release/19.x: [flang] fix C_PTR function result lowering (#100082) #100324
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@clementval What do you think about merging this PR to the release branch? |
llvmbot
added
flang
|
@llvm/pr-subscribers-flang-fir-hlfir Author: None (llvmbot) ChangesBackport 1ead51a Requested by: @jeanPerier Full diff: https://github.com/llvm/llvm-project/pull/100324.diff 3 Files Affected:
diff --git a/flang/lib/Optimizer/Builder/FIRBuilder.cpp b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
index 2961df96b3cab..fbe79d0e45e5a 100644
--- a/flang/lib/Optimizer/Builder/FIRBuilder.cpp
+++ b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
@@ -1541,21 +1541,44 @@ mlir::Value fir::factory::genMaxWithZero(fir::FirOpBuilder &builder,
zero);
}
+static std::pair<mlir::Value, mlir::Type>
+genCPtrOrCFunptrFieldIndex(fir::FirOpBuilder &builder, mlir::Location loc,
+ mlir::Type cptrTy) {
+ auto recTy = mlir::cast<fir::RecordType>(cptrTy);
+ assert(recTy.getTypeList().size() == 1);
+ auto addrFieldName = recTy.getTypeList()[0].first;
+ mlir::Type addrFieldTy = recTy.getTypeList()[0].second;
+ auto fieldIndexType = fir::FieldType::get(cptrTy.getContext());
+ mlir::Value addrFieldIndex = builder.create<fir::FieldIndexOp>(
+ loc, fieldIndexType, addrFieldName, recTy,
+ /*typeParams=*/mlir::ValueRange{});
+ return {addrFieldIndex, addrFieldTy};
+}
+
mlir::Value fir::factory::genCPtrOrCFunptrAddr(fir::FirOpBuilder &builder,
mlir::Location loc,
mlir::Value cPtr,
mlir::Type ty) {
- assert(mlir::isa<fir::RecordType>(ty));
- auto recTy = mlir::dyn_cast<fir::RecordType>(ty);
- assert(recTy.getTypeList().size() == 1);
- auto fieldName = recTy.getTypeList()[0].first;
- mlir::Type fieldTy = recTy.getTypeList()[0].second;
- auto fieldIndexType = fir::FieldType::get(ty.getContext());
- mlir::Value field =
- builder.create<fir::FieldIndexOp>(loc, fieldIndexType, fieldName, recTy,
- /*typeParams=*/mlir::ValueRange{});
- return builder.create<fir::CoordinateOp>(loc, builder.getRefType(fieldTy),
- cPtr, field);
+ auto [addrFieldIndex, addrFieldTy] =
+ genCPtrOrCFunptrFieldIndex(builder, loc, ty);
+ return builder.create<fir::CoordinateOp>(loc, builder.getRefType(addrFieldTy),
+ cPtr, addrFieldIndex);
+}
+
+mlir::Value fir::factory::genCPtrOrCFunptrValue(fir::FirOpBuilder &builder,
+ mlir::Location loc,
+ mlir::Value cPtr) {
+ mlir::Type cPtrTy = fir::unwrapRefType(cPtr.getType());
+ if (fir::isa_ref_type(cPtr.getType())) {
+ mlir::Value cPtrAddr =
+ fir::factory::genCPtrOrCFunptrAddr(builder, loc, cPtr, cPtrTy);
+ return builder.create<fir::LoadOp>(loc, cPtrAddr);
+ }
+ auto [addrFieldIndex, addrFieldTy] =
+ genCPtrOrCFunptrFieldIndex(builder, loc, cPtrTy);
+ auto arrayAttr =
+ builder.getArrayAttr({builder.getIntegerAttr(builder.getIndexType(), 0)});
+ return builder.create<fir::ExtractValueOp>(loc, addrFieldTy, cPtr, arrayAttr);
}
fir::BoxValue fir::factory::createBoxValue(fir::FirOpBuilder &builder,
@@ -1596,15 +1619,6 @@ fir::BoxValue fir::factory::createBoxValue(fir::FirOpBuilder &builder,
return fir::BoxValue(box, lbounds, explicitTypeParams);
}
-mlir::Value fir::factory::genCPtrOrCFunptrValue(fir::FirOpBuilder &builder,
- mlir::Location loc,
- mlir::Value cPtr) {
- mlir::Type cPtrTy = fir::unwrapRefType(cPtr.getType());
- mlir::Value cPtrAddr =
