Fix PAGImageView deadlock when size updates trigger flush

[Batxent]

Summary

• move PAGImageView animator updates out of the imageViewLock scope in setBounds: and setFrame:
• prevent update -> onAnimationFlush -> flush() from re-entering the same non-recursive mutex
• add inline comments to document the lock-ordering requirement

Root Cause

PAGImageView used to call [animator update] while still holding imageViewLock during size changes.

That update path can synchronously flow through:

update -> onAnimationFlush -> flush()

flush() also acquires imageViewLock, so the same thread can try to lock the same non-recursive mutex twice and deadlock.

Fix

Keep decoder reset and size-related state updates inside imageViewLock, but defer [animator update] until after the lock has been released.

This preserves the existing flush behavior while removing the re-entrant lock path.

Validation

• reproduced a deadlock when PAGImageView handled size changes and an immediate flush was triggered
• built libpag for iphoneos with the patch
• verified the deadlock no longer occurs with the patched binary

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 75.59%. Comparing base (f456817) to head (e8d31bd).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3348      +/-   ##
==========================================
- Coverage   75.60%   75.59%   -0.01%     
==========================================
  Files         510      510              
  Lines       35435    35435              
  Branches    11283    11283              
==========================================
- Hits        26789    26786       -3     
+ Misses       6356     6355       -1     
- Partials     2290     2294       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Batxent]

The main issue here is lock re-entrancy.

setBounds: / setFrame: previously called [animator update] while imageViewLock was still held.
That update path can synchronously flow back into PAGImageView::flush(), which also acquires imageViewLock.

So the effective path was:

setBounds/setFrame
-> lock imageViewLock
-> update
-> onAnimationFlush
-> flush
-> lock imageViewLock again

Because the mutex is non-recursive, this can deadlock on the same thread.

The fix is to keep the decoder reset under the lock, but defer [animator update] until after unlock.