`shutdown()` set `accepting = false` and `start()` refused for ever after, so
a queue was single-use. Any process that boots more than one app broke: a test
suite closing one harness and opening the next, a hot reload, a multi-tenant
host. The failure landed far from its cause -- the SECOND app to boot threw
WRN-QUEUE-CLOSED out of the dev server because an unrelated one had shut down
earlier in the same process. That is what turned six example-app security
tests red only when run alongside the rest of the suite.
Starting is an explicit intent to run, so it reopens the queue. `add()` keeps
its guard, so work offered to a queue that is shutting down is still refused.
Also migrates the example app's welcome-email queue to `defineQueue`, which
the new loader requires. It still used `defineJob`, so the loader refused it
and took the whole example app down -- 14 failures from one unmigrated file.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
B1: qualify each generated __wrn_api_check_* assertion name with a short
hash of the page's path (relative to app/, for reproducibility across
checkouts) so two pages declaring a same-named block no longer collide
with an identical type alias (TS2300).
B2: skip assertion emission for any block that is not client-mode, or
that has zero declared request fields. ssr sectioned blocks can never
declare a request and always fell back to Record<string, never>, whose
keyof is `string` -- making the key-exactness arm of AssertAssignable
evaluate to false unconditionally (TS2344) on every ssr sectioned block
regardless of correctness. Chose to skip both non-client blocks and
zero-field client blocks, since neither has anything meaningful to
assert type-safety about.
B3: only resolve the endpoint's input (query params / ctx.req.json())
when the endpoint declares an input schema. Previously the router-set
fix accidentally read the request body unconditionally, so a handler
with no input schema that parses the request itself hit
ERR_BODY_ALREADY_USED.
B4: run response/error bodies in client-mode api blocks through
eraseFunctionTypes, matching every other browser-bound body in
client-codegen.ts, so a TypeScript-only construct inside one (e.g. an
annotated locally-declared function) doesn't reach the .mjs artifact.
B5: only exclude "api" from state/prop destructuring in the generated
browser module when the page actually has client-mode api blocks (i.e.
there is a real `api` binding to shadow). Previously "api" was always
excluded, so a page with `state api` and no api blocks got an
undeclared `api` reference (ReferenceError) in client code.
B6: prefix each emitted assertion with `export`, so it isn't flagged as
an unused local under a downstream project's noUnusedLocals (TS6196).
B7: wrnexusCallApi now resolves with undefined for an ok 204/205
response, or an ok response with an empty/unparseable body, instead of
rejecting with "Response was not valid JSON" -- matching the spec's
failure table (error path only for non-2xx, network failure, or an
actually unparseable body on a non-empty response).
Regenerated examples/basic-app's generated types and editor bundles to
match. Confirmed the example's type gate still fails when an
unaccepted field is added to a request body, and passes cleanly
otherwise.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
directory.ts previously exported a plain (ctx: Context) => ... handler.
With that shape ApiInput<> resolved to unknown, so the generated
__wrn_api_check assertion for the demo page passed trivially even with
a field the endpoint does not accept -- the worked example did not
demonstrate the type safety it exists to demonstrate.
Rewrite directory.ts to use defineEndpoint with a schema (matching
typed-user.ts), which gives the generated assertion a real input type
to check against. Confirmed: adding an unaccepted field to the block's
request body now fails typecheck naming
__wrn_api_check_searchDirectory; removing it passes with zero net
diff.
Fix a real bug this surfaced: packages/core/src/endpoint.ts only read
its input from a second 'rawInput' argument, but the actual HTTP
router (packages/dev-server/src/runtime.ts handleApi) invokes route
handlers as handler(ctx) with no second argument. Every
defineEndpoint-based route -- including the pre-existing typed-user.ts
example -- silently received an empty/undefined input through the
real router (confirmed via curl: valid typed-user payloads were
rejected as 'Required'; directory's name filter matched every record
regardless of query). Fixed by having the endpoint wrapper parse the
request itself (query params for GET/HEAD, JSON body otherwise) when
no rawInput is explicitly supplied, while still honoring an explicit
rawInput for direct/unit-test callers.
Also update api-block-demo.wrn's response section: defineEndpoint
wraps handler output as { data: ... }, so the block's raw response
body is now { data: { users: [...] } } -- response reads
data.data.users instead of data.users.
Re-verified in a real browser after the endpoint rewrite and the
router fix: search returns exactly "Ajay, Asha", exactly one
POST /api/directory carrying x-csrf-token, and the error section
still runs cleanly (no exception, empty result) on a missing route.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- Add examples/basic-app/app/api/directory.ts and app/pages/api-block-demo.wrn
as the end-to-end worked example for typed api blocks (Task 6).
