The changelog had drifted into walls of text — v1.2.0-alpha.3's entry is a handful of paragraphs where a single bullet runs eleven lines and names the function that owned the bug. An operator cannot tell in ten seconds whether any of it bit them, which is the only job this file has. Adds a "How to write an entry" section to CHANGELOG.md as the rule: lead with what is user-visible and what is not, group by an area a user recognises rather than by subsystem or PR, one line per fix, say what was broken then what it does now, plain language over symbol names, no OC-* ids or file paths, counts in a summary line rather than on every bullet. Repository work that changes nothing observable gets at most a short block at the end. Shipped entries are left alone as history; the rule starts from the next release. Rewrites Unreleased to follow it, which also closes a real gap: that section documented B0/B1 repository plumbing and omitted all 62 operator-visible bug fixes from #1400 and #1402. Exactly backwards — the invisible half was written up and the half users would notice was not. A release cut from dev today would have shipped a changelog that mentioned a directory rename and not "banned users could still connect". docs/contributing.md's PR process now points at the rule, since that is where a contributor decides whether their change needs an entry. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
22 KiB
Contributing
How to set up the development environment and contribute to OwnCord.
Development Setup
Prerequisites
| Platform | Server | Client |
|---|---|---|
| Windows 10+ x64 | ✅ | ✅ |
| Linux x64 | ✅ | ✅ |
| Linux ARM64 | ✅ | ✅ (CI only) |
- Go 1.26+ (server)
- Node.js 24+ (client) — pinned in
Client/.nvmrc;engine-strictmakes a wrong major a hard failure, not a warning - Rust / Cargo (Tauri client — not needed for server-only work)
- Docker + Compose v2 (optional — alternative to building the server locally)
Available Commands
Root facade — one entry point
From the repository root. These orchestrate the per-stack commands below; they
are a convenience, not a replacement. Nothing here needs make, and everything
works the same on Windows, macOS and Linux.
| Command | Description |
|---|---|
npm run bootstrap |
npm ci in all three package roots |
npm run check |
Everything CI gates on: server, client, Rust |
npm run check:server |
Server only — build variants, vet, race, deadlock, lint, generated-output drift |
npm run check:client |
Client only — typecheck, lint, format, unit + integration tests |
npm run check:rust |
Tauri backend — cargo test --lib and clippy |
npm run check:docs |
Fail if a watched document contradicts the ledger's finding counts, or the ledger fails to render |
npm run format |
Prettier over the client, gofmt -w over the server |
npm run generate |
Regenerate protocol constants and the sqlc query layer |
npm run release:preflight |
check plus a client production build |
node scripts/run.mjs --list |
Print the exact command every task runs, and where |
Tools CI installs but you may not have — golangci-lint, sqlc — are skipped
with a printed reason rather than failing the run.
Working on the server only? You never need Node. The facade prints each command it runs and the directory it runs it in; those are the commands in the next section, and using them directly is equally correct.
Server (Go)
| Command | Description |
|---|---|
go build -o chatserver.exe -ldflags "-s -w" . |
Build server binary (Windows) |
CGO_ENABLED=0 go build -o chatserver -ldflags "-s -w" . |
Build server binary (Linux) |
go build -tags otel . |
Build with OpenTelemetry SDK (requires go get first — see Phase B) |
go build -tags wazero . |
Build with Wazero plugin runtime (requires go get first — see Phase C) |
go test ./... |
Run all server tests |
go test ./... -cover |
Run server tests with coverage |
go test -race ./... |
Run server tests with race detection |
Make targets (run from Server/):
| Command | Description |
|---|---|
make test |
Run the test suite the way CI does (-race, 20 min timeout) |
make test-deadlock |
Run the deadlock-detection pass CI also runs (-tags deadlock) |
make cover |
Per-package coverage (what CI uploads) + a function summary |
make cover-all |
Cross-package coverage — the honest number (also lists 0.0% functions) |
make sqlc-install |
Install the pinned sqlc version into $GOBIN |
make sqlc-generate |
Regenerate the type-safe Go query layer (db/dbgen/, SQLite engine) |
make sqlc-verify |
Fail if the committed dbgen output is stale (used by CI) |
make protocol-generate |
Regenerate the WS message-type constants (Go + TS) from protocol/schema.json |
