TimeCurveProtocolWarbowRefreshSection: missing Vitest coverage for permissionless write path

Brouie commented

2026-05-07 02:02:51 +00:00

(Migrated from gitlab.com)

Severity: P4 — test coverage

split out of #160 follow-up per @PlasticDigits — "all of these findings need to have issues opened for them".

context

TimeCurveProtocolWarbowRefreshSection.tsx (frontend/src/pages/timecurve/) is a permissionless write path covering candidate-load + on-chain podium refresh. landed via a648283 (#160 / EcoStrategy M-04) plus fff7946 (#170 P3 polish) plus #170 follow-up adjustments.

no Vitest test file exists for this component at HEAD 284c15c:

$ ls /srv/qa/repos/yieldomega/frontend/src/pages/timecurve/ | grep -iE "warbow.*refresh"
TimeCurveProtocolWarbowRefreshSection.tsx

sister flows in the same directory have tests (TimeCurveSimplePodiumSection.test.tsx, TimeCurveStakeAtLaunchSection.test.tsx, TimeCurveSubnav.test.tsx).

why it matters

permissionless write path with non-trivial state machine — fresh-load reset, pagination accumulation, distinct_sql_cap_hit OR-aggregation across pages, dedup vs podium hints + connected wallet, sale-ended early return, wrong-chain barrier, error UX. exactly the kind of surface where regressions ride in via unrelated refactors.

related #160 follow-up findings (separate tickets): #173 cosmetic double-call, #174 silent truncation, #172 CRITICAL gas-blowup design. each of these would benefit from regression coverage that does not exist today.

suggested coverage

minimum useful set:

loadFromIndexer happy path: paginates across next_offset, accumulates candidates, sets extras (apiTotal, podiumHintCount, distinct_sql_cap_hit OR-folded across pages)
loadFromIndexer indexer-returns-null branch (line 102): clears candidates + extras, sets loadErr
loadFromIndexer exception branch (line 131): friendlyRevertFromUnknown(e), candidates + extras cleared
loadFromIndexer malformed chunk.candidates (non-array): does not throw, ends with empty load + appropriate error state (covers #174 + the Array.isArray hardening if it lands)
loadFromIndexer reset-on-fresh-load: prior successful state cleared at top of new load (relevant when #160 finding 2 / "P3 reset-on-reload" lands)
runRefresh sale-ended branch: early return with #149 message
runRefresh chain-mismatch branch: early return with helper message, single invocation of chainMismatchWriteMessage (covers #173 once that fix lands)
runRefresh no-candidates branch: prompt to load first
runRefresh happy path: warBowRefreshCandidateAddresses calldata, writeContractAsync invocation, refetch hooks called on success

acceptance criteria

new test file frontend/src/pages/timecurve/TimeCurveProtocolWarbowRefreshSection.test.tsx
minimum cases listed above covered (or fewer, with rationale)
cd frontend && npm test -- --run passes including new file
no flake on repeat invocations

origin

split from #160 follow-up notes ("P4 — no Vitest unit/component test ships with the new section"). filing now per @PlasticDigits.

sister flow useTimeCurveArenaModel.tsx:2615-2641 is similarly untested per the original note. that is a separate scope and not covered by this ticket.

cc @PlasticDigits

**Severity:** P4 — test coverage split out of #160 follow-up per @PlasticDigits — "all of these findings need to have issues opened for them". ## context `TimeCurveProtocolWarbowRefreshSection.tsx` (`frontend/src/pages/timecurve/`) is a permissionless write path covering candidate-load + on-chain podium refresh. landed via `a648283` (#160 / EcoStrategy M-04) plus `fff7946` (#170 P3 polish) plus #170 follow-up adjustments. no Vitest test file exists for this component at HEAD `284c15c`: ``` $ ls /srv/qa/repos/yieldomega/frontend/src/pages/timecurve/ | grep -iE "warbow.*refresh" TimeCurveProtocolWarbowRefreshSection.tsx ``` sister flows in the same directory have tests (`TimeCurveSimplePodiumSection.test.tsx`, `TimeCurveStakeAtLaunchSection.test.tsx`, `TimeCurveSubnav.test.tsx`). ## why it matters permissionless write path with non-trivial state machine — fresh-load reset, pagination accumulation, `distinct_sql_cap_hit` OR-aggregation across pages, dedup vs podium hints + connected wallet, sale-ended early return, wrong-chain barrier, error UX. exactly the kind of surface where regressions ride in via unrelated refactors. related #160 follow-up findings (separate tickets): #173 cosmetic double-call, #174 silent truncation, #172 CRITICAL gas-blowup design. each of these would benefit from regression coverage that does not exist today. ## suggested coverage minimum useful set: - `loadFromIndexer` happy path: paginates across `next_offset`, accumulates candidates, sets extras (apiTotal, podiumHintCount, distinct_sql_cap_hit OR-folded across pages) - `loadFromIndexer` indexer-returns-null branch (line 102): clears candidates + extras, sets loadErr - `loadFromIndexer` exception branch (line 131): `friendlyRevertFromUnknown(e)`, candidates + extras cleared - `loadFromIndexer` malformed `chunk.candidates` (non-array): does not throw, ends with empty load + appropriate error state (covers #174 + the Array.isArray hardening if it lands) - `loadFromIndexer` reset-on-fresh-load: prior successful state cleared at top of new load (relevant when #160 finding 2 / "P3 reset-on-reload" lands) - `runRefresh` sale-ended branch: early return with #149 message - `runRefresh` chain-mismatch branch: early return with helper message, single invocation of `chainMismatchWriteMessage` (covers #173 once that fix lands) - `runRefresh` no-candidates branch: prompt to load first - `runRefresh` happy path: `warBowRefreshCandidateAddresses` calldata, `writeContractAsync` invocation, refetch hooks called on success ## acceptance criteria - [ ] new test file `frontend/src/pages/timecurve/TimeCurveProtocolWarbowRefreshSection.test.tsx` - [ ] minimum cases listed above covered (or fewer, with rationale) - [ ] `cd frontend && npm test -- --run` passes including new file - [ ] no flake on repeat invocations ## origin split from #160 follow-up notes ("P4 — no Vitest unit/component test ships with the new section"). filing now per @PlasticDigits. sister flow `useTimeCurveArenaModel.tsx:2615-2641` is similarly untested per the original note. that is a separate scope and not covered by this ticket. cc @PlasticDigits

