On Jan 30, 2012, at 9:26 PM, Nikolas Zimmermann wrote: > Am 30.01.2012 um 21:44 schrieb James Robinson: > >> I didn't see any regressions, pixel tests or otherwise, when running with >> this patch, nor did the bots. What platform / test configuration are you >> seeing issues on? > Well, that's exactly the problem, there are no regressions - if we break > repainting within svg, we can not notice it anymore due the changed painting > timing, as all tests use the seTimeout(0, doTest()) scheme, as I explained in > my initial mail. > Previously we repainted the document, then execute JS that eg. mutated a > rect, then triggered the final repainting using > layoutTestController.notifyDone() (speaking about pixel tests right now). > I should have stated more clearly, but this is only about pixel testing. > > I tried a crude hack, and disabled all dynamic repainting, and still most > tests pass, as the first painting now happens after the scripts all finished > executing, instead of inbetween as it used to happen. > >> This should not be an observable change in behavior. setTimeout() vs >> painting behavior has never been guaranteed in DRT or in actual browsers. > I'm not arguing that the way of testing is perfect, though its obvious that > it changed for DRT, and we now have a lot of useless svg tests, and can't > catch any regressions, even with pixel tests. > > Cheers, > Niko
Are we actually sure that the updateStyleForAllDocuments() changes caused this? Perhaps DRT/WTR need to do something now that they didn't do before. Simon _______________________________________________ webkit-dev mailing list [email protected] http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
