freshyo/docs/test reports/e2e-sanity-failed-cases-2026-10-10-1310.md
2026-10-10 15:00:50 +05:30

25 KiB
Raw Blame History

Freshyo E2E — failed test cases (detailed)

Companion to docs/e2e-sanity-test-plan.md (the plan) and docs/test reports/e2e-sanity-test-results-2026-10-10-1310.md (all results). This file expands only the FAIL cases with reproduction steps, evidence, root cause, and a suggested fix.

All evidence was produced on 2026-10-10 against the local stack: backend http://localhost:8787 (/api/trpc), disposable local D1, iOS simulator com.freshyotrial.app / in.freshyo.adminui, Chromium superadmin http://localhost:4175.

Legend: Severity — Critical / High / Medium / Low. Case IDs match the plan.


FC-01 — SEC-01: customer token can mutate global configuration

  • Status: ✅ Fixed (2026-10-10) — admin.const.* now requires a staff token (staffProcedure); updateConstants additionally rejects global-scope keys (globals are superadmin-only); the redundant getGlobalConstants/updateGlobalConstants/getLegacyConstants/updateLegacyConstants procedures were removed (no production UI caller). Regression test: apps/backend/tests/api/authorization-matrix.test.ts (staff-guarded block).
  • Severity: Critical
  • Case: SEC-01
  • Symptom: A signed-in customer can call the staff/admin global-config mutation and change app-wide constants.
  • Steps to reproduce:
    1. user.auth.login {identifier:"e2e_u2@test.com", password:"Pass@12345"} → customer token.
    2. admin.const.updateGlobalConstants {constants:[{key:"versionNum",value:"1.2.0"}]} with that token.
  • Expected: Rejected (staff-only).
  • Actual: 200 {success:true, updatedCount:1}.
  • Evidence: Response body from step 2; anonymous admin.product.getProducts → 401, customer admin.product.getProducts → 500 (Cannot read properties of null (reading 'regionId')).
  • Root cause: apps/backend/src/trpc/apis/admin-apis/apis/const.ts — updateGlobalConstants uses protectedProcedure (allows any authenticated user or staff) and never touches ctx.staffUser. protectedProcedure is defined in apps/backend/src/trpc/trpc-index.ts.
  • Suggested fix: Change the global-config procedures (getGlobalConstants, updateGlobalConstants, legacy ones) to superadminProcedure (global scope) or at minimum staffProcedure; remove the legacy admin global mutation or gate it behind superadmin.

FC-02 — SEC-04: order accepts a SKU from another region

  • Status: ✅ Fixed (2026-10-10) — placeOrderUtil now rejects an item whose product regionId ≠ the order region (409 "… is not available in this region").
  • Severity: High
  • Case: SEC-04
  • Symptom: A region-1 order can include a region-2 SKU.
  • Steps to reproduce:
    1. U2 with a region-1 address (877).
    2. user.order.placeOrder {regionId:1, addressId:877, paymentMethod:"cod", selectedItems:[{skuId:157 /* region-2 product */, quantity:1, slotId:2012}]}.
  • Expected: Rejected (SKU not in the ordering region).
  • Actual: 200, order 2456 created with regionId:1 and sku 157, priced at region-2's ₹149 + ₹20 delivery.
  • Evidence: Order 2456 in orders/order_items.
  • Root cause: apps/backend/src/trpc/apis/user-apis/apis/order.ts → placeOrderUtil resolves items via getOrderProductById(item.skuId) with no region filter. Address region is checked, but SKU/slot/bound region coherence is not.
  • Suggested fix: Validate each SKU's regionId (and its product's region) equals the order region; reject otherwise.

