[cxx-interop] [NFC] Stabilize interface order of Clang record members by j-hui · Pull Request #87026 · swiftlang/swift

j-hui · 2026-02-06T11:04:34Z

We may visit and thus import these members in an unpredictable order.
To avoid future churn for module interface test cases, sort the printed
module interface output according to some rough heuristics.

j-hui · 2026-02-06T11:16:34Z

@swift-ci please test

Xazax-hun

Overall looks good to me but I had a question and a small nit inline.

lib/AST/ASTPrinter.cpp

hnrklssn · 2026-02-06T20:30:18Z

test/Interop/C/bounds-safety/import-bounds-attributed-struct.swift

-// CHECK-NEXT:   init(a: UnsafeMutablePointer<Int{{[0-9]+}}>!, len: Int{{[0-9]+}})
-// CHECK-NEXT:   var a: UnsafeMutablePointer<Int{{[0-9]+}}>!
-// CHECK-NEXT:   var len: Int{{[0-9]+}}
+// CHECK-DAG:   init()


If this patch stabilises the order, I don't think we should use CHECK-DAG. It's significantly worse in terms of output when the checks fail.

For these, the order differs between C vs C++

I think it's because the default constructors are generated differently for C vs C++. C++ already has a notion of a default constructor, while in C it's entirely synthetic.

fair enough

test/Interop/Cxx/class/method/unsafe-projections.swift

test/Interop/Cxx/class/inheritance/virtual-methods-module-interface.swift

test/Interop/Cxx/stdlib/std-span-interface.swift

hnrklssn · 2026-02-06T20:36:01Z

test/Interop/Cxx/stdlib/std-span-interface.swift

+// CHECK-DAG:    /// This is an auto-generated wrapper for safer interop
+// CHECK-DAG:    @available(visionOS 1.0, tvOS 12.2, watchOS 5.2, iOS 12.2, macOS 10.14.4, *)
+// CHECK-DAG:    @_alwaysEmitIntoClient @_disfavoredOverload public mutating func methodWithSafeWrapper(_ s: Span<CInt>)
+// CHECK-DAG:    /// This is an auto-generated wrapper for safer interop
+// CHECK-DAG:    @available(visionOS 1.0, tvOS 12.2, watchOS 5.2, iOS 12.2, macOS 10.14.4, *)
+// CHECK-DAG:    @_lifetime(&self)
+// CHECK-DAG:    @_alwaysEmitIntoClient @_disfavoredOverload public mutating func getMutable(_ s: Span<CInt>) -> MutableSpan<CInt>


Also not a fan of this CHECK-DAG

These appear to be unstable. I'm not entirely sure why but I think it's because they are being sorted according to the filename of the macro-generated buffer.

urgh... can we sort macro expansion nodes according to the clang node the macro is expanded from?

Done, I think that worked. Fingers crossed for the CI runs

whelp. using SourceManager::getGeneratedSourceInfo() didn't work.

I hate to ask this, but would you be ok with punting this to follow-up work? I agree that the -DAG checks here is not great, but their "span" (no pun intended) is bounded by the following lines:

// CHECK-DAG: @_alwaysEmitIntoClient @_disfavoredOverload public mutating func getMutable(_ s: Span<CInt>) -> MutableSpan<CInt> // CHECK-NEXT: init() // CHECK-NEXT: mutating func methodWithSafeWrapper(_ s: ConstSpanOfInt) // CHECK-NEXT: mutating func getMutable(_ s: ConstSpanOfInt) -> SpanOfInt

I'm not familiar with how the -DAG matching algorithm is designed in lit but I would hope that having the CHECK-NEXTs here limit how much it needs to backtrack in the case of a non-match.

j-hui · 2026-02-06T21:20:43Z

test/Interop/Cxx/ergonomics/implicit-computed-properties-module-interface.swift

 // CHECK-NEXT: }

 // CHECK:      struct LongNameAllLower {
-// CHECK-NOT:     var


I have no idea what the intention of this original CHECK-NOT was supposed to be. It doesn't catch anything that consecutive CHECK-NEXTs would not catch.

hnrklssn

thanks for adding the extra QoL edits

j-hui · 2026-02-06T23:08:34Z

@swift-ci please test

j-hui · 2026-02-07T06:06:51Z

@swift-ci please test

j-hui · 2026-02-07T06:31:46Z

@swift-ci please test

We may visit and thus import these members in an unpredictable order. To avoid future churn for module interface test cases, sort the printed module interface output according to some rough heuristics.

j-hui · 2026-02-07T18:59:32Z

@swift-ci please test

eeckstein · 2026-02-09T12:33:46Z

test/Interop/Cxx/class/access/using-non-public-module-interface.swift

@@ -1,54 +1,54 @@
-// RUN: %target-swift-ide-test -print-module -module-to-print=UsingNonPublic -print-access -I %S/Inputs -source-filename=x -cxx-interoperability-mode=default -enable-experimental-feature ImportNonPublicCxxMembers | %FileCheck %s
+// RUN: %target-swift-ide-test -print-module -module-to-print=UsingNonPublic -print-access -I %S/Inputs -source-filename=x -cxx-interoperability-mode=default -enable-experimental-feature ImportNonPublicCxxMembers | tee %s.out | %FileCheck %s


@j-hui I guess the tee %s.out is just a left over from debugging. It produces the output file in the source directory. Can you please remove it?