make protocol-verify |
Fail if the committed protocol constants are stale (used by CI) |
make otel-up |
Start Jaeger (traces) + Prometheus (metrics) via Docker for local OTel development |
make otel-down |
Stop and remove the OTel dev containers |
Client (Tauri v2)
Build & dev
| Command | Description |
|---|---|
npm run dev |
Start Vite dev server with hot reload |
npm run build |
TypeScript check + Vite production build |
npm run tauri dev |
Launch Tauri app in dev mode |
npm run tauri build |
Build release installer (NSIS on Windows, AppImage+deb on Linux) |
Tests
| Command | Description |
|---|---|
npm test |
Run all tests (vitest) |
npm run test:unit |
Unit tests only |
npm run test:integration |
Integration tests only |
npm run test:contract |
Cross-component contract tests only |
npm run test:e2e |
Playwright E2E (mocked Tauri) |
npm run test:e2e:native |
Playwright E2E (real Tauri exe + CDP) |
npm run test:e2e:admin |
Playwright E2E (real Go server + SPA) |
npm run test:e2e:prod |
Playwright E2E (prod build) |
npm run test:e2e:ui |
Playwright UI mode |
npm run test:watch |
Vitest watch mode |
npm run test:coverage |
Coverage report |
npm run test:mutate |
Stryker mutation testing |
npm run test:mutate:dry |
Stryker dry-run (no mutations applied) |
npm run test:browser |
Vitest browser-mode tests |
Type checking, linting & formatting
| Command | Description |
|---|---|
npm run typecheck |
Full typecheck (all sources) |
npm run typecheck:build |
Typecheck build config only |
npm run lint |
oxlint + ESLint check (src/) |
npm run lint:fix |
ESLint auto-fix |
npm run lint:ox |
oxlint only (fast correctness checks) |
npm run format |
Prettier format (src/ + tests/) |
npm run format:check |
Prettier check only (no writes) |
npm run knip |
Dead code and unused export detection |
Git hooks (recommended)
Committed hooks in .githooks/ catch the most common CI failures locally. Enable once per clone (from the repo root):
npm run hooks:install # = git config core.hooksPath .githooks
| Hook | What it runs |
|---|---|
pre-commit |
gofmt + go vet (when Go files staged), oxlint + prettier + tsc --noEmit (when client TS staged), sqlc-verify / protocol-verify (when their inputs staged) |
pre-push |
Server build in all build-tag variants, client typecheck + type-aware ESLint. Set OWNCORD_PREPUSH_TESTS=1 to also run go test -race ./... |
Bypass with --no-verify or OWNCORD_SKIP_HOOKS=1 when needed — CI still enforces everything.
Neither hook needs make, and neither needs Node for the Go checks.
core.hooksPath is exclusive. Once set, Git resolves every hook against
.githooks/ and never looks in .git/hooks/ again. .githooks/ holds only
pre-commit and pre-push, so hooks:install silently disables any other
hook you installed there (post-commit, post-checkout, ...). Nothing warns
you. Put it under .githooks/ instead (untracked, so it stays yours), or skip
hooks:install and use npm run check before pushing.
Plugin Development
Plugins are WASM modules loaded at runtime when the server is built with -tags wazero.
The plugin ABI is experimental and carries no compatibility promise. The subsystem is disabled twice over — it compiles only under
-tags wazero(Server/plugin/sandbox_default.go), andplugins.enableddefaults tofalse(Server/config/config.go) — and the five exported functions may change or be removed in any release without a deprecation period.
See Server/plugin/examples/hello/README.md
for the ABI, the build command, and the pinned TinyGo/Go/Binaryen versions. That
file is the single source of truth for the plugin toolchain — this page used to
carry a second copy of the version table, and the two had already drifted apart
in wording.
The example's .wasm is not checked in: TinyGo embeds absolute host paths from
the building machine and offers no -trimpath, so its output is not
byte-reproducible and no CI job can verify it. Build it locally from the source
beside it.
Reporting problems, and where things go
Bugs, questions and vulnerabilities have three different destinations, and the difference matters most for the third.
| Kind | Where |
|---|---|
| A reproducible bug | Issues — the form asks for the environment detail needed to reproduce it |
| A question about setup or usage | Discussions → Q&A |
| An idea or feature suggestion | Discussions → Ideas — not an issue |
| A security vulnerability | Private advisory |
Security reporting
Never open a public issue, pull request, or discussion for a security bug. Use private security reporting; SECURITY.md is the canonical policy and states the response timeline, and docs/security.md says what stays private and for how long.
