llvm · mstorsjo · Jul 31, 2020
diff --git a/llvm/test/tools/llvm-rc/Inputs/split-path.rc b/llvm/test/tools/llvm-rc/Inputs/split-path.rc
@@ -0,0 +1 @@
+100 ICON "subdir" "/icon-new.ico"
diff --git a/llvm/test/tools/llvm-rc/Inputs/tokens-windres-invalid-concat.rc b/llvm/test/tools/llvm-rc/Inputs/tokens-windres-invalid-concat.rc
@@ -0,0 +1 @@
+"narrow" L"wide"
diff --git a/llvm/test/tools/llvm-rc/Inputs/tokens-windres.rc b/llvm/test/tools/llvm-rc/Inputs/tokens-windres.rc
@@ -0,0 +1,3 @@
+L"wide" " also wide",
+"narrow" " also narrow",
+L"wide" L" more wide",
diff --git a/llvm/test/tools/llvm-rc/split-path.test b/llvm/test/tools/llvm-rc/split-path.test
@@ -0,0 +1,6 @@
+; RUN: rm -rf %t
+; RUN: mkdir %t
+; RUN: cd %t
+; RUN: mkdir subdir
+; RUN: cp %p/Inputs/icon-new.ico subdir
+; RUN: llvm-windres --no-preprocess %p/Inputs/split-path.rc %t/split-path.res
diff --git a/llvm/test/tools/llvm-rc/windres-format.test b/llvm/test/tools/llvm-rc/windres-format.test
@@ -2,11 +2,11 @@
 ; from file suffixes.
 
 ; RUN: rm -f %t.res
-; RUN: llvm-windres --no-preprocess %p/Inputs/tag-stringtable-basic.rc %t.res
+; RUN: llvm-windres --no-preprocess %p/Inputs/tag-versioninfo.rc %t.res
 ; RUN: llvm-readobj %t.res | FileCheck %s --check-prefix=CHECK-RES
 
 ; RUN: rm -f %t.o
-; RUN: llvm-windres --no-preprocess -F pe-x86-64 %p/Inputs/tag-stringtable-basic.rc %t.o
+; RUN: llvm-windres --no-preprocess -F pe-x86-64 %p/Inputs/tag-versioninfo.rc %t.o
 ; RUN: llvm-readobj --coff-resources %t.o | FileCheck %s --check-prefix=CHECK-OBJ
 
 ; RUN: rm -f %t.obj
@@ -16,7 +16,7 @@
 ; Check that we can specify the input/output file types explicitly.
 ; Also check options for specifying the input/output file names.
 
-; RUN: cat %p/Inputs/tag-stringtable-basic.rc > %t-anonymous
+; RUN: cat %p/Inputs/tag-versioninfo.rc > %t-anonymous
 ; RUN: rm -f %t-anonymous2
 ; RUN: llvm-windres --no-preprocess -O res -J rc -o %t-anonymous2 -i %t-anonymous
 ; RUN: llvm-readobj %t-anonymous2 | FileCheck %s --check-prefix=CHECK-RES
@@ -25,7 +25,7 @@
 ; RUN: llvm-windres -F pe-x86-64 -J res -O coff -i%t-anonymous2 -o%t-anonymous3
 ; RUN: llvm-readobj --coff-resources %t-anonymous3 | FileCheck %s --check-prefix=CHECK-OBJ
 
-; CHECK-RES: Resource type (int): STRINGTABLE
+; CHECK-RES: Resource type (int):
 
 ; CHECK-OBJ: Format: COFF-x86-64
 ; CHECK-OBJ: Resources [

diff --git a/llvm/test/tools/llvm-rc/windres-prefix.test b/llvm/test/tools/llvm-rc/windres-prefix.test
@@ -10,7 +10,7 @@
 
 ; Check that the triple prefix also affects the output object file type.
 
-; RUN: %t/aarch64-w64-mingw32-windres --no-preprocess %p/Inputs/tag-stringtable-basic.rc %t.o
+; RUN: %t/aarch64-w64-mingw32-windres --no-preprocess %p/Inputs/tag-versioninfo.rc %t.o
 ; RUN: llvm-readobj --coff-resources %t.o | FileCheck %s --check-prefix=CHECK-OBJ
 
 ; CHECK-OBJ: Format: COFF-ARM64

diff --git a/llvm/test/tools/llvm-rc/windres-target.test b/llvm/test/tools/llvm-rc/windres-target.test
@@ -17,13 +17,13 @@
 
 ; Check the actual written object types.
 