- fir::factory::genCPtrOrCFunptrAddr(builder, loc, cPtr, cPtrTy);
- return builder.create<fir::LoadOp>(loc, cPtrAddr);
-}
-
mlir::Value fir::factory::createNullBoxProc(fir::FirOpBuilder &builder,
mlir::Location loc,
mlir::Type boxType) {
diff --git a/flang/lib/Optimizer/Transforms/AbstractResult.cpp b/flang/lib/Optimizer/Transforms/AbstractResult.cpp
index 3906aa553cb34..ff37310224e85 100644
--- a/flang/lib/Optimizer/Transforms/AbstractResult.cpp
+++ b/flang/lib/Optimizer/Transforms/AbstractResult.cpp
@@ -59,14 +59,16 @@ static mlir::FunctionType getNewFunctionType(mlir::FunctionType funcTy,
/*resultTypes=*/{});
}
+static mlir::Type getVoidPtrType(mlir::MLIRContext *context) {
+ return fir::ReferenceType::get(mlir::NoneType::get(context));
+}
+
/// This is for function result types that are of type C_PTR from ISO_C_BINDING.
/// Follow the ABI for interoperability with C.
static mlir::FunctionType getCPtrFunctionType(mlir::FunctionType funcTy) {
- auto resultType = funcTy.getResult(0);
- assert(fir::isa_builtin_cptr_type(resultType));
- llvm::SmallVector<mlir::Type> outputTypes;
- auto recTy = mlir::dyn_cast<fir::RecordType>(resultType);
- outputTypes.emplace_back(recTy.getTypeList()[0].second);
+ assert(fir::isa_builtin_cptr_type(funcTy.getResult(0)));
+ llvm::SmallVector<mlir::Type> outputTypes{
+ getVoidPtrType(funcTy.getContext())};
return mlir::FunctionType::get(funcTy.getContext(), funcTy.getInputs(),
outputTypes);
}
@@ -109,15 +111,11 @@ class CallConversion : public mlir::OpRewritePattern<Op> {
saveResult.getTypeparams());
llvm::SmallVector<mlir::Type> newResultTypes;
- // TODO: This should be generalized for derived types, and it is
- // architecture and OS dependent.
bool isResultBuiltinCPtr = fir::isa_builtin_cptr_type(result.getType());
- Op newOp;
- if (isResultBuiltinCPtr) {
- auto recTy = mlir::dyn_cast<fir::RecordType>(result.getType());
- newResultTypes.emplace_back(recTy.getTypeList()[0].second);
- }
+ if (isResultBuiltinCPtr)
+ newResultTypes.emplace_back(getVoidPtrType(result.getContext()));
+ Op newOp;
// fir::CallOp specific handling.
if constexpr (std::is_same_v<Op, fir::CallOp>) {
if (op.getCallee()) {
@@ -175,7 +173,7 @@ class CallConversion : public mlir::OpRewritePattern<Op> {
FirOpBuilder builder(rewriter, module);
mlir::Value saveAddr = fir::factory::genCPtrOrCFunptrAddr(
builder, loc, save, result.getType());
- rewriter.create<fir::StoreOp>(loc, newOp->getResult(0), saveAddr);
+ builder.createStoreWithConvert(loc, newOp->getResult(0), saveAddr);
}
op->dropAllReferences();
rewriter.eraseOp(op);
@@ -210,42 +208,52 @@ class ReturnOpConversion : public mlir::OpRewritePattern<mlir::func::ReturnOp> {
mlir::PatternRewriter &rewriter) const override {
auto loc = ret.getLoc();
rewriter.setInsertionPoint(ret);
- auto returnedValue = ret.getOperand(0);
- bool replacedStorage = false;
- if (auto *op = returnedValue.getDefiningOp())
- if (auto load = mlir::dyn_cast<fir::LoadOp>(op)) {
- auto resultStorage = load.getMemref();
- // The result alloca may be behind a fir.declare, if any.
- if (auto declare = mlir::dyn_cast_or_null<fir::DeclareOp>(
- resultStorage.getDefiningOp()))
- resultStorage = declare.getMemref();
- // TODO: This should be generalized for derived types, and it is
- // architecture and OS dependent.