FC-03 — SEC-05: order accepts unassigned / out-of-stock / closed-slot items

  • Status: ✅ Fixed (2026-10-10) — placeOrderUtil now rejects isDeleted/isOutOfStock/isSuspended items (400), requires a non-null slotId for scheduled orders, and validates each slot: exists, region match (409), isActive, freezeTime/deliveryTime in the future, and every ordered skuId present in slot.skuIds (400).
  • Severity: High
  • Case: SEC-05
  • Symptom: The backend accepts items it should reject.
  • Steps to reproduce:
    • (a) Unassigned SKU: placeOrder {regionId:1, selectedItems:[{skuId:1, quantity:1, slotId:2012}]} where sku 1 is not in slot 2012 → 200, order 2457.
    • (b) Out-of-stock: admin.product.toggleOutOfStock{id:155} → placeOrder {skuId:155, slotId:2012} → 200, order 2458 (stock restored afterward).
  • Expected: Both rejected.
  • Actual: Both create orders.
  • Evidence: Orders 2457, 2458; schema still rejects non-positive/non-integer quantity.
  • Root cause: placeOrderUtil (order.ts) never checks isOutOfStock / isSuspended / isDeleted, nor that the SKU is a member of the selected slot, nor slot active/frozen state (only capacity is checked).
  • Suggested fix: In placeOrderUtil, for each item assert: SKU exists, belongs to region, not OOS/suspended/deleted, is assigned to the chosen slot, and the slot is active and not frozen.

FC-04 — SEC-06: user-based coupon is not restricted to its user

  • Status: ✅ Fixed (2026-10-10) — validateAndGetCoupon now loads applicableUsers and rejects a user-based coupon for anyone not listed (unless isApplyForAll): This coupon is not applicable to your account.
  • Severity: High
  • Case: SEC-06
  • Symptom: A coupon created for U1 can be applied by U2.
  • Steps to reproduce:
    1. Admin: coupon.create {couponCode:"E2EU1ONLY", discountPercent:10, isUserBased:true, applicableUsers:[388], maxLimitForUser:3}.
    2. U2: placeOrder {…, couponId:46}.
  • Expected: Rejected (coupon is for user 388 only).
  • Actual: 200, order 2463 discounted.
  • Evidence: Order 2463; coupon 46 has isUserBased:true.
  • Root cause: packages/db_helper_sqlite/src/user-apis/coupon.ts → validateAndGetUserCoupon does not consult applicableUsers. (Expiry is enforced.)
  • Suggested fix: Reject when coupon.isUserBased && !applicableUsers.includes(userId).

FC-05 — SEC-07: public vendor packaging mutation + expired-snippet leak

  • Severity: High
  • Case: SEC-07
  • Symptom: Anyone can toggle order-item packaging; expired snippets still list orders.
  • Steps to reproduce:
    1. Admin creates a vendor snippet for slot 2012 (E2ESNIP).
    2. Anonymous admin.vendorSnippets.updateOrderItemPackaging {orderItemId:8237, is_packaged:true} → 200 (toggled U1's order item).
    3. Admin creates an expired snippet (validTill:"2020-01-01"); anonymous admin.vendorSnippets.getOrdersBySnippetAndSlot {snippetCode:"E2ESNIPEXP", slotId:2012} → 200 with customer names + orders.
  • Expected: Packaging edits require vendor authority; expired snippets return nothing.
  • Actual: Anonymous mutation succeeds; expired snippet leaks orders.
  • Evidence: Both responses above (item 8237 toggled true→false).
  • Root cause: apps/backend/src/trpc/apis/admin-apis/apis/vendor-snippets.ts — updateOrderItemPackaging is publicProcedure with the auth check commented out; getOrdersBySnippetAndSlot never checks validTill (unlike getOrdersBySnippet).
  • Suggested fix: Require a valid snippetCode (unexpired) and authorize the mutation against the order's slot/snippet; enforce expiry in the slot-variant query too.

FC-06 — BL-02: minRegularOrderValue is not a real minimum

  • Status: ✅ Resolved (2026-10-10) — the constant was removed. It was never a hard minimum (nothing enforced it) and is no longer used as the delivery threshold, so it was dead. Removed from both const registries, the essential-consts payload, the web-ui hook, the admin/superadmin bound editors, the seeds, and the tests. No minimum-order rule is enforced (matching the customer app).
  • Severity: Medium
  • Case: BL-02
  • Symptom: Orders below the configured minimum are accepted.
  • Steps to reproduce: placeOrder with subtotal ₹199 while bound A1 minRegularOrderValue:300.
  • Expected: Blocked if it is a hard minimum (policy needs confirmation).
  • Actual: 200 (orders 2457/2458).
  • Evidence: Orders below 300 accepted.
  • Root cause: placeOrderUtil uses minRegularOrderValue only as the free-delivery threshold, never as a minimum-order gate; the client uses freeDeliveryThreshold instead.
  • Suggested fix: Decide the contract (minimum vs informational). If a minimum, enforce it in placeOrder and surface it in the cart; if not, rename/document it and stop using it as the fee threshold.

