Update Compute integration to 0.5.2

[Kyle-Ye]

Summary

• update the default Compute source and binary version to 0.5.2 and use the matching binary checksum
• adapt the Compute shims to the updated input and graph tracing APIs
• restore the weak-attribute hashable conformance when consuming the binary distribution

Validation

• validated OpenSwiftUI UI test target compilation against both the Compute 0.5.2 source and binary distributions

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 32.17%. Comparing base (480e30e) to head (823c82e).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #236   +/-   ##
=======================================
  Coverage   32.17%   32.17%           
=======================================
  Files          71       71           
  Lines        2601     2601           
=======================================
  Hits          837      837           
  Misses       1764     1764           

Flag	Coverage Δ
macos	10.15% <ø> (ø)
ubuntu	12.03% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Kyle-Ye]

This is strange due to a current Compute release issue. I really do not want to add it. cc @jcmosc

1. Compute removed such conformance in 3c955fe3cff8ae29fa5a66ae087c0763d7f1d1c4. (I did not got the reason. But this is a small thing and I add patch it here if Compute do have a need to remove it.)
2. The weird thing is the 0.5.1 tag.

The source code point to b66dc81e2d16687b83769c492cd22ab699f28c0b which does not contain 3c955fe3cff8ae29fa5a66ae087c0763d7f1d1c4. So I do not need this patch.

But the 0.5.1 binary looks to be build by main after merge 3c955fe3cff8ae29fa5a66ae087c0763d7f1d1c4 which is 1fbf330926ffcebf74a9ea7e6dcd7957098f85c1 and includes 3c955fe3cff8ae29fa5a66ae087c0763d7f1d1c4. So there is a mismatch here.

[jcmosc]

I didn't realise the GitHub workflow was merging with main and then creating the XCFramework artifact. I've bumped the library to 0.5.2 with a new GitHub workflow that creates the artifact directly from the tagged commit instead of the merged PR.

Commit: jcmosc/Compute@fcea97e
New checksum: sha256:0eca53a3620776cffcc3d2047d11efaf893b53078b95e59e793f86caa3f2c169

Both source and binary should work now without your patch.

As for why the removal of the Hashable conformance - it's because my compatibility tests were failing due to linker error:

Undefined symbols for architecture arm64:
  "protocol conformance descriptor for __C.AGWeakAttribute : Swift.Hashable in AttributeGraph", referenced from:
      lazy protocol witness table accessor for type __C.AGWeakAttribute and conformance __C.AGWeakAttribute : Swift.Hashable in AttributeGraph in WeakAttributeTests.swift.o

You can also see the difference when calculating the layout string for AnyWeakAttribute:

With Hashable/Equatable conformance:

== IAGWeakAttribute, 8 bytes ==
(layout #:length 18 #:address 0x1009d3514
   (== #:size 8 #:type IAGWeakAttribute))

Without Hashable/Equatable conformance, verified by compatiblity testing:

== __Unnamed_struct__details, 8 bytes ==
(layout #:length 35 #:address 0x1009d3514
   (== #:size 4 #:type IAGAttribute)
   (== #:size 4 #:type UInt32))

[Kyle-Ye]

Undefined symbols for architecture arm64:
"protocol conformance descriptor for __C.AGWeakAttribute : Swift.Hashable in AttributeGraph", referenced from:
lazy protocol witness table accessor for type __C.AGWeakAttribute and conformance __C.AGWeakAttribute : Swift.Hashable in AttributeGraph in WeakAttributeTests.swift.o

This is not expected. Which platform and OS version are you trying to link with? @jcmosc

[jcmosc]

I'm on macOS 26.5.2

[Kyle-Ye]

I'm on macOS 26.6. And this AGExample in DPF works fine for me by using Equatable or Hashable conformance of AGWeakAttribute.

[Kyle-Ye]

We do not have such issue on 0.5.2 now. But the next version in main will got those conformance removed and we'll need to deal with it in the next release. And I still do not agree remove is a right choice.

[tuist]

🛠️ Tuist Run Report 🛠️

Tests 🧪

Scheme	Status	Cache hit rate	Tests	Skipped	Ran	Commit
OpenAttributeGraph-Package	✅	0 %	90	0	90	ae67115b8

Builds 🔨

Scheme	Status	Duration	Commit
OpenAttributeGraph	✅	10.0s	36e6fff51
OpenAttributeGraph-Package	✅	1m 7s	36e6fff51