Add vfs::FileSystem::getRealPath

[ioeric]

Summary

And change FileManager::getCanonicalName to use getRealPath.

[ioeric]

• Add vfs::FileSystem::getRealPath; change getCanonicalName to use getRealPath

[bkramer]

Looks good. Please watch the windows buildbots carefully after landing this.

lib/Basic/FileManager.cpp

537

PATH_MAX is not a standard thing and probably not there on windows. I'd just hardcode it to 256.

[ioeric]

• s/PATH_MAX/4096/

lib/Basic/FileManager.cpp

537

PATH_MAX on unit is 4096. Let's go with 4096.

include/clang/Basic/VirtualFileSystem.h

public:
 
   /// Set the working directory. This will affect all following operations on
   /// this file system and may propagate down for nested file systems.
   virtual std::error_code setCurrentWorkingDirectory(const Twine &Path) = 0;
 
   /// Get the working directory of this file system.
   virtual llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const = 0;
 
+  /// Gets real path of \p Path e.g. collapse all . and .. patterns, resolve
+  /// symlinks. For real file system, this uses `llvm::sys::fs::real_path`.
+  /// This returns errc::operation_not_permitted if not implemented by subclass.
+  virtual std::error_code getRealPath(const Twine &Path,
+                                      SmallVectorImpl<char> &Output) const;
+
   /// Check whether a file exists. Provided for convenience.
   bool exists(const Twine &Path);
 
   /// Make \a Path an absolute path.
   ///
   /// Makes \a Path absolute using the current directory if it is not already.
   /// An empty \a Path will result in the current directory.
   ///

lib/Basic/FileManager.cpp

StringRef FileManager::getCanonicalName(const DirectoryEntry *Dir) {
   // FIXME: use llvm::sys::fs::canonical() when it gets implemented
   llvm::DenseMap<const DirectoryEntry *, llvm::StringRef>::iterator Known
     = CanonicalDirNames.find(Dir);
   if (Known != CanonicalDirNames.end())
     return Known->second;
 
   StringRef CanonicalName(Dir->getName());
 
-#ifdef LLVM_ON_UNIX
-  char CanonicalNameBuf[PATH_MAX];
-  if (realpath(Dir->getName().str().c_str(), CanonicalNameBuf))
+  SmallString<PATH_MAX> CanonicalNameBuf;
+  if (!FS->getRealPath(Dir->getName(), CanonicalNameBuf))
     CanonicalName = StringRef(CanonicalNameBuf).copy(CanonicalNameStorage);
-#else
-  SmallString<256> CanonicalNameBuf(CanonicalName);
-  llvm::sys::fs::make_absolute(CanonicalNameBuf);
-  llvm::sys::path::native(CanonicalNameBuf);
-  // We've run into needing to remove '..' here in the wild though, so
-  // remove it.
-  // On Windows, symlinks are significantly less prevalent, so removing
-  // '..' is pretty safe.
-  // Ideally we'd have an equivalent of `realpath` and could implement
-  // sys::fs::canonical across all the platforms.
-  llvm::sys::path::remove_dots(CanonicalNameBuf, /* remove_dot_dot */ true);
-  CanonicalName = StringRef(CanonicalNameBuf).copy(CanonicalNameStorage);
-#endif
 
   CanonicalDirNames.insert(std::make_pair(Dir, CanonicalName));
   return CanonicalName;
 }
 
 void FileManager::PrintStats() const {
   llvm::errs() << "\n*** File Manager Stats:\n";
   llvm::errs() << UniqueRealFiles.size() << " real files found, "

lib/Basic/VirtualFileSystem.cpp

std::error_code FileSystem::makeAbsolute(SmallVectorImpl<char> &Path) const {
 
   auto WorkingDir = getCurrentWorkingDirectory();
   if (!WorkingDir)
     return WorkingDir.getError();
 
   return llvm::sys::fs::make_absolute(WorkingDir.get(), Path);
 }
 
+std::error_code FileSystem::getRealPath(const Twine &Path,
+                                        SmallVectorImpl<char> &Output) const {
+  return errc::operation_not_permitted;
+}
+
 bool FileSystem::exists(const Twine &Path) {
   auto Status = status(Path);
   return Status && Status->exists();
 }
 
 #ifndef NDEBUG
 static bool isTraversalComponent(StringRef Component) {
   return Component.equals("..") || Component.equals(".");

 class RealFileSystem : public FileSystem {
 public:
   ErrorOr<Status> status(const Twine &Path) override;
   ErrorOr<std::unique_ptr<File>> openFileForRead(const Twine &Path) override;
   directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
 
   llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override;
   std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
+  std::error_code getRealPath(const Twine &Path,
+                              SmallVectorImpl<char> &Output) const override;
 };
 
 } // namespace
 
 ErrorOr<Status> RealFileSystem::status(const Twine &Path) {
   sys::fs::file_status RealStatus;
   if (std::error_code EC = sys::fs::status(Path, RealStatus))
     return EC;
std::error_code RealFileSystem::setCurrentWorkingDirectory(const Twine &Path) {
   // guaranteeing that all subsequent relative path operations work
   // on the same path the original chdir resulted in. This makes a
   // difference for example on network filesystems, where symlinks might be
   // switched during runtime of the tool. Fixing this depends on having a
   // file system abstraction that allows openat() style interactions.
   return llvm::sys::fs::set_current_path(Path);
 }
 
+std::error_code
+RealFileSystem::getRealPath(const Twine &Path,
+                            SmallVectorImpl<char> &Output) const {
+  return llvm::sys::fs::real_path(Path, Output);
+}
+
 IntrusiveRefCntPtr<FileSystem> vfs::getRealFileSystem() {
   static IntrusiveRefCntPtr<FileSystem> FS = new RealFileSystem();
   return FS;
 }
 
 namespace {
 
 class RealFSDirIter : public clang::vfs::detail::DirIterImpl {