- Add **/*.generated.api-checks.ts to .prettierignore: this generated file
must match the CLI's raw output byte-for-byte for check:generated-types,
and prettier was reformatting it.
- Fix packages/csr/test/api-call.test.ts: no-unsafe-function-type lint error
from the raw Function type, uncovered while running the full gate.
- Rebuild editors/vscode bundles (packages/compiler and packages/syntax
changed in Tasks 1-5).
- Regenerate docs/public-api-0.8.json (additive: isIdentPart, isIdentStart,
skipLiteralOrComment newly exported from packages/syntax).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
AssertAssignable was one-directional ([Actual] extends [Expected]), so a
block declaring a field the contract doesn't accept passed silently
(TypeScript's excess-property check only applies to fresh object
literals, not conditional-type extends). Add a key-exactness check
(Exclude<keyof Actual, keyof Expected> extends never) alongside the
assignability check. Guard it with 'unknown extends Expected' so
untyped (no defineEndpoint contract) routes still only warn, per the
existing behaviour, instead of being forced to fail on every declared
field.
Add packages/cli/test/api-block-types.test.ts cases that regenerate a
fixture and run the real TypeScript compiler (via bunx tsc) over the
generated output, asserting on its diagnostics rather than on emitted
text: matching fields compile clean; a wrong-typed field, an extra
field (the Finding-A regression guard), and a missing required field
all fail, each pointing at the offending block's __wrn_api_check_*
line.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
skipLibCheck exempts .d.ts contents from being checked, so assertions
written inside wrnexus.generated.d.ts were never evaluated by tsc.
Emit them into wrnexus.generated.api-checks.ts instead, referencing
the WRNexusGenerated namespace's helper types (which stay in the
.d.ts). Track the new generated file in check:generated-types.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds a scratch page (kept intentionally, per controller ruling on Task 8)
and one end-to-end test that drives the real language server over LSP
stdio against that page's real text, verifying completion (with the
seo-block negative case tested via a simulated '<' keystroke and
mutation-verified against html-regions.ts), hover, folding ranges,
linked editing, and wrn/tagComplete all work together on realistic
content.
Regenerates routes.gen.ts and wrnexus.generated.d.ts for the new page's
route, required by check:generated-types.
Two checks from the original brief (auto-close-tag insertion and Emmet
Tab-expansion) require a live VS Code Extension Development Host and
are documented as outstanding manual verification in
.superpowers/sdd/2026-08-18-wrn-html-editing/task-8-report.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The same generate command emitted ? one run and $1 the next, which looked
like non-determinism. It is not: postgres uses $1 placeholders where
sqlite and mysql use ?, and the driver comes from the active profile, so
building under a different profile rewrites this committed file.
The header now records the dialect it was generated for, making the flip
visible in the diff and explaining check:generated-types failures instead
of leaving them looking like random churn.
Worth deciding separately: a committed artifact whose contents depend on
the active profile will keep drifting. Either generate per dialect, or
stop committing it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The generator's placeholder style is not deterministic across runs: the
same command emitted $1 once and ? the next time, depending on the
database dialect active in the environment. Restoring the committed
output and reverting my earlier regeneration, which was environment
churn rather than an intended change.
Worth a look on its own: a generator whose output depends on ambient
environment makes check:generated-types environment-sensitive.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Regenerated output for the committed query generator: positional
placeholders now render as $1 rather than ?.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The island pieces existed but nothing connected .wrn compilation to island
emission. Now:
- codegen emits a data-wrn-island placeholder for component tags bound to
.tsx imports, keeping .wrn components on the normal mount path
- the dev pipeline and static build resolve island imports, thread the
names into codegen, and build the bundles
- collectScripts adds /__wrnexus/islands.js only when island markup is
present, so island-free pages still ship nothing
- island routes classify as static-interactive via hasIslands
Three bugs found by driving a real page in the browser:
1. The mount runtime was never built anywhere, so the bootstrap 404'd and
no island mounted.
2. Building the runtime separately from the islands gave each its own copy
of React: "Cannot read properties of null (reading 'useState')". The
runtime is now an entrypoint of the same build so React stays in one
shared chunk. The existing single-React test only compared bundles
within one build and could not see across build boundaries.
