From 6423f226b4ec549371f2a435f5817c3d7e1984f4 Mon Sep 17 00:00:00 2001 From: Manuel Huber Date: Thu, 26 Oct 2023 14:26:49 +0200 Subject: [PATCH] Also use unload event to clean up session lock (#11800) Signed-off-by: Manuel Huber Co-authored-by: Florian Duros --- src/utils/SessionLock.ts | 41 ++++++++++++++++++++++++---------------- 1 file changed, 25 insertions(+), 16 deletions(-) diff --git a/src/utils/SessionLock.ts b/src/utils/SessionLock.ts index 76780b543c..2c68720947 100644 --- a/src/utils/SessionLock.ts +++ b/src/utils/SessionLock.ts @@ -171,6 +171,27 @@ export async function getSessionLock(onNewInstance: () => Promise): Promis } } + // handler for pagehide and unload events, used later + function onPagehideEvent(): void { + // only remove the ping if we still think we're the owner. Otherwise we could be removing someone else's claim! + if (lockServicer !== null) { + prefixedLogger.debug("page hide: clearing our claim"); + window.clearInterval(lockServicer); + window.localStorage.removeItem(SESSION_LOCK_CONSTANTS.STORAGE_ITEM_PING); + window.localStorage.removeItem(SESSION_LOCK_CONSTANTS.STORAGE_ITEM_OWNER); + lockServicer = null; + } + + // It's worth noting that, according to the spec, the page might come back to life again after a pagehide. + // + // In practice that's unlikely because Element is unlikely to qualify for the bfcache, but if it does, + // this is probably the best we can do: we certainly don't want to stop the user loading any new tabs because + // Element happens to be in a bfcache somewhere. + // + // So, we just hope that we aren't in the middle of any crypto operations, and rely on `onStorageEvent` kicking + // in soon enough after we resume to tell us if another tab woke up while we were asleep. + } + async function releaseLock(): Promise { // tell the app to shut down await onNewInstance(); @@ -239,23 +260,11 @@ export async function getSessionLock(onNewInstance: () => Promise): Promis window.addEventListener("storage", onStorageEvent); // also add a listener to clear our claims when our tab closes or navigates away - window.addEventListener("pagehide", (event) => { - // only remove the ping if we still think we're the owner. Otherwise we could be removing someone else's claim! - if (lockServicer !== null) { - prefixedLogger.debug("page hide: clearing our claim"); - window.localStorage.removeItem(SESSION_LOCK_CONSTANTS.STORAGE_ITEM_PING); - window.localStorage.removeItem(SESSION_LOCK_CONSTANTS.STORAGE_ITEM_OWNER); - } + window.addEventListener("pagehide", onPagehideEvent); - // It's worth noting that, according to the spec, the page might come back to life again after a pagehide. - // - // In practice that's unlikely because Element is unlikely to qualify for the bfcache, but if it does, - // this is probably the best we can do: we certainly don't want to stop the user loading any new tabs because - // Element happens to be in a bfcache somewhere. - // - // So, we just hope that we aren't in the middle of any crypto operations, and rely on `onStorageEvent` kicking - // in soon enough after we resume to tell us if another tab woke up while we were asleep. - }); + // The pagehide event is called unreliably on Firefox, so additionally add an unload handler. + // https://bugzilla.mozilla.org/show_bug.cgi?id=1854492 + window.addEventListener("unload", onPagehideEvent); return true; }