1039009 - Handle floating children of Ruby frames

Status: RESOLVED FIXED

(Core :: Layout: Floats, defect)

Description

[sgbowen]

We should make sure that children of ruby frames which are floats are laid out reasonably and that this behavior is documented in the spec.

Comment 1

[David Baron :dbaron:]

The spec might not need to say much, if anything.  But the the trick code-wise is making sure that the floats behave as though all of the ruby markup is like inlines and none of it is like blocks (even though we're implementing ruby-text-container as a block); they should behave as floats anchored within inline content do.

It might be sufficient to just ensure the ruby-text-container is a block without NS_BLOCK_FLOAT_MGR set.

Comment 2

[Xidorn Quan [:xidorn] UTC+11]

I guess spec need to specify the equivalent inline anchor position of floats in ruby annotation. I don't think we need to worry about it now, though.

Comment 3

[Xidorn Quan [:xidorn] UTC+11]

Attachment: patch

Comment 4

[David Baron :dbaron:]

What behavior does this implement?

Comment 5

[Xidorn Quan [:xidorn] UTC+11]

The anchor position is the current position in the base line layout. Since an annotation is always reflowed before its base, any float in the annotation is positioned as if it is at the begining of the ruby pair (or ruby segment for spans).

e.g.

  <rb>hello world</rb><rt>hello<float/> world</rt>

would be identical to

  <float/><rb>hello world</rb><rt>hello world</rt>

Comment 6

[David Baron :dbaron:]

Comment on attachment 8534129 [details] [diff] [review]
patch

As long as the float gets pushed down if its anchor point gets pushed down (which I think it should), that sounds fine.

It might be good to have a reftest in addition to the crashtest that you have.

Comment 7

[Xidorn Quan [:xidorn] UTC+11]

This behavior is not stable yet. I have a plan to reflow ruby base before its ruby text for implementing ruby-merge: auto. (not decide, just thinking) In that way, the behavior here will change. As the exact behavior has not been specified by the spec, I'd prefer not to have a reftest at this point. Does it sound good?

Comment 8

[David Baron :dbaron:]

I don't think that should make a difference for simple cases -- the only ways it influences the position are:
 (1) ensuring the float is on the same line
 (2) relative positions when there are multiple floats in the same line

so I think you should be able to test cases where there's only one float in the line.

Comment 9

[Xidorn Quan [:xidorn] UTC+11]

try:
https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=d3df0b040125

Comment 10

[Xidorn Quan [:xidorn] UTC+11]

inbound:
https://hg.mozilla.org/integration/mozilla-inbound/rev/c10540d36073

Comment 11

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/c10540d36073