-; RUN: llvm-windres --no-preprocess -F i686-w64-mingw32 %p/Inputs/tag-stringtable-basic.rc %t.o
+; RUN: llvm-windres --no-preprocess -F i686-w64-mingw32 %p/Inputs/tag-versioninfo.rc %t.o
 ; RUN: llvm-readobj --coff-resources %t.o | FileCheck %s --check-prefixes=CHECK-OBJ,CHECK-OBJ-I686
-; RUN: llvm-windres --no-preprocess -F x86_64-w64-mingw32 %p/Inputs/tag-stringtable-basic.rc %t.o
+; RUN: llvm-windres --no-preprocess -F x86_64-w64-mingw32 %p/Inputs/tag-versioninfo.rc %t.o
 ; RUN: llvm-readobj --coff-resources %t.o | FileCheck %s --check-prefixes=CHECK-OBJ,CHECK-OBJ-X86-64
-; RUN: llvm-windres --no-preprocess -F armv7-w64-mingw32 %p/Inputs/tag-stringtable-basic.rc %t.o
+; RUN: llvm-windres --no-preprocess -F armv7-w64-mingw32 %p/Inputs/tag-versioninfo.rc %t.o
 ; RUN: llvm-readobj --coff-resources %t.o | FileCheck %s --check-prefixes=CHECK-OBJ,CHECK-OBJ-ARMV7
-; RUN: llvm-windres --no-preprocess -F aarch64-w64-mingw32 %p/Inputs/tag-stringtable-basic.rc %t.o
+; RUN: llvm-windres --no-preprocess -F aarch64-w64-mingw32 %p/Inputs/tag-versioninfo.rc %t.o
 ; RUN: llvm-readobj --coff-resources %t.o | FileCheck %s --check-prefixes=CHECK-OBJ,CHECK-OBJ-AARCH64
 
 ; CHECK-OBJ-I686: Format: COFF-i386

diff --git a/llvm/test/tools/llvm-rc/windres-tokenizer.test b/llvm/test/tools/llvm-rc/windres-tokenizer.test
@@ -0,0 +1,13 @@
+; RUN: not llvm-windres --no-preprocess --verbose %p/Inputs/tokens-windres.rc %t.res | FileCheck %s
+; llvm-windres fails on this sample because it is an invalid resource file
+; script. We silence the error message and just analyze the output.
+
+; CHECK:  String: L"wide also wide"
+; CHECK-NEXT:  Comma: ,
+; CHECK-NEXT:  String: "narrow also narrow"
+; CHECK-NEXT:  Comma: ,
+; CHECK-NEXT:  String: L"wide more wide"
+
+; RUN: not llvm-windres --no-preprocess --verbose %p/Inputs/tokens-windres-invalid-concat.rc %t.res 2>&1 | FileCheck %s --check-prefix=CONCAT-ERROR
+
+; CONCAT-ERROR: Error parsing file: Cannot concatenate a wide string L"wide" after a narrow one "narrow"
diff --git a/llvm/tools/llvm-rc/ResourceScriptToken.cpp b/llvm/tools/llvm-rc/ResourceScriptToken.cpp
@@ -85,7 +85,9 @@ namespace {
 
 class Tokenizer {
 public:
-  Tokenizer(StringRef Input) : Data(Input), DataLength(Input.size()), Pos(0) {}
+  Tokenizer(StringRef Input, StringSaver &Saver, bool IsWindres)
+      : Data(Input), DataLength(Input.size()), Pos(0), Saver(Saver),
+        IsWindres(IsWindres) {}
 
   Expected<std::vector<RCToken>> run();
 
@@ -141,8 +143,14 @@ class Tokenizer {
   // an identifier describing a block start or end.
   void processIdentifier(RCToken &token) const;
 
