1 //===- VirtualFileSystem.cpp - Virtual File System Layer --------*- C++ -*-===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 // This file implements the VirtualFileSystem interface.
10 //===----------------------------------------------------------------------===//
11
12 #include "clang/Basic/VirtualFileSystem.h"
13 #include "llvm/ADT/DenseMap.h"
14 #include "llvm/ADT/STLExtras.h"
15 #include "llvm/ADT/StringExtras.h"
16 #include "llvm/ADT/StringSet.h"
17 #include "llvm/ADT/iterator_range.h"
18 #include "llvm/Support/Errc.h"
19 #include "llvm/Support/MemoryBuffer.h"
20 #include "llvm/Support/Path.h"
21 #include "llvm/Support/YAMLParser.h"
22 #include <atomic>
23 #include <memory>
24
25 using namespace clang;
26 using namespace clang::vfs;
27 using namespace llvm;
28 using llvm::sys::fs::file_status;
29 using llvm::sys::fs::file_type;
30 using llvm::sys::fs::perms;
31 using llvm::sys::fs::UniqueID;
32
Status(const file_status & Status)33 Status::Status(const file_status &Status)
34 : UID(Status.getUniqueID()), MTime(Status.getLastModificationTime()),
35 User(Status.getUser()), Group(Status.getGroup()), Size(Status.getSize()),
36 Type(Status.type()), Perms(Status.permissions()), IsVFSMapped(false) {}
37
Status(StringRef Name,StringRef ExternalName,UniqueID UID,sys::TimeValue MTime,uint32_t User,uint32_t Group,uint64_t Size,file_type Type,perms Perms)38 Status::Status(StringRef Name, StringRef ExternalName, UniqueID UID,
39 sys::TimeValue MTime, uint32_t User, uint32_t Group,
40 uint64_t Size, file_type Type, perms Perms)
41 : Name(Name), UID(UID), MTime(MTime), User(User), Group(Group), Size(Size),
42 Type(Type), Perms(Perms), IsVFSMapped(false) {}
43
equivalent(const Status & Other) const44 bool Status::equivalent(const Status &Other) const {
45 return getUniqueID() == Other.getUniqueID();
46 }
isDirectory() const47 bool Status::isDirectory() const {
48 return Type == file_type::directory_file;
49 }
isRegularFile() const50 bool Status::isRegularFile() const {
51 return Type == file_type::regular_file;
52 }
isOther() const53 bool Status::isOther() const {
54 return exists() && !isRegularFile() && !isDirectory() && !isSymlink();
55 }
isSymlink() const56 bool Status::isSymlink() const {
57 return Type == file_type::symlink_file;
58 }
isStatusKnown() const59 bool Status::isStatusKnown() const {
60 return Type != file_type::status_error;
61 }
exists() const62 bool Status::exists() const {
63 return isStatusKnown() && Type != file_type::file_not_found;
64 }
65
~File()66 File::~File() {}
67
~FileSystem()68 FileSystem::~FileSystem() {}
69
70 ErrorOr<std::unique_ptr<MemoryBuffer>>
getBufferForFile(const llvm::Twine & Name,int64_t FileSize,bool RequiresNullTerminator,bool IsVolatile)71 FileSystem::getBufferForFile(const llvm::Twine &Name, int64_t FileSize,
72 bool RequiresNullTerminator, bool IsVolatile) {
73 auto F = openFileForRead(Name);
74 if (!F)
75 return F.getError();
76
77 return (*F)->getBuffer(Name, FileSize, RequiresNullTerminator, IsVolatile);
78 }
79
80 //===-----------------------------------------------------------------------===/
81 // RealFileSystem implementation
82 //===-----------------------------------------------------------------------===/
83
84 namespace {
85 /// \brief Wrapper around a raw file descriptor.
