CI Only the CI linux-up-ci job covers Linux: parity up, parity status and parity test on testdata/hello against the CI fixture image, on a Linux Docker daemon. No hand-run Linux checklist row exists

The Windows run found three bugs, all fixed, and each fix is in the CHANGELOG:

  • parity test failed a healthy app when Podman's connection to its machine dropped for a moment (exited 125);
  • parity up came up "healthy" on a database volume made with other passwords, with every page saying it could not reach MySQL;
  • on Podman the proxy dropped the app about a minute after up, because HAProxy asked Docker's DNS address.

The checklist tool#

From the repository root, with Go and the engine running, once per engine (MAINTAINING, "Running the checklist"):

go run ./tools/checklist -engine podman
go run ./tools/checklist -engine docker
go run ./tools/checklist -engine podman -note "Rosetta off"     (a Mac, after turning Rosetta off in Podman Desktop)

It builds parity from the checkout, copies testdata/hello into a scratch folder (parity/checklist in your user cache folder, which Docker Desktop and Podman's machines share; -work moves it) under the name parity-checklist, so it never meets an app of yours, and in each mode (-modes mount,bake by default) runs, in order: check, image get (only when Citadel's image is not here yet), up, test at once and again 30, 60, 90 and 120 seconds after up (-again), status (the mirror must say "Matches Citadel."), and down --volumes. A failed up still ends with down --volumes. Every command's whole output goes to checklist.log in the scratch folder. It ends by printing the row for the table above; paste it, and add what a person saw that the tool cannot. Exit 0 every step passed, 1 a step failed, 2 it could not run (no engine running, a build that failed). It changes nothing outside the scratch folder and the engine, and pushes nothing. -parity <file> tries a downloaded release binary instead of building one; -keep leaves the last mode running to look at.

Still owed#

Two things stand between 1.0 and its tag: a Mac and a tar file.

  • The Apple Silicon run (README, "Road to 1.0"): one real parity up and parity test on an Apple Silicon Mac, with Podman and with Docker. The two commands that run the whole check on that Mac and print the row for MAINTAINING's table:

    go run ./tools/checklist -engine podman
    go run ./tools/checklist -engine docker
    
  • Citadel's 8.4r5 image tar (README, "Road to 1.0"): the one outside dependency. Until its tar exists, 1.0 offers 8.5r3 only.

1.0 waits on these two.

Commands you can run#

With only Docker, the image CI uses, with the Go caches in a named volume so nothing downloads twice (MAINTAINING, "Build and test"):

docker run --rm -v "${PWD}:/src" -w /src -v stack-go-cache:/go -e GOCACHE=/go/build-cache golang:1.26.11-trixie@sha256:fa340cb59ebeb1d53ac1cccfc334201d938fe5f3458003cccb767096d8e032b8 sh -c "go vet ./... && go test ./... && go run ./tools/build"