1343171Sdim//===- VirtualFileSystem.h - Virtual File System Layer ----------*- C++ -*-===//
2343171Sdim//
3353358Sdim// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4353358Sdim// See https://llvm.org/LICENSE.txt for license information.
5353358Sdim// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6343171Sdim//
7343171Sdim//===----------------------------------------------------------------------===//
8343171Sdim//
9343171Sdim/// \file
10343171Sdim/// Defines the virtual file system interface vfs::FileSystem.
11343171Sdim//
12343171Sdim//===----------------------------------------------------------------------===//
13343171Sdim
14343171Sdim#ifndef LLVM_SUPPORT_VIRTUALFILESYSTEM_H
15343171Sdim#define LLVM_SUPPORT_VIRTUALFILESYSTEM_H
16343171Sdim
17343171Sdim#include "llvm/ADT/IntrusiveRefCntPtr.h"
18343171Sdim#include "llvm/ADT/None.h"
19343171Sdim#include "llvm/ADT/Optional.h"
20343171Sdim#include "llvm/ADT/SmallVector.h"
21343171Sdim#include "llvm/ADT/StringRef.h"
22343171Sdim#include "llvm/ADT/Twine.h"
23343171Sdim#include "llvm/Support/Chrono.h"
24343171Sdim#include "llvm/Support/ErrorOr.h"
25343171Sdim#include "llvm/Support/FileSystem.h"
26343171Sdim#include "llvm/Support/Path.h"
27343171Sdim#include "llvm/Support/SourceMgr.h"
28343171Sdim#include <cassert>
29343171Sdim#include <cstdint>
30343171Sdim#include <ctime>
31343171Sdim#include <memory>
32343171Sdim#include <stack>
33343171Sdim#include <string>
34343171Sdim#include <system_error>
35343171Sdim#include <utility>
36343171Sdim#include <vector>
37343171Sdim
38343171Sdimnamespace llvm {
39343171Sdim
40343171Sdimclass MemoryBuffer;
41343171Sdim
42343171Sdimnamespace vfs {
43343171Sdim
44343171Sdim/// The result of a \p status operation.
45343171Sdimclass Status {
46343171Sdim  std::string Name;
47343171Sdim  llvm::sys::fs::UniqueID UID;
48343171Sdim  llvm::sys::TimePoint<> MTime;
49343171Sdim  uint32_t User;
50343171Sdim  uint32_t Group;
51343171Sdim  uint64_t Size;
52343171Sdim  llvm::sys::fs::file_type Type = llvm::sys::fs::file_type::status_error;
53343171Sdim  llvm::sys::fs::perms Perms;
54343171Sdim
55343171Sdimpublic:
56343171Sdim  // FIXME: remove when files support multiple names
57343171Sdim  bool IsVFSMapped = false;
58343171Sdim
59343171Sdim  Status() = default;
60343171Sdim  Status(const llvm::sys::fs::file_status &Status);
61353358Sdim  Status(const Twine &Name, llvm::sys::fs::UniqueID UID,
62343171Sdim         llvm::sys::TimePoint<> MTime, uint32_t User, uint32_t Group,
63343171Sdim         uint64_t Size, llvm::sys::fs::file_type Type,
64343171Sdim         llvm::sys::fs::perms Perms);
65343171Sdim
66343171Sdim  /// Get a copy of a Status with a different name.
67353358Sdim  static Status copyWithNewName(const Status &In, const Twine &NewName);
68343171Sdim  static Status copyWithNewName(const llvm::sys::fs::file_status &In,
69353358Sdim                                const Twine &NewName);
70343171Sdim
71343171Sdim  /// Returns the name that should be used for this file or directory.
72343171Sdim  StringRef getName() const { return Name; }
73343171Sdim
74343171Sdim  /// @name Status interface from llvm::sys::fs
75343171Sdim  /// @{
76343171Sdim  llvm::sys::fs::file_type getType() const { return Type; }
77343171Sdim  llvm::sys::fs::perms getPermissions() const { return Perms; }
78343171Sdim  llvm::sys::TimePoint<> getLastModificationTime() const { return MTime; }
79343171Sdim  llvm::sys::fs::UniqueID getUniqueID() const { return UID; }
80343171Sdim  uint32_t getUser() const { return User; }
81343171Sdim  uint32_t getGroup() const { return Group; }
82343171Sdim  uint64_t getSize() const { return Size; }
83343171Sdim  /// @}
84343171Sdim  /// @name Status queries
85343171Sdim  /// These are static queries in llvm::sys::fs.
86343171Sdim  /// @{
87343171Sdim  bool equivalent(const Status &Other) const;
88343171Sdim  bool isDirectory() const;
89343171Sdim  bool isRegularFile() const;
90343171Sdim  bool isOther() const;
91343171Sdim  bool isSymlink() const;
92343171Sdim  bool isStatusKnown() const;
93343171Sdim  bool exists() const;
94343171Sdim  /// @}
95343171Sdim};
96343171Sdim
97343171Sdim/// Represents an open file.
98343171Sdimclass File {
99343171Sdimpublic:
100343171Sdim  /// Destroy the file after closing it (if open).
101343171Sdim  /// Sub-classes should generally call close() inside their destructors.  We
102343171Sdim  /// cannot do that from the base class, since close is virtual.
103343171Sdim  virtual ~File();
104343171Sdim
105343171Sdim  /// Get the status of the file.
