Fix PR5633 by making the preprocessor handle the case where we can

stat a file but where mmaping it fails.  In this case, we emit an
error like:
t.c:1:10: fatal error: error opening file '../../foo.h'

instead of "cannot find file".



git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@90110 91177308-0d34-0410-b5e6-96231b3b80d8

13 files changed, 49 insertions, 26 deletions

diff --git a/include/clang/Basic/DiagnosticLexKinds.td b/include/clang/Basic/DiagnosticLexKinds.td
index 7f3f4ea1fc..39123d9b37 100644
--- a/include/clang/Basic/DiagnosticLexKinds.td
+++ b/include/clang/Basic/DiagnosticLexKinds.td
@@ -170,6 +170,8 @@ def ext_pp_counter : Extension<
 def err_pp_invalid_directive : Error<"invalid preprocessing directive">;
 def err_pp_hash_error : Error<"#error%0">;
 def err_pp_file_not_found : Error<"'%0' file not found">, DefaultFatal;
+def err_pp_error_opening_file : Error<
+  "error opening file '%0'">, DefaultFatal;
 def err_pp_empty_filename : Error<"empty filename">;
 def err_pp_include_too_deep : Error<"#include nested too deeply">;
 def err_pp_expects_filename : Error<"expected \"FILENAME\" or <FILENAME>">;
diff --git a/include/clang/Lex/Lexer.h b/include/clang/Lex/Lexer.h
index c2db4d357c..52bf194883 100644
--- a/include/clang/Lex/Lexer.h
+++ b/include/clang/Lex/Lexer.h
@@ -78,7 +78,7 @@ public:
   /// with the specified preprocessor managing the lexing process.  This lexer
   /// assumes that the associated file buffer and Preprocessor objects will
   /// outlive it, so it doesn't take ownership of either of them.
-  Lexer(FileID FID, Preprocessor &PP);
+  Lexer(FileID FID, const llvm::MemoryBuffer *InputBuffer, Preprocessor &PP);
 
   /// Lexer constructor - Create a new raw lexer object.  This object is only
   /// suitable for calls to 'LexRawToken'.  This lexer assumes that the text
@@ -89,7 +89,8 @@ public:
   /// Lexer constructor - Create a new raw lexer object.  This object is only
   /// suitable for calls to 'LexRawToken'.  This lexer assumes that the text
   /// range will outlive it, so it doesn't take ownership of it.
-  Lexer(FileID FID, const SourceManager &SM, const LangOptions &Features);
+  Lexer(FileID FID, const llvm::MemoryBuffer *InputBuffer,
+        const SourceManager &SM, const LangOptions &Features);
 
   /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
   /// _Pragma expansion.  This has a variety of magic semantics that this method
diff --git a/include/clang/Lex/Preprocessor.h b/include/clang/Lex/Preprocessor.h
index 1c0036e3ed..edd34b7189 100644
--- a/include/clang/Lex/Preprocessor.h
+++ b/include/clang/Lex/Preprocessor.h
@@ -329,9 +329,9 @@ public:
   void EnterMainSourceFile();
 
   /// EnterSourceFile - Add a source file to the top of the include stack and
-  /// start lexing tokens from it instead of the current buffer.  If isMainFile
-  /// is true, this is the main file for the translation unit.
-  void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir);
+  /// start lexing tokens from it instead of the current buffer.  Return true
+  /// on failure.
+  bool EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir);
 
   /// EnterMacro - Add a Macro to the top of the include stack and start lexing
   /// tokens from it instead of the current buffer.  Args specifies the
diff --git a/lib/Frontend/CacheTokens.cpp b/lib/Frontend/CacheTokens.cpp
index 98126c89eb..7296246df2 100644
--- a/lib/Frontend/CacheTokens.cpp
+++ b/lib/Frontend/CacheTokens.cpp
@@ -482,7 +482,8 @@ void PTHWriter::GeneratePTH(const std::string *MainFile) {
     if (!B) continue;
 
     FileID FID = SM.createFileID(FE, SourceLocation(), SrcMgr::C_User);
-    Lexer L(FID, SM, LOpts);
+    const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
+    Lexer L(FID, FromFile, SM, LOpts);
     PM.insert(FE, LexTokens(L));
   }
 
diff --git a/lib/Frontend/DiagChecker.cpp b/lib/Frontend/DiagChecker.cpp
index 26bb6ccfa7..e7a66b1729 100644
--- a/lib/Frontend/DiagChecker.cpp
+++ b/lib/Frontend/DiagChecker.cpp
@@ -149,7 +149,8 @@ static void FindExpectedDiags(Preprocessor &PP,
   FileID FID = PP.getSourceManager().getMainFileID();
 
   // Create a lexer to lex all the tokens of the main file in raw mode.
-  Lexer RawLex(FID, PP.getSourceManager(), PP.getLangOptions());
+  const llvm::MemoryBuffer *FromFile = PP.getSourceManager().getBuffer(FID);
+  Lexer RawLex(FID, FromFile, PP.getSourceManager(), PP.getLangOptions());
 
