Extra AlloyBrowserHostImpl::DestroyBrowser nullptr checks

[amaitland]

In `AlloyBrowserHostImpl::DestroyBrowser` a user is reporting that `platform_delegate_` is a nullptr. Looks like it's fairly reproducible.

Call stack at https://github.com/cefsharp/CefSharp/is ... 1873906342

Basically `CefBrowserHost:CloseBrowser(true)` is being called within `DoClose`. I can modify the example code to avoid making this call, it's difficult to stop the user from doing this though. I might be able to hack around it. I can go into more technical detail as to the reason this happens if it's relevant, I know this is light on details.

My question is, would adding a null ptr check to AlloyBrowserHostImpl::DestroyBrowser be acceptable? There are lots of other places in AlloyBrowserHostImpl that have checks.

Code: Select all

if (platform_delegate_) {
    platform_delegate_->BrowserDestroyed(this);
  }


Or some other check?

[magreenblatt]

What is the call stack for the crash?

EDIT: Never mind, I see it in the CefSharp issue.

[magreenblatt]

These shutdown ordering issues can be tricky. I’ll look at it properly this week and get back to you.

[amaitland]

Thanks for the amazingly fast response!