mirror of
https://github.com/luxfi/safe-wallet.git
synced 2026-07-27 06:55:00 +00:00
TimandGitHub
495d3cc054
* tests(mobile): add e2e coverage for WalletConnectGate flows Adds Maestro tests for the two non-signing gate states on the review-and-execute screen: - wc-gate-reconnect.yml: expired session renders "Reconnect wallet to continue"; tapping restores the session and reveals Execute transaction. - wc-gate-switch-network.yml: wrong-network state renders "Switch network to continue"; tapping clears the flag and reveals Execute transaction. Wires up the e2e control surface: - WalletConnectContext.e2e.tsx: mock reconnect() and switchNetworkIfNeeded() now flip walletConnectE2eState so the gate transitions correctly. - TestCtrls: e2eWcGateReconnect and e2eWcGateWrongNetwork buttons. - connectSignerSetup: setupWcGateReconnect / setupWcGateWrongNetwork seed a pending-tx safe with a WC signer in the desired gate state. Pin isows@1.0.7 in package.json resolutions. Part of WA-1858. * tests(mobile): extract shared review-and-execute nav for wc-gate tests Both wc-gate YAML tests duplicated ~30 lines navigating from app start to the Review-and-Execute screen of the first pending tx. Move that navigation into apps/mobile/e2e/utils/setup/open-review-and-execute.yml, parameterised by SETUP_BUTTON_ID, and have both gate tests `runFlow` it. The seeding button id is now the only piece of per-test configuration. * tests(mobile): assert wc-gate dismissal by id, not by neighbouring text The gate-dismissed assertion used a regex matching either "Execute transaction" or "Insufficient funds". Those are distinct semantic states (funded vs unfunded signer); coupling them in one assertion lets a funds-error regression silently pass the gate test. Replace the text regex with `extendedWaitUntil: notVisible` on the gate button id. That waits for re-render and asserts the gate is gone in one step, decoupling the test from the underlying execute-button copy. * tests(mobile): drop unread fields from WC e2e state overrides WalletConnectGate and useWalletConnectStatus only read `address`, `isWrongNetwork`, and `isWalletConnectSigner`. The setup helpers and the `reconnect` mock were also setting `hasProvider` and `walletInfo` — unobservable from any production code path, so they only added noise. - setupWcGateReconnect: drop the override entirely; post-reset defaults already model an expired session. - setupWcGateWrongNetwork: drop hasProvider and walletInfo. - WalletConnectContext.e2e.tsx reconnect mock: drop hasProvider. * tests(mobile): add `wc-gate` tag to connect-signer suite The two child flows (wc-gate-reconnect.yml, wc-gate-switch-network.yml) declare a `wc-gate` tag, but the parent __suite__.yml didn't surface it. Tag-based runs (`maestro test --include-tags wc-gate`) now also pick up the aggregated suite, matching repo convention. * tests(mobile): annotate wcSigner literal with explicit Signer type The literal relied on inference plus `as const` on `type`. With an explicit Signer annotation, future widening or narrowing of the discriminated union in signersSlice will surface here at compile time instead of failing silently in the e2e flow. The `as const` workaround is no longer needed. * tests(mobile): reuse setupPendingTxSafe from WC gate setup setupWcGateBase duplicated the pending-tx safe scaffolding already in setupHelpers.ts (base config, signer registration, active safe/signer). The only delta was the signer's `type: 'walletconnect'` and the extra setExecutionMethod dispatch. Extend setupPendingTxSafe with an optional `options.signer` override. Existing five private-key callers are unchanged. setupWcGateBase now constructs the WC Signer literal and delegates everything else to the shared helper. * docs(mobile): document `e2e<PascalCase>` exception for TestCtrls triggers The repo-wide testID convention is kebab-case, but every TestCtrls scenario-trigger button uses `e2eConnectSignerOwner`-style camelCase. That divergence was undocumented, so new contributors would either churn the convention or get pushed to match it without context. Codify the exception: product testIDs stay kebab-case; the hidden e2e-build-only TestCtrls triggers use `e2e<PascalCase>`. * docs: document `isows@1.0.7` resolution and add resolutions log The root package.json `resolutions` block had 12 entries with zero recorded rationale, so each new reviewer has to investigate why a pin exists. Add `docs/resolutions.md` as the canonical home for that rationale and seed it with the `isows@1.0.7` entry — a benign dedup (no CVE) added to align the transitive version under viem. Future contributors are asked to update the log alongside any new resolutions entry.
654 B
654 B
Yarn resolutions rationale
The root package.json resolutions block forces specific
versions of transitive dependencies. Each entry has a reason — without one
it's impossible to tell, months later, whether a pin is still load-bearing
or safe to drop. Add a short note here whenever a new entry is added.
Entries
isows: 1.0.7— dedup the transitive version that ships under viem (1.0.6 → 1.0.7 is a maintenance bump, no CVE; the pin avoids two copies in the bundle when other packages resolve isows differently).
When you add an entry to
resolutions, add the matching line here in the same commit.