hs.ui.window().video() crashes on .show() — Swift generic-metadata init failure in AVKit VideoPlayer (macOS 27 beta)

[yosbeda]

First, thank you, @cmsj, for #149 and for shipping the video element so
quickly in #157. Really appreciate it.

What's happening

Calling .video() on an hs.ui.window() and then .show()-ing the window
reliably crashes the app. Reproduced with a minimal script built directly
from this repo's own docstring example for .video():

const clip = HSVideo.fromURLs(["/path/to/any/file.mp4"]);

hs.ui.window({ x: 100, y: 100, w: 640, h: 360 })
    .video(clip)
        .frame({ w: 640, h: 360 })
    .show();

clip.play();

What's been ruled out

• Not specific to any particular video file. Reproduced with an original
  H.264/portrait source and with a completely vanilla ffmpeg-generated test
  clip (640x480, 30fps, H.264 Main, AAC). Same crash either way.
• Not caused by surrounding app code. Reproduced with a from-scratch script
  containing nothing but the four lines above, no shared helpers, no layout
  containers, no window styling beyond the defaults.
• Not a threading issue. The entire crashing call stack is one unbroken
  synchronous frame chain on com.apple.main-thread, rooted in
  NSApplication sendEvent:.

Crash signature

SIGABRT via abort(), through swift::fatalError, getSuperclassMetadata,
_swift_initClassMetadataImpl, _AVKit_SwiftUI, ViewResponderFilter. init(inputs:view:), PlatformViewRepresentableFeature.modifyViewOutputs,
NSViewRepresentable._makeView, NSHostingView.viewDidMoveToWindow, and
finally NSWindow.setContentView:.

The failure is a Swift runtime generic-metadata instantiation error
(swift_getTypeByMangledName/getSuperclassMetadata) while resolving a
type needed by AVKit's VideoPlayer, which on macOS is backed by
NSViewRepresentable. It fires the moment the hosting view is attached to
the window.

Environment

• Hammerspoon 2 0.0.9 (same commit as PR Add a video element to hs.ui. Closes #149 #157's merge, 7a2dcab. App
  reports itself as version 1.2, build 87)
• macOS 27.0 (26A5388g), beta
• Apple Silicon (Macmini9,1)

A note on how this report was put together

This report, the diagnostic scripts, and the analysis above were produced
with AI assistance (Claude). I have not personally reviewed or verified the
Swift-level reasoning about why this happens (the generic-metadata /
NSViewRepresentable explanation). I can't read the Swift source myself. I
can vouch for the observable facts: the exact reproduction steps above, and
that I personally triggered the crash multiple times. Please treat the
"why" portion as a starting hypothesis from AI analysis of the crash trace,
not as something I've confirmed at the code level myself.

[cmsj]

@yosbeda is this still happening with the 27 Release Candidate?