QC test report

This page records what Optivo is verified against and what the last full run produced. It is written for anyone who needs evidence rather than assurances: a reviewer, a merchant evaluating the app, or us, six months from now, trying to remember what was covered.

Every result below comes from a run executed on the date given. Nothing here is a claim about code that was not exercised — cases that are not automated are listed as such, in their own section.

Run summary

Run date: 28 October 2026 · Environment: Docker Compose (node:22-alpine, PostgreSQL 16, Redis 7)

Gate	Scope	Result
Application suite	47 suites, 1,459 assertions (Vitest)	1,459 passed, 0 failed
Cart Transform function	33 tests (Rust, cargo test)	33 passed, 0 failed
Type checking	react-router typegen && tsc --noEmit, strict mode	Clean
Lint	ESLint across the repo	Clean
Storefront bundle budget	Core payload on every product page	15,489 B / 15,872 B gzipped

Total: 1,492 automated assertions, all passing. The application suite takes about 4 minutes 40 seconds; the function suite is effectively instant.

The bundle budget is at 97.6% of its ceiling

383 bytes of headroom remain on the core storefront payload. The budget is enforced by scripts/build-widget.mjs and fails the build when exceeded, so this cannot regress silently — but the next feature that touches the widget will need to pay for itself. Lazy chunks (date picker, upload, HTML) are well within their own budgets and load only when a set uses them.

Test cases by area

Pricing accuracy — the category’s worst failure mode

The defining risk for an app of this kind is a product page previewing one price and checkout charging another. It is tested as a first-class concern rather than as a side effect.

ID	Case	Verified by	Result
QC-01	A flat fee, percentage, per-character, formula or tiered rule produces the same figure in the widget and in the function	priceAccuracy, pricingRoundTrip	Pass
QC-02	228 pricing permutations across rule types, quantities and option combinations	pricing	Pass
QC-03	Formula evaluation, including precedence, division and malformed input	formula (40), Rust formula.rs	Pass
QC-04	Quantity tier boundaries — the first unit, the tier edge, and above the last tier	tiered	Pass
QC-05	Checkout pricing against fixtures shared byte-for-byte with the widget	checkoutPricing, Rust pricing::matches_every_shared_fixture	Pass
QC-06	A choice surcharge survives the JSON round trip without floating-point drift	Rust wire.rs	Pass
QC-07	An absent, tampered or unknown-version metafield charges nothing rather than guessing	Rust wire.rs	Pass
QC-08	Individually invalid rules are skipped while the remaining valid rules still price	Rust wire.rs	Pass

QC-05 and QC-06 are the pair that matter most: the same fixtures are evaluated by the TypeScript widget and by the Rust function, and the two must agree.

Option types

ID	Case	Verified by	Result
QC-09	All input types render, validate and serialise their answer	newOptionTypes, optionBuilder	Pass
QC-10	Static types (Heading, Paragraph, Divider, HTML) carry no answer and accept no pricing rule	staticOptionTypes (36)	Pass
QC-11	Choice rows — radio, dropdown, checkbox, single and multiple selection	choiceRows (38)	Pass
QC-12	Colour swatch parsing and palette handling	colourPalette (23)	Pass
QC-13	Custom-field patterns cannot be used to mount a ReDoS	patternSafety (40)	Pass
QC-14	HTML blocks are rebuilt from an allowlist; scripts, handlers and javascript: URLs do not survive	patternSafety, staticOptionTypes	Pass
QC-15	Per-option style customisation applies without leaking between options	optionStyle (18)	Pass
QC-16	Plan gating is enforced when an option is added, never when it renders or prices	optionAccess, planFeatures	Pass

Assignment and publishing

ID	Case	Verified by	Result
QC-17	Broad scopes (all, tag, type, vendor, collection) resolve to the right product set	assignmentScopes, scopeWorker	Pass
QC-18	A product created after a scope was matched still receives its options	scopeBackfill, webhooks	Pass
QC-19	Publishing writes the pricing config to the product metafield; a draft set never reaches the storefront	metafieldSync, configCache	Pass
QC-20	The Cart Transform function is registered on install and removed on uninstall	cartTransformRegister	Pass