   // Return comments as tokens, this is how we find expected diagnostics.
   RawLex.SetCommentRetentionState(true);
diff --git a/lib/Frontend/FrontendActions.cpp b/lib/Frontend/FrontendActions.cpp
index 6cb3928d51..8092b715e3 100644
--- a/lib/Frontend/FrontendActions.cpp
+++ b/lib/Frontend/FrontendActions.cpp
@@ -192,7 +192,8 @@ void DumpRawTokensAction::ExecuteAction() {
   SourceManager &SM = PP.getSourceManager();
 
   // Start lexing the specified input file.
-  Lexer RawLex(SM.getMainFileID(), SM, PP.getLangOptions());
+  const llvm::MemoryBuffer *FromFile = SM.getBuffer(SM.getMainFileID());
+  Lexer RawLex(SM.getMainFileID(), FromFile, SM, PP.getLangOptions());
   RawLex.SetKeepWhitespaceMode(true);
 
   Token RawTok;
diff --git a/lib/Frontend/RewriteMacros.cpp b/lib/Frontend/RewriteMacros.cpp
index b5d59c0aa4..0bcbd4ff0e 100644
--- a/lib/Frontend/RewriteMacros.cpp
+++ b/lib/Frontend/RewriteMacros.cpp
@@ -65,7 +65,8 @@ static void LexRawTokensFromMainFile(Preprocessor &PP,
 
   // Create a lexer to lex all the tokens of the main file in raw mode.  Even
   // though it is in raw mode, it will not return comments.
-  Lexer RawLex(SM.getMainFileID(), SM, PP.getLangOptions());
+  const llvm::MemoryBuffer *FromFile = SM.getBuffer(SM.getMainFileID());
+  Lexer RawLex(SM.getMainFileID(), FromFile, SM, PP.getLangOptions());
 
   // Switch on comment lexing because we really do want them.
   RawLex.SetCommentRetentionState(true);
diff --git a/lib/Frontend/VerifyDiagnosticsClient.cpp b/lib/Frontend/VerifyDiagnosticsClient.cpp
index 2891aec504..99ec910be0 100644
--- a/lib/Frontend/VerifyDiagnosticsClient.cpp
+++ b/lib/Frontend/VerifyDiagnosticsClient.cpp
@@ -164,12 +164,14 @@ static void FindExpectedDiags(Preprocessor &PP,
                               DiagList &ExpectedNotes) {
   // Create a raw lexer to pull all the comments out of the main file.  We don't
   // want to look in #include'd headers for expected-error strings.
-  FileID FID = PP.getSourceManager().getMainFileID();
-  if (PP.getSourceManager().getMainFileID().isInvalid())
+  SourceManager &SM = PP.getSourceManager();
+  FileID FID = SM.getMainFileID();
+  if (SM.getMainFileID().isInvalid())
     return;
 
   // Create a lexer to lex all the tokens of the main file in raw mode.
-  Lexer RawLex(FID, PP.getSourceManager(), PP.getLangOptions());
+  const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
+  Lexer RawLex(FID, FromFile, SM, PP.getLangOptions());
 
   // Return comments as tokens, this is how we find expected diagnostics.
   RawLex.SetCommentRetentionState(true);
diff --git a/lib/Lex/Lexer.cpp b/lib/Lex/Lexer.cpp
index f4a4432410..52a7a04567 100644
--- a/lib/Lex/Lexer.cpp
+++ b/lib/Lex/Lexer.cpp
@@ -95,13 +95,11 @@ void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
 /// with the specified preprocessor managing the lexing process.  This lexer
 /// assumes that the associated file buffer and Preprocessor objects will
 /// outlive it, so it doesn't take ownership of either of them.
-Lexer::Lexer(FileID FID, Preprocessor &PP)
+Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
   : PreprocessorLexer(&PP, FID),
     FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
     Features(PP.getLangOptions()) {
 
-  const llvm::MemoryBuffer *InputFile = PP.getSourceManager().getBuffer(FID);
-  
   InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
             InputFile->getBufferEnd());
 
@@ -129,9 +127,9 @@ Lexer::Lexer(SourceLocation fileloc, const LangOptions &features,
 /// Lexer constructor - Create a new raw lexer object.  This object is only
 /// suitable for calls to 'LexRawToken'.  This lexer assumes that the text
 /// range will outlive it, so it doesn't take ownership of it.
-Lexer::Lexer(FileID FID, const SourceManager &SM, const LangOptions &features)
+Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
+             const SourceManager &SM, const LangOptions &features)
   : FileLoc(SM.getLocForStartOfFile(FID)), Features(features) {
-  const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
 
   InitLexer(FromFile->getBufferStart(), FromFile->getBufferStart(),
             FromFile->getBufferEnd());
@@ -163,7 +161,8 @@ Lexer *Lexer::Create_PragmaLexer(SourceLocation SpellingLoc,
 
   // Create the lexer as if we were going to lex the file normally.
   FileID SpellingFID = SM.getFileID(SpellingLoc);
-  Lexer *L = new Lexer(SpellingFID, PP);
+  const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
+  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
 
