fix: Stop shared values on plain DPs retaining their owners (backport #26123) - #26142

GitHub Issue: closes unoplatform/kahua-private#570

PR Type:

🐞 Bugfix

What changed? 🚀

A shared value from a ResourceDictionary, such as a {StaticResource} from App.xaml, can be set on a custom dependency property registered with plain PropertyMetadata, for example from a control Style setter. Uno then tracks the value as a "bindable child" of the control. On master, a shared Style or Brush used this way keeps its first owner control, and that control's DataContext, alive for as long as the dictionary lives. For App.xaml, that is the app's lifetime.

The first commit is the reporter's patch, kept as written. Every change on top of it is a separate commit.

The reporter's fix (781c253)

1. Weak owner link (root-cause fix). A shareable value (Style, Brush, Transition, …) recorded its owner in a strong _associatedParent field. That field was cleared only when the property was reassigned or a second owner appeared, so a single owner was rooted. The link is now weak, like the mentor link WinUI uses. A collected owner counts as no owner.
2. Templates and styles take no DataContext. WinUI's FrameworkTemplate and Style have no DataContext. Their stores now opt out of inheritance context when they are constructed, so no path can cache a DataContext on them. They are also no longer tracked as bindable children, matching the existing exclusion on the ValueDoesNotInheritDataContext path.
3. Release cleared and replaced values. Setting the property to null, a UIElement or a non-DependencyObject value now releases the previous child. The DataContext the owner pushed into it is cleared.

Follow-ups from review

• Releasing a shared child broke the other holders (9965ecb). Item 3 cleared a released child's DataContext even when another owner, or another property of the same owner, still held the child with the same DataContext. That holder's bindings went blank. This regression came with item 3. Each child now counts the bindable slots holding it, and its DataContext is cleared only when the last one lets it go.
  • Cost: one int field per DependencyObject.
  • The count isn't decremented when an owner is collected while it still holds the child. In that case, the child's DataContext is not cleared on a later release, as before this PR.
• No allocation per association (62d8ca9). The weak link allocated a finalizable WeakReference each time a shareable value got an owner, for example on every visual-state brush swap. It now borrows the owner's pooled SelfWeakReference, as the templated-parent link does.
• ALC unload sweep (ffb0d3d, defensive hardening). If an owner was collected before its AssemblyLoadContext unloaded, the sweep saw no parent and kept the DataContext that owner had pushed. A collected owner now counts as a cleared association, as the strong reference used to guarantee.
• Tests, comments that still described the strong field, and a catch in the sample's async void handler.

Behavior differences with WinUI

I checked the new tests on the native WinUI head in a local throwaway run, with the #if HAS_UNO guards removed. 11 of 15 pass there. These match WinUI:

• the leak tests for a shared Style, Brush and DataTemplate, inside or outside a dictionary;
• a template held in a collection;
• the binding and ContentTemplate regression guards;
• the two-owner tests: a child released by one holder keeps the DataContext of the remaining holder.

Uno is stricter than WinUI on purpose in two cases, because both keep the view model from being retained:

• WinUI keeps a released child following its former owner's DataContext. Uno detaches it.
• WinUI keeps the view model alive through a released child's binding while the child is alive. Uno clears it.

One test fails on WinUI because of how it is written: WinUI can't apply a Style that targets a control defined only in C#.

As in the reporter's patch, both test files stay #if HAS_UNO.

Known gap

A shared Brush held outside any ResourceDictionary (a static field or a code cache) is still left with its removed owner's view model. The owner itself is now collected. On master, both leaked. Dictionary values are covered by the existing IsResourceDictionaryItem guard, and templates and styles by item 2. Clearing that cached DataContext on unload would also need a re-push on reload, so it is tracked separately.

Validation

• Runtime, Skia desktop (Win32):
  • The reporter's 8 tests (Given_SharedTemplate_DataContext) and 7 new ones (Given_PlainDP_SharedValue) pass.
  • Without the fix, 5 of the reporter's 8 fail on master, which matches the issue. A Style or DataTemplate held outside a dictionary also leaks on master (checked with equivalent probe tests).
  • The 3 two-owner tests fail with the reporter's commit alone and pass with the follow-up. The same scenario, as a probe test, passes on master.
  • Regression run: 1008 passed, 0 failed, 5 skipped, 1 inconclusive (WebGPU). It covers Given_FrameworkTemplate, Given_Style, Given_UIElement, Given_ContentControl, Given_ResourceDictionary, Given_DependencyObjectCollection, Given_CollectibleAlcAssociations, Given_Image_Release, Given_BindingMemoryLeak, the template and DataContext suites, Given_xBind, and the Windows_UI_Xaml_Data and Windows_UI_Xaml_Media namespaces.
• Runtime, native WinUI (WinAppSDK): the local throwaway run described above.
• Unit tests: Uno.UI.UnitTests has no new failures. The 13 Given_Calendar failures are the existing time-zone baseline.
• Reporter's repro, against this build: the shared Style and shared Brush scenarios go from leaking the control and its view model on 7.1.0-dev.21 to nothing retained.
• Compile: a strict desktop build (analyzers on) and dotnet format whitespace are clean.
• Not run locally: WASM and mobile; CI will cover them.

The fix also applies to release/stable/6.7 and servicing/6.8. Those branches keep this code in DependencyObjectStore*.cs, so the backports will need manual ports.

PR Checklist ✅

• 🧪 Added Runtime tests, UI tests, or a manual test sample (for bug fixes / features, if applicable)
• 📚 Docs have been added/updated following the documentation template (for bug fixes / features)
• 🖼️ Validated PR Screenshots Compare Test Run results.
• ❗ Contains NO breaking changes
• 👀 Reviewed 2 other open pull requests (optional but appreciated!)

🤖 Generated with Claude Code