+  Expected<RCToken> mergeStringTokens(const RCToken &FirstToken,
+                                      const RCToken &SecondToken);
+
   StringRef Data;
   size_t DataLength, Pos;
+
+  StringSaver &Saver;
+  bool IsWindres;
 };
 
 void Tokenizer::skipCurrentLine() {
@@ -188,6 +196,16 @@ Expected<std::vector<RCToken>> Tokenizer::run() {
         return getStringError("Integer invalid or too large: " +
                               Token.value().str());
       }
+    } else if (TokenKind == Kind::String && !Result.empty() && IsWindres) {
+      RCToken &Prev = Result.back();
+      if (Prev.kind() == Kind::String) {
+        Expected<RCToken> MergedToken = mergeStringTokens(Prev, Token);
+        if (!MergedToken)
+          return MergedToken.takeError();
+        // Remove the old token and replace the new one with the merged token.
+        Result.pop_back();
+        Token = *MergedToken;
+      }
     }
 
     Result.push_back(Token);
@@ -356,12 +374,43 @@ void Tokenizer::processIdentifier(RCToken &Token) const {
     Token = RCToken(Kind::BlockEnd, Name);
 }
 
+Expected<RCToken> Tokenizer::mergeStringTokens(const RCToken &FirstToken,
+                                               const RCToken &SecondToken) {
+  assert(FirstToken.kind() == Kind::String &&
+         SecondToken.kind() == Kind::String);
+  StringRef First = FirstToken.value();
+  StringRef Second = SecondToken.value();
+  bool IsFirstLong = First.starts_with_insensitive("L");
+  bool IsSecondLong = Second.starts_with_insensitive("L");
+  //  "aaa"  "bbb" =  "aaabbb"
+  // L"aaa" L"bbb" = L"aaabbb"
+  // L"aaa"  "bbb" = L"aaabbb"
+  //  "aaa" L"bbb" = <error>
+  if (IsSecondLong && !IsFirstLong)
+    return getStringError("Cannot concatenate a wide string " + Twine(Second) +
+                          " after a narrow one " + Twine(First));
+  if (IsFirstLong)
+    First = First.drop_front();
+  if (IsSecondLong)
+    Second = Second.drop_front();
+  bool FirstUnquoted = First.consume_front("\"") && First.consume_back("\"");
+  bool SecondUnquoted = Second.consume_front("\"") && Second.consume_back("\"");
+  assert(FirstUnquoted && SecondUnquoted);
+  (void)FirstUnquoted;
+  (void)SecondUnquoted;
+
+  StringRef MergedString =
+      Saver.save(Twine(IsFirstLong ? "L" : "") + "\"" + First + Second + "\"");
+  return RCToken(Kind::String, MergedString);
+}
+
 } // anonymous namespace
 
 namespace llvm {
 
-Expected<std::vector<RCToken>> tokenizeRC(StringRef Input) {
-  return Tokenizer(Input).run();
+Expected<std::vector<RCToken>> tokenizeRC(StringRef Input, StringSaver &Saver,
+                                          bool IsWindres) {
+  return Tokenizer(Input, Saver, IsWindres).run();
 }
 
 } // namespace llvm
diff --git a/llvm/tools/llvm-rc/ResourceScriptToken.h b/llvm/tools/llvm-rc/ResourceScriptToken.h
@@ -26,6 +26,7 @@
 
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Support/Error.h"
+#include "llvm/Support/StringSaver.h"
 
 #include <cstdint>
 #include <map>
@@ -74,7 +75,8 @@ class RCToken {
 // Tokens returned by this function hold only references to the parts
 // of the Input. Memory buffer containing Input cannot be freed,
 // modified or reallocated.
-Expected<std::vector<RCToken>> tokenizeRC(StringRef Input);
+Expected<std::vector<RCToken>> tokenizeRC(StringRef Input, StringSaver &Saver,
+                                          bool IsWindres);
 
 } // namespace llvm
 

diff --git a/llvm/tools/llvm-rc/llvm-rc.cpp b/llvm/tools/llvm-rc/llvm-rc.cpp
@@ -635,7 +635,11 @@ void doRc(std::string Src, std::string Dest, RcOptions &Opts,
   StringRef Contents = FileContents->getBuffer();
 
   std::string FilteredContents = filterCppOutput(Contents);
-  std::vector<RCToken> Tokens = ExitOnErr(tokenizeRC(FilteredContents));
+
+  BumpPtrAllocator Alloc;
+  StringSaver Saver(Alloc);
+  std::vector<RCToken> Tokens =
+      ExitOnErr(tokenizeRC(FilteredContents, Saver, Opts.IsWindres));
 
   if (Opts.BeVerbose) {
     const Twine TokenNames[] = {