Deduplicate #undef directives imported from multiple modules.

No functionality change, but deeply-importing module files are smaller and
faster now.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@232140 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/clang/Serialization/ASTReader.h b/include/clang/Serialization/ASTReader.h
index 951a21016f7802e560d838d037f784a0df2b62ef..d926c2176acc7c2ab32889e8d18bae6435f8693e 100644 (file)
--- a/include/clang/Serialization/ASTReader.h
+++ b/include/clang/Serialization/ASTReader.h
@@ -539,6 +539,14 @@ private:
   /// been loaded.
   std::vector<MacroInfo *> MacrosLoaded;
 
+  typedef std::pair<IdentifierInfo *, serialization::SubmoduleID>
+      LoadedMacroInfo;
+
+  /// \brief A set of #undef directives that we have loaded; used to
+  /// deduplicate the same #undef information coming from multiple module
+  /// files.
+  llvm::DenseSet<LoadedMacroInfo> LoadedUndefs;
+
   typedef ContinuousRangeMap<serialization::MacroID, ModuleFile *, 4>
     GlobalMacroMapType;
 
@@ -1844,7 +1852,8 @@ public:
   serialization::IdentifierID getGlobalIdentifierID(ModuleFile &M,
                                                     unsigned LocalID);
 
-  ModuleMacroInfo *getModuleMacro(const PendingMacroInfo &PMInfo);
+  ModuleMacroInfo *getModuleMacro(IdentifierInfo *II,
+                                  const PendingMacroInfo &PMInfo);
 
   void resolvePendingMacro(IdentifierInfo *II, const PendingMacroInfo &PMInfo);
 

diff --git a/lib/Serialization/ASTReader.cpp b/lib/Serialization/ASTReader.cpp
index da5a0e68e93aa388f6015339a58dfb0372d772f1..2ee4975a46883bfda8eda078c3077ed82d9ef7f8 100644 (file)
--- a/lib/Serialization/ASTReader.cpp
+++ b/lib/Serialization/ASTReader.cpp
@@ -826,7 +826,7 @@ IdentifierInfo *ASTIdentifierLookupTrait::ReadData(const internal_key_type& k,
         uint32_t LocalMacroID =
             endian::readNext<uint32_t, little, unaligned>(d);
         DataLen -= 4;
-        if (LocalMacroID == 0xdeadbeef) break;
+        if (LocalMacroID == (uint32_t)-1) break;
         LocalMacroIDs.push_back(LocalMacroID);
       }
     }
@@ -1807,7 +1807,7 @@ struct ASTReader::ModuleMacroInfo {
 };
 
 ASTReader::ModuleMacroInfo *
-ASTReader::getModuleMacro(const PendingMacroInfo &PMInfo) {
+ASTReader::getModuleMacro(IdentifierInfo *II, const PendingMacroInfo &PMInfo) {
   ModuleMacroInfo Info;
 
   uint32_t ID = PMInfo.ModuleMacroData.MacID;
@@ -1815,6 +1815,11 @@ ASTReader::getModuleMacro(const PendingMacroInfo &PMInfo) {
     // Macro undefinition.
     Info.SubModID = getGlobalSubmoduleID(*PMInfo.M, ID >> 1);
     Info.MI = nullptr;
+
+    // If we've already loaded the #undef of this macro from this module,
+    // don't do so again.
+    if (!LoadedUndefs.insert(std::make_pair(II, Info.SubModID)).second)
+      return nullptr;
   } else {
     // Macro definition.
     GlobalMacroID GMacID = getGlobalMacroID(*PMInfo.M, ID >> 1);
@@ -1848,7 +1853,7 @@ void ASTReader::resolvePendingMacro(IdentifierInfo *II,
 
   // Module Macro.
 
-  ModuleMacroInfo *MMI = getModuleMacro(PMInfo);
+  ModuleMacroInfo *MMI = getModuleMacro(II, PMInfo);
   if (!MMI)
     return;
 

diff --git a/lib/Serialization/ASTWriter.cpp b/lib/Serialization/ASTWriter.cpp
index 3f8db194d068e2b087dd5aa4a383d396c8f61591..15e7de48c9be76d46b60064629a38cb865dca750 100644 (file)
--- a/lib/Serialization/ASTWriter.cpp
+++ b/lib/Serialization/ASTWriter.cpp
@@ -3347,6 +3347,7 @@ public:
     using namespace llvm::support;
     endian::Writer<little> LE(Out);
 
+    assert((uint16_t)DataLen == DataLen && (uint16_t)KeyLen == KeyLen);
     LE.write<uint16_t>(DataLen);
     // We emit the key length after the data length so that every
     // string is preceded by a 16-bit length. This matches the PTH
@@ -3425,7 +3426,7 @@ public:
           }
           emitMacroOverrides(Out, getOverriddenSubmodules(MD, Scratch));
         }
-        LE.write<uint32_t>(0xdeadbeef);
+        LE.write<uint32_t>((uint32_t)-1);
       }
     }