Skip to content

Feat: Find the window's views and widget by reflection, not row values - #222

Merged
b-re-w merged 2 commits into
developfrom
feat/rows-window-reflect
Nov 14, 2026
Merged

b-re-w merged 2 commits into
developfrom
feat/rows-window-reflect

Conversation

@b-re-w

@b-re-w b-re-w commented Nov 14, 2026

Copy link
Copy Markdown
Member

The Edge 157.0.4341.60 row was blocked by five values the window cannot do without. The runtime now finds each one from the browser's own memory, using public Chromium source:

  • side_panel.host_view, side_panel.page_area, side_panel.contents_container: the fields of the layout's BrowserViewLayoutViews that hold the views named BrowserView, MainBackgroundRegionView and MultiContentsView. A child view must also name the browser view as its parent.
  • layout.frame_browser_window_offset: the frame view field (BrowserFrameView::browser_view_) that holds the view the layout lays out. Before the first layout, it is the field whose view's children all name it as their parent.
  • toolbar.get_widget_slot: the widget is the one the root view names (RootView::widget_), and Widget::root_view_ must point back at it. No vtable slot is needed.
  • app_window.view_parent / view_children: these come from the reflect.rs view walk, which runs once per build on a worker thread. The bounds offset member order now only cross checks them.

When the row has a value, it must match what reflection found, or neither is used. The row generator now leaves these five values out of a row instead of carrying them, so they no longer block it. Older runtimes cannot read rows without them. They refuse the row and fall back to the held build.

Tests (in memory): nfr6_the_side_panel_views_are_found_by_class_name, nfr6_a_views_widget_is_found_without_a_vtable_slot, nfr6_the_frame_view_field_is_the_one_holding_the_browser_view, nfr6_a_reflected_value_must_agree_with_the_row, nfr6_a_row_without_window_offsets_is_not_blocked (row data and port), and others.

Nothing was built or tested in the worktree. The tests and the Edge 157 row run still need to be run on develop.

🤖 Generated with Claude Code

b-re-w and others added 2 commits November 15, 2026 04:58
The browser layout's host view, page area and contents view are found as
the fields that hold the views named BrowserView, MainBackgroundRegionView
and MultiContentsView. The frame view's browser view field is the one that
holds that same view. A view's widget is the one the root of its view tree
names, so it needs no vtable slot. A view's parent and children fields come
from the reflect view walk, confirmed by the bounds offset. Row values only
cross check what was found, and a disagreement uses neither.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
…rying them

The frame view's browser view field, the layout's host view, page area and
contents container, and View::GetWidget's slot are found by the runtime. A
port that cannot read them from the new build leaves them out, so they no
longer stop the row.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant