Fix Virtualize End anchoring after initial provider load

[ilonatommy]

Virtualize<TItem> with AnchorMode.End could remain at the beginning of the list after the initial ItemsProvider result populated an empty viewport.

Description

• An initially empty End-mode viewport is already at the bottom - wasAtBottomLastRender should be set conditionally.
• Apply pending spacer styles before reading scrollHeight to pin the viewport. This is required for WebAssembly render mode, where observer callbacks can run after the synchronous refresh.

For test reviewers:

The tests need to be deterministic, so using observer gate for timing the actions was necessary. Without the gate, the first IntersectionObserver callback does two things:

1. Records that the empty viewport is at the bottom.
2. Starts the initial provider request.

This creates a timing dependency. If the callback records the empty viewport's bottom state before the provider result is rendered, it can mask the bug and make an ordinary E2E test pass even on the unfixed implementation.

The test therefore temporarily holds Virtualize spacer observer callbacks and invokes RefreshDataAsync() explicitly. This produces the relevant initial 0 -> N provider transition before any observer callback can establish the bottom state. It then verifies that AnchorMode.End pins the populated viewport by itself.

Fixes #69302

[Copilot]

🔵 Needs a closer look

The E2E test setup must be corrected to exercise the intended transition safely and prevent callbacks from referencing deleted state.

Pull request overview

Fixes Virtualize<TItem> AnchorMode.End anchoring after the initial ItemsProvider load populates an empty viewport.

Changes:

• Corrects initial bottom-state tracking.
• Applies pending spacer styles before measuring scroll geometry.
• Adds refresh support and deterministic E2E coverage.

File summaries

File	Summary
src/Components/Web.JS/src/Virtualize.ts	Updates End anchoring and layout synchronization.
src/Components/test/testassets/BasicTestApp/VirtualizationAnchorMode.razor	Adds a refresh-data test control.
src/Components/test/E2ETest/Tests/VirtualizationTest.cs	Adds initial-load anchoring coverage; two moderate findings remain, each with 1 vote, regarding zero-count requests and observer callback cleanup.

Review details

Suppressed comments (2)

src/Components/test/E2ETest/Tests/VirtualizationTest.cs:5502

• Because the wrapper is installed before this reload, no OnSpacerBeforeVisible callback reaches Virtualize to initialize _visibleItemCapacity. The test provider honors request.Count with Take(request.Count), so this explicit refresh starts with a zero-count request and renders no items; scrollHeight - clientHeight remains zero and the assertion cannot exercise the intended 0→N transition. Seed a nonzero request capacity without allowing the bottom-tracking callback to run, or otherwise adjust the setup before calling RefreshDataAsync.

            Browser.Exists(By.Id("reload-with-initial-index")).Click();
            Browser.True(() => Convert.ToInt64(js.ExecuteScript(
                "return window.__virtualizeObserverCallbacks.length;"), CultureInfo.InvariantCulture) > 0);

            Browser.Exists(By.Id("refresh-data")).Click();
            Browser.Contains("Refreshed data", () => Browser.Exists(By.Id("status")).Text);
            Browser.True(() => GetMaximumScrollTop(js, container) > 0);

src/Components/test/E2ETest/Tests/VirtualizationTest.cs:5528

• Restoring the global constructor does not change the observer instance created from this wrapper. Its callback still calls window.__virtualizeObserverCallbacks.push(...), so deleting that property can make any later spacer notification throw a TypeError while the mounted Virtualize is still observing (and can pollute subsequent tests). Disable the wrapper for existing observers or keep a safe queue until the observer/page is disposed.

                window.IntersectionObserver = window.__nativeIntersectionObserver;
                delete window.__nativeIntersectionObserver;
                delete window.__virtualizeObserverCallbacks;

• Files reviewed: 3/3 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[ilonatommy]

• Because the wrapper is installed before this reload, no OnSpacerBeforeVisible callback reaches Virtualize to initialize _visibleItemCapacity. The test provider honors request.Count with Take(request.Count), so this explicit refresh starts with a zero-count request and renders no items; scrollHeight - clientHeight remains zero and the assertion cannot exercise the intended 0→N transition. Seed a nonzero request capacity without allowing the bottom-tracking callback to run, or otherwise adjust the setup before calling RefreshDataAsync.

Not true, request.Count might be but TotalItemCount=1000 still created large spacer. The test fails without the fix with the exactly issue reported.
The test does not render real rows before its first assertion, but that is intentional.

• Restoring the global constructor does not change the observer instance created from this wrapper. Its callback still calls window.__virtualizeObserverCallbacks.push(...), so deleting that property can make any later spacer notification throw a TypeError while the mounted Virtualize is still observing (and can pollute subsequent tests). Disable the wrapper for existing observers or keep a safe queue until the observer/page is disposed.

Looks valid.

[Copilot]

🔵 Needs a closer look

Lifecycle-sensitive observer ordering across Server and WebAssembly warrants final human validation.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 0 new
• Review effort level: Balanced