This assertion was added in bug 363247 and I'm hitting it a lot. ###!!! ASSERTION: bits don't correspond to style change reason: 'aIntrinsicDirty != eStyleChange || aBitToAdd == NS_FRAME_IS_DIRTY', file /Users/jruderman/central/layout/base/nsPresShell.cpp, line 3126

It's an eStyleChange with NS_FRAME_HAS_DIRTY_CHILDREN, yeah. The caller is nsMathMLContainerFrame::ReLayoutChildren (which always does eStyleChange) and its caller is nsMathMLmathBlockFrame::InsertFrames.

Given that this function is marking ancestors with NS_FRAME_IS_DIRTY, it seems like it's effectively saying the frame is dirty.

It's not marking ancestors of the frame it passes to FrameNeedsReflow with NS_FRAME_IS_DIRTY, though.

Er, right... I'd looked a bit more since comment 2... but the difference seems incredibly minor, and I don't see any reason why an operation on a child frame should cause a frame many frames up the tree to get NS_FRAME_HAS_DIRTY_CHILDREN rather than NS_FRAME_IS_DIRTY. If we need to do this much damage, we may as well be consistent about it.

I could live with that, I guess.

(Full reftests pass, although I don't know how meaningful that is for MathML dynamic change handling.)

Comment on attachment 361411 [details] [diff] [review] patch I think I'd be fine with a followup bug to remove those XXX comments and dirty bit sets they're about.

Filed bug 477915 to follow up.