- if (fir::isa_builtin_cptr_type(returnedValue.getType())) {
- rewriter.eraseOp(load);
- auto module = ret->getParentOfType<mlir::ModuleOp>();
- FirOpBuilder builder(rewriter, module);
- mlir::Value retAddr = fir::factory::genCPtrOrCFunptrAddr(
- builder, loc, resultStorage, returnedValue.getType());
- mlir::Value retValue = rewriter.create<fir::LoadOp>(
- loc, fir::unwrapRefType(retAddr.getType()), retAddr);
- rewriter.replaceOpWithNewOp<mlir::func::ReturnOp>(
- ret, mlir::ValueRange{retValue});
- return mlir::success();
- }
- resultStorage.replaceAllUsesWith(newArg);
- replacedStorage = true;
- if (auto *alloc = resultStorage.getDefiningOp())
- if (alloc->use_empty())
- rewriter.eraseOp(alloc);
+ mlir::Value resultValue = ret.getOperand(0);
+ fir::LoadOp resultLoad;
+ mlir::Value resultStorage;
+ // Identify result local storage.
+ if (auto load = resultValue.getDefiningOp<fir::LoadOp>()) {
+ resultLoad = load;
+ resultStorage = load.getMemref();
+ // The result alloca may be behind a fir.declare, if any.
+ if (auto declare = resultStorage.getDefiningOp<fir::DeclareOp>())
+ resultStorage = declare.getMemref();
+ }
+ // Replace old local storage with new storage argument, unless
+ // the derived type is C_PTR/C_FUN_PTR, in which case the return
+ // type is updated to return void* (no new argument is passed).
+ if (fir::isa_builtin_cptr_type(resultValue.getType())) {
+ auto module = ret->getParentOfType<mlir::ModuleOp>();
+ FirOpBuilder builder(rewriter, module);
+ mlir::Value cptr = resultValue;
+ if (resultLoad) {
+ // Replace whole derived type load by component load.
+ cptr = resultLoad.getMemref();
+ rewriter.setInsertionPoint(resultLoad);
}
- // The result storage may have been optimized out by a memory to
- // register pass, this is possible for fir.box results, or fir.record
- // with no length parameters. Simply store the result in the result storage.
- // at the return point.
- if (!replacedStorage)
- rewriter.create<fir::StoreOp>(loc, returnedValue, newArg);
- rewriter.replaceOpWithNewOp<mlir::func::ReturnOp>(ret);
+ mlir::Value newResultValue =
+ fir::factory::genCPtrOrCFunptrValue(builder, loc, cptr);
+ newResultValue = builder.createConvert(
+ loc, getVoidPtrType(ret.getContext()), newResultValue);
+ rewriter.setInsertionPoint(ret);
+ rewriter.replaceOpWithNewOp<mlir::func::ReturnOp>(
+ ret, mlir::ValueRange{newResultValue});
+ } else if (resultStorage) {
+ resultStorage.replaceAllUsesWith(newArg);
+ rewriter.replaceOpWithNewOp<mlir::func::ReturnOp>(ret);
+ } else {
+ // The result storage may have been optimized out by a memory to
+ // register pass, this is possible for fir.box results, or fir.record
+ // with no length parameters. Simply store the result in the result
+ // storage. at the return point.
+ rewriter.create<fir::StoreOp>(loc, resultValue, newArg);
+ rewriter.replaceOpWithNewOp<mlir::func::ReturnOp>(ret);
+ }
+ // Delete result old local storage if unused.
+ if (resultStorage)
+ if (auto alloc = resultStorage.getDefiningOp<fir::AllocaOp>())
+ if (alloc->use_empty())
+ rewriter.eraseOp(alloc);
return mlir::success();
}
@@ -263,8 +271,6 @@ class AddrOfOpConversion : public mlir::OpRewritePattern<fir::AddrOfOp> {
mlir::PatternRewriter &rewriter) const override {
auto oldFuncTy = mlir::cast<mlir::FunctionType>(addrOf.getType());
mlir::FunctionType newFuncTy;
- // TODO: This should be generalized for derived types, and it is
- // architecture and OS dependent.
