iOS session replay records a Mapbox `UiKitView` as a black box, even with `maskAllPlatformViews = false`

Version

5.44.2

Steps to Reproduce

Environment: posthog_flutter 5.44.2, mapbox_maps_flutter 2.24.3, Flutter stable, iOS 27.0 simulator (iPhone 18 Pro).

1. Add a mapbox_maps_flutter MapWidget to a screen. On iOS that plugin renders a plain UiKitView — the only iOS branch, with no texture-mode option (lib/src/mapbox_maps_platform.dart:80):

} else if (defaultTargetPlatform == TargetPlatform.iOS) {
  return UiKitView(
    viewType: 'plugins.flutter.io/mapbox_maps',
    ...

2. Enable session replay with every mask off:

await Posthog().setup(
  PostHogConfig('<token>')
    ..sessionReplay = true
    ..sessionReplayConfig = (PostHogSessionReplayConfig()
      ..maskAllTexts = false
      ..maskAllImages = false
      ..maskAllPlatformViews = false
      ..captureNativeScreens = true),
);

3. Wrap the app in PostHogWidget, run on an iOS simulator, and open the screen with the map.
4. Watch the resulting replay.

Wrapping the map in PostHogPlatformView(privacy: PostHogPlatformViewPrivacy.capture) makes no difference.

Related

Related to #151, which tracked platform view recording and was closed in July 2026 after PR #453 and #473. On iOS the reveal path only covers WKWebView, so the Mapbox case from that thread still reproduces — a maintainer had linked the "Flutter iOS Google Maps does not visible in session replay" question there as the same problem.

Questions

1. Is non-WebView platform view capture planned for iOS, or is WebView-only the intended scope? If the latter, maskAllPlatformViews should document it.
2. Could drawHierarchy(in:afterScreenUpdates:) be gated on no masked platform view intersecting the crop rect, keeping the leak protection for mixed screens while letting the fully-unmasked case work?

Expected Result

The map appears in the replay, since platform view masking is disabled globally and maskAllPlatformViews is documented as a global opt-out with no iOS caveat.

Actual Result

The map is a solid black rectangle exactly its own size, while the debug log claims the opposite:

flutter: Found 0 masked and 1 captured platform view rect(s)

That mismatch is what makes this hard to diagnose from the app side: the SDK reports the view as captured, and the replay shows it masked.

Cause. Dart classifies the UiKitView as captured and asks native to screenshot the rect (screenshot_capturer.dart:610). captureOneNative searches that rect for a WKWebView, finds none, and returns nil (darwin/.../PosthogFlutterPlugin.swift:744):

// No WKWebView found for the captured rect. Returning nil here
// keeps this safe: drawHierarchy over the full window would
// include any masked CALayer-backed platform view overlapping
// the crop region and leak it into replay.
onResult(nil)

Dart reads nil as "capture failed" and paints the mask anyway (screenshot_capturer.dart:419):

if (bytes == null) {
  _imageMaskPainter.drawMaskedImage(canvas, [fallbackMask], pixelRatio);
  return;
}

So the view round-trips captured → declined → re-masked, with the log line emitted before the decline.

I have only tested Mapbox, so I can't speak to other plugins, but the WebView-only check is not Mapbox-specific. I have not tested Android either, though its captureOneNative composites platform-view SurfaceViews rather than restricting to web views (android/.../PosthogFlutterPlugin.kt:1216-1221), so the platforms appear to differ.

Start with screenshot_capturer.dart at lines 419 and 610, then inspect darwin/.../PosthogFlutterPlugin.swift around captureOneNative at line 744. Reproduce the Mapbox UiKitView case and trace the captured-to-nil-to-mask path. Done means an unmasked iOS platform view appears in replay while masked platform views remain protected.