Bug 206337 – REGRESSION (r251110): Crash on https://developer.apple.com/tutorials/swiftui/creating-and-combining-views

• Status: RESOLVED
• Resolution: FIXED
• Component: Forms

[Ryosuke Niwa]

WebKit crashes with the following backtrace when visiting https://developer.apple.com/tutorials/swiftui/creating-and-combining-views.
0   WebCore                       	0x00000001b88af7b0 WebCore::RadioButtonGroups::hasCheckedButton(WebCore::HTMLInputElement const&) const + 136
1   WebCore                       	0x00000001b88af74c WebCore::RadioButtonGroups::hasCheckedButton(WebCore::HTMLInputElement const&) const + 36
2   WebCore                       	0x00000001b8ae7310 WebCore::RadioInputType::matchesIndeterminatePseudoClass() const + 48
3   ???                           	0x0000000caf0068a0 0 + 54475647136
4   WebCore                       	0x00000001b9238b0c WebCore::Style::ElementRuleCollector::collectMatchingRulesForList(WTF::Vector<WebCore::Style::RuleData, 1ul, WTF::CrashOnOverflow, 16ul> const*, WebCore::Style::MatchRequest const&) + 384
5   WebCore                       	0x00000001b9238840 WebCore::Style::ElementRuleCollector::collectMatchingRules(WebCore::Style::MatchRequest const&) + 740
6   WebCore                       	0x00000001b9239930 WebCore::Style::ElementRuleCollector::matchUARules() + 104
7   WebCore                       	0x00000001b9239f60 WebCore::Style::ElementRuleCollector::matchAllRules(bool, bool) + 40
8   WebCore                       	0x00000001b9250990 WebCore::Style::Resolver::styleForElement(WebCore::Element const&, WebCore::RenderStyle const*, WebCore::RenderStyle const*, WebCore::RuleMatchingBehavior, WebCore::SelectorFilter const*) + 488
9   WebCore                       	0x00000001b925b9e4 WebCore::Style::TreeResolver::styleForElement(WebCore::Element&, WebCore::RenderStyle const&) + 228
10  WebCore                       	0x00000001b925bd20 WebCore::Style::TreeResolver::resolveElement(WebCore::Element&) + 112
11  WebCore                       	0x00000001b925cd9c WebCore::Style::TreeResolver::resolveComposedTree() + 1388
12  WebCore                       	0x00000001b925d8f0 WebCore::Style::TreeResolver::resolve() + 700

<rdar://problem/58441241>

[Ryosuke Niwa]

Created attachment 387895 [details]
Fixes the bug

[Geoffrey Garen]

Comment on attachment 387895 [details]
Fixes the bug

r=me

[Wenson Hsieh]

Comment on attachment 387895 [details]
Fixes the bug

View in context: https://bugs.webkit.org/attachment.cgi?id=387895&action=review

> LayoutTests/fast/forms/match-pseudo-on-radio-before-finalizing-tree-insertion-crash.html:18
> +script.textContent = 'document.getElementById("result").textContent = getComputedStyle(input).color ? "PASS" : "PASS"';

Nit - this could just be document.getElementById("result").textContent = "PASS";

[Ryosuke Niwa]

Comment on attachment 387895 [details]
Fixes the bug

View in context: https://bugs.webkit.org/attachment.cgi?id=387895&action=review

>> LayoutTests/fast/forms/match-pseudo-on-radio-before-finalizing-tree-insertion-crash.html:18
>> +script.textContent = 'document.getElementById("result").textContent = getComputedStyle(input).color ? "PASS" : "PASS"';
> 
> Nit - this could just be document.getElementById("result").textContent = "PASS";

Ah, no, no, getComputedStyle(input).color is what triggers the crash.
Without it, the test won't do anything useful.
Perhaps I should split into a separate line to make it clear.

[Wenson Hsieh]

Comment on attachment 387895 [details]
Fixes the bug

View in context: https://bugs.webkit.org/attachment.cgi?id=387895&action=review

>>> LayoutTests/fast/forms/match-pseudo-on-radio-before-finalizing-tree-insertion-crash.html:18
>>> +script.textContent = 'document.getElementById("result").textContent = getComputedStyle(input).color ? "PASS" : "PASS"';
>> 
>> Nit - this could just be document.getElementById("result").textContent = "PASS";
> 
> Ah, no, no, getComputedStyle(input).color is what triggers the crash.
> Without it, the test won't do anything useful.
> Perhaps I should split into a separate line to make it clear.

Oh, I see! Yep, I think that splitting it out clarifies things.

[Ryosuke Niwa]

Comment on attachment 387895 [details]
Fixes the bug

View in context: https://bugs.webkit.org/attachment.cgi?id=387895&action=review

> LayoutTests/fast/forms/match-pseudo-on-radio-before-finalizing-tree-insertion-crash.html:26
> +

Oh weird, this doesn't have the second input which makes this test hang in the trunk.
Will fix that before landing it.

[Ryosuke Niwa]

Committed r254722: <https://trac.webkit.org/changeset/254722>