534793 - Inconsistent heights with -moz-inline-box, <table style="height: 200%;">

Status: NEW

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: dynamic

Comment 1

[Jesse Ruderman]

Attachment: static

Comment 2

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Seems like the second time around, the box's Layout isn't bothering to call Layout() on the table (in a case where that Layout looks like it would produce a different result).

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

The question is why, in nsSprocketLayout::Layout, NS_SUBTREE_DIRTY(child) was false (such that |layout| could end up false).  Probably it got cleared during some of the intrinsic width computation.  In other words, this seems like a problem similar to the problems we had within block layout before the reflow branch, when we were trying to use one set of dirty bits but also doing intrinsic size computation using layout.

That said, it seems like we really don't need the child to be dirty; we just need to reflow it with the vertical and horizontal resize bits set.  (Maybe we should make the intrinsic size computation do another reflow to set the child back to its old state?)