window.postMessage 保证事件的顺序吗？

发布于 2024-12-06 09:39:23 字数 363 浏览 6 评论 0原文

window.onmessage = ...
window.postMessage('1', '*');
window.postMessage('2', '*');

是否 postMessage (http://www.whatwg.org/specs/web-apps/current-work/multipage/webappapis.html#queue-a-task）保证事件的顺序？

原文

window.onmessage = ...
window.postMessage('1', '*');
window.postMessage('2', '*');

Does postMessage (http://www.whatwg.org/specs/web-apps/current-work/multipage/webappapis.html#queue-a-task) guarantee the order of events?

分享到QQ

分享到微博

如果你对这篇内容有疑问，欢迎到本站社区发帖提问参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。

发布评论

需要登录才能够评论，你可以免费注册一个本站的账号。

孤蝉 2024-12-13 09:39:23

我不知道，尽管规范似乎表明它没有做出这样的保证（这令人惊讶）。很明显，一旦将 MessageEvent 添加到接收端的 任务队列 中，尽管 MessageEvent 创建和分派，它的顺序仍会保持不变。与原始 postMessage 调用异步，因此理论上您可能会出现以下情况：

主线程：

window.postMessage('1', '*'); --> thread spawned to create MessageEvent
window.postMessage('2', '*'); --> new thread spawned for another MessageEvent

如果线程管理系统允许第二个 postMessage 在第一个线程设法分派MessageEvent，并且由于任何不幸的原因允许较新的线程执行（稀释的优先级反转），再次在第一个设法分派之前，那么您确实会以相反的顺序收到这些消息。

尽管规范中可能有其他地方为这些异步执行提供了更多上下文并排除了这种情况 - 但我找不到它。

I don't know, although the wording of the spec seems to suggest it doesn't make such guarantees (which is surprising). It's clear that once the MessageEvent is added to the task queue on the receiving end, then it's order is maintained, although the MessageEvent creation and dispatch are asynchronous to the original postMessage call, so theoretically it appears that you could have the following situation:

main thread:

window.postMessage('1', '*'); --> thread spawned to create MessageEvent
window.postMessage('2', '*'); --> new thread spawned for another MessageEvent

If the thread management system allowed the second postMessage to execute before the first thread managed to dispatch the MessageEvent, and for whatever unlucky reason allowed that newer thread to execute (a diluted priority inversion), again before the first managed to dispatch, then you would indeed receive those messages in the reverse order.

Although there might be some other place in the spec that provides more context for these asynchronous executions and rules out this case - I couldn't find it.

回复收藏 0 原文