DWARF type hashing: Handle multiple (including recursive) references to the same type

This uses a map, keeping the type DIE numbering separate from the DIEs
themselves - alternatively we could do things the way GCC does if we
want to add an integer to the DIE type to record the numbering there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193105 91177308-0d34-0410-b5e6-96231b3b80d8

lib/CodeGen/AsmPrinter/DIEHash.cpp

@@ -196,22 +196,30 @@ void DIEHash::hashAttribute(AttrEntry Attr) {
   // 7.27s3
   // ... An attribute that refers to another type entry T is processed as
   // follows:
-  // a) If T is in the list of [previously hashed types], use the letter 'R' as
-  // the marker and use the unsigned LEB128 encoding of [the index of T in the
-  // list] as the attribute value; otherwise,
-
-  // [TODO: implement clause (a)]
-
   if (const DIEEntry *EntryAttr = dyn_cast<DIEEntry>(Value)) {
     DIE *Entry = EntryAttr->getEntry();
+    unsigned &DieNumber = Numbering[Entry];
+    if (DieNumber) {
+      // a) If T is in the list of [previously hashed types], use the letter
+      // 'R' as the marker
+      addULEB128('R');
 
-    // b) use the letter 'T' as a the marker, ...
+      addULEB128(Desc->getAttribute());
+
+      // and use the unsigned LEB128 encoding of [the index of T in the
+      // list] as the attribute value;
+      addULEB128(DieNumber);
+      return;
+    }
+
+    // otherwise, b) use the letter 'T' as a the marker, ...
     addULEB128('T');
 
     addULEB128(Desc->getAttribute());
 
     // ... process the type T recursively by performing Steps 2 through 7, and
     // use the result as the attribute value.
+    DieNumber = Numbering.size();
     computeHash(Entry);
     return;
   }
@@ -321,7 +329,6 @@ void DIEHash::addAttributes(DIE *Die) {
 // given in section 7.27 of the DWARF4 standard. It is the md5 hash of a
 // flattened description of the DIE.
 void DIEHash::computeHash(DIE *Die) {
-
   // Append the letter 'D', followed by the DWARF tag of the DIE.
   addULEB128('D');
   addULEB128(Die->getTag());
@@ -377,6 +384,8 @@ uint64_t DIEHash::computeDIEODRSignature(DIE *Die) {
 /// with the inclusion of the full CU and all top level CU entities.
 // TODO: Initialize the type chain at 0 instead of 1 for CU signatures.
 uint64_t DIEHash::computeCUSignature(DIE *Die) {
+  Numbering.clear();
+  Numbering[Die] = 1;
 
   // Hash the DIE.
   computeHash(Die);
@@ -396,6 +405,8 @@ uint64_t DIEHash::computeCUSignature(DIE *Die) {
 /// with the inclusion of additional forms not specifically called out in the
 /// standard.
 uint64_t DIEHash::computeTypeSignature(DIE *Die) {
+  Numbering.clear();
+  Numbering[Die] = 1;
 
   if (DIE *Parent = Die->getParent())
     addParentContext(Parent);

lib/CodeGen/AsmPrinter/DIEHash.h

@@ -11,6 +11,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "llvm/ADT/DenseMap.h"
 #include "llvm/Support/MD5.h"
 
 namespace llvm {
@@ -125,5 +126,6 @@ private:
 
 private:
   MD5 Hash;
+  DenseMap<DIE*, unsigned> Numbering;
 };
 }

unittests/CodeGen/DIEHashTest.cpp

@@ -92,7 +92,6 @@ TEST(TypeWithMember, DIEHash) {
   DIE *Member = new DIE(dwarf::DW_TAG_member);
   DIEString MemberStr(&Four, "member");
   Member->addValue(dwarf::DW_AT_name, dwarf::DW_FORM_strp, &MemberStr);
-  // type
   DIEInteger Zero(0);
   Member->addValue(dwarf::DW_AT_data_member_location, dwarf::DW_FORM_data1, &Zero);
 
@@ -112,4 +111,62 @@ TEST(TypeWithMember, DIEHash) {
 
   ASSERT_EQ(0x5646aa436b7e07c6ULL, MD5Res);
 }
+
+TEST(ReusedType, DIEHash) {
+  DIE Unnamed(dwarf::DW_TAG_structure_type);
+  DIEInteger Eight(8);
+  Unnamed.addValue(dwarf::DW_AT_byte_size, dwarf::DW_FORM_data1, &Eight);
+
+  DIE *Mem1 = new DIE(dwarf::DW_TAG_member);
+  DIEInteger Four(4);
+  DIEString Mem1Str(&Four, "mem1");
+  Mem1->addValue(dwarf::DW_AT_name, dwarf::DW_FORM_strp, &Mem1Str);
+  DIEInteger Zero(0);
+  Mem1->addValue(dwarf::DW_AT_data_member_location, dwarf::DW_FORM_data1, &Zero);
+
+  Unnamed.addChild(Mem1);
+
+  DIE *Mem2 = new DIE(dwarf::DW_TAG_member);
+  DIEString Mem2Str(&Four, "mem2");
+  Mem2->addValue(dwarf::DW_AT_name, dwarf::DW_FORM_strp, &Mem2Str);
+  Mem2->addValue(dwarf::DW_AT_data_member_location, dwarf::DW_FORM_data1, &Four);
+
+  Unnamed.addChild(Mem2);
+
+  DIE Int(dwarf::DW_TAG_base_type);
+  DIEString IntStr(&Four, "int");
+  Int.addValue(dwarf::DW_AT_name, dwarf::DW_FORM_strp, &IntStr);
+  Int.addValue(dwarf::DW_AT_byte_size, dwarf::DW_FORM_data1, &Four);
+  DIEInteger Five(5);
+  Int.addValue(dwarf::DW_AT_encoding, dwarf::DW_FORM_data1, &Five);
+
+  DIEEntry IntRef(&Int);
+  Mem1->addValue(dwarf::DW_AT_type, dwarf::DW_FORM_ref4, &IntRef);
+  Mem2->addValue(dwarf::DW_AT_type, dwarf::DW_FORM_ref4, &IntRef);
+
+  uint64_t MD5Res = DIEHash().computeTypeSignature(&Unnamed);
+
+  ASSERT_EQ(0x3a7dc3ed7b76b2f8ULL, MD5Res);
+}
+
+TEST(RecursiveType, DIEHash) {
+  DIE Foo(dwarf::DW_TAG_structure_type);
+  DIEInteger One(1);
+  Foo.addValue(dwarf::DW_AT_byte_size, dwarf::DW_FORM_data1, &One);
+  DIEString FooStr(&One, "foo");
+  Foo.addValue(dwarf::DW_AT_name, dwarf::DW_FORM_strp, &FooStr);
+
+  DIE *Mem = new DIE(dwarf::DW_TAG_member);
+  DIEString MemStr(&One, "mem");
+  Mem->addValue(dwarf::DW_AT_name, dwarf::DW_FORM_strp, &MemStr);
+  DIEEntry FooRef(&Foo);
+  Mem->addValue(dwarf::DW_AT_type, dwarf::DW_FORM_ref4, &FooRef);
+  // DW_AT_external and DW_AT_declaration are ignored anyway, so skip them.
+
+  Foo.addChild(Mem);
+
+  uint64_t MD5Res = DIEHash().computeTypeSignature(&Foo);
+
+  ASSERT_EQ(0x73d8b25aef227b06ULL, MD5Res);
+}
 }