if (oldFuncTy.getNumResults() != 0 &&
fir::isa_builtin_cptr_type(oldFuncTy.getResult(0)))
newFuncTy = getCPtrFunctionType(oldFuncTy);
@@ -298,8 +304,6 @@ class AbstractResultOpt
// Convert function type itself if it has an abstract result.
auto funcTy = mlir::cast<mlir::FunctionType>(func.getFunctionType());
if (hasAbstractResult(funcTy)) {
- // TODO: This should be generalized for derived types, and it is
- // architecture and OS dependent.
if (fir::isa_builtin_cptr_type(funcTy.getResult(0))) {
func.setType(getCPtrFunctionType(funcTy));
patterns.insert<ReturnOpConversion>(context, mlir::Value{});
diff --git a/flang/test/Fir/abstract-results.fir b/flang/test/Fir/abstract-results.fir
index 82f1cd33073fd..93e63dc657f0c 100644
--- a/flang/test/Fir/abstract-results.fir
+++ b/flang/test/Fir/abstract-results.fir
@@ -87,8 +87,8 @@ func.func @boxfunc_callee() -> !fir.box<!fir.heap<f64>> {
// FUNC-BOX: return
}
-// FUNC-REF-LABEL: func @retcptr() -> i64
-// FUNC-BOX-LABEL: func @retcptr() -> i64
+// FUNC-REF-LABEL: func @retcptr() -> !fir.ref<none>
+// FUNC-BOX-LABEL: func @retcptr() -> !fir.ref<none>
func.func @retcptr() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}> {
%0 = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}> {bindc_name = "rec", uniq_name = "_QFrecErec"}
%1 = fir.load %0 : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>
@@ -98,12 +98,14 @@ func.func @retcptr() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__addres
// FUNC-REF: %[[FIELD:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
// FUNC-REF: %[[ADDR:.*]] = fir.coordinate_of %[[ALLOC]], %[[FIELD]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
// FUNC-REF: %[[VAL:.*]] = fir.load %[[ADDR]] : !fir.ref<i64>
- // FUNC-REF: return %[[VAL]] : i64
+ // FUNC-REF: %[[CAST:.*]] = fir.convert %[[VAL]] : (i64) -> !fir.ref<none>
+ // FUNC-REF: return %[[CAST]] : !fir.ref<none>
// FUNC-BOX: %[[ALLOC:.*]] = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}> {bindc_name = "rec", uniq_name = "_QFrecErec"}
// FUNC-BOX: %[[FIELD:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
// FUNC-BOX: %[[ADDR:.*]] = fir.coordinate_of %[[ALLOC]], %[[FIELD]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
// FUNC-BOX: %[[VAL:.*]] = fir.load %[[ADDR]] : !fir.ref<i64>
- // FUNC-BOX: return %[[VAL]] : i64
+ // FUNC-BOX: %[[CAST:.*]] = fir.convert %[[VAL]] : (i64) -> !fir.ref<none>
+ // FUNC-BOX: return %[[CAST]] : !fir.ref<none>
}
// FUNC-REF-LABEL: func private @arrayfunc_callee_declare(
@@ -311,8 +313,8 @@ func.func @test_address_of() {
}
-// FUNC-REF-LABEL: func.func private @returns_null() -> i64
-// FUNC-BOX-LABEL: func.func private @returns_null() -> i64
+// FUNC-REF-LABEL: func.func private @returns_null() -> !fir.ref<none>
+// FUNC-BOX-LABEL: func.func private @returns_null() -> !fir.ref<none>
func.func private @returns_null() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
// FUNC-REF-LABEL: func @test_address_of_cptr
@@ -323,12 +325,12 @@ func.func @test_address_of_cptr() {
fir.call @_QMtest_c_func_modPsubr(%1) : (() -> ()) -> ()
return
- // FUNC-REF: %[[VAL_0:.*]] = fir.address_of(@returns_null) : () -> i64
- // FUNC-REF: %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (() -> i64) -> (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>)
+ // FUNC-REF: %[[VAL_0:.*]] = fir.address_of(@returns_null) : () -> !fir.ref<none>
+ // FUNC-REF: %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (() -> !fir.ref<none>) -> (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>)
// FUNC-REF: %[[VAL_2:.*]] = fir.convert %[[VAL_1]] : (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> (() -> ())
// FUNC-REF: fir.call @_QMtest_c_func_modPsubr(%[[VAL_2]]) : (() -> ()) -> ()