FC-07 — BL-03 / BL-06: client and backend delivery-fee formulas diverge

  • Status: ✅ Fixed (2026-10-10) — the backend now uses the customer app's model: threshold key freeDeliveryThreshold (not minRegularOrderValue) and the post-discount basis (finalTotal = subtotal − discount; delivery = finalTotal < threshold ? charge : 0).
  • Severity: High
  • Cases: BL-03, BL-06
  • Symptom: The amount the customer sees differs from what the backend saves.
  • Steps to reproduce:
    • BL-06 (different keys): bound A2 min=300, free=800, charge=200; subtotal ₹548. Backend order 2459 → totalAmount:548, deliveryCharge:0; client preview → 548<800 → ₹200 delivery → ₹748.
    • BL-03 (different discount basis): bound A1 threshold 300; subtotal ₹349, flat ₹100 coupon. Backend order 2461 → totalAmount:249, deliveryCharge:0; client preview → post-coupon 249<300 → ₹20 → ₹269.
  • Expected: Identical rule in cart, checkout, persisted order, and admin.
  • Actual: ₹200 / ₹20 discrepancies.
  • Evidence: Orders 2459, 2461.
  • Root cause: apps/user-ui/components/cart-page.tsx uses deliveryCharge = (subtotal − discount) < freeDeliveryThreshold ? charge : 0. placeOrderUtil (order.ts) uses expectedDeliveryCharge = subtotal(pre-discount) < minRegularOrderValue ? charge : 0. Different key and different discount basis.
  • Suggested fix: One shared pricing function (server-authoritative). Use the same threshold key and the same pre/post-discount basis on both sides; ideally the client renders the server quote.

FC-08 — D10: percentage discount is applied to subtotal + delivery fee

  • Status: ✅ Fixed (2026-10-10) — the coupon discount is now computed on the cart subtotal only; delivery is decided afterwards and added on top (no longer discounted).
  • Severity: High
  • Case: BL-05 (and visible in SEC-06 order 2463)
  • Symptom: A % coupon discounts the delivery fee too.
  • Steps to reproduce: subtotal ₹199 + delivery ₹20, 10% coupon → order 2463 totalAmount:197.1.
  • Expected: Discount on the item subtotal only → ₹199×0.9 + ₹20 = ₹199.1.
  • Actual: (199+20)×0.9 = ₹197.1.
  • Evidence: Order 2463; also 15% case 2480 (₹186.15 = 219×0.85).
  • Root cause: placeOrderUtil passes subOrderTotalWithDelivery into applyDiscountToUserOrder for the first suborder.
  • Suggested fix: Apply the coupon to the item subtotal, then add delivery; keep the same order on the client.

FC-09 — CP-05: SKU-restricted coupon is not enforced

  • Status: ⚪ Not a defect for now (2026-10-10) — SKU-specific coupons are not created today; coupons are intended to apply to all SKUs. skuIds is deliberately left unenforced. (Revisit if SKU-scoped coupons are ever introduced.)
  • Severity: High
  • Case: CP-05
  • Symptom: A coupon limited to sku 155 discounts a cart containing only sku 156.
  • Steps to reproduce: coupon.create {couponCode:"E2ESKU155", discountPercent:10, skuIds:[155]} → placeOrder with only sku 156 → 200, order 2474 (10% off 349).
  • Expected: Rejected.
  • Actual: Applied.
  • Evidence: Order 2474.
  • Root cause: validateAndGetUserCoupon does not enforce skuIds / applicable products.
  • Suggested fix: Validate cart items against the coupon's product/SKU scope server-side.