This repository is public, so a commit message, a PR description and a branch name are all disclosure channels. If you are fixing something you believe is a security problem, say so in the private advisory first and let the fix be coordinated — do not describe the weakness in the public change that repairs it. The same applies to weaknesses in the repository's own automation and settings, not only to bugs in the server or client.
Branch and PR model
This section is the single source of truth for the branch model. Everywhere
else -- the root README.md, CLAUDE.md, the PR template -- summarises it and
links here rather than restating it.
dev-- the integration branch. All contributions targetdev.main-- releases only.devis merged tomainfor a release, and release tags are cut frommain.
dev is protected and PR-only: direct pushes are rejected, twelve status checks
are required, required_approving_review_count is 0, and the rule is enforced
on admins. So a PR is self-mergeable once CI is green, but no commit reaches
dev without CI having run on it. Settings and rationale live in
docs/plans/b0-dev-branch-protection.sh.
Two consequences worth knowing before you open a PR:
- The Docker and Tauri Full Build jobs are gated on
mainand report as skipped on a PR intodev. That is expected, not a failure. - Squash merge, and a conventional commit subject on the squashed commit.
Branch Naming
feature/<name>-- new featuresfix/<name>-- bug fixesdocs/<name>-- documentation changes
Commit Format
Use conventional commits:
feat: add thread support to channels
fix: prevent duplicate WebSocket connections
refactor: extract permission checks into middleware
docs: update quick-start guide
test: add integration tests for invite flow
chore: bump Go dependencies
perf: cache role permissions in memory
ci: add lint step to GitHub Actions
For anything non-trivial the body carries the reasoning, not a restatement of
the diff: what was wrong, why the obvious fix is wrong, what was done, concrete
numbers, and a Verified: paragraph proving both directions — that the defect
was present before and is absent after.
End with an explicit Not included: line naming adjacent scope you
deliberately left out, and why. A written deferral is a deliverable: it is what
separates considered-and-declined from silently-missed, and it means adjacent
work you spotted mid-change does not have to become either scope creep or a
blocking question. Put it in the commit that noticed it.
Pull Request Process
See Branch and PR model above for what to branch from and target.
- Branch from
dev - Open the PR against
dev - All twelve required checks must pass --
devis protected, so a red PR cannot merge - Request code review
- Squash merge, conventional commit subject
If your change is one an operator would notice, add a CHANGELOG.md entry under
## Unreleased. That file's "How to write an entry" section is the rule, and
it is not optional styling: scannable lists grouped by user-facing area, one line
per fix, what was broken then what it does now. No walls of text, no OC-* ids,
no file paths.
Testing
The client suite enforces 70% coverage thresholds in vitest.config.ts;
the Go suite has deliberately no floor (T-2026-07-25-19) — use make cover-all
to see the honest cross-package number. Follow a test-driven workflow and never
lower a threshold to make a change fit.
Tiers
| Tier | Command | CI job | Blocking |
|---|---|---|---|
Client/tests/unit |
npm run test:unit |
Client Unit Tests | yes |
Client/tests/integration |
npm run test:integration |
Client Unit Tests | yes |
Client/tests/contract |
npm run test:contract |
Client Unit Tests | yes |
Client/tests/browser |
npm run test:browser |
— | no |
Client/tests/e2e |
npm run test:e2e |
Client E2E (Playwright) | yes |
Client/tests/e2e @parity |
— | Client E2E (parity subset, blocking) | yes |
Client/tests/e2e/native |
npm run test:e2e:native |
— | no |
Client/tests/e2e/admin |
npm run test:e2e:admin |
Admin Panel E2E (real server, non-blocking) | no |
Server/**/*_test.go |
make test |
Server Build & Test | yes |
Client/src-tauri |
cargo test --lib |
Rust Unit Tests | yes |
npm test — not npm run test:unit — is what CI runs and what
npm run check:client invokes, so it is the command that covers
tests/contract.
What belongs in tests/contract
A test is a contract test when its assertions read, import or execute an
artifact owned by a different top-level component (Server/, Client/, root
protocol/) than the one its runner lives in. A comment referencing the other
side does not count.
- Placement follows capability, not ownership. A contract test lives in the
tier whose runtime can execute or parse the artifact. If the owning component
can execute it, it stays in that component's own suite —
Server/updater/tauri_key_contract_test.goreadsClient/src-tauri/tauri.conf.jsonand stays in Go, because Go parses JSON fine and the assertion is about a server constant. - Ownership is declared in the name, never in the directory. The file name
and the top-level
describe/Testname must name the owned artifact's path. - A contract test may only live in a blocking tier. A non-blocking job is
not coverage.
