http://trac.webkit.org/changeset/118913 seems to have caused webarchive/test-xml-stylesheet.xml to fail: - <key>WebSubresources</key> - <array> - <dict> - <key>WebResourceData</key> - <string>#test { - background: green; - } - </string> - <key>WebResourceMIMEType</key> - <string>text/css</string> - <key>WebResourceResponse</key> - <dict> - <key>MIMEType</key> - <string>text/css</string> - <key>URL</key> - <string>file:///LayoutTests/webarchive/resources/test-xml-stylesheet.css</string> - <key>expectedContentLength</key> - <integer>31</integer> - </dict> - <key>WebResourceURL</key> - <string>file:///LayoutTests/webarchive/resources/test-xml-stylesheet.css</string> - </dict> - </array> I will check in failing results.
<rdar://problem/11559199>
- addSubresourceAttributeURLs(urls); + if (!isElementNode()) + return; + static_cast<const Element*>(this)->collectSubresourceURLsFromAttributes(urls); This means that we're not collecting from ProcessingInstructions anyway. An OVERRIDE annotation would have caught that. I think that the patch should be rolled out. Given that these functions are collecting URLs from more than just element attributes, new names seem misleading (and that's how the regression got introduced anyway). Some kind of renaming (with OVERRIDE) seems appropriate, but it seems easier to do in a new patch.
s/anyway/any more/
Rolling out in https://bugs.webkit.org/show_bug.cgi?id=87891
Please don't roll out large patches for such minor reasons. This would have been a trivial fix.
Antti, you may have not read comment 2. The idea of this patch was wrong.