106343171Sdim  virtual llvm::ErrorOr<Status> status() = 0;
107343171Sdim
108343171Sdim  /// Get the name of the file
109343171Sdim  virtual llvm::ErrorOr<std::string> getName() {
110343171Sdim    if (auto Status = status())
111343171Sdim      return Status->getName().str();
112343171Sdim    else
113343171Sdim      return Status.getError();
114343171Sdim  }
115343171Sdim
116343171Sdim  /// Get the contents of the file as a \p MemoryBuffer.
117343171Sdim  virtual llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>>
118343171Sdim  getBuffer(const Twine &Name, int64_t FileSize = -1,
119343171Sdim            bool RequiresNullTerminator = true, bool IsVolatile = false) = 0;
120343171Sdim
121343171Sdim  /// Closes the file.
122343171Sdim  virtual std::error_code close() = 0;
123343171Sdim};
124343171Sdim
125343171Sdim/// A member of a directory, yielded by a directory_iterator.
126343171Sdim/// Only information available on most platforms is included.
127343171Sdimclass directory_entry {
128343171Sdim  std::string Path;
129360784Sdim  llvm::sys::fs::file_type Type = llvm::sys::fs::file_type::type_unknown;
130343171Sdim
131343171Sdimpublic:
132343171Sdim  directory_entry() = default;
133343171Sdim  directory_entry(std::string Path, llvm::sys::fs::file_type Type)
134343171Sdim      : Path(std::move(Path)), Type(Type) {}
135343171Sdim
136343171Sdim  llvm::StringRef path() const { return Path; }
137343171Sdim  llvm::sys::fs::file_type type() const { return Type; }
138343171Sdim};
139343171Sdim
140343171Sdimnamespace detail {
141343171Sdim
142343171Sdim/// An interface for virtual file systems to provide an iterator over the
143343171Sdim/// (non-recursive) contents of a directory.
144343171Sdimstruct DirIterImpl {
145343171Sdim  virtual ~DirIterImpl();
146343171Sdim
147343171Sdim  /// Sets \c CurrentEntry to the next entry in the directory on success,
148343171Sdim  /// to directory_entry() at end,  or returns a system-defined \c error_code.
149343171Sdim  virtual std::error_code increment() = 0;
150343171Sdim
151343171Sdim  directory_entry CurrentEntry;
152343171Sdim};
153343171Sdim
154343171Sdim} // namespace detail
155343171Sdim
156343171Sdim/// An input iterator over the entries in a virtual path, similar to
157343171Sdim/// llvm::sys::fs::directory_iterator.
158343171Sdimclass directory_iterator {
159343171Sdim  std::shared_ptr<detail::DirIterImpl> Impl; // Input iterator semantics on copy
160343171Sdim
161343171Sdimpublic:
162343171Sdim  directory_iterator(std::shared_ptr<detail::DirIterImpl> I)
163343171Sdim      : Impl(std::move(I)) {
164343171Sdim    assert(Impl.get() != nullptr && "requires non-null implementation");
165343171Sdim    if (Impl->CurrentEntry.path().empty())
166343171Sdim      Impl.reset(); // Normalize the end iterator to Impl == nullptr.
167343171Sdim  }
168343171Sdim
169343171Sdim  /// Construct an 'end' iterator.
170343171Sdim  directory_iterator() = default;
171343171Sdim
172343171Sdim  /// Equivalent to operator++, with an error code.
173343171Sdim  directory_iterator &increment(std::error_code &EC) {
174343171Sdim    assert(Impl && "attempting to increment past end");
175343171Sdim    EC = Impl->increment();
176343171Sdim    if (Impl->CurrentEntry.path().empty())
177343171Sdim      Impl.reset(); // Normalize the end iterator to Impl == nullptr.
178343171Sdim    return *this;
179343171Sdim  }
180343171Sdim
181343171Sdim  const directory_entry &operator*() const { return Impl->CurrentEntry; }
182343171Sdim  const directory_entry *operator->() const { return &Impl->CurrentEntry; }
183343171Sdim
184343171Sdim  bool operator==(const directory_iterator &RHS) const {
185343171Sdim    if (Impl && RHS.Impl)
186343171Sdim      return Impl->CurrentEntry.path() == RHS.Impl->CurrentEntry.path();
187343171Sdim    return !Impl && !RHS.Impl;
188343171Sdim  }
189343171Sdim  bool operator!=(const directory_iterator &RHS) const {
190343171Sdim    return !(*this == RHS);
191343171Sdim  }
192343171Sdim};
193343171Sdim
194343171Sdimclass FileSystem;
195343171Sdim
196343171Sdimnamespace detail {
197343171Sdim
198343171Sdim/// Keeps state for the recursive_directory_iterator.
199343171Sdimstruct RecDirIterState {
200343171Sdim  std::stack<directory_iterator, std::vector<directory_iterator>> Stack;
201343171Sdim  bool HasNoPushRequest = false;
202343171Sdim};
203343171Sdim
204343171Sdim} // end namespace detail
205343171Sdim
206343171Sdim/// An input iterator over the recursive contents of a virtual path,
207343171Sdim/// similar to llvm::sys::fs::recursive_directory_iterator.
208343171Sdimclass recursive_directory_iterator {
209343171Sdim  FileSystem *FS;
210343171Sdim  std::shared_ptr<detail::RecDirIterState>
211343171Sdim      State; // Input iterator semantics on copy.