Admin Panel E2Eiscontinue-on-error: true(.github/workflows/ci.yml), so it is ineligible however well it fits topically — until it graduates. Client/is one component: its TypeScript frontend and its thin Rust backend insrc-tauri/are the same side of the boundary, so atests/unittest that readssrc-tauri/tauri.conf.jsonis an ordinary unit test. The same goes for a Go test reading its own package's embedded assets (Server/admin/perm_grid_test.go).
E2E is runtime coupling rather than artifact coupling; it stays in tests/e2e.
If a tier ever gains a runner of its own, model its anti-vacuity guard on
Server/invariants/invariants_test.go — it fails loudly when a configured scope
resolves to nothing, rather than passing on an empty set.
Code Style
- TypeScript: See Client Architecture
- Go:
gofmt+golangci-lint, standard library preferred - Rust:
cargo fmt+cargo clippy, minimal code (native APIs only)
Dependency Policy
The policy behind what the lockfiles already enforce (decided 2026-08-05, closing audit findings 2026-04-07 #8 / DC-11):
-
Lockfiles are authoritative.
package-lock.json,go.sumandCargo.lockpin every transitive dependency; CI installs only from them (npm ci, module/registry verification — never a barenpm installin CI or hooks).package.jsonkeeps ordinary caret ranges: exact-pinning it would duplicate what the lockfile does while making every security patch a manual edit. -
Upgrades arrive as reviewed PRs, not ambient drift. Dependabot runs weekly per ecosystem (
.github/dependabot.yml) with semver-major updates ignored across the board — majors are adopted deliberately, by a human, reading the changelog. Peer-coupled groups (vitest/@vitest/*,@stryker-mutator/*) update as one PR so exact peer pins cannot wedge. -
Security gates run on every PR:
npm audit --omit=dev --audit-level=high(shipped deps only — dev-tooling advisories are triaged in the workflow comment instead of blocking on unfixable pins),govulncheckfor Go,cargo auditfor Rust, andkniprefuses unused client dependencies outright. -
Version skew is pinned at the toolchain level too:
Client/.nvmrc, everyactions/setup-nodein CI, and anenginesblock in all threepackage.jsonfiles say Node 24 — withengine-strict=truein each package's.npmrc, so a wrong major fails the install instead of warning.Server/sqlc.versionpins sqlc, Go pins viago.mod(GOTOOLCHAIN=auto), and GitHub Actions are SHA-pinned with Dependabot bumping the pins. The one deliberate exception is the plugin toolchain: TinyGo and Binaryen are documented rather than file-pinned, because no gate installs them and nothing would read the pin — the example plugin's README is their single source of truth. -
Three package roots, not an npm workspace — measured 2026-08-26 (npm 11.17, Node 26), not decided on principle. Making
/,/Clientand/tools/mcp-introspectnpm workspaces buys one 298 KB lockfile instead of three (17 KB / 253 KB / 42 KB) and dedupes 614 resolved packages to 582 — 32 packages, 5.2%. Client install time is unchanged: 5642 ms against 5667 ms. The things you would expect to break do not:npm ciinsideClient/still exits 0,npm run <script>still resolves the hoisted binaries (npm prepends every ancestornode_modules/.bintoPATH), andengine-strictstill fails the install on a wrong Node major. The costs that are real:- Ten CI steps key on
cache-dependency-path: Client/package-lock.json— six inci.yml, four in the tag-only, CI-ungatedrelease.yml. That file stops existing, and four of the ten have no gate that would catch it. - Repository Hygiene installs root-only on purpose (prettier is all it
needs). Under workspaces that grows 970 ms → 6172 ms and 39 → 318
packages, unless every call site gains
--workspaces=false— the mitigation works (1112 ms, 38 packages) but has to be remembered forever. - One lockfile puts all three npm Dependabot groups back into the same file.
They rewrite three disjoint files today; undoing that reinstates the
merge-then-rebase-then-re-run-CI storm the grouping comment at the top of
.github/dependabot.ymlexists to prevent.
Thirty-two deduped packages does not pay for that. The roots stay separate,
npm run bootstrapinstalls all three, and Dependabot covers all three. - Ten CI steps key on