FeatureFlagStore.set() was try/finally with no catch. A rejected
PUT /admin/flags/{key} escaped into the `void this.store.set(...)` call site
as an unhandled promise rejection; the finally-block reload then snapped the
control back to its old value. The admin saw a toggle that silently refused
to move, with no error rendered anywhere and nothing in the state.
set() now folds through the existing runSubmit helper and returns
Result<string, void>, reloading either way so the state still reflects the
server. The page awaits it and renders the failure in an app-alert.
Found by the CQRS-light pass (CQ-002/CQ-004) as one of three mutations that
reach the raw ApiClient without producing a Result — the baseline's BL-007
inventory had missed all three.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
authGuard and capabilityGuard were duplicated byte-for-byte across both
apps, along with their specs — 57 of the 211 duplicated lines BL-002
measured in the two auth contexts, the largest block after session.store.ts.
They are not actor-specific. They ask "is anyone logged in" and "may they do
X", never "who are you or how did you get here". ADR-0002 §3's non-sharing
decision scopes to identity and login flow — Principal, DigiD vs employee
SSO — and a route guard is neither; §Consequences names auth.guard.ts only
as a seam that localises the change, not as something that must be
duplicated.
Moves both to libs/shared/src/application/auth.guard.ts, reading SESSION_PORT
instead of an app-local SessionStore. The port gains one member,
isAuthenticated: Signal<boolean> — free, because both SessionStores already
expose exactly that (session.store.ts:40) and both apps already register
{ provide: SESSION_PORT, useExisting: SessionStore }. The seam existed; it
was just narrower than what it already carried.
Each app keeps a re-export at @auth/auth.guard so app.routes.ts is untouched
— routing asks the auth context for its guards, which is the direction the
boundary should read. The two identical specs collapse into one, plus a case
asserting the guard resolves through the port.
Deliberately NOT merged: session.store.ts, session.ts, digid.adapter.ts,
login-form.component.ts, login.page.ts. Those are identical only because
ADR-C-004 (Session -> Principal) was never executed. Merging them would make
a citizen DigiD/BSN login the backoffice's shared login.
Measured with tools/baseline-scan.mjs: ssp/auth duplicated lines 211 -> 151,
bhp/auth 86.8% -> 82.5%, repo-wide 7.1% -> 6.6%. Both guard clone pairs drop
out of the top-clones list. What remains is exactly the three files
ADR-C-004 should differentiate.
behaviour-spec.mdx regenerated (the spec moved libraries).
Verified: lint, typecheck, dep:check (0 violations, 224 modules), prettier,
ng build --localize for both apps, and 407 tests passing across all four
projects.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Four close-outs and their README rows. The behaviour spec is regenerated
once here rather than per-track — it derives from every test name in the
repo, so any track running it would have conflicted with the other three.
Records two findings the arc surfaced but did not cause: the /brief/preview
staleness for non-DemoOwner identities (blocking per-spec identity isolation
in brief-v2.spec.ts), and that WP-72/73 had to share a commit because both
edit Program.cs — separate execution waves prevented build collisions but
did not produce separable diffs.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
ADR-0001's own canonical "config value" example was unenforced: GET
/intake/policy echoed ScholingThreshold, but no request DTO carried a
scholing answer, so the server had nothing to re-validate. A crafted
POST could skip a requirement the wizard presents as mandatory.
IntakePolicy.RejectIncompleteScholing is the authority — three-valued
completeness (below threshold an answer is required; "nee" is legal and
still submits; punten only belong to a followed scholing), living in the
class that owns the constant so scripts/check-seam.sh keeps guarding the
FE/BE literal pair. Both submit paths call it; a violation 400s with
ProblemDetails and leaves the aanvraag a Concept. Gated on
Type == "intake" (the endpoint's switch lumps herregistratie with
intake, which has no scholing question), and guarded by `reject is null`
so a zero-uren submission is still decided on its merits.
Also fixes a live FE bug in the same rule: validateStep required punten
whenever scholingGevolgd was 'ja' regardless of lageUren, while the
template renders those fields only when lageUren — so answering 'ja'
then raising uren either blocked the user on an invisible field or
emitted aanvullendeScholing: undefined alongside punten. punten now
derives from aanvullendeScholing, so that combination is unrepresentable
in ValidIntake.
Note: EndpointTests' Worked_hours_submission_succeeds was itself
asserting the vulnerable payload ({ uren: 40 }, no answer) and needed a
complete answer added; the zero-hours rows are the ordering regression
net and are unmodified.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Gherkin/Cucumber was considered and rejected for business-readable BDD
scenarios: step-binding by runtime string match undoes the compile-time
guarantees WP-70 just added, and needs two frameworks for .NET+TS with
no non-technical co-author in view. Instead scripts/gen-behaviour-spec.mjs
(modeled on the existing gen-snippets.mjs) extracts every describe/it
and [Fact]/[Theory] name straight from the real suites into
libs/shared/docs/behaviour-spec.mdx, gated for drift in CI exactly like
gen-snippets/gen-api — the page can never diverge from the tests because
it's generated from them, and test names stay the single source of truth.
scripts/check-seam.sh guards the one FE/BE rule duplication most likely
to silently diverge: IntakePolicy.cs's ScholingThreshold vs
intake.machine.ts's SCHOLING_THRESHOLD_DEFAULT, two unlinked literals
pinned separately in each side's own tests but never against each other.
package.json/CI wiring for both (gen:behaviour-spec, check:seam) shipped
in the prior commit alongside the typecheck gate, since all three touch
the same few config files.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
ehoandClaude Opus 5