[Mlir-commits] [mlir] [MLIR][LLVM] Avoid creating invalid DICompositeTypes in import (PR #70797)

[Christian Ulmann]

https://github.com/Dinistro created https://github.com/llvm/llvm-project/pull/70797

This commit ensures that the debug info import skips `DICompositeTypes` that have an array type tag and failed to translate the base type. This is necessary because array `DICompositeTypes` require a base type to be valid.
Note that this is currently not verified in LLVM, instead it leads to an explosion of the `ASMPrinter`.

>From 1ca5494d26060456603ae8a98c9e5426ba7f4431 Mon Sep 17 00:00:00 2001
From: Christian Ulmann <christian.ulmann at nextsilicon.com>
Date: Tue, 31 Oct 2023 12:51:34 +0000
Subject: [PATCH] [MLIR][LLVM] Avoid creating invalid DICompositeTypes in
 import

This commit ensures that the debug info import skips `DICompositeTypes`
that have an array type tag and failed to translate the base type. This
is necessary because array `DICompositeTypes` require a base type to be
valid.
Note that this is currently not verified in LLVM, instead it leads to
an explosion of the `ASMPrinter`.
---
 mlir/lib/Target/LLVMIR/DebugImporter.cpp     | 10 ++++-
 mlir/test/Target/LLVMIR/Import/debug-info.ll | 39 +++++++++++++++++---
 2 files changed, 42 insertions(+), 7 deletions(-)

diff --git a/mlir/lib/Target/LLVMIR/DebugImporter.cpp b/mlir/lib/Target/LLVMIR/DebugImporter.cpp
index a3e81d0dd87a43a..36a03ec71c0c6c6 100644
--- a/mlir/lib/Target/LLVMIR/DebugImporter.cpp
+++ b/mlir/lib/Target/LLVMIR/DebugImporter.cpp
@@ -13,6 +13,7 @@
 #include "mlir/IR/Location.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/ScopeExit.h"
+#include "llvm/BinaryFormat/Dwarf.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DebugInfoMetadata.h"
 #include "llvm/IR/Metadata.h"
@@ -69,11 +70,16 @@ DICompositeTypeAttr DebugImporter::translateImpl(llvm::DICompositeType *node) {
   // TODO: Support debug metadata with cyclic dependencies.
   if (llvm::is_contained(elements, nullptr))
     elements.clear();
+  DITypeAttr baseType = translate(node->getBaseType());
+  // Arrays require a base type, otherwise the debug metadata is considered to
+  // be malformed.
+  if (node->getTag() == llvm::dwarf::DW_TAG_array_type && !baseType)
+    return nullptr;
   return DICompositeTypeAttr::get(
       context, node->getTag(), getStringAttrOrNull(node->getRawName()),
       translate(node->getFile()), node->getLine(), translate(node->getScope()),
-      translate(node->getBaseType()), flags.value_or(DIFlags::Zero),
-      node->getSizeInBits(), node->getAlignInBits(), elements);
+      baseType, flags.value_or(DIFlags::Zero), node->getSizeInBits(),
+      node->getAlignInBits(), elements);
 }
 
 DIDerivedTypeAttr DebugImporter::translateImpl(llvm::DIDerivedType *node) {
diff --git a/mlir/test/Target/LLVMIR/Import/debug-info.ll b/mlir/test/Target/LLVMIR/Import/debug-info.ll
index 2c4fb2c213cf3e2..08e60deb699d8d1 100644
--- a/mlir/test/Target/LLVMIR/Import/debug-info.ll
+++ b/mlir/test/Target/LLVMIR/Import/debug-info.ll
@@ -159,7 +159,7 @@ define void @derived_type() !dbg !3 {
 