212343171Sdim
213343171Sdimpublic:
214343171Sdim  recursive_directory_iterator(FileSystem &FS, const Twine &Path,
215343171Sdim                               std::error_code &EC);
216343171Sdim
217343171Sdim  /// Construct an 'end' iterator.
218343171Sdim  recursive_directory_iterator() = default;
219343171Sdim
220343171Sdim  /// Equivalent to operator++, with an error code.
221343171Sdim  recursive_directory_iterator &increment(std::error_code &EC);
222343171Sdim
223343171Sdim  const directory_entry &operator*() const { return *State->Stack.top(); }
224343171Sdim  const directory_entry *operator->() const { return &*State->Stack.top(); }
225343171Sdim
226343171Sdim  bool operator==(const recursive_directory_iterator &Other) const {
227343171Sdim    return State == Other.State; // identity
228343171Sdim  }
229343171Sdim  bool operator!=(const recursive_directory_iterator &RHS) const {
230343171Sdim    return !(*this == RHS);
231343171Sdim  }
232343171Sdim
233343171Sdim  /// Gets the current level. Starting path is at level 0.
234343171Sdim  int level() const {
235343171Sdim    assert(!State->Stack.empty() &&
236343171Sdim           "Cannot get level without any iteration state");
237343171Sdim    return State->Stack.size() - 1;
238343171Sdim  }
239343171Sdim
240343171Sdim  void no_push() { State->HasNoPushRequest = true; }
241343171Sdim};
242343171Sdim
243343171Sdim/// The virtual file system interface.
244343171Sdimclass FileSystem : public llvm::ThreadSafeRefCountedBase<FileSystem> {
245343171Sdimpublic:
246343171Sdim  virtual ~FileSystem();
247343171Sdim
248343171Sdim  /// Get the status of the entry at \p Path, if one exists.
249343171Sdim  virtual llvm::ErrorOr<Status> status(const Twine &Path) = 0;
250343171Sdim
251343171Sdim  /// Get a \p File object for the file at \p Path, if one exists.
252343171Sdim  virtual llvm::ErrorOr<std::unique_ptr<File>>
253343171Sdim  openFileForRead(const Twine &Path) = 0;
254343171Sdim
255343171Sdim  /// This is a convenience method that opens a file, gets its content and then
256343171Sdim  /// closes the file.
257343171Sdim  llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>>
258343171Sdim  getBufferForFile(const Twine &Name, int64_t FileSize = -1,
259343171Sdim                   bool RequiresNullTerminator = true, bool IsVolatile = false);
260343171Sdim
261343171Sdim  /// Get a directory_iterator for \p Dir.
262343171Sdim  /// \note The 'end' iterator is directory_iterator().
263343171Sdim  virtual directory_iterator dir_begin(const Twine &Dir,
264343171Sdim                                       std::error_code &EC) = 0;
265343171Sdim
266343171Sdim  /// Set the working directory. This will affect all following operations on
267343171Sdim  /// this file system and may propagate down for nested file systems.
268343171Sdim  virtual std::error_code setCurrentWorkingDirectory(const Twine &Path) = 0;
269343171Sdim
270343171Sdim  /// Get the working directory of this file system.
271343171Sdim  virtual llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const = 0;
272343171Sdim
273343171Sdim  /// Gets real path of \p Path e.g. collapse all . and .. patterns, resolve
274343171Sdim  /// symlinks. For real file system, this uses `llvm::sys::fs::real_path`.
275343171Sdim  /// This returns errc::operation_not_permitted if not implemented by subclass.
276343171Sdim  virtual std::error_code getRealPath(const Twine &Path,
277343171Sdim                                      SmallVectorImpl<char> &Output) const;
278343171Sdim
279343171Sdim  /// Check whether a file exists. Provided for convenience.
280343171Sdim  bool exists(const Twine &Path);
281343171Sdim
282343171Sdim  /// Is the file mounted on a local filesystem?
283343171Sdim  virtual std::error_code isLocal(const Twine &Path, bool &Result);
284343171Sdim
285343171Sdim  /// Make \a Path an absolute path.
286343171Sdim  ///
287343171Sdim  /// Makes \a Path absolute using the current directory if it is not already.
288343171Sdim  /// An empty \a Path will result in the current directory.
289343171Sdim  ///
290343171Sdim  /// /absolute/path   => /absolute/path
291343171Sdim  /// relative/../path => <current-directory>/relative/../path
292343171Sdim  ///
293343171Sdim  /// \param Path A path that is modified to be an absolute path.
294343171Sdim  /// \returns success if \a path has been made absolute, otherwise a
295343171Sdim  ///          platform-specific error_code.
296360784Sdim  virtual std::error_code makeAbsolute(SmallVectorImpl<char> &Path) const;
297343171Sdim};
298343171Sdim
299343171Sdim/// Gets an \p vfs::FileSystem for the 'real' file system, as seen by
300343171Sdim/// the operating system.
301353358Sdim/// The working directory is linked to the process's working directory.
302353358Sdim/// (This is usually thread-hostile).
303343171SdimIntrusiveRefCntPtr<FileSystem> getRealFileSystem();
304343171Sdim
305353358Sdim/// Create an \p vfs::FileSystem for the 'real' file system, as seen by
306353358Sdim/// the operating system.
307353358Sdim/// It has its own working directory, independent of (but initially equal to)
308353358Sdim/// that of the process.
309353358Sdimstd::unique_ptr<FileSystem> createPhysicalFileSystem();
310353358Sdim
311343171Sdim/// A file system that allows overlaying one \p AbstractFileSystem on top
312343171Sdim/// of another.
