Regression from https://bugs.webkit.org/show_bug.cgi?id=90646. I incorrectly analyzed the issue with Robert Hogan's negative margin patch and fooled myself into putting back in an incorrect minimum width check from long ago. What should have happened in the test case I patched is that the overflow element should shrink to 0. The issue was that it exposed a bug in our clearing algorithm where you could need a relayout even if you didn't actually move. This only occurs because the floats list is getting changed mid-layout because of negative margins.
Created attachment 151752 [details] Patch
Created attachment 151775 [details] Patch New patch for EWS.
Created attachment 151993 [details] Patch
Comment on attachment 151993 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=151993&action=review > Source/WebCore/ChangeLog:23 > + Reviewed by NOBODY (OOPS!). This should go above the description. > Source/WebCore/ChangeLog:25 > + No new tests required, since the test in fast/block/float is now correctly covering the issue. But was that test failing before?
Fixed in r122489.
*** Bug 90965 has been marked as a duplicate of this bug. ***
*** Bug 91036 has been marked as a duplicate of this bug. ***