Run development on the dev server's code, not a build's cache

A production deploy leaves its service worker on the device, and it
claims every tab. It serves /.bunext/public/* cache first without
knowing which build asked, while every cache-busting mechanism the app
has (the precache gate, /build-info.json, the loading page) is
production only. So switching to `bun run dev` kept serving the old
build, vendor files included, and wrote the dev files back into the
production build's cache.

Development now clears that worker and every cache it wrote
(`clear-development-service-worker.ts`, from ServiceWorkerRegistrar,
which did nothing in development before), reloads once into the
uncontrolled tab, and /sw.js answers with a stub that unregisters
itself so Web Push and the loading page can't register it back.
This commit is contained in:
2026-10-21 15:57:29 +01:00
parent 8bf6bedec5
commit 361b8add9b
4 changed files with 174 additions and 3 deletions
@@ -7,13 +7,30 @@ import Span from "@moduletrace/twui/layout/Span";
import Button from "@moduletrace/twui/layout/Button";
import { RefreshCw } from "lucide-react";
import { PwaServiceWorkerPath } from "@/src/data/pwa-data";
import clearDevelopmentServiceWorker from "@/src/functions/frontend/clear-development-service-worker";
const UpdateCheckMs = 60 * 60 * 1000;
/**
* Development only: waits out a `clients.claim` that raced the clear, so
* the clearing reload doesn't come back controlled by the old worker.
*/
const DevelopmentClearRetryMs = 3000;
/**
* Development only: guards against reloading in a loop when the worker
* can't be cleared (a restricted profile, say).
*/
const DevelopmentClearAttemptsKey = "cobble-dev-sw-clear";
const DevelopmentClearMaxAttempts = 3;
/**
* Registers the app's service worker (production only, so development
* HMR is never cached) and offers a reload once a new version is
* installed. It never reloads on its own: a bot reply may be running.
*
* In development it does the opposite: it clears the worker and the
* caches a production deploy left on the device, so the dev server's
* code is never served from a build's cache (see
* `clear-development-service-worker.ts`).
*/
export default function ServiceWorkerRegistrar() {
const [update_ready, setUpdateReady] = useState(false);
@@ -78,6 +95,71 @@ export default function ServiceWorkerRegistrar() {
};
}, []);
useEffect(() => {
if (process.env.NODE_ENV == "production") return;
if (typeof navigator == "undefined") return;
if (!("serviceWorker" in navigator)) return;
let cancelled = false;
let retry: ReturnType<typeof setTimeout> | undefined;
/**
* A worker claimed this tab while clearing or right after it:
* clear again, then reload into the uncontrolled tab. Without
* this, the reload would land back on the cached build.
*/
const onControllerChange = () => {
if (cancelled) return;
retry = setTimeout(() => {
if (cancelled) return;
clearDevelopmentServiceWorker()
.then((cleared) => {
if (!cancelled && cleared) window.location.reload();
})
.catch(() => undefined);
}, DevelopmentClearRetryMs);
};
navigator.serviceWorker.addEventListener("controllerchange", onControllerChange);
clearDevelopmentServiceWorker()
.then((cleared) => {
if (cancelled) return;
if (!cleared) {
window.sessionStorage.removeItem(DevelopmentClearAttemptsKey);
return;
}
/**
* This load is still the old worker's code: one reload
* takes the dev server's, twice at most.
*/
const attempts = Number(
window.sessionStorage.getItem(DevelopmentClearAttemptsKey) || 0,
);
if (attempts >= DevelopmentClearMaxAttempts) return;
window.sessionStorage.setItem(
DevelopmentClearAttemptsKey,
String(attempts + 1),
);
window.location.reload();
})
.catch(() => undefined);
return () => {
cancelled = true;
if (retry) clearTimeout(retry);
navigator.serviceWorker.removeEventListener(
"controllerchange",
onControllerChange,
);
};
}, []);
useEffect(() => {
if (!reloading) return;