313343171Sdim///
314343171Sdim/// Consists of a stack of >=1 \p FileSystem objects, which are treated as being
315343171Sdim/// one merged file system. When there is a directory that exists in more than
316343171Sdim/// one file system, the \p OverlayFileSystem contains a directory containing
317343171Sdim/// the union of their contents.  The attributes (permissions, etc.) of the
318343171Sdim/// top-most (most recently added) directory are used.  When there is a file
319343171Sdim/// that exists in more than one file system, the file in the top-most file
320343171Sdim/// system overrides the other(s).
321343171Sdimclass OverlayFileSystem : public FileSystem {
322343171Sdim  using FileSystemList = SmallVector<IntrusiveRefCntPtr<FileSystem>, 1>;
323343171Sdim
324343171Sdim  /// The stack of file systems, implemented as a list in order of
325343171Sdim  /// their addition.
326343171Sdim  FileSystemList FSList;
327343171Sdim
328343171Sdimpublic:
329343171Sdim  OverlayFileSystem(IntrusiveRefCntPtr<FileSystem> Base);
330343171Sdim
331343171Sdim  /// Pushes a file system on top of the stack.
332343171Sdim  void pushOverlay(IntrusiveRefCntPtr<FileSystem> FS);
333343171Sdim
334343171Sdim  llvm::ErrorOr<Status> status(const Twine &Path) override;
335343171Sdim  llvm::ErrorOr<std::unique_ptr<File>>
336343171Sdim  openFileForRead(const Twine &Path) override;
337343171Sdim  directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
338343171Sdim  llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override;
339343171Sdim  std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
340343171Sdim  std::error_code isLocal(const Twine &Path, bool &Result) override;
341343171Sdim  std::error_code getRealPath(const Twine &Path,
342343171Sdim                              SmallVectorImpl<char> &Output) const override;
343343171Sdim
344343171Sdim  using iterator = FileSystemList::reverse_iterator;
345343171Sdim  using const_iterator = FileSystemList::const_reverse_iterator;
346353358Sdim  using reverse_iterator = FileSystemList::iterator;
347353358Sdim  using const_reverse_iterator = FileSystemList::const_iterator;
348343171Sdim
349343171Sdim  /// Get an iterator pointing to the most recently added file system.
350343171Sdim  iterator overlays_begin() { return FSList.rbegin(); }
351343171Sdim  const_iterator overlays_begin() const { return FSList.rbegin(); }
352343171Sdim
353353358Sdim  /// Get an iterator pointing one-past the least recently added file system.
354343171Sdim  iterator overlays_end() { return FSList.rend(); }
355343171Sdim  const_iterator overlays_end() const { return FSList.rend(); }
356353358Sdim
357353358Sdim  /// Get an iterator pointing to the least recently added file system.
358353358Sdim  reverse_iterator overlays_rbegin() { return FSList.begin(); }
359353358Sdim  const_reverse_iterator overlays_rbegin() const { return FSList.begin(); }
360353358Sdim
361353358Sdim  /// Get an iterator pointing one-past the most recently added file system.
362353358Sdim  reverse_iterator overlays_rend() { return FSList.end(); }
363353358Sdim  const_reverse_iterator overlays_rend() const { return FSList.end(); }
364343171Sdim};
365343171Sdim
366343171Sdim/// By default, this delegates all calls to the underlying file system. This
367343171Sdim/// is useful when derived file systems want to override some calls and still
368343171Sdim/// proxy other calls.
369343171Sdimclass ProxyFileSystem : public FileSystem {
370343171Sdimpublic:
371343171Sdim  explicit ProxyFileSystem(IntrusiveRefCntPtr<FileSystem> FS)
372343171Sdim      : FS(std::move(FS)) {}
373343171Sdim
374343171Sdim  llvm::ErrorOr<Status> status(const Twine &Path) override {
375343171Sdim    return FS->status(Path);
376343171Sdim  }
377343171Sdim  llvm::ErrorOr<std::unique_ptr<File>>
378343171Sdim  openFileForRead(const Twine &Path) override {
379343171Sdim    return FS->openFileForRead(Path);
380343171Sdim  }
381343171Sdim  directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override {
382343171Sdim    return FS->dir_begin(Dir, EC);
383343171Sdim  }
384343171Sdim  llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override {
385343171Sdim    return FS->getCurrentWorkingDirectory();
386343171Sdim  }
387343171Sdim  std::error_code setCurrentWorkingDirectory(const Twine &Path) override {
388343171Sdim    return FS->setCurrentWorkingDirectory(Path);
389343171Sdim  }
390343171Sdim  std::error_code getRealPath(const Twine &Path,
391343171Sdim                              SmallVectorImpl<char> &Output) const override {
392343171Sdim    return FS->getRealPath(Path, Output);
393343171Sdim  }
394343171Sdim  std::error_code isLocal(const Twine &Path, bool &Result) override {
395343171Sdim    return FS->isLocal(Path, Result);
396343171Sdim  }
397343171Sdim
398343171Sdimprotected:
399343171Sdim  FileSystem &getUnderlyingFS() { return *FS; }
400343171Sdim
401343171Sdimprivate:
402343171Sdim  IntrusiveRefCntPtr<FileSystem> FS;
403343171Sdim
404343171Sdim  virtual void anchor();
405343171Sdim};
406343171Sdim
407343171Sdimnamespace detail {
408343171Sdim
409343171Sdimclass InMemoryDirectory;
410343171Sdimclass InMemoryFile;
411343171Sdim
412343171Sdim} // namespace detail
413343171Sdim
414343171Sdim/// An in-memory file system.