FC-10 — CP-11: coupons are not region-scoped

  • Status: ⚪ Not a defect for now (2026-10-10) — region-scoped coupons are not wanted; coupons are global. No region filter is added. (Revisit if regional coupons are ever required.)
  • Severity: High
  • Case: CP-11
  • Symptom: A region-1 coupon discounts a region-2 order.
  • Steps to reproduce: U2 region-2 address (883) + region-2 sku 157 + region-1 coupon 45 → 200, order 2477 total ₹79.
  • Expected: Per region policy (reject if region-scoped).
  • Actual: Applied.
  • Evidence: Order 2477.
  • Root cause: coupon storage/validation (coupons, validateAndGetUserCoupon, user coupon fetches) carries no region filter.
  • Suggested fix: Decide global-vs-regional coupon policy; if regional, store region_id and filter both retrieval and validation.

FC-11 — PD-13: deleting a store orphans its products

  • Status: ✅ Fixed (2026-10-10) — deleteStore now counts the store's products and refuses to delete while any exist (plain error: "Cannot delete a store that still has products…"); the store_id → NULL step was removed, so products can no longer be orphaned.
  • Severity: High
  • Case: PD-13
  • Symptom: deleteStore succeeds even when the store has products.
  • Steps to reproduce: admin.store.deleteStore {storeId:1} (store 1 had 41 products) → 200 "Store deleted successfully".
  • Expected: Blocked (or products reassigned).
  • Actual: 41 products left with store_id = NULL.
  • Evidence: DB count of product_info where store_id is null jumped to 41 (restored manually).
  • Root cause: apps/backend/src/trpc/apis/admin-apis/apis/store.ts deleteStore and packages/db_helper_sqlite deleteStoreFromDb have no dependency guard/reassignment.
  • Suggested fix: Reject deletion when products reference the store, or require an explicit reassignment target; never leave products with a null store.

FC-12 — DS-02: deleting a product tag fails (FK)

  • Status: ✅ Fixed (2026-10-10) — deleteProductTag now deletes the product_tags junction rows first (in a transaction) before the tag row, so a tag with products deletes cleanly. No FK-cascade migration needed.
  • Severity: High
  • Case: DS-02
  • Symptom: Tag delete throws a raw DB error when the tag has products.
  • Steps to reproduce: create tag 12 with products 155/156 → admin.product.deleteProductTag {id:12}.
  • Expected: Tag + junction rows removed.
  • Actual: 500 Failed query: delete from "product_tag_info" ….
  • Evidence: product_tags still had row (product 155 → tag 12); FK ON DELETE no action.
  • Root cause: deleteProductTagInDb deletes product_tag_info without first deleting product_tags junction rows.
  • Suggested fix: Delete junction rows (or use ON DELETE CASCADE) before the tag row.

FC-13 — SL-04: closed (frozen) slot is accepted by placeOrder

  • Status: ✅ Fixed (2026-10-10) — the slot validation added for FC-03 rejects a slot whose freezeTime (or deliveryTime) has passed (400).
  • Severity: High
  • Case: SL-04
  • Symptom: An order can be placed into a slot whose freeze time has passed.
  • Steps to reproduce: create slot 2015 with freezeTime:"2020-01-01" → placeOrder {skuId:155, slotId:2015} → 200, order 2469.
  • Expected: Rejected (orders closed).
  • Actual: Accepted.
  • Evidence: Order 2469.
  • Root cause: placeOrder checks capacity (getUserSlotCapacityStatus) but not slot active/freeze state; the client only filters slots.
  • Suggested fix: Server-side reject when slot.freezeTime <= now or !slot.isActive.

FC-14 — CT-07: cart does not reflect an out-of-stock item

  • Status: ⚪ Not a defect (false flag) (2026-10-10) — reproduced and disproven. The cart does implement the CT-07 behavior: cart-page.tsx renders an "Out of Stock" badge (:962), filters OOS lines out of the checkout payload (:1247), and blocks checkout with "Items Unavailable" when every line is OOS (:1264). The E2E only observed the pre-propagation window: after toggleOutOfStock the app keeps the previous availability_version_num until its essentialConsts refetchInterval (60 s) fires, so the cart shows the line normally for up to ~1 minute, then correctly shows "Out of Stock". The report's stated root cause (cart-page ignores isOutOfStock) was inaccurate.
  • Severity: High
  • Case: CT-07
  • Repro (2026-10-10): cart held sku 155 ("E2E Chicken 500 g", qty 6, ₹1194); toggleOutOfStock{id:155} bumped availability av-17 → av-18 (published av-18 shows isOutOfStock:true); stopApp → launchApp → open cart → line normal; ~70 s later (the refetch interval) → red "Out of Stock" badge. Stock restored afterwards.
  • Evidence (original E2E): Published availability.json for region 1 av-15 showed sku 155 isOutOfStock:true; the cart screenshot was taken inside the staleness window.
  • Residual (out of scope, by decision): the ~60 s propagation delay is cache-refresh latency, not a cart-logic bug.

