Margin collapse and float/clear interactions are right up there with Unix signals and Event Tracing for Windows as some of the most evil things ever invented…

To elaborate: I don’t think any browser handles all the cases right, 20 years after the spec was developed.

*mumbles something about the various open issues on CSS2 around margin collapsing* At some point, hopefully later this year, I'll have the time to sit down and try and actually get us to the point of having _something_ closer to reality spec'd.

What I’d like more than anything is some sort of guidance from the spec as to when I have to fall back to “guess where a float/cleared block is; if wrong then redo layout”. All implementations except WeasyPrint I’m aware of do this in varying circumstances…

I’m not sure that CSS 2.1 margin-collapse/clear interactions are practically implementable as spec’d and it’d be very nice to know what approximation browsers actually go with.

I'm pretty certain the answer is they're not.

As a serious point: the more margin collapsing tests we can get into WPT the better, in my view. There's been a fair few from LayoutNG, and I'd be more than happy to see a bunch from Layout 2020.