415343171Sdimclass InMemoryFileSystem : public FileSystem {
416343171Sdim  std::unique_ptr<detail::InMemoryDirectory> Root;
417343171Sdim  std::string WorkingDirectory;
418343171Sdim  bool UseNormalizedPaths = true;
419343171Sdim
420343171Sdim  /// If HardLinkTarget is non-null, a hardlink is created to the To path which
421343171Sdim  /// must be a file. If it is null then it adds the file as the public addFile.
422343171Sdim  bool addFile(const Twine &Path, time_t ModificationTime,
423343171Sdim               std::unique_ptr<llvm::MemoryBuffer> Buffer,
424343171Sdim               Optional<uint32_t> User, Optional<uint32_t> Group,
425343171Sdim               Optional<llvm::sys::fs::file_type> Type,
426343171Sdim               Optional<llvm::sys::fs::perms> Perms,
427343171Sdim               const detail::InMemoryFile *HardLinkTarget);
428343171Sdim
429343171Sdimpublic:
430343171Sdim  explicit InMemoryFileSystem(bool UseNormalizedPaths = true);
431343171Sdim  ~InMemoryFileSystem() override;
432343171Sdim
433343171Sdim  /// Add a file containing a buffer or a directory to the VFS with a
434343171Sdim  /// path. The VFS owns the buffer.  If present, User, Group, Type
435343171Sdim  /// and Perms apply to the newly-created file or directory.
436343171Sdim  /// \return true if the file or directory was successfully added,
437343171Sdim  /// false if the file or directory already exists in the file system with
438343171Sdim  /// different contents.
439343171Sdim  bool addFile(const Twine &Path, time_t ModificationTime,
440343171Sdim               std::unique_ptr<llvm::MemoryBuffer> Buffer,
441343171Sdim               Optional<uint32_t> User = None, Optional<uint32_t> Group = None,
442343171Sdim               Optional<llvm::sys::fs::file_type> Type = None,
443343171Sdim               Optional<llvm::sys::fs::perms> Perms = None);
444343171Sdim
445343171Sdim  /// Add a hard link to a file.
446343171Sdim  /// Here hard links are not intended to be fully equivalent to the classical
447343171Sdim  /// filesystem. Both the hard link and the file share the same buffer and
448343171Sdim  /// status (and thus have the same UniqueID). Because of this there is no way
449343171Sdim  /// to distinguish between the link and the file after the link has been
450343171Sdim  /// added.
451343171Sdim  ///
452343171Sdim  /// The To path must be an existing file or a hardlink. The From file must not
453343171Sdim  /// have been added before. The To Path must not be a directory. The From Node
454343171Sdim  /// is added as a hard link which points to the resolved file of To Node.
455343171Sdim  /// \return true if the above condition is satisfied and hardlink was
456343171Sdim  /// successfully created, false otherwise.
457343171Sdim  bool addHardLink(const Twine &From, const Twine &To);
458343171Sdim
459343171Sdim  /// Add a buffer to the VFS with a path. The VFS does not own the buffer.
460343171Sdim  /// If present, User, Group, Type and Perms apply to the newly-created file
461343171Sdim  /// or directory.
462343171Sdim  /// \return true if the file or directory was successfully added,
463343171Sdim  /// false if the file or directory already exists in the file system with
464343171Sdim  /// different contents.
465343171Sdim  bool addFileNoOwn(const Twine &Path, time_t ModificationTime,
466343171Sdim                    llvm::MemoryBuffer *Buffer, Optional<uint32_t> User = None,
467343171Sdim                    Optional<uint32_t> Group = None,
468343171Sdim                    Optional<llvm::sys::fs::file_type> Type = None,
469343171Sdim                    Optional<llvm::sys::fs::perms> Perms = None);
470343171Sdim
471343171Sdim  std::string toString() const;
472343171Sdim
473343171Sdim  /// Return true if this file system normalizes . and .. in paths.
474343171Sdim  bool useNormalizedPaths() const { return UseNormalizedPaths; }
475343171Sdim
476343171Sdim  llvm::ErrorOr<Status> status(const Twine &Path) override;
477343171Sdim  llvm::ErrorOr<std::unique_ptr<File>>
478343171Sdim  openFileForRead(const Twine &Path) override;
479343171Sdim  directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
480343171Sdim
481343171Sdim  llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override {
482343171Sdim    return WorkingDirectory;
483343171Sdim  }
484343171Sdim  /// Canonicalizes \p Path by combining with the current working
485343171Sdim  /// directory and normalizing the path (e.g. remove dots). If the current
486343171Sdim  /// working directory is not set, this returns errc::operation_not_permitted.
487343171Sdim  ///
488343171Sdim  /// This doesn't resolve symlinks as they are not supported in in-memory file
489343171Sdim  /// system.
490343171Sdim  std::error_code getRealPath(const Twine &Path,
491343171Sdim                              SmallVectorImpl<char> &Output) const override;
492343171Sdim  std::error_code isLocal(const Twine &Path, bool &Result) override;
493343171Sdim  std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
494343171Sdim};
495343171Sdim
496343171Sdim/// Get a globally unique ID for a virtual file or directory.
