[webkit-reviews] review canceled: [Bug 226426] [iOS] UI process crashes when deallocating WKWebView in a script message handler during an active touch event : [Attachment 430114] Patch
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Sun May 30 07:59:31 PDT 2021
Wenson Hsieh <wenson_hsieh at apple.com> has canceled Wenson Hsieh
<wenson_hsieh at apple.com>'s request for review:
Bug 226426: [iOS] UI process crashes when deallocating WKWebView in a script
message handler during an active touch event
https://bugs.webkit.org/show_bug.cgi?id=226426
Attachment 430114: Patch
https://bugs.webkit.org/attachment.cgi?id=430114&action=review
--- Comment #3 from Wenson Hsieh <wenson_hsieh at apple.com> ---
Comment on attachment 430114
--> https://bugs.webkit.org/attachment.cgi?id=430114
Patch
View in context: https://bugs.webkit.org/attachment.cgi?id=430114&action=review
>> Source/WebKit/ChangeLog:22
>> + Possible state.
>
> This doesn’t seem like it would fix the problem described above reliably. All
we need for the crash to happen again is for someone to keep WebPageProxy alive
by holding a ref to it after WebPageProxy’s weak pointer to the page client is
null.
>
> To fix the problem described above, we need to check the page client for
null, not the WebPageProxy.
>
> I don’t think this indirect fix is nearly as good. However, the idea that we
should only capture a weak reference seems good in and of itself. It’s just
insufficient to fix the "web page proxy could outlive the page client" issue.
Makes sense — I'll keep the weak capture, but add a null check for
`m_pageClient` as well.
More information about the webkit-reviews
mailing list