In strict mode we draw the table row background to wide. It is a leftover from some ancient days, we are off exactly by the cellspacing we introduce in the paint method. All reflow debuuging shows to me that we the comment in the code is no longer valid. testcase + patch will follow
r=dbaron. I think there were other bugs like this from a taking a statement in CSS2 too literally (the statement didn't mean what it said when saying that certain backgrounds should "cover the entire table"). This is really a duplicate of bug 15931, anyway :-). Do all the backgrounds in http://www.people.fas.harvard.edu/~dbaron/css/test/sec170501 http://www.people.fas.harvard.edu/~dbaron/css/test/sec170501a http://www.people.fas.harvard.edu/~dbaron/css/test/sec170501b now look right or are there other bugs still around too?
See in particular Bert's comment in http://lists.w3.org/Archives/Public/www-style/1999Jul/0083.html which is in reply to one of the posts mentioned in bug 15931.
fix checked in