497343171Sdimllvm::sys::fs::UniqueID getNextVirtualUniqueID();
498343171Sdim
499343171Sdim/// Gets a \p FileSystem for a virtual file system described in YAML
500343171Sdim/// format.
501343171SdimIntrusiveRefCntPtr<FileSystem>
502343171SdimgetVFSFromYAML(std::unique_ptr<llvm::MemoryBuffer> Buffer,
503343171Sdim               llvm::SourceMgr::DiagHandlerTy DiagHandler,
504343171Sdim               StringRef YAMLFilePath, void *DiagContext = nullptr,
505343171Sdim               IntrusiveRefCntPtr<FileSystem> ExternalFS = getRealFileSystem());
506343171Sdim
507343171Sdimstruct YAMLVFSEntry {
508343171Sdim  template <typename T1, typename T2>
509343171Sdim  YAMLVFSEntry(T1 &&VPath, T2 &&RPath)
510343171Sdim      : VPath(std::forward<T1>(VPath)), RPath(std::forward<T2>(RPath)) {}
511343171Sdim  std::string VPath;
512343171Sdim  std::string RPath;
513343171Sdim};
514343171Sdim
515343171Sdimclass VFSFromYamlDirIterImpl;
516343171Sdimclass RedirectingFileSystemParser;
517343171Sdim
518343171Sdim/// A virtual file system parsed from a YAML file.
519343171Sdim///
520343171Sdim/// Currently, this class allows creating virtual directories and mapping
521343171Sdim/// virtual file paths to existing external files, available in \c ExternalFS.
522343171Sdim///
523343171Sdim/// The basic structure of the parsed file is:
524343171Sdim/// \verbatim
525343171Sdim/// {
526343171Sdim///   'version': <version number>,
527343171Sdim///   <optional configuration>
528343171Sdim///   'roots': [
529343171Sdim///              <directory entries>
530343171Sdim///            ]
531343171Sdim/// }
532343171Sdim/// \endverbatim
533343171Sdim///
534343171Sdim/// All configuration options are optional.
535360784Sdim///   'case-sensitive': <boolean, default=(true for Posix, false for Windows)>
536343171Sdim///   'use-external-names': <boolean, default=true>
537343171Sdim///   'overlay-relative': <boolean, default=false>
538343171Sdim///   'fallthrough': <boolean, default=true>
539343171Sdim///
540343171Sdim/// Virtual directories are represented as
541343171Sdim/// \verbatim
542343171Sdim/// {
543343171Sdim///   'type': 'directory',
544343171Sdim///   'name': <string>,
545343171Sdim///   'contents': [ <file or directory entries> ]
546343171Sdim/// }
547343171Sdim/// \endverbatim
548343171Sdim///
549343171Sdim/// The default attributes for virtual directories are:
550343171Sdim/// \verbatim
551343171Sdim/// MTime = now() when created
552343171Sdim/// Perms = 0777
553343171Sdim/// User = Group = 0
554343171Sdim/// Size = 0
555343171Sdim/// UniqueID = unspecified unique value
556343171Sdim/// \endverbatim
557343171Sdim///
558343171Sdim/// Re-mapped files are represented as
559343171Sdim/// \verbatim
560343171Sdim/// {
561343171Sdim///   'type': 'file',
562343171Sdim///   'name': <string>,
563343171Sdim///   'use-external-name': <boolean> # Optional
564343171Sdim///   'external-contents': <path to external file>
565343171Sdim/// }
566343171Sdim/// \endverbatim
567343171Sdim///
568343171Sdim/// and inherit their attributes from the external contents.
569343171Sdim///
570343171Sdim/// In both cases, the 'name' field may contain multiple path components (e.g.
571343171Sdim/// /path/to/file). However, any directory that contains more than one child
572343171Sdim/// must be uniquely represented by a directory entry.
573343171Sdimclass RedirectingFileSystem : public vfs::FileSystem {
574343171Sdimpublic:
575343171Sdim  enum EntryKind { EK_Directory, EK_File };
576343171Sdim
577343171Sdim  /// A single file or directory in the VFS.