Yes I'm very sorry, it is being removed here: #87075

j-hui requested review from Xazax-hun, egorzhdan, hamishknight, hborla, hnrklssn, rintaro, slavapestov, susmonteiro, tshortli and xedin as code owners February 6, 2026 11:04

j-hui force-pushed the sort-members branch from dee5d8b to 1814f34 Compare February 6, 2026 11:16

Xazax-hun reviewed Feb 6, 2026

View reviewed changes

lib/AST/ASTPrinter.cpp Outdated Show resolved Hide resolved

j-hui force-pushed the sort-members branch from 1814f34 to b005afc Compare February 6, 2026 20:23

j-hui requested a review from Xazax-hun February 6, 2026 20:23

j-hui force-pushed the sort-members branch from b005afc to 66c9559 Compare February 6, 2026 20:24

hnrklssn reviewed Feb 6, 2026

View reviewed changes

test/Interop/Cxx/class/method/unsafe-projections.swift Outdated Show resolved Hide resolved

Xazax-hun approved these changes Feb 6, 2026

View reviewed changes

hnrklssn reviewed Feb 6, 2026

View reviewed changes

test/Interop/Cxx/class/inheritance/virtual-methods-module-interface.swift Show resolved Hide resolved

hnrklssn reviewed Feb 6, 2026

View reviewed changes

test/Interop/Cxx/stdlib/std-span-interface.swift Outdated Show resolved Hide resolved

hnrklssn reviewed Feb 6, 2026

View reviewed changes

j-hui commented Feb 6, 2026

View reviewed changes

j-hui force-pushed the sort-members branch from 66c9559 to f7dbf44 Compare February 6, 2026 22:06

hnrklssn approved these changes Feb 6, 2026

View reviewed changes

j-hui force-pushed the sort-members branch from f7dbf44 to a00d7a4 Compare February 6, 2026 23:08

j-hui force-pushed the sort-members branch from a00d7a4 to 41286df Compare February 7, 2026 04:27

j-hui force-pushed the sort-members branch from 41286df to f27b12e Compare February 7, 2026 06:31

j-hui enabled auto-merge February 7, 2026 06:31

[cxx-interop] [NFC] Stabilize interface order of Clang record members

9dc1174

We may visit and thus import these members in an unpredictable order. To avoid future churn for module interface test cases, sort the printed module interface output according to some rough heuristics.

j-hui force-pushed the sort-members branch from f27b12e to 9dc1174 Compare February 7, 2026 18:59

j-hui mentioned this pull request Feb 7, 2026

[MemberLoading] Split storage and non-storage member loading #87065

Merged

j-hui merged commit 9f2f857 into swiftlang:main Feb 8, 2026
4 of 5 checks passed

finagolfin mentioned this pull request Feb 9, 2026

[android] Fix CI for recent trunk pulls #87075

Merged

eeckstein reviewed Feb 9, 2026

View reviewed changes

		@@ -1,54 +1,54 @@
		// RUN: %target-swift-ide-test -print-module -module-to-print=UsingNonPublic -print-access -I %S/Inputs -source-filename=x -cxx-interoperability-mode=default -enable-experimental-feature ImportNonPublicCxxMembers \| %FileCheck %s
		// RUN: %target-swift-ide-test -print-module -module-to-print=UsingNonPublic -print-access -I %S/Inputs -source-filename=x -cxx-interoperability-mode=default -enable-experimental-feature ImportNonPublicCxxMembers \| tee %s.out \| %FileCheck %s

Conversation

j-hui commented Feb 6, 2026

Uh oh!

j-hui commented Feb 6, 2026

Uh oh!

Xazax-hun left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

hnrklssn left a comment

Choose a reason for hiding this comment

Uh oh!

j-hui commented Feb 6, 2026

Uh oh!

j-hui commented Feb 7, 2026

Uh oh!

j-hui commented Feb 7, 2026

Uh oh!

j-hui commented Feb 7, 2026

Uh oh!

Uh oh!

eeckstein Feb 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

eeckstein Feb 9, 2026 •

edited

Loading