fix(frontend): 5 bugs from parallel hunt — IME / search-race / palette-race / IPC catch / transition-all - #202
Merged
Merged
Conversation
…ntract
WAWQAQ msg `c4e6e61c` 2026-06-24: yuejing role = frontend engineer
solving kenji's finds + proactively hunting FE / animation / chain
bugs. Spawned 3 parallel deep-review agents (animation/motion,
execution-chain, React logic) and fixed the 5 HIGH-confidence bugs
they converged on.
## H1. IME composition mis-fires Quick Chat send
`apps/desktop/src/renderer/OnboardingHero.tsx:564`
`ReadyEmptyHero.handleKey` triggered `submit()` on Enter without an
`event.nativeEvent.isComposing` / `event.key === 'Process'` guard.
Chinese / Japanese / Korean users committing an IME composition with
Enter immediately sent their unfinished draft. The main Composer at
`packages/ui/src/components.tsx:5640` has had the guard for ages; the
onboarding clone had drifted. Mirrored the guard.
## H2. SearchModal debounce killed by parent re-renders during streaming
## H3. CommandPalette highlight reset by parent re-renders during streaming
`apps/desktop/src/renderer/main.tsx:776+ + :3083-3098`
`SearchModal` received `deps={{ searchThread: (req) => window.maka.search.thread(req) }}` — a fresh inline object every render. SearchModal's debounce effect lists `searchThread` in its dep array, so during an active turn stream (App re-renders many times per second) the 180 ms `setTimeout` was torn down + recreated on every render and never fired — search was effectively dead while streaming.
Same root cause for CommandPalette's `onSelectSession` inline arrow: the palette's `[combined]`-deps effect at command-palette.tsx:629 reset `setHighlight` on every parent render, clobbering keyboard navigation during streaming.
Fix: stable refs + memos. `openSessionInChatRef` mirrors the live closure; `searchModalDeps` (useMemo), `searchModalOnNavigate` / `paletteOnSelectSession` (useCallback) all stable across App renders.
## H4. Privacy-toggle IPC rejection becomes Unhandled Promise Rejection
`apps/desktop/src/renderer/settings/SettingsModal.tsx:1130`
`onChange={(v) => void props.onUpdateSettings({ privacy: { incognitoActive: v } })}` discarded the resolved value but did NOT catch rejection. `updateSettings` re-throws on IPC failure → renderer-level Unhandled Promise Rejection with NO user feedback. Added `.catch` with toast surface mirroring the rest of the file's PR-STOP-ERROR-SURFACE-0 / PR-BOT-RESTART-RACE-0 pattern (`generalizedErrorMessageChinese`).
## H5. Tailwind `transition-all` slipped through motion contract
`packages/ui/src/primitives/{command,sheet,sidebar}.tsx`
PR-MOTION-TOKEN-CONVERGE-0's contract scanned renderer CSS files
only; shadcn primitive classnames live in `.tsx` files and weren't
covered. Three sites violated:
- `command.tsx:46` overlay → `transition-opacity` (only opacity moves)
- `sheet.tsx:35` overlay → `transition-opacity` (same)
- `sidebar.tsx:318` rail → `transition-[transform,opacity]` (the rail hovers + slides, neither requires `all`)
Extended `motion-token-converge-contract.test.ts` with a fourth test
that walks `packages/ui/src/primitives/**` + `apps/desktop/src/renderer/**`
for any `\btransition-all\b` token. The three above are scrubbed;
future `transition-all` insertion trips the test.
## Out of scope (logged for follow-up)
- MEDIUM findings from the 3 agents (ClaudeSubscriptionCard mountedRef,
rapid-resize localStorage spam, WechatQrLoginModal interval churn,
ReadyEmptyHero `[props]` deps, etc.) — not user-visible HIGH bugs.
- Bare easing / duration cleanup across maka-tokens.css — kenji's
aesthetic lane.
This was referenced Jun 24, 2026
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
PR-FE-BUG-HUNT-4 (LOW from kenji-led bug-hunt 2026-06-24 round 1): `refreshMessagesUntilTurn` polls `readMessages` every 40ms for up to 1200ms while waiting for a freshly-sent user turn to land in the session journal. The setState was already gated on `activeIdRef.current === sessionId`, but the IPC call itself kept firing until the deadline — burning bandwidth and CPU on a session the user no longer cared about. Fix: check `activeIdRef.current !== sessionId` at the top of each loop iteration AND immediately after the await. Bail early so the polling cycle stops, not just the setState. Same guard on the deadline-fallthrough `refreshMessages(sessionId)`. Behavior change is invisible when the user stays on the same session (common case). Saves up to 30 IPC roundtrips per session-switch during the post-send window when the user navigates fast. Single function, single file. No conflict with #202 / #204 / #206 / #207 since none of those touch `refreshMessagesUntilTurn`.