578343171Sdim  class Entry {
579343171Sdim    EntryKind Kind;
580343171Sdim    std::string Name;
581343171Sdim
582343171Sdim  public:
583343171Sdim    Entry(EntryKind K, StringRef Name) : Kind(K), Name(Name) {}
584343171Sdim    virtual ~Entry() = default;
585343171Sdim
586343171Sdim    StringRef getName() const { return Name; }
587343171Sdim    EntryKind getKind() const { return Kind; }
588343171Sdim  };
589343171Sdim
590343171Sdim  class RedirectingDirectoryEntry : public Entry {
591343171Sdim    std::vector<std::unique_ptr<Entry>> Contents;
592343171Sdim    Status S;
593343171Sdim
594343171Sdim  public:
595343171Sdim    RedirectingDirectoryEntry(StringRef Name,
596343171Sdim                              std::vector<std::unique_ptr<Entry>> Contents,
597343171Sdim                              Status S)
598343171Sdim        : Entry(EK_Directory, Name), Contents(std::move(Contents)),
599343171Sdim          S(std::move(S)) {}
600343171Sdim    RedirectingDirectoryEntry(StringRef Name, Status S)
601343171Sdim        : Entry(EK_Directory, Name), S(std::move(S)) {}
602343171Sdim
603343171Sdim    Status getStatus() { return S; }
604343171Sdim
605343171Sdim    void addContent(std::unique_ptr<Entry> Content) {
606343171Sdim      Contents.push_back(std::move(Content));
607343171Sdim    }
608343171Sdim
609343171Sdim    Entry *getLastContent() const { return Contents.back().get(); }
610343171Sdim
611343171Sdim    using iterator = decltype(Contents)::iterator;
612343171Sdim
613343171Sdim    iterator contents_begin() { return Contents.begin(); }
614343171Sdim    iterator contents_end() { return Contents.end(); }
615343171Sdim
616343171Sdim    static bool classof(const Entry *E) { return E->getKind() == EK_Directory; }
617343171Sdim  };
618343171Sdim
619343171Sdim  class RedirectingFileEntry : public Entry {
620343171Sdim  public:
621343171Sdim    enum NameKind { NK_NotSet, NK_External, NK_Virtual };
622343171Sdim
623343171Sdim  private:
624343171Sdim    std::string ExternalContentsPath;
625343171Sdim    NameKind UseName;
626343171Sdim
627343171Sdim  public:
628343171Sdim    RedirectingFileEntry(StringRef Name, StringRef ExternalContentsPath,
629343171Sdim                         NameKind UseName)
630343171Sdim        : Entry(EK_File, Name), ExternalContentsPath(ExternalContentsPath),
631343171Sdim          UseName(UseName) {}
632343171Sdim
633343171Sdim    StringRef getExternalContentsPath() const { return ExternalContentsPath; }
634343171Sdim
635343171Sdim    /// whether to use the external path as the name for this file.
636343171Sdim    bool useExternalName(bool GlobalUseExternalName) const {
637343171Sdim      return UseName == NK_NotSet ? GlobalUseExternalName
638343171Sdim                                  : (UseName == NK_External);
639343171Sdim    }
640343171Sdim
641343171Sdim    NameKind getUseName() const { return UseName; }
642343171Sdim
643343171Sdim    static bool classof(const Entry *E) { return E->getKind() == EK_File; }
644343171Sdim  };
645343171Sdim
646343171Sdimprivate:
647343171Sdim  friend class VFSFromYamlDirIterImpl;
648343171Sdim  friend class RedirectingFileSystemParser;
649343171Sdim
650360784Sdim  bool shouldUseExternalFS() const {
651360784Sdim    return ExternalFSValidWD && IsFallthrough;
652360784Sdim  }
653360784Sdim
654360784Sdim  // In a RedirectingFileSystem, keys can be specified in Posix or Windows
655360784Sdim  // style (or even a mixture of both), so this comparison helper allows
656360784Sdim  // slashes (representing a root) to match backslashes (and vice versa).  Note
657360784Sdim  // that, other than the root, patch components should not contain slashes or
658360784Sdim  // backslashes.
659360784Sdim  bool pathComponentMatches(llvm::StringRef lhs, llvm::StringRef rhs) const {
660360784Sdim    if ((CaseSensitive ? lhs.equals(rhs) : lhs.equals_lower(rhs)))
661360784Sdim      return true;
662360784Sdim    return (lhs == "/" && rhs == "\\") || (lhs == "\\" && rhs == "/");
663360784Sdim  }
664360784Sdim
665343171Sdim  /// The root(s) of the virtual file system.
666343171Sdim  std::vector<std::unique_ptr<Entry>> Roots;
667343171Sdim
668360784Sdim  /// The current working directory of the file system.
669360784Sdim  std::string WorkingDirectory;
670360784Sdim
671360784Sdim  /// Whether the current working directory is valid for the external FS.
672360784Sdim  bool ExternalFSValidWD = false;
673360784Sdim
674343171Sdim  /// The file system to use for external references.
675343171Sdim  IntrusiveRefCntPtr<FileSystem> ExternalFS;
676343171Sdim
677343171Sdim  /// If IsRelativeOverlay is set, this represents the directory
678343171Sdim  /// path that should be prefixed to each 'external-contents' entry
679343171Sdim  /// when reading from YAML files.
680343171Sdim  std::string ExternalContentsPrefixDir;
681343171Sdim
682343171Sdim  /// @name Configuration
683343171Sdim  /// @{
684343171Sdim
685343171Sdim  /// Whether to perform case-sensitive comparisons.
686343171Sdim  ///
687343171Sdim  /// Currently, case-insensitive matching only works correctly with ASCII.
688360784Sdim  bool CaseSensitive =
689360784Sdim#ifdef _WIN32
690360784Sdim      false;
691360784Sdim#else
692360784Sdim      true;
693360784Sdim#endif
694343171Sdim
695343171Sdim  /// IsRelativeOverlay marks whether a ExternalContentsPrefixDir path must
696343171Sdim  /// be prefixed in every 'external-contents' when reading from YAML files.
697343171Sdim  bool IsRelativeOverlay = false;
698343171Sdim
699343171Sdim  /// Whether to use to use the value of 'external-contents' for the
700343171Sdim  /// names of files.  This global value is overridable on a per-file basis.
701343171Sdim  bool UseExternalNames = true;
702343171Sdim
703343171Sdim  /// Whether to attempt a file lookup in external file system after it wasn't
704343171Sdim  /// found in VFS.
705343171Sdim  bool IsFallthrough = true;
706343171Sdim  /// @}
707343171Sdim
708343171Sdim  /// Virtual file paths and external files could be canonicalized without "..",
709343171Sdim  /// "." and "./" in their paths. FIXME: some unittests currently fail on
710343171Sdim  /// win32 when using remove_dots and remove_leading_dotslash on paths.
