On about half of loads, the testcase triggers these four assertions: ###!!! ASSERTION: this shouldn't happen anymore: 'NS_UNCONSTRAINEDSIZE != aComputedWidth && NS_UNCONSTRAINEDSIZE != aAvailWidth', file mozilla/layout/generic/nsHTMLReflowState.cpp, line 1892 ###!!! ASSERTION: shouldn't use unconstrained widths anymore: '(mFrameType == NS_CSS_FRAME_TYPE_INLINE && !frame->IsFrameOfType(nsIFrame::eReplaced)) || frame->GetType() == nsGkAtoms::textFrame || mComputedWidth != NS_UNCONSTRAINEDSIZE', file mozilla/layout/generic/nsHTMLReflowState.cpp, line 295 ###!!! ASSERTION: unconstrained widths no longer supported: 'aContainingBlockWidth != NS_UNCONSTRAINEDSIZE', file mozilla/layout/base/nsLayoutUtils.cpp, line 1601 ###!!! ASSERTION: wrong kind of child frame: 'aIsBlock == f->GetStyleDisplay()->IsBlockOutside()', file mozilla/layout/generic/nsLineBox.cpp, line 71 I'm guessing this is a style system bug because the (reduced) testcase is full of stuff like style="", "* { }", inherit, and -moz-initial.

Hmm. "display: -moz-initial" is pretty broken right now; it gives semi-random values depending on what other rules are involved. The patch for bug 80887 seems to fix this bug over here.

WFM, Mac trunk debug.