RESOLVED WORKSFORME272248
Crash in WebKit::WebDataListSuggestionPicker::close
https://bugs.webkit.org/show_bug.cgi?id=272248
Summary Crash in WebKit::WebDataListSuggestionPicker::close
Michael Catanzaro
Reported 2024-04-05 14:45:19 PDT
Created attachment 470776 [details] Full backtrace I'm not able to reproduce this, but here's a crash that occurred after clicking Back from https://www.brother-usa.com/home to https://pwg.org/printers/. Looks like I must have had a search input focused, but that's not sufficient to make it crash. #0 WebKit::WebPage::identifier (this=0x0) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebKit/WebProcess/WebPage/WebPage.h:439 #1 WebKit::WebDataListSuggestionPicker::close (this=0x7fb0229a5460) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebKit/WebProcess/WebCoreSupport/WebDataListSuggestionPicker.cpp:62 #2 0x00007fb03bb3ea27 in WebCore::TextFieldInputType::closeSuggestions (this=0x7fb0222c8400) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebCore/html/TextFieldInputType.cpp:1012 #3 WebCore::TextFieldInputType::~TextFieldInputType (this=0x7fb0222c8400) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebCore/html/TextFieldInputType.cpp:87 #4 0x00007fb03bb366fe in WebCore::SearchInputType::~SearchInputType (this=0x7fb0222c8400) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebCore/html/SearchInputType.h:41 #5 WebCore::SearchInputType::~SearchInputType (this=0x7fb0222c8400) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebCore/html/SearchInputType.h:41 #6 0x00007fb03ba72a1f in std::default_delete<WebCore::InputType>::operator() (__ptr=0x7fb0229a5460, this=<optimized out>) at /usr/bin/../lib/gcc/x86_64-unknown-linux-gnu/13.2.0/../../../../include/c++/13.2.0/bits/unique_ptr.h:99 #7 WTF::RefCounted<WebCore::InputType, std::default_delete<WebCore::InputType> >::deref (this=0x7fb0229a5468) at WTF/Headers/wtf/RefCounted.h:220 #8 WTF::DefaultRefDerefTraits<WebCore::InputType>::derefIfNotNull (ptr=0x7fb0229a5460) at WTF/Headers/wtf/Ref.h:62 #9 WTF::RefPtr<WebCore::InputType, WTF::RawPtrTraits<WebCore::InputType>, WTF::DefaultRefDerefTraits<WebCore::InputType> >::~RefPtr (this=0x7f97be010c58) at WTF/Headers/wtf/RefPtr.h:60 #10 WebCore::HTMLInputElement::~HTMLInputElement (this=0x7f97be010b60) at /buildstream/gnome/sdk/webkitgtk-6.0.bst/Source/WebCore/html/HTMLInputElement.cpp:158 Full backtrace attached.
Attachments
Full backtrace (18.00 KB, text/plain)
2024-04-05 14:45 PDT, Michael Catanzaro
no flags
Michael Catanzaro
Comment 1 2024-04-05 14:47:36 PDT
OK, found a reproducer and it's very simple. Visit https://pwg.org/printers/, click on the "Make, model, etc." search entry, press Ctrl+W to close the page. It will crash.
Michael Catanzaro
Comment 2 2024-04-05 14:55:03 PDT
Well the problem seems clear enough. Since 272284@main WebDataListSuggestionPicker keeps a WeakRef<WebPage> m_page, expecting it to become nullptr. But everywhere it uses m_page, it assumes that it will never be nullptr. This is inconsistent.
Alexey Proskuryakov
Comment 3 2024-04-08 10:29:39 PDT
FWIW, the crash doesn't reproduce in Safari. Perhaps we just destroy the process without running the destructors?
Michael Catanzaro
Comment 4 2024-04-08 13:44:47 PDT
(In reply to Alexey Proskuryakov from comment #3) > FWIW, the crash doesn't reproduce in Safari. Perhaps we just destroy the > process without running the destructors? Doubt it. You have to fire event handlers when the page is closed, for example. Probably something else is different. The weak pointer misuse is a cross-platform issue and looks easy to fix. The only challenge here will be figuring out if this is testable and writing a test if so.
Michael Catanzaro
Comment 5 2024-04-08 14:05:23 PDT
(Well, I'm also uncomfortable with the WebPage itself being destroyed while JS code is still executing. But that's apparently expected.)
Radar WebKit Bug Importer
Comment 6 2024-04-12 14:46:14 PDT
Michael Catanzaro
Comment 7 2024-05-30 14:14:50 PDT
It's no longer possible to reproduce this due to bug #274927, a UI process crash that occurs first.
Michael Catanzaro
Comment 8 2024-05-30 16:34:04 PDT
After fixing that crash, I still can't reproduce this, so I'm going to assume it is fixed and not worth investigating further.
Note You need to log in before you can comment on or make changes to this bug.