   // Now that the lexer is created, change the start/end locations so that we
   // just lex the subsection of the file that we want.  This is lexing from a
diff --git a/lib/Lex/PPDirectives.cpp b/lib/Lex/PPDirectives.cpp
index dc7d95e701..9caca339be 100644
--- a/lib/Lex/PPDirectives.cpp
+++ b/lib/Lex/PPDirectives.cpp
@@ -16,6 +16,7 @@
 #include "clang/Lex/HeaderSearch.h"
 #include "clang/Lex/MacroInfo.h"
 #include "clang/Lex/LexDiagnostic.h"
+#include "clang/Basic/FileManager.h"
 #include "clang/Basic/SourceManager.h"
 #include "llvm/ADT/APInt.h"
 using namespace clang;
@@ -1111,7 +1112,9 @@ void Preprocessor::HandleIncludeDirective(Token &IncludeTok,
   }
 
   // Finally, if all is good, enter the new file!
-  EnterSourceFile(FID, CurDir);
+  if (EnterSourceFile(FID, CurDir))
+    Diag(FilenameTok, diag::err_pp_error_opening_file)
+      << std::string(SourceMgr.getFileEntryForID(FID)->getName());
 }
 
 /// HandleIncludeNextDirective - Implements #include_next.
diff --git a/lib/Lex/PPLexerChange.cpp b/lib/Lex/PPLexerChange.cpp
index d5d6a6e972..8a61d7b9c2 100644
--- a/lib/Lex/PPLexerChange.cpp
+++ b/lib/Lex/PPLexerChange.cpp
@@ -64,7 +64,7 @@ PreprocessorLexer *Preprocessor::getCurrentFileLexer() const {
 
 /// EnterSourceFile - Add a source file to the top of the include stack and
 /// start lexing tokens from it instead of the current buffer.
-void Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir) {
+bool Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir) {
   assert(CurTokenLexer == 0 && "Cannot #include a file inside a macro!");
   ++NumEnteredSourceFiles;
 
@@ -72,10 +72,19 @@ void Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir) {
     MaxIncludeStackDepth = IncludeMacroStack.size();
 
   if (PTH) {
-    if (PTHLexer *PL = PTH->CreateLexer(FID))
-      return EnterSourceFileWithPTH(PL, CurDir);
+    if (PTHLexer *PL = PTH->CreateLexer(FID)) {
+      EnterSourceFileWithPTH(PL, CurDir);
+      return false;
+    }
   }
-  EnterSourceFileWithLexer(new Lexer(FID, *this), CurDir);
+  
+  // Get the MemoryBuffer for this FID, if it fails, we fail.
+  const llvm::MemoryBuffer *InputFile = getSourceManager().getBuffer(FID);
+  if (InputFile == 0)
+    return true;
+  
+  EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir);
+  return false;
 }
 
 /// EnterSourceFileWithLexer - Add a source file to the top of the include stack
diff --git a/lib/Rewrite/HTMLRewrite.cpp b/lib/Rewrite/HTMLRewrite.cpp
index b4bf419bc5..342b0e6ef5 100644
--- a/lib/Rewrite/HTMLRewrite.cpp
+++ b/lib/Rewrite/HTMLRewrite.cpp
@@ -353,7 +353,8 @@ void html::SyntaxHighlight(Rewriter &R, FileID FID, const Preprocessor &PP) {
   RewriteBuffer &RB = R.getEditBuffer(FID);
 
   const SourceManager &SM = PP.getSourceManager();
-  Lexer L(FID, SM, PP.getLangOptions());
+  const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
+  Lexer L(FID, FromFile, SM, PP.getLangOptions());
   const char *BufferStart = L.getBufferStart();
 
   // Inform the preprocessor that we want to retain comments as tokens, so we
@@ -444,7 +445,8 @@ void html::HighlightMacros(Rewriter &R, FileID FID, const Preprocessor& PP) {
   const SourceManager &SM = PP.getSourceManager();
   std::vector<Token> TokenStream;
 
-  Lexer L(FID, SM, PP.getLangOptions());
+  const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
+  Lexer L(FID, FromFile, SM, PP.getLangOptions());
 
   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
   // macros.
diff --git a/lib/Rewrite/TokenRewriter.cpp b/lib/Rewrite/TokenRewriter.cpp
index 0effbb18b8..789d53f4af 100644
--- a/lib/Rewrite/TokenRewriter.cpp
+++ b/lib/Rewrite/TokenRewriter.cpp
@@ -23,7 +23,8 @@ TokenRewriter::TokenRewriter(FileID FID, SourceManager &SM,
   ScratchBuf.reset(new ScratchBuffer(SM));
 
   // Create a lexer to lex all the tokens of the main file in raw mode.
-  Lexer RawLex(FID, SM, LangOpts);
+  const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
+  Lexer RawLex(FID, FromFile, SM, LangOpts);
 
   // Return all comments and whitespace as tokens.
   RawLex.SetKeepWhitespaceMode(true);