Brouie commented

2026-05-07 02:06:00 +00:00

(Migrated from gitlab.com)

mentioned in issue #176

Brouie commented

2026-05-07 02:09:58 +00:00

(Migrated from gitlab.com)

mentioned in issue #160

Brouie commented

2026-05-07 03:19:40 +00:00

(Migrated from gitlab.com)

mentioned in merge request !377

Brouie commented

2026-05-07 04:13:13 +00:00

(Migrated from gitlab.com)

mentioned in merge request !378

Brouie commented

2026-05-07 23:32:21 +00:00

(Migrated from gitlab.com)

@PlasticDigits — did a deep-dive on the test scope before writing any code. surfaced one finding worth surfacing before i commit to a path.

the finding

there is no in-repo precedent for testing a wagmi-coupled component. every existing .test.tsx uses renderToStaticMarkup (server-side, no hooks). zero hits for vi.mock("wagmi", @testing-library/react, or @vitest-environment per-file switches across the whole frontend.

closest analogue is referralRegisterWalletSession.test.ts:16-22 which mocks ONE function from wagmi/actions. TimeCurveProtocolWarbowRefreshSection would need the full wagmi hook surface (useAccount, useChainId, useReadContracts, useWriteContract) plus wagmi/actions.waitForTransactionReceipt plus the writeContractWithGasBuffer helper plus a per-file env switch to happy-dom or jsdom to fire button clicks. that's net-new test infrastructure for this codebase.

branch inventory at HEAD `efb6e0b`

loadFromIndexer (lines 91-141) — 9 distinct branches: happy single-page, happy multi-page pagination + apiTotal/podiumHintCount captured on first chunk + capHit OR-aggregated, indexer-returns-null, malformed candidates non-array, pagination guard cap at 50, throw-in-loop with friendlyRevertFromUnknown, finally-clears-loading, undefined-vs-null next_offset both break, empty-pages-after-success.

runRefresh (lines 143-191) — 8 distinct branches: saleEnded short-circuit (#149), chain-mismatch short-circuit, address-undefined short-circuit, wbReads-not-loaded short-circuit, no-candidates-loaded short-circuit, happy submit with calldata + softCap 12_000_000n + onEstimateRevert "rethrow" + waitForTransactionReceipt + dual refetch, write rejection with friendlyRevertFromUnknown, useEffect-clears-txErr-on-chain-correction.

plus 6 render-side branches (warbowPodiumFinalized fallback, idxExtras conditional rows, saleEnded placeholder, refresh button disabled cascade, load button label flip, error StatusMessage variants).

20 test cases mapped end-to-end. 7-8 reachable today via renderToStaticMarkup. the rest depend on testing-library + dom env, OR a refactor.

three paths

(A) refactor. extract loadFromIndexer and runRefresh into pure dependency-injected helpers in a sibling .ts file (deps: fetchTimecurveWarbowRefreshCandidates, writeContractWithGasBuffer, waitForTransactionReceipt, the friendly-error helper, etc.). component file becomes thinner — just hook glue + render. unit-test the helpers like writeContractWithGasBuffer.test.ts does. consistent with existing patterns. no new dev dependencies. larger PR diff, but adjacent code stays clean.

(B) add testing infrastructure. install @testing-library/react, set per-file env via // @vitest-environment happy-dom, mock the full wagmi hook surface in one block, write component-integration tests. smaller diff, but introduces a new testing pattern that diverges from the rest of the repo. one new dev dependency.