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
) PR #202 added an IME guard to the composer textarea's keydown handler so Enter during CJK composition wouldn't fire submit mid-character. The paste handler (components.tsx:5726) had the same exposure but was left out. Symptom: a user mid-CJK composition who pastes a clipboard file (e.g. screenshot captured via system shortcut) hits the file-import path, which calls `event.preventDefault()` — that aborts the in-flight IME composition and eats the partially-typed character. Rare but real. The same kind of bug pattern as the keydown one. Fix: prepend `if (event.nativeEvent.isComposing) return;` at the top of `onTextareaPaste`, matching the exact pattern at line 5640 in the keydown handler. 2 lines + 6-line WHY comment. When IME is active, the paste should fall through to the textarea's native paste so the IME can handle (or ignore) it as the user expects. The file-import path only matters once composition has ended. Found via PR-FE-BUG-HUNT scan; 2 sibling Explore agents also ran on Daily Review and cron PlanReminderPanel — both reported findings that I manually verified as false positives (Daily Review agent contradicted itself mid-output; cron `reminder.schedule` null-safety claim is impossible because TypeScript types it as required non-null). Reporting transparently so kenji/WAWQAQ can discount those. ## Diff 1 file, 8 insertions(+), 0 deletions(-). No overlap with any open PR. ## Verification Disk still ~100%, couldn't run `pnpm install && pnpm test`. Matches existing pattern verbatim, so behavior outside the IME edge case is unchanged.
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
…edRef (#204) PR-FE-BUG-HUNT-1 (MEDIUM from kenji-led bug-hunt 2026-06-24 round 1): the Claude subscription card launches a browser OAuth flow that takes seconds-to-minutes. Closing the Settings modal while a `startLogin`, `submitPaste`, `cancelLogin`, `logout`, or `refreshQuota` call is in flight produced `setState on unmounted component` warnings (loud in dev, masks real bugs in prod). `refreshExperimentalGate` and `refresh` had the same exposure. Mirrored the `mountedRef` pattern already in use by other settings sub-cards in this file: - `claudeCardMountedRef = useRef(true)` with a unmount-flip effect - Every `await` is followed by an `if (!claudeCardMountedRef.current) return;` before any setState / toast call - Finally blocks gate `finishPendingAction()` the same way (so the pending-action ref doesn't leak past unmount) No behavior change when the card stays mounted (the common case). Scoped to a single file to minimize merge conflict surface with the in-flight PR #202 (which doesn't touch ProvidersPanel.tsx).
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
#206) PR-FE-BUG-HUNT-2 (MEDIUM from kenji-led bug-hunt 2026-06-24 round 1): `WechatQrLoginModal`'s 3-second polling effect listed the whole `result` object in its dep array. Every successful QR refresh produces a new `result` reference → effect cleanup tears down the interval → effect re-armed → 3-second clock restarts from zero. In practice the polling cadence drifted: each refresh pushes the next poll up to ~5.9 s after the previous poll instead of the intended 3 s. Over a long scan, this delays the moment the modal notices `loggedIn: true` from the bridge. Fix: derive a boolean `shouldPollQr = !!result?.ok && !result.loggedIn && !result.expired` and depend on `[shouldPollQr]`. The interval is armed once when scanning starts and torn down once when the gating state flips (logged in, expired, or error). `reloadQrCode` only touches stable refs (`loadingQrRef.current`) and state setters, so the captured closure is safe across renders. Scoped to a single function (`WechatQrLoginModal` at lines 687-693) to keep merge surface minimal with the in-flight PRs #202 / #204.