Storefront widget

ID	Case	Verified by	Result
QC-21	Widget rendering, selection state and the running price preview	widget	Pass
QC-22	Theme app extension renders only on product templates	themeAppExtension (29), themeEmbed (33)	Pass
QC-23	App Proxy responses are correctly shaped and HMAC-verified	proxyResponse	Pass
QC-24	Rate limiting buckets per shop and per client IP, so one busy shop cannot exhaust another’s quota	rateLimit (12)	Pass
QC-25	Core bundle stays within 15,872 B gzipped on every product page	build:widget	Pass

Lifecycle, privacy and billing

ID	Case	Verified by	Result
QC-26	app/uninstalled removes the app’s records and the config written to products	webhooks, reinstall	Pass
QC-27	Reinstalling after an uninstall produces a clean, working install	reinstall	Pass
QC-28	The three mandatory privacy webhooks handle disclosure and deletion according to their purpose	webhooks, webhookJson	Pass
QC-29	Webhook handlers accept real Shopify payload shapes, not hand-written approximations	webhookJson	Pass
QC-30	Plan changes, cancellation and lapse are reflected in entitlements	billing, planFeatures	Pass
QC-31	Operator panel access control and credential hashing	operatorPanel, operatorCrypto	Pass

Admin experience

ID	Case	Verified by	Result
QC-32	Option set index, settings and dashboard render with real data shapes	optionSetIndex, optionSetSettings, dashboard	Pass
QC-33	Onboarding reflects the true app-embed state read from the published theme	onboarding	Pass
QC-34	Form intents route to the right action and reject unknown intents	formIntentFields	Pass
QC-35	Polaris components receive valid props (guards against silent UI breakage on upgrade)	polarisProps	Pass

What is not automated

Stated plainly, because a QC report that implies total coverage is worse than no report at all.

• A real storefront purchase. The preview-versus-charge agreement is proven against shared fixtures in both implementations (QC-05), not by driving a browser through a live checkout. The five-minute manual run on App testing is what closes that gap, and it is worth running against any theme you intend to support.
• Theme compatibility across the ecosystem. The extension is verified against Online Store 2.0 app-embed behaviour, not against individual commercial themes. Vintage themes are unsupported by design.
• File upload. The option type is marked Coming soon and is not selectable, so its storage paths (uploads, customerUploads, storage, storageIntegration, shopifyFiles — all passing) are exercised in isolation rather than end to end.
• Order-line recording. orders/create is not subscribed, because it requires the read_orders scope and adding a scope forces every existing merchant to re-consent. Customer selections still reach the order as native line item properties; the app simply keeps no copy.
• Load and concurrency. Rate limiting is unit-tested (QC-24); the app has not been load-tested at scale.

Defects found and fixed in this cycle

Found	Defect	Status
28 Oct 2026	Documentation search was disabled site-wide: astro build left import.meta.env.DEV true, so Starlight rendered “Search is only available in production builds” on the published site, over a complete Pagefind index	Fixed — the build script now pins NODE_ENV=production
27 Oct 2026	The help centre 404’d at every path because the nginx vhost on the host predated the static-docs routing	Fixed — vhost updated and reloaded
25 Oct 2026	ssl_stapling warnings and an http2 directive incompatible with nginx 1.24	Fixed — HTTP/2 set as a listen parameter

Reproducing this run

Everything above runs inside Docker. No step needs anything installed on the host beyond Docker itself.

docker compose up -d
docker compose exec app npm run test         # 1,459 assertions
docker compose exec functions cargo test     # 33 tests
docker compose exec app npm run typecheck
docker compose exec app npm run lint
docker compose exec app npm run build:widget # enforces the bundle budget

The same five gates run on every pull request and on every push to main, and the deploy is gated on them. A release that fails any one of them does not reach production.