FC-15 — AU-07: profile update is broken (ctx.req.header)

  • Status: ✅ Fixed (2026-10-10) — Context.req is now typed as the fetch Request the adapter passes, and updateProfile reads ctx.req.headers.get('authorization'). The optional email was relaxed too: an empty email means "no change", and the client only requires it when registering.
  • Severity: High
  • Case: AU-07
  • Symptom: updateProfile always errors.
  • Steps to reproduce: user.auth.updateProfile {name:"E2E U2 Renamed", bio:"hello bio"} with a valid token.
  • Expected: Profile updated.
  • Actual: 500 "ctx.req.header is not a function".
  • Evidence: Response above; UI Edit Profile also blocked earlier by a required-email validation.
  • Root cause: apps/backend/src/trpc/apis/user-apis/apis/auth.ts line ~342 calls ctx.req.header('authorization'), but ctx.req is a fetch Request (only .headers.get).
  • Suggested fix: Read the header via ctx.req.headers.get('authorization') (or drop the token echo). Relax the client's email requirement if email is optional.

FC-16 — AU-08: password change has no effect (duplicate creds row)

  • Status: ✅ Fixed (2026-10-10) — upsertUserPassword now selects then updates-or-inserts (updating all rows for the user, repairing existing duplicates), and a uniqueIndex on user_creds.user_id was added to the schema. Existing duplicate rows (users 379, 908, 1311) must be deduped before the index migration is applied.
  • Severity: High
  • Case: AU-08
  • Symptom: After updatePassword, the old password still works and the new one fails.
  • Steps to reproduce: fresh user U5 → user.auth.updatePassword {password:"NewPass@456"} → login with old (works) / new (fails).
  • Expected: New password works, old fails.
  • Actual: Old logs in (200), new fails (500).
  • Evidence: user_creds for the user contains two rows (ids 26, 27) after the change.
  • Root cause: upsertUserAuthPasswordInDb inserts a new row instead of updating the existing one; login reads the stale row.
  • Suggested fix: Update the existing user_creds row (unique on user_id) rather than inserting; add a unique constraint if missing.

FC-17 — OR-12: no order idempotency (duplicate submits)

  • Status: 🟡 Mitigated client-side (2026-10-10) — the Place Order button is disabled while a submit is in flight, and handlePlaceOrder early-returns via a synchronous useRef guard (plus isPending), so an accidental double-tap cannot submit twice. Server-side idempotency was intentionally deferred: placeOrder still has no idempotency key, so a direct API replay or network retry can still create duplicates.
  • Severity: Medium
  • Case: OR-12
  • Symptom: Retrying/submitting twice creates duplicate orders.
  • Steps to reproduce: call the same placeOrder twice with identical params.
  • Expected: One order group (idempotent) or a clear dedupe.
  • Actual: Two calls created 4 orders (2465–2468).
  • Evidence: Order ids above.
  • Root cause: placeOrder has no idempotency key; only the client has a loading lock.
  • Suggested fix: Accept a client-generated idempotency key and short-circuit duplicate submissions.

FC-18 — AD-06: address region/bound coherence not validated

  • Severity: Medium
  • Case: AD-06
  • Symptom: An address can reference a bound from a different region.
  • Steps to reproduce: user.address.createAddress {regionId:2, boundId:1} (bound 1 is region 1's) → 200, address 882 stored region 2 / bound 1.
  • Expected: Rejected (bound must belong to the region).
  • Actual: Accepted.
  • Evidence: Address 882.
  • Root cause: apps/backend/src/trpc/apis/user-apis/apis/address.ts create/update do not check that boundId belongs to regionId. Compounding this, AddressForm submits the current browsing bound, so the client can send an incoherent pair.
  • Suggested fix: Validate bound.regionId === regionId server-side; have AddressForm derive the bound from the chosen address region, not the browsing region.