(C) defer. park #175 until either path A or B comes up naturally. ticket stays open as a known coverage gap. P4 severity supports this.

my read

(A) is the cleanest fit — same idiom as writeContractWithGasBuffer which we just landed in !378. extracted helpers are reusable for future work and trivial to mock at the boundary. (B) introduces test-shape divergence in the codebase. (C) is fine if you want to push #175 down the priority list, but worth flagging that the ticket as filed assumed (B) shape.

happy to ship (A) — would land as a refactor MR (component split + helpers + ~12-15 helper unit tests + ~7-8 render tests for the component shell). ETA reasonable from here. flag if you'd rather (B) or (C).

@PlasticDigits — did a deep-dive on the test scope before writing any code. surfaced one finding worth surfacing before i commit to a path. ## the finding there is no in-repo precedent for testing a wagmi-coupled component. every existing `.test.tsx` uses `renderToStaticMarkup` (server-side, no hooks). zero hits for `vi.mock("wagmi"`, `@testing-library/react`, or `@vitest-environment` per-file switches across the whole frontend. closest analogue is `referralRegisterWalletSession.test.ts:16-22` which mocks ONE function from `wagmi/actions`. `TimeCurveProtocolWarbowRefreshSection` would need the full wagmi hook surface (`useAccount`, `useChainId`, `useReadContracts`, `useWriteContract`) plus `wagmi/actions.waitForTransactionReceipt` plus the `writeContractWithGasBuffer` helper plus a per-file env switch to `happy-dom` or `jsdom` to fire button clicks. that's net-new test infrastructure for this codebase. ## branch inventory at HEAD efb6e0b `loadFromIndexer` (lines 91-141) — 9 distinct branches: happy single-page, happy multi-page pagination + apiTotal/podiumHintCount captured on first chunk + capHit OR-aggregated, indexer-returns-null, malformed candidates non-array, pagination guard cap at 50, throw-in-loop with friendlyRevertFromUnknown, finally-clears-loading, undefined-vs-null next_offset both break, empty-pages-after-success. `runRefresh` (lines 143-191) — 8 distinct branches: saleEnded short-circuit (#149), chain-mismatch short-circuit, address-undefined short-circuit, wbReads-not-loaded short-circuit, no-candidates-loaded short-circuit, happy submit with calldata + softCap 12_000_000n + onEstimateRevert "rethrow" + waitForTransactionReceipt + dual refetch, write rejection with friendlyRevertFromUnknown, useEffect-clears-txErr-on-chain-correction. plus 6 render-side branches (warbowPodiumFinalized fallback, idxExtras conditional rows, saleEnded placeholder, refresh button disabled cascade, load button label flip, error StatusMessage variants). 20 test cases mapped end-to-end. 7-8 reachable today via `renderToStaticMarkup`. the rest depend on testing-library + dom env, OR a refactor. ## three paths (A) **refactor.** extract `loadFromIndexer` and `runRefresh` into pure dependency-injected helpers in a sibling `.ts` file (deps: `fetchTimecurveWarbowRefreshCandidates`, `writeContractWithGasBuffer`, `waitForTransactionReceipt`, the friendly-error helper, etc.). component file becomes thinner — just hook glue + render. unit-test the helpers like `writeContractWithGasBuffer.test.ts` does. consistent with existing patterns. no new dev dependencies. larger PR diff, but adjacent code stays clean. (B) **add testing infrastructure.** install `@testing-library/react`, set per-file env via `// @vitest-environment happy-dom`, mock the full wagmi hook surface in one block, write component-integration tests. smaller diff, but introduces a new testing pattern that diverges from the rest of the repo. one new dev dependency. (C) **defer.** park #175 until either path A or B comes up naturally. ticket stays open as a known coverage gap. P4 severity supports this. ## my read (A) is the cleanest fit — same idiom as `writeContractWithGasBuffer` which we just landed in !378. extracted helpers are reusable for future work and trivial to mock at the boundary. (B) introduces test-shape divergence in the codebase. (C) is fine if you want to push #175 down the priority list, but worth flagging that the ticket as filed assumed (B) shape. happy to ship (A) — would land as a refactor MR (component split + helpers + ~12-15 helper unit tests + ~7-8 render tests for the component shell). ETA reasonable from here. flag if you'd rather (B) or (C).

PlasticDigits (Migrated from gitlab.com) closed this issue

2026-05-09 04:56:17 +00:00

PlasticDigits commented

2026-05-09 04:56:18 +00:00

(Migrated from gitlab.com)

Closing as we are redesigning the warbow podium to be set by trusted governance instead of onchain due to issues from warbow point decreases

Rows
Columns

TimeCurveProtocolWarbowRefreshSection: missing Vitest coverage for permissionless write path #175

context

why it matters

suggested coverage

acceptance criteria

origin

the finding

branch inventory at HEAD efb6e0b

three paths

my read

branch inventory at HEAD `efb6e0b`