Fix: Mark ancestors dirty in mutation methods for proper style traversal

[joeleaver]

Summary

This PR fixes a bug where DOM mutations via set_attribute, set_node_text, and clear_attribute would not result in visual updates during continuous operations (like slider dragging), even though the correct damage/hint flags were being set.

The Problem

The Stylo style traversal uses the dirty_descendants flag on ancestor nodes to determine which subtrees need restyling. When a node needs restyling, the traversal algorithm walks down from the root but only descends into subtrees where an ancestor has dirty_descendants = true.

The mutation methods (set_attribute, set_node_text, clear_attribute) were correctly setting:

• RestyleHint::restyle_subtree() on the node
• ALL_DAMAGE on the node

However, they were not calling mark_ancestors_dirty(), which sets the dirty_descendants flag on all ancestor nodes up to the root.

Without this flag, the style traversal would skip the modified nodes entirely, leaving their damage unprocessed until something else (like a hover state change) triggered a proper traversal of that subtree.

Why hover() works correctly

The hover() and unhover() methods on Node call set_restyle_hint(), which internally calls mark_ancestors_dirty():

pub fn set_restyle_hint(&self, hint: RestyleHint) {
    if let Some(element_data) = self.stylo_element_data.borrow_mut().as_mut() {
        element_data.hint.insert(hint);
    }
    // Mark all ancestors as having dirty descendants so the style traversal
    // will visit this node's subtree
    self.mark_ancestors_dirty();  // <-- This is the key!
}

This is why hover state changes would "fix" the visual state - they would trigger a proper traversal that would also pick up pending damage from earlier mutations.

The Fix

Added mark_ancestors_dirty() calls to:

• set_node_text() - after insert_damage()
• set_attribute() - after setting hint/damage on node and parent
• clear_attribute() - after setting hint/damage

Testing

Tested with a slider widget that updates inline styles during drag. Before this fix, the slider bar/thumb would not move visually during drag (only on mouse release or when cursor moved over a different element). After this fix, the slider updates in real-time during drag.

Impact

This is a correctness fix with minimal performance impact. The mark_ancestors_dirty() function is O(tree depth) and early-exits when it finds an ancestor that already has the flag set, so repeated mutations to the same subtree have minimal overhead.

[nicoburns]

@joeleaver This looks like the right fix where it's fixed, but I think there are a few more places where the damage needs to be set. Namely when adding and removing nodes.

Specifically I thinkremove_node, remove_and_drop_node, remove_and_drop_all_children, and add_children_to_parent. And perhaps process_added_subtree (but it doesn't need to recurse up the hierarchy, it can just set it on the specific node as it's already doing a downward recursive traversal, and the parent is already covered).

But you may want to use your own judgement, because it's definitely tricky stuff this.

---

Btw, is there a reason why there are now 2 PRs? Could this and #332 be combined into one?

[joeleaver]

Sorry, #332 covers this, this was made in error.

[nicoburns]

(#332 appears not to contain the second commit from this PR atm)