[HLSL] Allow narrowing in initialization lists

[llvm-beanz]

HLSL's initialization lists are extremely generous about allowing conversions. This change demotes the C++11 warning to the legacy warning when in HLSL mode.

Required for #56067

[llvmbot]

@llvm/pr-subscribers-hlsl

@llvm/pr-subscribers-clang

Author: Chris B (llvm-beanz)

Changes

HLSL's initialization lists are extremely generous about allowing conversions. This change demotes the C++11 warning to the legacy warning when in HLSL mode.

Required for #56067

---

Full diff: https://github.com/llvm/llvm-project/pull/108035.diff

2 Files Affected:

• (modified) clang/lib/Sema/SemaInit.cpp (+1-1)
• (modified) clang/test/AST/HLSL/vector-constructors.hlsl (+2-1)

diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index 7dc17187524621..8a32114a1ab313 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -9537,7 +9537,7 @@ static void DiagnoseNarrowingInInitList(Sema &S,
                       unsigned ConstRefDiagID, unsigned WarnDiagID) {
     unsigned DiagID;
     auto &L = S.getLangOpts();
-    if (L.CPlusPlus11 &&
+    if (L.CPlusPlus11 && !L.HLSL &&
         (!L.MicrosoftExt || L.isCompatibleWithMSVC(LangOptions::MSVC2015)))
       DiagID = IsConstRef ? ConstRefDiagID : DefaultDiagID;
     else
diff --git a/clang/test/AST/HLSL/vector-constructors.hlsl b/clang/test/AST/HLSL/vector-constructors.hlsl
index 905f11d9223248..9161ad110df1f1 100644
--- a/clang/test/AST/HLSL/vector-constructors.hlsl
+++ b/clang/test/AST/HLSL/vector-constructors.hlsl
@@ -1,4 +1,5 @@
-// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -x hlsl -ast-dump -o - %s | FileCheck %s
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -ast-dump -o - %s | FileCheck %s
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -ast-dump -o - %s | FileCheck %s
 
 typedef float float2 __attribute__((ext_vector_type(2)));
 typedef float float3 __attribute__((ext_vector_type(3)));

[bogner]

Can we test this?

[llvm-beanz]

Can we test this?

It is tested. The modified test case adds a run line with -std=hlsl202x which fails without this change due to narrowing conversion in an initializer list.

[coopp]

looks fine. Curious, why is this extra RUN: line added? Doesn't the other one already contain the same parameters -ast-dump -o - %s | FileCheck %s ?

RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -ast-dump -o - %s | FileCheck %s

[llvm-beanz]

looks fine. Curious, why is this extra RUN: line added? Doesn't the other one already contain the same parameters -ast-dump -o - %s | FileCheck %s ?

RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -ast-dump -o - %s | FileCheck %s

The diff is a bit awkward to read, but the second run line adds -std=hlsl202x to explicitly run under 202x, which (before this change) treated narrowing conversion in initializer lists as an error.