In bug 1117983 parts 1, 2 and 3, I make {inline,block}-size parse the keywords that width takes -- min-content, -moz-available, etc. -- but convert them to unset if they happen to get set on height. We should undo this change to nsRuleNode::ComputePositionData once we support these keywords on height.

(In reply to Cameron McCormack (:heycam) (away April 3-6) from comment #0) > We should undo > this change to nsRuleNode::ComputePositionData once we support these > keywords on height. This isn't quite right IMHO: we don't want to support these keywords on height as such, but rather on inline-size (whether width or height) and not on block-size (whether height or width). Bug 1106669 already made nsLayoutUtils::IntrinsicForContainer work on inline-size instead of width, so I'm not sure that this hack was ever necessary.

So rather than remove the hack, I think we just need to adjust it. Per some email conversation, I think we want to do this: > 1. adjust the “hack” to allow the keywords to be set on > nsStylePosition::mHeight (and prevent them being set on mWidth) > depending on the writing mode, rather than always allowing them on > mWidth and not on mheight > > 2. allow the keywords to be parsed as ‘height’ property values > > 3. disallow the keywords on ‘block-size’

(heads-up: I'm backlogged on reviews at the moment [catching up, but still have some big grid ones to get through]. If you can find another reviewer, you may have have a faster review turnaround.)

Comment on attachment 8624596 [details] [diff] [review] Support -moz-min-content etc. keywords in {,min-,max-}inline-size and their physical equivalents. OK, thanks for the heads up Daniel. bz you good to take this one?

Comment on attachment 8624596 [details] [diff] [review] Support -moz-min-content etc. keywords in {,min-,max-}inline-size and their physical equivalents. >+ const bool vertical = aContext->StyleVisibility()->mWritingMode != >+ NS_STYLE_WRITING_MODE_HORIZONTAL_TB; Could you MOZ_ASSERT that the value is one of the ones we know about (so we'll catch it if we add a new value)? r=me with that.