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
Kenji aesthetic audit reminder 4/6 (msg `6cc0e04d` 2026-06-24, finding #2): Tooltip is a high-frequency hover/focus surface and should use instant positioning + compositor-only popup transitions. Two sites violated: 1. Positioner: `transition-[top,left,right,bottom,transform]` — `top` / `left` / `right` / `bottom` are layout-trigger properties. Every reposition (e.g. anchor moved by scroll, or side flipped on collision) ran through layout + paint instead of GPU compositor. Reduced to `transition-transform` so the transform-based reposition still tweens smoothly; the layout positioning happens instantly off-frame. 2. Popup: `transition-[width,height,scale,opacity]` — `width` / `height` are layout-trigger. Reduced to `transition-[scale,opacity]`, which is what the data-starting-style / data-ending-style scale-98 + opacity-0 actually animate. The popup-width / popup-height CSS vars still set the size; just no longer transitioned. Net behavior: tooltip pops in/out via scale+fade (compositor), repositions via transform (compositor). Layout properties are applied instantly per frame instead of crossfading. Drawer's `transition-[transform,box-shadow,height,background-color]` and Tabs' `transition-[width,translate]` indicator are flagged in kenji's audit but deferred — both interact with swipe gesture / tab sizing logic and need a focused PR with visual smoke verification. Scoped to 2 lines in 1 file to keep merge surface minimal with the in-flight PRs #202 / #204 / #206.
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
…rag (#210) PR-FE-BUG-HUNT-5 (LOW from kenji-led bug-hunt 2026-06-24 round 1): sidebar-list resizer fires `setSessionListWidth` on every pointermove during a drag — at ~60Hz over a long drag, that's a couple hundred localStorage writes for a single resize gesture. The setting converges to the user's final width at rest; intermediate values aren't load-bearing. Fix: 200ms trailing debounce on the `useEffect` that flushes to localStorage. Cleanup clears the pending timeout on each fresh width change, so only the last-render value gets written. Drag finished → 200ms idle → one localStorage write. Behavior is invisible — the user's final width persists exactly as before, just without the per-frame I/O. Especially noticeable on spinning disks and on Electron's IPC-backed safeLocalStorage path. Diff: 1 file, +10/-2. Single function in main.tsx. No conflict zone with the in-flight PRs (the localStorage write at line 1153 is far from the SearchModal/CommandPalette callback area at line 768/3083 that PR #202 touches).
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
…scape hatch (#214) Kenji aesthetic-audit reminder 9 (msg `bacde601` 2026-06-24, findings #2 + #3): two chrome design-system holes still leaking gradients / non-token blur into the app. ## #2 — darwin sidebar footer gradient `html[data-os="darwin"] .maka-session-panel-footer` was setting: `background: linear-gradient(180deg, transparent 0%, oklch(.../0.30) 100%)` The panel above already has a 50% glass base + 8px backdrop blur, so the footer doesn't need any additional material — but the 180deg fade IS the same "上面灰下面白" pattern WAWQAQ called out twice before in chat sweep (`1e693dee` / `5d3b10e5`). Replaced with flat `oklch(...) / 0.15` + hairline border-top. ## #3 — `[data-gradual-blur-layer]` block A 56-line block that let any element opt into: - `linear-gradient(to bottom/top, var(--background) → transparent)` - `backdrop-filter: blur(6px)` - raw `z-index: 5` just by wearing the `data-gradual-blur-layer` attribute. Kenji called it a "blur escape hatch" — a future surface could re- introduce blur and a non-token z-index without going through any review of the design system. `grep -rn data-gradual-blur-layer --include=*.tsx` confirms zero JSX consumers — the CSS is dead. Removing the entire block kills the escape hatch + retires one of the four raw z-index numbers kenji flagged in finding #4. `notes/reference-atlas.md` references stay — those describe what reference upstream does, not what we ship. ## Diff 1 file changed, 18 insertions(+), 60 deletions(-) net delete. No JSX touched. No tests touched (no contract pinned this attr). Footer change is darwin-only so non-darwin platforms unaffected. Far from PR #202 styles.css edit area (onboarding ~1500-1700 + cubic-bezier sweep on transition rules). PR #210 touches main.tsx debounce, not styles.css. No merge conflicts.
