[Webkit-unassigned] [Bug 98021] Wrong display with "tr:nth-child(even) td" and missing <tbody>

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=98021


Takashi Sakamoto <tasak at google.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tasak at google.com




--- Comment #2 from Takashi Sakamoto <tasak at google.com>  2012-10-03 04:10:40 PST ---
Hello,

This bug is caused by sibling style cache.
If some positional rules are applied to the given element, the element cannot use sibling style cache.
To know whether there exists any positional rule or not, sibling style cache looks at the parent style of the element, i.e. childrenAffectedByPositinalRules. The value is set in SelectorChecker::checkSelector.

So,
<style>
table tr:nth-child(even) td { background: #eee; }
</style>
<table>
  <tr><th></th></tr>
  <tr><td></td></tr>
</table>

while resolving styles for the first <tr><th></th></tr>, since <th> doesn't match any rules, <tr>'s style will be created but its parent's childrenAffectedByPositionalRules will be false.

In the next line, i.e. <tr><td></td></tr>, the <tr>'s style is the same as the first <tr>'s style, because the <tr>'s parent's childenAffectedByPositionalRules is false.

Since<td> matches the rightmost selector of "table tr:nth-child(even) td", checkSelector checks whether there exists any ancestor which matches "tr:nth-child(even)" or not. In checkSelector,

        case CSSSelector::PseudoNthChild:
...
                if (m_mode == ResolvingStyle) {
                    RenderStyle* childStyle = context.elementStyle ? context.elementStyle : element->renderStyle();
...
                        childStyle->setChildIndex(count);
...

to avoid calculating child position again and again, the <tr>'s style is updated to have the <tr>'s position. However, the <tr>'s style is the same as the previous sibling <tr>.

Best regards,
Takashi Sakamoto

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.