[iOS] New Arch: PagerView renders a blank page when reactViewController() is nil on first didMoveToWindow

[IsaacIsrael]

Environment

Package	Version
react-native-pager-view	8.0.0 – 8.0.2 (current master)
react-native	0.85.x
Architecture	New Architecture (Fabric), iOS

Description

On the New Architecture iOS path, PagerView intermittently renders a permanently blank page — no content is visible even though children are mounted with valid frames, and switching pages does not recover it.

It is timing-dependent, so it reproduces inconsistently — most often when the pager is mounted as part of a screen that is still being attached (e.g. behind a navigation transition, inside a freshly-pushed screen or a tab).

Root cause

In ios/PagerViewProvider.swift, setupView() assigns self.hostingController before confirming that reactViewController() is non-nil:

private func setupView() {
  if self.hostingController != nil {
    return
  }

  self.hostingController = UIHostingController(   // assigned unconditionally
    rootView: PagerView(props: props, delegate: delegate),
    ignoreSafeArea: true
  )
  if let hostingController, let parentViewController = reactViewController() {
    // addChild / addSubview / pinEdges happen ONLY here
  }
}

setupView() is called from didMoveToWindow(). On the first pass reactViewController() can be nil (the view is in a window but its owning view controller isn't attached yet). When that happens:

1. The if let … reactViewController() block is skipped, so the SwiftUI hosting view is never added as a subview and never pinned → its frame stays .zero → the page is blank.
2. self.hostingController is now non-nil, so the next didMoveToWindow() (when the VC is available) hits the if self.hostingController != nil { return } early-return and setup never retries.

Net result: a blank pager for the lifetime of the view.

Evidence (device logs)

With temporary logging in PagerViewProvider / PageChildViewController:

setupView creating hostingController ... reactVC=0          // reactViewController() was nil
didMoveToWindow hasWindow=0 ...                              // detached
didMoveToWindow hasWindow=1 provider.bounds={390,754} ...    // re-attached, VC now exists
setupView SKIPPED (hostingController already exists)         // never retries
provider.layoutSubviews bounds={390,653} hostingView.frame={0,0,0,0}   // host never sized
provider.layoutSubviews bounds={390,700} hostingView.frame={0,0,0,0}   // stays blank

provider.bounds is correct while hostingController.view.frame stays {0,0,0,0}.

Reproducible steps

1. iOS, New Architecture enabled.
2. Render a PagerView inside a screen that is pushed/attached (so reactViewController() is nil on the first didMoveToWindow).
3. The page area is blank (intermittently); switching pages does not help.

Proposed fix

Guard on reactViewController() before creating/assigning the hosting controller (so a failed pass leaves hostingController nil and can retry), and retry setupView() from layoutSubviews() so attachment happens as soon as the VC hierarchy is ready.

PR: #1088

[ivannzme]

Tried the #1088 approach (guard setupView() on reactViewController() + retry from layoutSubviews()) on 8.0.0, RN 0.83.6, react-native-screens 4.23.0, New Arch, iOS. It fixes the blank-on-first-mount case here, but there's a second failure mode in the same area it does not cover:

When a PagerView screen is pushed over (another native-stack screen on top) and popped back, the pager shows the correct page but its horizontal swipe is dead. Taps and vertical scroll inside the pages still work. First mount is fine; only the push/pop round trip breaks it.

I went further and also tore the UIHostingController down in didMoveToWindow when window == nil and recreated it on return. Logs confirm the host is destroyed on navigate-away and a fresh UIHostingController is created on return, yet the freshly-created host still cannot be swiped. So at least for this case the dead gesture is not explained by the hosting-controller lifecycle alone.

I also saw several attach/detach cycles per navigation (repeated didMoveToWindow window=true/false, hostingController == nil on consecutive window=true), which looks like the screen being remounted as new instances rather than reattached. Can provide a repro on request.

[ivannzme]

Following up on my earlier comment to correct it, so it doesn't mislead anyone: the "second failure mode" I described (PagerView swipe dead after pushing another screen and popping back) turned out to be a bug in our own app, not in pager-view.

Our content screen remounts on back-navigation (and PiP restore), and our swipe-list state was consumed-once and cleared on mount, so on return the screen rendered a single non-paged view. There was no PagerView to swipe in the first place. The native host logs I posted were transition churn from that remount, not evidence of a lost gesture. Once we kept that state across the remount, swiping works again on both iOS and Android.

So please disregard my earlier "recreating the host doesn't fix it" note as evidence about pager-view. The actual issue reported here (#1089, blank page when reactViewController() is nil on the first didMoveToWindow) still stands as filed; I just don't have evidence of a separate gesture-loss bug. Apologies for the noise.