The suite drove a containerized freight API, so every code change needed an
image rebuild before a test could see it, and there was no way to attach a
debugger. Files also ran strictly in sequence against one shared database,
which is the root of the warm-stack gotchas the README documents: stowaway
paid bookings climbing back aboard, a short consist on the fifth file.
The freight app now boots inside each vitest worker from dist/, and each
worker owns a whole shard of the topology - its own database, payment API,
gateway mock and broker vhost - so nothing mutable is shared and files run
in parallel. Full suite drops from roughly 20 minutes to 196s at 4 shards.
- main.ts exports createFreightApp() so the harness applies the same prefix,
pipes, filters and interceptors as production instead of replaying them by
hand; self-start is guarded by require.main so the Dockerfile CMD still boots
- booking-window tick cadence is env-driven (BOOKING_WINDOW_TICK_CRON), */1 in
the suite, */10 unchanged in production
- prepare-shards.mjs seeds a template database (boot seeders, then the SQL
fixtures that depend on them) and clones it per shard; it.mjs re-clones on
every run, so each run is hermetic
- gateway mock and payment API are generated per shard: the mock keeps modes
and orders process-global and 20 of 25 specs reset it in beforeAll, and the
inbound CBE bill query has to reach one specific shard's app
- poll() samples every 250ms instead of 2000ms, keeping the caller's deadline
- authz.it.ts seeds its own invoice; it previously read another spec's leftover
and returned early, which silently passed on a pristine database
Known: an unlocked MAX(sequence_no)+1 in train-scheduling.service.ts races
under concurrent allocation and leaves a short consist, so 1-3 specs fail
intermittently. Pre-existing and reproduces at the production tick cadence.
Namespaces the OAuth landing path in all three places it exists: the API's
ack controller, both web apps' routes, and the redirect_uri env values.
A bare /callback claimed a generic top-level path in every app for one
provider's redirect.
The API side needed care. The ack controller moves to @Controller
('fayda/callback'), and the global-prefix exclusion has to name that exact
path — setGlobalPrefix's exclude is an exact route match, not a subtree, so
excluding "fayda" would have left /fayda/callback served at
/api/fayda/callback and 404ing at the registered redirect_uri, while
reading as though it covered everything under /fayda. Naming the full path
also keeps /api/fayda/verification/* prefixed, which every client calls.
Also drops a stale comment on the portal's callback route describing the
popup that no longer exists, and records why the route is public: behind
RequireAuth the onboarding gate redirects to /portal before the code+state
exchange can run.
NOT verified at runtime — this changes route registration, so boot the API
and confirm GET /fayda/callback answers un-prefixed and
/api/fayda/verification/start still resolves before relying on it.
Deploying this requires registering the new redirect_uri with eSignet
first; FAYDA_WEB_REDIRECT_URI, FAYDA_PORTAL_REDIRECT_URI and any mobile
client must be updated in step or verification breaks with a redirect_uri
mismatch.
The onboarding journey had been failing at the company step for a while:
that step was restructured into StepSection cards, so its TIN and VAT
fields no longer have <label> elements and the label-based fill() helper
could not find them. Match on aria-label instead, which also sidesteps the
"0012345678" placeholder both fields share.
Two further staleness bugs were hiding behind that one, both in fill()
itself. It chained clear() into type() on a subject captured beforehand, so
a step-persist PATCH resolving between the two detached it; and re-querying
by the captured id was no better, because the fields remount rather than
re-render and Mantine mints a fresh generated id when they do. Resolve
label -> for -> element afresh for each action. fillPhone gets the same
treatment.
Tighten what the journey proves about Fayda. It verified out-of-band and
then never checked the result reached the UI, so assert the panel renders
fayda-mock's own payload — name, phone and email — which only holds if it
travelled Fayda -> API -> UI, and cross-check ownerFaydaSub on the company,
since that sub is what locks the owner's fields server-side. The PoA step
now asserts no file input exists while the PoA is unverified, covering the
DARS gating; asserted structurally so rewording the document setting cannot
turn a regression into a passing test.
The verification popup was opened after the /start round-trip, by which
point the click's user activation is spent — iOS Safari blocks it outright,
so mobile customers could never verify. Replace the popup with a full-page
redirect: the panel stashes {subject, returnTo} in sessionStorage and
navigates the tab to eSignet, and /callback completes the code+state
exchange itself before returning the user where they were.
This drops the postMessage listener, the popup-closed poller and the
pop-up-blocked branch. onVerified goes with them: the app boots fresh on
the way back, so the target page refetches rather than being pushed to.
Hide the DARS delegation upload until the PoA is Fayda-verified. The paper
authorises the representative the verification names, so it has nothing to
authorise before one exists — and it has to stop being required while
hidden, or the save blocks on a control the customer cannot see. Freight
forwarders are still held to having a PoA by the step's verification gate
and by the API. This also removes the one thing the redirect could not
carry across: a staged File cannot be serialised to sessionStorage, and
there is now never one pending before verification.
Unsaved text typed since the last step-save is still lost on redirect; the
wizard's per-step persistence covers everything already advanced past.
The sheet attests EDR has taken custody. Export custody happens at
warehouse receipt, so the shared received-with-GRN gate now runs even
when wagons are already allocated. Receipt-row query also accepts the
legacy notes GRN fallback, matching the loading gate.
Top Returned Containers table ignored the EDR/Customer tab because
the backend never persisted which truck type performed the return.
Added returned_by column + DTO/entity field, wired create payload to
send it, and filtered the table by the active tab.
- Updated OverviewContractsTabPanel to include a new donut chart for freight type distribution.
- Modified OverviewOperationsTabPanel to improve data visualization with additional charts and refactored data handling.
- Introduced CreateScheduleWindowFields component for configuring booking windows in train scheduling.
- Added new API endpoints for allocation candidates and booking allocation in trainScheduling.service.
- Enhanced BookingRequestsPage to support allocation of paid bookings with a modal for selecting alternative dates.
- Updated QUERY_KEYS and URLS constants to accommodate new operations and features.
- Improved type definitions for overview and train scheduling to support new functionalities.
Intercity bookings ride the wagons freed by earlier unloads (e.g. import
containers uncoupled at Dire Dawa). loadBooking now auto-allocates a
DOMESTIC booking onto on-train slots whose cargo has all departed —
greedy in consist order by capacity, container numbers copied for the
marshalling tally. Falls back to unallocated load when nothing is free.