FC-19 — RG-02: region creation is not atomic on D1

  • Status: ✅ Fixed (2026-10-10) — createRegionWithBounds now validates the bounds before touching the DB (≥1 bound, unique priorities) so a bad payload never reaches the inserts, and wraps the post-region inserts in a try/catch that deletes the region + its child rows if anything fails. Reproduced (leftover region 5), then re-verified: the duplicate-priority create now leaves no region behind.
  • Severity: Medium
  • Case: RG-02
  • Symptom: A failed region create leaves a half-created region.
  • Steps to reproduce: superadmin.regions.create with two bounds at priority 1.
  • Expected: Error, no region persisted.
  • Actual: 500 "Each bound priority must be unique…" and region E2E-DupPriority exists with 0 bounds / 0 consts.
  • Evidence: Region row 4 with no bounds (removed manually).
  • Root cause: packages/db_helper_sqlite/src/superadmin-apis/regions.ts createRegionWithBounds wraps inserts in db.transaction, but on D1 the region insert is not rolled back when the bound insert violates the unique priority index.
  • Suggested fix: Validate bounds (unique priorities, valid geometry) before inserting, or use an atomic D1 batch that fails as a unit.

FC-20 — AU-03 (and D8): error statuses flattened to HTTP 500

  • Severity: Low
  • Case: AU-03 (same class as the cross-region getProductById 500)
  • Symptom: Auth/validation/not-found failures return 500 instead of 4xx.
  • Steps to reproduce: user.auth.login with a wrong password → 500; unknown user → 500; admin.product.getProductById cross-region → 500 "Product not found".
  • Expected: 401 / 404 respectively.
  • Actual: 500.
  • Evidence: Response statuses above.
  • Root cause: ApiError carries an HTTP status but apps/backend/src/app.ts onError / the tRPC error mapping only maps TRPCError codes, so ApiError falls through to 500.
  • Suggested fix: Translate ApiError (and thrown Error with a status) to the matching HTTP status in the global error handler / tRPC error formatter.

Summary

Ref Case Severity One-line
FC-01 SEC-01 Critical Customer token mutates global consts — ✅ fixed
FC-02 SEC-04 High Cross-region SKU accepted — ✅ fixed
FC-03 SEC-05 High Unassigned / OOS / closed-slot items accepted — ✅ fixed
FC-04 SEC-06 High User-based coupon used by another user — ✅ fixed
FC-05 SEC-07 High Public packaging mutation + expired-snippet leak
FC-06 BL-02 Medium Minimum order not enforced — ✅ resolved (constant removed)
FC-07 BL-03/06 High Client vs backend delivery-fee divergence — ✅ fixed
FC-08 BL-05 High % coupon discounts the delivery fee — ✅ fixed
FC-09 CP-05 High SKU-restricted coupon not enforced — ⚪ not a defect for now
FC-10 CP-11 High Coupons not region-scoped — ⚪ not a defect for now
FC-11 PD-13 High Store delete orphans products — ✅ fixed
FC-12 DS-02 High Tag delete fails (FK) — ✅ fixed
FC-13 SL-04 High Frozen slot accepted — ✅ fixed
FC-14 CT-07 High Cart ignores out-of-stock — ⚪ not a defect (false flag)
FC-15 AU-07 High Profile update broken — ✅ fixed
FC-16 AU-08 High Password change ineffective — ✅ fixed
FC-17 OR-12 Medium No order idempotency — 🟡 client-side guard (server deferred)
FC-18 AD-06 Medium Incoherent address region/bound
FC-19 RG-02 Medium Non-atomic region create — ✅ fixed
FC-20 AU-03 Low Errors return 500 instead of 4xx

Suggested fix order: FC-01 → FC-02/03/13 (order validation) → FC-07/08 (pricing) → FC-04/09/10 (coupons) → FC-05 (vendor auth) → FC-16/15 (account) → FC-11/12/19 → FC-17/18/20. (FC-09/FC-10/FC-14 closed as false flags.)