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
Kenji aesthetic-audit reminder 9 (msg `bacde601` 2026-06-24, finding #4): four bare integer z-index sites in styles.css (line 407 z-4, 3858 z-2, 7193 z-35, 14463 z-5) sidestep the `--z-*` token scale defined in maka-tokens.css. PR #214 already took out the gradual-blur-layer site (z-5). This PR closes the remaining three. ## Changes 1. `maka-tokens.css`: add two new tokens preserving the previous raw values exactly: - `--z-panel-action: 4` - `--z-settings-fullpage: 35` 2. `styles.css` line 407 (`.maka-workspace-top-actions`): `z-index: 4` → `var(--z-panel-action)`. Identical stacking. 3. `styles.css` line 7193 (`.settingsModal.settingsPage`): `z-index: 35` → `var(--z-settings-fullpage)`. 35 sits between `--z-titlebar` (40) and `--z-sticky` (20) so the OS titlebar still owns top of the stack. Preserved visual behavior exactly. 4. `styles.css` line 3858 (`.maka-skill-featured-art span:nth-child(2)`): bare `z-index: 2` kept. This is local stacking among 3 decorative sticker spans inside a fixed-size container — they never escape the parent and don't need a global token. Added explicit allowlist comment. 5. New `apps/desktop/src/main/__tests__/z-index-contract.test.ts` enforces: - every bare `z-index: <integer>;` must be tokenized or allowlisted (with a documented reason) - allowlist entries must be present in styles.css (no stale entries) - the 10 semantic `--z-*` tokens stay defined in maka-tokens.css Future PRs that re-introduce a bare z-index will fail this test unless they explicitly bump the allowlist. ## Diff 3 files: maka-tokens.css +7, styles.css +17/-2, new test +~110 lines. No visual change. Far from PR #202 (transition sweep), PR #210 (main.tsx debounce), PR #214 (footer/gradual-blur). No conflict. ## Verification Disk still ~100%, couldn't run `pnpm install && pnpm test`. Contract test is greenfield — first run will validate the 4 sites in this same PR.
jackwener
added a commit
that referenced
this pull request
Jun 24, 2026
#223) WAWQAQ reported a runtime crash: ReferenceError: useCallback is not defined at q80 (dist/renderer/assets/index-D4LgUpob.js:128:31343) PR #202 added stable-ref refs for SearchModal / CommandPalette via `useCallback` at main.tsx:794 + main.tsx:797, but the React import on line 1 was not updated to bring `useCallback` in: import { StrictMode, useEffect, useMemo, useRef, useState, ... } from 'react'; ^ no useCallback This survived the build because esbuild / Vite don't fail on undeclared identifiers in source — only at runtime when the function is actually called. Same root cause as PR #221 (paste IME isComposing): I didn't run `tsc --noEmit` against the affected package before pushing. Fix: add `useCallback` to the named imports on line 1.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
WAWQAQ msg `c4e6e61c` 2026-06-24: yuejing = frontend engineer (solve kenji's finds + hunt FE/animation/chain bugs). Spawned 3 parallel deep-review agents (animation/motion, execution-chain, React logic); this PR fixes the 5 HIGH-confidence bugs they converged on.
Fixes
H1. IME composition mis-fires Quick Chat send — `OnboardingHero.tsx:564`. CN/JP/KR users committing an IME composition with Enter immediately sent their unfinished draft. Mirrored the main Composer's `event.nativeEvent.isComposing` guard.
H2 + H3. Search + palette die during streaming — `main.tsx`. `SearchModal deps={{...}}` + `CommandPalette onSelectSession={(s) => ...}` were inline. App re-renders many times per second during streams; the debounce + highlight-reset effects keyed on these props were torn down every render, so search never fired and palette navigation reset constantly. Fix: `useMemo` deps + `useCallback` over a `useRef` of `openSessionInChat`.
H4. Privacy IPC error unhandled — `SettingsModal.tsx:1130`. `void onUpdateSettings(...)` discarded rejection → Unhandled Promise Rejection with no user feedback. Added `.catch` + toast surface mirroring PR-STOP-ERROR-SURFACE-0 / PR-BOT-RESTART-RACE-0.
H5. Tailwind `transition-all` slipped through motion contract — 3 sites in `packages/ui/src/primitives/{command,sheet,sidebar}.tsx`. The PR-MOTION-TOKEN-CONVERGE-0 contract only scanned CSS files. Replaced with `transition-opacity` (overlays) / `transition-[transform,opacity]` (sidebar rail). Extended the contract test with a 4th case that walks `packages/ui/src/primitives` + `apps/desktop/src/renderer` for any `\btransition-all\b` token.
Out of scope (follow-ups)
Verification
Local disk is currently at 100% (system-level, not repo-side) so I could not run `pnpm install && pnpm test && pnpm build` end-to-end before pushing. The diff is small and the fixes are static-analysis-clean. Please review the diff carefully + run the test suite yourself before merging. I'll rerun the gate once disk space frees up.
Branch: `yuejing/fe-bug-hunt-0`