- // FUNC-BOX: %[[VAL_0:.*]] = fir.address_of(@returns_null) : () -> i64
- // FUNC-BOX: %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (() -> i64) -> (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>)
+ // FUNC-BOX: %[[VAL_0:.*]] = fir.address_of(@returns_null) : () -> !fir.ref<none>
+ // FUNC-BOX: %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (() -> !fir.ref<none>) -> (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>)
// FUNC-BOX: %[[VAL_2:.*]] = fir.convert %[[VAL_1]] : (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> (() -> ())
// FUNC-BOX: fir.call @_QMtest_c_func_modPsubr(%[[VAL_2]]) : (() -> ()) -> ()
}
@@ -380,18 +382,20 @@ func.func @test_indirect_calls_return_cptr(%arg0: () -> ()) {
// FUNC-REF: %[[VAL_0:.*]] = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}> {bindc_name = ".result"}
// FUNC-REF: %[[VAL_1:.*]] = fir.convert %[[ARG0]] : (() -> ()) -> (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>)
- // FUNC-REF: %[[VAL_2:.*]] = fir.convert %[[VAL_1]] : (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> (() -> i64)
- // FUNC-REF: %[[VAL_3:.*]] = fir.call %[[VAL_2]]() : () -> i64
+ // FUNC-REF: %[[VAL_2:.*]] = fir.convert %[[VAL_1]] : (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> (() -> !fir.ref<none>)
+ // FUNC-REF: %[[VAL_3:.*]] = fir.call %[[VAL_2]]() : () -> !fir.ref<none>
// FUNC-REF: %[[VAL_4:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
// FUNC-REF: %[[VAL_5:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_4]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
- // FUNC-REF: fir.store %[[VAL_3]] to %[[VAL_5]] : !fir.ref<i64>
+ // FUNC-REF: %[[CAST:.*]] = fir.convert %[[VAL_3]] : (!fir.ref<none>) -> i64
+ // FUNC-REF: fir.store %[[CAST]] to %[[VAL_5]] : !fir.ref<i64>
// FUNC-BOX: %[[VAL_0:.*]] = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}> {bindc_name = ".result"}
// FUNC-BOX: %[[VAL_1:.*]] = fir.convert %[[ARG0]] : (() -> ()) -> (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>)
- // FUNC-BOX: %[[VAL_2:.*]] = fir.convert %[[VAL_1]] : (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> (() -> i64)
- // FUNC-BOX: %[[VAL_3:.*]] = fir.call %[[VAL_2]]() : () -> i64
+ // FUNC-BOX: %[[VAL_2:.*]] = fir.convert %[[VAL_1]] : (() -> !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> (() -> !fir.ref<none>)
+ // FUNC-BOX: %[[VAL_3:.*]] = fir.call %[[VAL_2]]() : () -> !fir.ref<none>
// FUNC-BOX: %[[VAL_4:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
// FUNC-BOX: %[[VAL_5:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_4]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
- // FUNC-BOX: fir.store %[[VAL_3]] to %[[VAL_5]] : !fir.ref<i64>
+ // FUNC-BOX: %[[CAST:.*]] = fir.convert %[[VAL_3]] : (!fir.ref<none>) -> i64
+ // FUNC-BOX: fir.store %[[CAST]] to %[[VAL_5]] : !fir.ref<i64>
}
// ----------------------- Test GlobalOp rewrite ------------------------
|
pawosm-arm
approved these changes
Jul 24, 2024
Functions returning C_PTR were lowered to function returning intptr (i64 on 64bit arch). This caused conflicts when these functions were defined as returning !fir.ref<none>/llvm.ptr in other compiler generated contexts (e.g., malloc). Lower them to return !fir.ref<none>. This should deal with llvm#97325 and llvm#98644. (cherry picked from commit 1ead51a)
|
@jeanPerier (or anyone else). If you would like to add a note about this fix in the release notes (completely optional). Please reply to this comment with a one or two sentence description of the fix. When you are done, please add the release:note label to this PR. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Backport 1ead51a
Requested by: @jeanPerier