Spec change: http://www.w3.org/TR/2015/WD-css-grid-1-20150806/#abspos "If the placement only contains a grid span, replace it with the two 'auto' lines in that axis." CSSWG minutes: https://lists.w3.org/Archives/Public/www-style/2015May/0280.html

Actually, the whole paragraph is: "If the placement only contains a grid span, replace it with the two auto lines in that axis. (This happens when both grid-placement properties in an axis contributed a span originally, and §9.2.1 Grid Placement Conflict Handling caused the second span to be ignored.)" So what they mean here (I think) is that "span */ span *", which is error corrected by 9.2.1 into "auto/ span *", should instead be treated as "auto/ auto". An explicit "auto / span *" should still be treated as spanning from the start padding edge (called "line 0" in the spec) to some line in the grid.

Comment on attachment 8648377 [details] [diff] [review] fix + tests Review of attachment 8648377 [details] [diff] [review]: ----------------------------------------------------------------- r=me, just one nit on the assertion: ::: layout/generic/nsGridContainerFrame.cpp @@ +685,5 @@ > > LineRange r = ResolveLineRange(aStart, aEnd, aLineNameList, aAreaStart, > aAreaEnd, aExplicitGridEnd, aStyle); > + if (r.IsAuto()) { > + MOZ_ASSERT(aStart.mHasSpan && aEnd.mHasSpan); Could you include a message with the assertion here? If this starts failing someday, it'd be helpful to have a human-readable hint for why we currently think it should hold in this particular case. (Note that the asserted condition doesn't hold for 'auto' ResolveLineRange() return values in general - I think we only expect it to hold here because we already handled all of the other cases for which we'd expect it to return an auto-positioned range (?))

> Could you include a message with the assertion here? Sure.