fix(splash): move handoff to backend app://ready listener (closes #42)#43
Merged
Conversation
v1.1.0's splash → main window handoff lived in the frontend
(src/main.tsx): two rAFs after ReactDOM.render(), then
`getCurrentWindow().show()` + `splash.close()` over IPC. On Linux
WebKitGTK 2.52+ (Fedora 44, recent Arch), the heavy first-launch
init (sqlx migrations + SQLite pool open + library scan) took
longer than the rAF window, so the show() either failed silently
or revealed an empty webview — splash hung forever, main window
never appeared. Ubuntu showed the same race on first launch only
(second launch cached the init, won the race).
Move the handoff to native code:
- src-tauri/src/lib.rs: setup() now installs an `app://ready`
event listener that calls a new `reveal_main_close_splash`
helper (show main → focus → close splash, same ordering as
before). A 15 s fallback timer fires the same helper if the
event never arrives (frontend crash before render).
- src/main.tsx: revealMainWindow becomes signalReady — keep the
double rAF to let React commit + the compositor paint at least
one frame, then `emit("app://ready")`. No more IPC show/close.
- src-tauri/tauri.conf.json + public/splash.html: drop
`transparent: true` on the splash window (and switch the body
background to opaque #121212). Transparency forced an alpha-
capable EGL config that some WebKitGTK builds reject, adding
a second EGL failure mode on top of the timing bug.
Tested on Windows 11 (no regression), pending Linux confirmation
on Fedora 44 + Ubuntu via CI-built bundles.
github-actions
Bot
added
scope: frontend
|
Warning Rate limit exceeded
To continue reviewing without waiting, purchase usage credits in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (10)
📝 WalkthroughWalkthroughCe PR remplace le handoff frontend synchronisé par un signal : le frontend émet ChangesSplash handoff event-driven
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src-tauri/src/lib.rs`:
- Around line 385-403: Les deux chemins (la closure sur app.listen et la tâche
de fallback) marquent le handoff via handoff_done_for_event / handoff_done
(swap(true,...)) avant d'appeler reveal_main_close_splash, ce qui empêche tout
retry si reveal échoue; changez la logique pour n'écrire le flag qu'après un
reveal réussi: appelez reveal_main_close_splash(&handoff_handle) (et
&fallback_handle) d'abord, gérez/propaguez son éventuel Result/erreur, et
n'exécutez swap(true, Ordering::SeqCst) que si le reveal a réussi (ou utilisez
compare_exchange pour atomiquement marquer comme fait après succès); en cas
d'échec, loggez l'erreur et ne marquez pas le handoff comme terminé pour
permettre au fallback/événement de retenter.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 645807c7-0747-4699-adce-1664e6be61e3
📒 Files selected for processing (4)
public/splash.htmlsrc-tauri/src/lib.rssrc-tauri/tauri.conf.jsonsrc/main.tsx
💤 Files with no reviewable changes (1)
- src-tauri/tauri.conf.json
Previously both the `app://ready` listener and the 15s fallback timer marked the handoff as done *before* calling the reveal helper. If main.show() or splash.close() failed transiently (e.g. a brief WebKitGTK glitch on first reveal), there was no path back to a working UI — the flag stayed `true`, the listener short-circuited, the fallback was already past, and the user was stuck on an eternal splash. Make `reveal_main_close_splash` return a `bool` (true iff both ops succeeded), and have both call-sites roll the `done` flag back on failure so the other path (or a subsequent ready re-emission) can retry. Per coderabbit review on PR #43.
WebKitGTK 2.52 suspends requestAnimationFrame callbacks while a window is hidden. Since the main window is created with `visible: false` until the backend reveals it, the 2-rAF dance in signalReady() never fired, deadlocking the handoff for the full 15 s safety-net window. useEffect runs after the first React commit, which is the real guarantee we needed — the compositor paints as part of the reveal itself.
…n windows Without `transparent: true`, the WebView2 control on Windows shows its default white background until splash.html is parsed and the CSS `background: #121212` applies, producing a brief white flash. Setting the native window backgroundColor paints the OS-level window dark before WebView2 emits its first frame.
