The objective of fast/css/long-css-selector-matches.html test is whether stack overflow happens or not if a long css selector is used. To make stack overflow happen in old implementation of CSSStyleSelector, we had to increase a CSS selector's depth level at some level. But that increases a execution time of that test. So we have a dilemma. The more and more we increate a depth level, stack overflow becomes to happen more likely in bad implementation, but a timeout of test becomes to happen more likely in some platforms. In fact, the test in chromium in test_shell causes TIMEOUT, so we had to mark that test 'PASS TIMEOUT' in chromium/test_expectations.txt: https://trac.webkit.org/changeset/70123. It is not ideal situation that marking it 'PASS TIMEOUT' in only some platforms. We have to come to compromise about a depth level so that it doesn't cause TIMEOUT.
Created attachment 71254 [details] make-depth-level-smaller
Hi reviewers, if you have any suggestions about how to deal this test, any opinions are welcome. I am not pretty sure that making a depth level smaller is a good idea.
Comment on attachment 71254 [details] make-depth-level-smaller I've cleared the review flag because the original change was reverted in http://trac.webkit.org/changeset/70209