711343171Sdim  bool UseCanonicalizedPaths =
712343171Sdim#ifdef _WIN32
713343171Sdim      false;
714343171Sdim#else
715343171Sdim      true;
716343171Sdim#endif
717343171Sdim
718360784Sdim  RedirectingFileSystem(IntrusiveRefCntPtr<FileSystem> ExternalFS);
719343171Sdim
720343171Sdim  /// Looks up the path <tt>[Start, End)</tt> in \p From, possibly
721343171Sdim  /// recursing into the contents of \p From if it is a directory.
722343171Sdim  ErrorOr<Entry *> lookupPath(llvm::sys::path::const_iterator Start,
723343171Sdim                              llvm::sys::path::const_iterator End,
724343171Sdim                              Entry *From) const;
725343171Sdim
726343171Sdim  /// Get the status of a given an \c Entry.
727343171Sdim  ErrorOr<Status> status(const Twine &Path, Entry *E);
728343171Sdim
729343171Sdimpublic:
730343171Sdim  /// Looks up \p Path in \c Roots.
731343171Sdim  ErrorOr<Entry *> lookupPath(const Twine &Path) const;
732343171Sdim
733343171Sdim  /// Parses \p Buffer, which is expected to be in YAML format and
734343171Sdim  /// returns a virtual file system representing its contents.
735343171Sdim  static RedirectingFileSystem *
736343171Sdim  create(std::unique_ptr<MemoryBuffer> Buffer,
737343171Sdim         SourceMgr::DiagHandlerTy DiagHandler, StringRef YAMLFilePath,
738343171Sdim         void *DiagContext, IntrusiveRefCntPtr<FileSystem> ExternalFS);
739343171Sdim
740343171Sdim  ErrorOr<Status> status(const Twine &Path) override;
741343171Sdim  ErrorOr<std::unique_ptr<File>> openFileForRead(const Twine &Path) override;
742343171Sdim
743343171Sdim  std::error_code getRealPath(const Twine &Path,
744343171Sdim                              SmallVectorImpl<char> &Output) const override;
745343171Sdim
746343171Sdim  llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override;
747343171Sdim
748343171Sdim  std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
749343171Sdim
750343171Sdim  std::error_code isLocal(const Twine &Path, bool &Result) override;
751343171Sdim
752360784Sdim  std::error_code makeAbsolute(SmallVectorImpl<char> &Path) const override;
753360784Sdim
754343171Sdim  directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
755343171Sdim
756343171Sdim  void setExternalContentsPrefixDir(StringRef PrefixDir);
757343171Sdim
758343171Sdim  StringRef getExternalContentsPrefixDir() const;
759343171Sdim
760360784Sdim  void dump(raw_ostream &OS) const;
761360784Sdim  void dumpEntry(raw_ostream &OS, Entry *E, int NumSpaces = 0) const;
762343171Sdim#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
763343171Sdim  LLVM_DUMP_METHOD void dump() const;
764343171Sdim#endif
765343171Sdim};
766343171Sdim
767343171Sdim/// Collect all pairs of <virtual path, real path> entries from the
768343171Sdim/// \p YAMLFilePath. This is used by the module dependency collector to forward
769343171Sdim/// the entries into the reproducer output VFS YAML file.
770343171Sdimvoid collectVFSFromYAML(
771343171Sdim    std::unique_ptr<llvm::MemoryBuffer> Buffer,
772343171Sdim    llvm::SourceMgr::DiagHandlerTy DiagHandler, StringRef YAMLFilePath,
773343171Sdim    SmallVectorImpl<YAMLVFSEntry> &CollectedEntries,
774343171Sdim    void *DiagContext = nullptr,
775343171Sdim    IntrusiveRefCntPtr<FileSystem> ExternalFS = getRealFileSystem());
776343171Sdim
777343171Sdimclass YAMLVFSWriter {
778343171Sdim  std::vector<YAMLVFSEntry> Mappings;
779343171Sdim  Optional<bool> IsCaseSensitive;
780343171Sdim  Optional<bool> IsOverlayRelative;
781343171Sdim  Optional<bool> UseExternalNames;
782343171Sdim  std::string OverlayDir;
783343171Sdim
784343171Sdimpublic:
785343171Sdim  YAMLVFSWriter() = default;
786343171Sdim
787343171Sdim  void addFileMapping(StringRef VirtualPath, StringRef RealPath);
788343171Sdim
789343171Sdim  void setCaseSensitivity(bool CaseSensitive) {
790343171Sdim    IsCaseSensitive = CaseSensitive;
791343171Sdim  }
792343171Sdim
793343171Sdim  void setUseExternalNames(bool UseExtNames) { UseExternalNames = UseExtNames; }
794343171Sdim
795343171Sdim  void setOverlayDir(StringRef OverlayDirectory) {
796343171Sdim    IsOverlayRelative = true;
797343171Sdim    OverlayDir.assign(OverlayDirectory.str());
798343171Sdim  }
799343171Sdim
800343171Sdim  const std::vector<YAMLVFSEntry> &getMappings() const { return Mappings; }
801343171Sdim
802343171Sdim  void write(llvm::raw_ostream &OS);
803343171Sdim};
804343171Sdim
805343171Sdim} // namespace vfs
806343171Sdim} // namespace llvm
807343171Sdim
808343171Sdim#endif // LLVM_SUPPORT_VIRTUALFILESYSTEM_H
809