UniscribeHelper::draw() should not ASSERT that ScriptTextOut() always succeed. It even has logic to deal with this kind of failures.
Created attachment 53302 [details] patch
For background, we probably used to have a DCHECK here with is a debug-only assert to help catch errors. However, there are legitimate reasons this function may fail, especially in the Chrome sandbox, which is what the existing recovery code is supposed to do.
Comment on attachment 53302 [details] patch r=me
Landed as r57595.
fast/dom/Window/HTMLFrameSetElement-window-eventListener-attributes.html has been failing consistently on SnowLeopard since this checkin: http://build.webkit.org/results/SnowLeopard%20Intel%20Leaks/r57603%20(5978)/fast/dom/Window/HTMLFrameSetElement-window-eventListener-attributes-stderr.txt ASSERTION FAILED: m_wrapper || !m_jsFunction (/Volumes/Data/WebKit-BuildSlave/snowleopard-intel-leaks/build/WebCore/bindings/js/JSEventListener.h:83 JSC::JSObject* WebCore::JSEventListener::jsFunction(WebCore::ScriptExecutionContext*) const) That's the same ASSERT as bug 36779, except now its failing every time.
Nevermind, this test is flakey and has been failing for much longer.