 ; CHECK-DAG: #[[INT:.+]] = #llvm.di_basic_type<tag = DW_TAG_base_type, name = "int">
 ; CHECK-DAG: #[[FILE:.+]] = #llvm.di_file<"debug-info.ll" in "/">
-; CHECK-DAG: #[[COMP1:.+]] = #llvm.di_composite_type<tag = DW_TAG_array_type, name = "array1", line = 10, sizeInBits = 128, alignInBits = 32>
+; CHECK-DAG: #[[COMP1:.+]] = #llvm.di_composite_type<tag = DW_TAG_array_type, name = "array1", line = 10, baseType = #[[INT]], sizeInBits = 128, alignInBits = 32>
 ; CHECK-DAG: #[[COMP2:.+]] = #llvm.di_composite_type<{{.*}}, file = #[[FILE]], scope = #[[FILE]], baseType = #[[INT]]>
 ; CHECK-DAG: #[[COMP3:.+]] = #llvm.di_composite_type<{{.*}}, flags = Vector, elements = #llvm.di_subrange<count = 4 : i64>>
 ; CHECK-DAG: #[[COMP4:.+]] = #llvm.di_composite_type<{{.*}}, flags = Vector, elements = #llvm.di_subrange<lowerBound = 0 : i64, upperBound = 4 : i64, stride = 1 : i64>>
@@ -179,10 +179,10 @@ define void @composite_type() !dbg !3 {
 !4 = !DISubroutineType(types: !5)
 !5 = !{!7, !8, !9, !10, !18}
 !6 = !DIBasicType(name: "int")
-!7 = !DICompositeType(tag: DW_TAG_array_type, name: "array1", line: 10, size: 128, align: 32)
+!7 = !DICompositeType(tag: DW_TAG_array_type, name: "array1", line: 10, size: 128, align: 32, baseType: !6)
 !8 = !DICompositeType(tag: DW_TAG_array_type, name: "array2", file: !2, scope: !2, baseType: !6)
-!9 = !DICompositeType(tag: DW_TAG_array_type, name: "array3", flags: DIFlagVector, elements: !13)
-!10 = !DICompositeType(tag: DW_TAG_array_type, name: "array4", flags: DIFlagVector, elements: !14)
+!9 = !DICompositeType(tag: DW_TAG_array_type, name: "array3", flags: DIFlagVector, elements: !13, baseType: !6)
+!10 = !DICompositeType(tag: DW_TAG_array_type, name: "array4", flags: DIFlagVector, elements: !14, baseType: !6)
 !11 = !DISubrange(count: 4)
 !12 = !DISubrange(lowerBound: 0, upperBound: 4, stride: 1)
 !13 = !{!11}
@@ -192,7 +192,7 @@ define void @composite_type() !dbg !3 {
 !15 = !DISubrange(count: !16)
 !16 = !DILocalVariable(scope: !3, name: "size")
 !17 = !{!15}
-!18 = !DICompositeType(tag: DW_TAG_array_type, name: "unsupported_elements", flags: DIFlagVector, elements: !17)
+!18 = !DICompositeType(tag: DW_TAG_array_type, name: "unsupported_elements", flags: DIFlagVector, elements: !17, baseType: !6)
 
 ; // -----
 
@@ -558,3 +558,32 @@ define void @func_in_module(ptr %arg) !dbg !8 {
 !2 = !DIFile(filename: "debug-info.ll", directory: "/")
 !8 = distinct !DISubprogram(name: "func_in_module", scope: !10, file: !2, unit: !1);
 !10 = !DIModule(scope: !2, name: "module", configMacros: "bar", includePath: "/", apinotes: "/", file: !2, line: 42, isDecl: true)
+
+; // -----
+
+; Verifies that array types that have an unimportable base type are removed to
+; avoid producing invalid IR.
+; CHECK: #[[DI_LOCAL_VAR:.+]] = #llvm.di_local_variable<
+; CHECK-NOT: type =
+
+; CHECK-LABEL: @array_with_cyclic_base_type
+define i32 @array_with_cyclic_base_type(ptr %0) !dbg !3 {
+  call void @llvm.dbg.value(metadata ptr %0, metadata !4, metadata !DIExpression()), !dbg !7
+  ret i32 0
+}
+
+; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
+declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+
+!llvm.module.flags = !{!0}
+!llvm.dbg.cu = !{!1}
+
+!0 = !{i32 2, !"Debug Info Version", i32 3}
+!1 = distinct !DICompileUnit(language: DW_LANG_C, file: !2)
+!2 = !DIFile(filename: "debug-info.ll", directory: "/")
+!3 = distinct !DISubprogram(name: "func", scope: !2, file: !2, line: 46, scopeLine: 48, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !1)
+!4 = !DILocalVariable(name: "op", arg: 5, scope: !3, file: !2, line: 47, type: !5)
+!5 = !DICompositeType(tag: DW_TAG_array_type, size: 42, baseType: !6)
+!6 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !5)
+!7 = !DILocation(line: 0, scope: !3)