Add Layout 2020 box tree support for Flexbox, behind a pref

[SimonSapin]

CC #26639

Layout support will come in future PRs. This one generates a zero-size fragment with no content.

[highfive]

Heads up! This PR modifies the following files:

• @emilio: components/style/properties/longhands/position.mako.rs, components/style/values/specified/box.rs

[nox]

@bors-servo r+

[bors-servo]

📌 Commit 1ca05db has been approved by nox

[bors-servo]

⌛ Testing commit 1ca05db with merge 833dca9...

[bors-servo]

💔 Test failed - status-taskcluster

[SimonSapin]

Failures are legit, I messed up the cfg for serialization of display: inline-table. Pushed a fixup commit for easier review, I’ll squash it before landing.

I’ve also pushed other commits for sprinkling some rayon on.

[SimonSapin]

@bors-servo r=nox

[bors-servo]

📌 Commit 5fa3312 has been approved by nox

[bors-servo]

⌛ Testing commit 5fa3312 with merge af723c3...

[bors-servo]

💔 Test failed - status-taskcluster

[SimonSapin]

This PR adds a FlexLevel variant to the LayoutBox enum. PR #26741 which landed in the meantime added code that matches on LayoutBox. So the merge does not conflict but fails to compile.

The fix is straightforward:

    fixup! Flex items in the box tree

diff --git components/layout_2020/flow/root.rs components/layout_2020/flow/root.rs
index 61f6ca888b..79285767d3 100644
--- components/layout_2020/flow/root.rs
+++ components/layout_2020/flow/root.rs
@@ -10,6 +10,7 @@ use crate::display_list::stacking_context::{
 };
 use crate::dom_traversal::{iter_child_nodes, Contents, NodeExt};
 use crate::element_data::LayoutBox;
+use crate::flexbox::FlexLevelBox;
 use crate::flow::construct::ContainsFloats;
 use crate::flow::float::FloatBox;
 use crate::flow::inline::InlineLevelBox;
@@ -119,6 +120,7 @@ impl BoxTree {
         enum UpdatePoint {
             AbsolutelyPositionedBlockLevelBox(ArcRefCell<BlockLevelBox>),
             AbsolutelyPositionedInlineLevelBox(ArcRefCell<InlineLevelBox>),
+            AbsolutelyPositionedFlexLevelBox(ArcRefCell<FlexLevelBox>),
         }
 
         fn update_point<'dom, Node>(
@@ -188,6 +190,16 @@ impl BoxTree {
                         },
                         _ => return None,
                     },
+                    LayoutBox::FlexLevel(flex_level_box) => match &*flex_level_box.borrow() {
+                        FlexLevelBox::OutOfFlowAbsolutelyPositionedBox(_)
+                            if box_style.position.is_absolutely_positioned() =>
+                        {
+                            UpdatePoint::AbsolutelyPositionedFlexLevelBox(
+                                flex_level_box.clone(),
+                            )
+                        },
+                        _ => return None,
+                    },
                 };
             Some((primary_style.clone(), display_inside, update_point))
         }
@@ -217,6 +229,12 @@ impl BoxTree {
                                 out_of_flow_absolutely_positioned_box,
                             );
                     },
+                    UpdatePoint::AbsolutelyPositionedFlexLevelBox(flex_level_box) => {
+                        *flex_level_box.borrow_mut() =
+                            FlexLevelBox::OutOfFlowAbsolutelyPositionedBox(
+                                out_of_flow_absolutely_positioned_box,
+                            );
+                    },
                 }
                 return true;
             }

[SimonSapin]

@bors-servo r=nox

[bors-servo]

📌 Commit b9069d4 has been approved by nox

[bors-servo]

⌛ Testing commit b9069d4 with merge 3d6fed8...

[bors-servo]

☀️ Test successful - status-taskcluster
Approved by: nox
Pushing 3d6fed8 to master...

[Manishearth]

is this something that should be cached?

[SimonSapin]

It could be, but I’m not sure that would be a win. I’d guess that white-space only text nodes are typically small (at most a couple newlines plus indentation in non-minified HTML) so this should run quickly, and in other cases we’ll typically find a non-white-space character not far from the start where Iterator::all will short-circuit and we’ll also return quickly.