The Now Playing share PNG looked blurry / pixelated because the canvas pulled `track.artwork_path_2x` first (a 128×128 thumbnail from src-tauri/src/thumbnails.rs). On the 1080×1080 card that meant upscaling 4.5× for the 580 px cover slot and ~10× for the 1320 px backdrop — both visibly soft, the backdrop especially mushy because the old code "faked" a blur by drawing the cover huge over the canvas + a dark wash. - Prefer `track.artwork_path` (the original cover the scanner extracted from the audio file, typically 500–1500 px) and fall back to the thumbnails only when the original is missing. - Set `imageSmoothingQuality = "high"` for the residual upscale to 580 px so even small originals stay sharp instead of bilinear-soft. - Switch the backdrop from the upscale + wash hack to a real `ctx.filter = "blur(60px) brightness(0.6)"`, which both Tauri's Windows WebView2 (Chromium) and Linux/macOS WebKitGTK 2.40+ support natively. Cleaner backdrop, no quality loss on the foreground cover.
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src-tauri/src/lib.rs`:
- Around line 709-728: In reveal_main_close_splash, avoid closing the splash
when the main window wasn't shown: only call splash.close() after main.show()
succeeded (or main existed and set_focus completed); e.g. track success with a
flag (or reuse ok) set to true only when app.get_webview_window("main") returns
Some and main.show() did not Err, and move or gate the splash.close() call
behind that flag so splash is not closed on main.show() failure or when main is
missing.
- Around line 401-412: Le fallback actuel dans le bloc
tauri::async_runtime::spawn (utilisant fallback_done, reveal_main_close_splash
et fallback_handle) est one-shot et perd la course avec ReadySignal; remplace la
logique sleep unique par une boucle de retry bornée (p.ex. N tentatives) qui
attend l'intervalle (15s ou un backoff), vérifie fallback_done.swap(true,
Ordering::SeqCst) avant chaque tentative, appelle
reveal_main_close_splash(&fallback_handle) et si elle échoue remet fallback_done
à false et continue la boucle jusqu'à épuisement des tentatives ; si
reveal_main_close_splash réussit ou fallback_done était déjà true, break la
boucle; ainsi le fallback retente lui-même de façon limitée au lieu d'être
consommé une seule fois.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 6ba3005e-4524-4498-8e6a-d394877ced17
📒 Files selected for processing (4)
src-tauri/src/lib.rssrc-tauri/tauri.conf.jsonsrc/lib/nowPlayingCard.tssrc/main.tsx
…veal failure Two coderabbit findings on the splash handoff: 1. `reveal_main_close_splash` was still calling `splash.close()` even when the main window was missing or `main.show()` had failed. On that path the user ended up with zero visible windows. Early- return before touching the splash if the main reveal didn't actually succeed. 2. The fallback timer was one-shot: a single 15 s sleep, one reveal attempt, done. `ReadySignal` also emits `app://ready` only once at mount. If the timer narrowly won the race AND the reveal then failed, both triggers were spent and the user was stuck. Replace the one-shot sleep with a bounded retry loop: up to 10 attempts with a 250 ms backoff between failures. Permanent failure escalates to a `tracing::error!` so it shows up in diagnostics.
The inline ReadySignal component in src/main.tsx triggered react-refresh/only-export-components — Fast Refresh requires a file to export *only* components, and main.tsx has the ReactDOM.createRoot + i18n bootstrap side effects alongside. Move ReadySignal to src/components/common/ReadySignal.tsx with the same useEffect-based emit. Identical runtime behaviour, just cleaner module boundaries. main.tsx now imports it like any other component.
github-actions
Bot
added
scope: docs
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.
Summary
Hotfix for #42 — v1.1.0 splash screen never closes on Linux (Fedora COPR + Arch AUR every launch, Ubuntu first launch only). Root cause analysis: a timing race between the frontend's rAF-driven splash close and WebKitGTK 2.52's slow first-launch webview init.
Fix in three layers
setup()now installs anapp.listen("app://ready", …)that calls a newreveal_main_close_splash(&AppHandle)helper (show main → focus → close splash, in that order). Atokio::time::sleep(15s)fallback fires the same helper if the event never arrives (frontend crash before render).revealMainWindowbecomessignalReady: keep the double rAF so React commits + the compositor paints at least one frame, thenemit("app://ready"). No more IPC show/close from JS.transparent: true(already documented as risky in a memory note for AppImage on WebKitGTK 2.52+). Body background switches to opaque#121212to match the design.Why this works
Test plan
bun run typecheck+cargo clippy --all-targets— both cleanthrow new Error()inApp.tsxand verifying the splash still goes awayFollow-up
Once merged, this is a
fix:commit so release-please will roll a v1.1.1 release-please PR automatically. Merging that PR cuts the v1.1.1 tag → release.yml builds + fans out to AUR / Winget / COPR / apt-publish. The Linux regression banner on the v1.1.0 release notes can be amended at that point.Summary by CodeRabbit
Style
#121212.Améliorations