Created attachment 451172 [details] testcase All of these should look identical: <u style="all: revert">foo</u> <u style="text-decoration: revert">foo</u> <u style="all: revert-layer">foo</u> <u style="text-decoration: revert-layer">foo</u> <u>foo</u> But only the last one has the underline set in UA origin. That's because text-decoration is low-priority / deferred / applied in parse order. `revert` and `revert-layer` don't seem to work on these properties: https://webkit-search.igalia.com/webkit/rev/30ea950a1b2947097300016afbe68f78665b0e43/Source/WebCore/style/PropertyCascade.cpp#41-73
<rdar://problem/88935655>
Created attachment 453668 [details] Patch
This patch modifies the imported WPT tests. Please ensure that any changes on the tests (not coming from a WPT import) are exported to WPT. Please see https://trac.webkit.org/wiki/WPTExportProcess
Created attachment 454635 [details] Patch
Just changed the test to use /html/resources/common.js instead of hardcoding a list of elements.
Created attachment 454640 [details] Patch
Comment on attachment 454640 [details] Patch I forgot that all-prop-revert-layer-noop.html passes after this.
Committed r291260 (248412@main): <https://commits.webkit.org/248412@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 454640 [details].
Comment on attachment 454640 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=454640&action=review > Source/WebCore/style/PropertyCascade.cpp:176 > + m_deferredPropertiesIndices.set(id, m_deferredProperties.size()); It is not ok to add a HashMap (and a heap allocation) to this generic code path when it is only needed and used in the rare rollback case. The code should be structured differently.
In https://bugs.webkit.org/show_bug.cgi?id=238260 this HashMap is used in sensible way to eliminate the m_deferredProperties member. In that context this patch makes more sense. Still, it is not clear it should be a HashMap.