3. Island props arrived as attribute strings, so start={3} was "3" and
incrementing produced "31" then "311". Props now follow JSX semantics:
{…} parses as JSON, quoted values stay strings, and a runtime
expression is a WRN-ISLAND-PROPS build error rather than a silent
wrong value.
island-codegen.ts no longer imports @wrnexus/core. Compiler modules are
bundled into the Node-only VS Code extension, which contains no other
packages, so a runtime import of core broke the editor compiler; the two
helpers are implemented locally and the core dependency is dropped again.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two guards protect the core promise: a route with no islands emits no
assets at all, and a page with several islands keeps React in one shared
chunk.
buildIslands now writes a generated entry per island instead of passing
component sources directly. Two islands sharing a source deduped to a
single entrypoint, and output order is not guaranteed to match input
order, so island names could bind to the wrong bundle.
Island modules are excluded from the editor compiler bundle: it globs
packages/compiler/src, and island-bundle.ts calls Bun.build while
island-codegen.ts imports @wrnexus/core — neither belongs in a Node-only
VS Code artifact.
Integration assertions share one build. bun test interferes with
Bun.build's module reads after several build calls in one process, while
the same calls succeed repeatedly outside the runner; production is
unaffected.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Section, SectionHeader and PublicPageShell move onto wire-* classes with
local style blocks, variants as data attributes. SectionHeader loses 38
utility lines, and Section stops spending one line per variant and colour
pairing: tinted and solid now select on two attributes. PageHeader was already
on the convention and only needed the audit.
examples/basic-app/app/pages/layout.wrn composes the whole set into one page.
Building it surfaced a library-wide bug. Ten custom properties were referenced
by components and defined by nothing: --wire-color-focus, --wire-color-surface-soft,
--wire-color-on-danger, --wire-color-surface-subtle and the input-* family,
plus hover and contrast for every semantic colour except primary and
secondary. An undefined custom property does not warn, it resolves to nothing,
so focus rings drew with no colour and every soft surface rendered
transparent -- 27 components referenced surface-soft alone. They are derived
in the theme now, and a test checks every token a component references against
the rendered theme CSS rather than the source, since most are generated.
The semantic spread also had to move ahead of the primary and secondary
entries so the palette keeps winning for those two.
Known and unresolved: LayoutSplitter emits its sizeChange output and the
component does fire it, but a parent binding on the tag is not invoked. The
tour page therefore points at the handle aria-valuenow rather than wiring a
handler that would never update. Outputs work elsewhere, so this is narrower
than an outputs-are-broken problem and needs its own investigation.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
global.css defined its own fixed palette -- bg 0b1020, text e7ecff -- while
Wire UI surfaces follow the theme tokens. Switching to light turned the cards
light and left this text light with them, so the sign-in form rendered at a
contrast of about 1.1 and could not be read. The palette now derives from the
wire tokens, and the body wash is tinted from the primary token rather than a
fixed blue. Measured on the login form: light goes from 1.1 to 17.7, dark
stays at 18.2.
Anything an application hardcodes has to be themed as well, or it only ever
looks right in one mode.
The navigation tour is also reachable now: a Navigation entry in the site nav,
translated in both locales, and the page adopts the public layout so there is
a way back out of it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
examples/basic-app/app/pages/navigation.wrn puts all nine navigation
components in a single console shell instead of showing each alone: Navbar
with a nested dropdown, MegaMenu beside it, Breadcrumb, Sidebar as a rail that
becomes a Drawer, Tabs backed by a query parameter, a Stepper wizard,
Pagination, Scrollspy following the article, and Nav in the footer.
Three framework limits shaped the layout and are written into the page rather
than hidden:
- object props are held in state and bound, because a brace at the start of
an attribute is read as an interpolation
- a shared function on a page is compiled standalone and cannot see page
state by name, so state is passed as arguments
- component props and slot content render once and do not track page state,
so anything that has to react lives in page scope; Tabs reports the
selection and the page owns what is shown
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
DataTable replaces the 20-line Table scaffold entirely: columns, sorting,
filtering, pagination, selection, bulk actions, comparison layout, sticky
first column, custom HTML cells, and a remote source driven by a `request`
output rather than a function prop (props travel as HTML attributes, so a
function arrives as its own source text).
Toaster replaces the hand-rolled status div: tone icons, actions, hover
pause/resume and a progress bar.
Overlays audit -- Modal and Drawer declared aria-modal="true" but nothing
ever moved focus into the panel, so the @keydown handler on their root
never ran and closeOnEscape did nothing. Focus, focus restore, a Tab trap
and a body scroll lock now live in the reactive runtime, shared by both.
ContextMenu placed pointer menus by subtracting a guessed 340x420 from the
viewport, which pushed every menu that was not that size away from the
pointer; it now positions at the pointer and lets the anchored clamp pull
it back once it can be measured.
The reactive runtime size budget moves 150k -> 175k to cover anchored
overlays, dialog behaviour, the toaster and the DataTable client half.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ClintchizandClaude Opus 5