86 class RealFile : public File {
87 int FD;
88 Status S;
89 friend class RealFileSystem;
RealFile(int FD)90 RealFile(int FD) : FD(FD) {
91 assert(FD >= 0 && "Invalid or inactive file descriptor");
92 }
93
94 public:
95 ~RealFile() override;
96 ErrorOr<Status> status() override;
97 ErrorOr<std::unique_ptr<MemoryBuffer>>
98 getBuffer(const Twine &Name, int64_t FileSize = -1,
99 bool RequiresNullTerminator = true,
100 bool IsVolatile = false) override;
101 std::error_code close() override;
102 void setName(StringRef Name) override;
103 };
104 } // end anonymous namespace
~RealFile()105 RealFile::~RealFile() { close(); }
106
status()107 ErrorOr<Status> RealFile::status() {
108 assert(FD != -1 && "cannot stat closed file");
109 if (!S.isStatusKnown()) {
110 file_status RealStatus;
111 if (std::error_code EC = sys::fs::status(FD, RealStatus))
112 return EC;
113 Status NewS(RealStatus);
114 NewS.setName(S.getName());
115 S = std::move(NewS);
116 }
117 return S;
118 }
119
120 ErrorOr<std::unique_ptr<MemoryBuffer>>
getBuffer(const Twine & Name,int64_t FileSize,bool RequiresNullTerminator,bool IsVolatile)121 RealFile::getBuffer(const Twine &Name, int64_t FileSize,
122 bool RequiresNullTerminator, bool IsVolatile) {
123 assert(FD != -1 && "cannot get buffer for closed file");
124 return MemoryBuffer::getOpenFile(FD, Name, FileSize, RequiresNullTerminator,
125 IsVolatile);
126 }
127
128 // FIXME: This is terrible, we need this for ::close.
129 #if !defined(_MSC_VER) && !defined(__MINGW32__)
130 #include <unistd.h>
131 #include <sys/uio.h>
132 #else
133 #include <io.h>
134 #ifndef S_ISFIFO
135 #define S_ISFIFO(x) (0)
136 #endif
137 #endif
close()138 std::error_code RealFile::close() {
139 if (::close(FD))
140 return std::error_code(errno, std::generic_category());
141 FD = -1;
142 return std::error_code();
143 }
144
setName(StringRef Name)145 void RealFile::setName(StringRef Name) {
146 S.setName(Name);
147 }
148
149 namespace {
150 /// \brief The file system according to your operating system.
151 class RealFileSystem : public FileSystem {
152 public:
153 ErrorOr<Status> status(const Twine &Path) override;
154 ErrorOr<std::unique_ptr<File>> openFileForRead(const Twine &Path) override;
155 directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
156 };
157 } // end anonymous namespace
158
status(const Twine & Path)159 ErrorOr<Status> RealFileSystem::status(const Twine &Path) {
160 sys::fs::file_status RealStatus;
161 if (std::error_code EC = sys::fs::status(Path, RealStatus))
162 return EC;
163 Status Result(RealStatus);
164 Result.setName(Path.str());
165 return Result;
166 }
167
168 ErrorOr<std::unique_ptr<File>>
openFileForRead(const Twine & Name)169 RealFileSystem::openFileForRead(const Twine &Name) {
170 int FD;
171 if (std::error_code EC = sys::fs::openFileForRead(Name, FD))
172 return EC;
173 std::unique_ptr<File> Result(new RealFile(FD));
174 Result->setName(Name.str());
175 return std::move(Result);
176 }
177
getRealFileSystem()178 IntrusiveRefCntPtr<FileSystem> vfs::getRealFileSystem() {
179 static IntrusiveRefCntPtr<FileSystem> FS = new RealFileSystem();
180 return FS;
181 }
182
183 namespace {
184 class RealFSDirIter : public clang::vfs::detail::DirIterImpl {
185 std::string Path;
186 llvm::sys::fs::directory_iterator Iter;
187 public:
RealFSDirIter(const Twine & _Path,std::error_code & EC)188 RealFSDirIter(const Twine &_Path, std::error_code &EC)
189 : Path(_Path.str()), Iter(Path, EC) {
190 if (!EC && Iter != llvm::sys::fs::directory_iterator()) {
191 llvm::sys::fs::file_status S;
192 EC = Iter->status(S);
193 if (!EC) {
194 CurrentEntry = Status(S);
195 CurrentEntry.setName(Iter->path());
196 }
197 }
198 }
199
increment()200 std::error_code increment() override {
201 std::error_code EC;
202 Iter.increment(EC);
203 if (EC) {
204 return EC;
205 } else if (Iter == llvm::sys::fs::directory_iterator()) {
206 CurrentEntry = Status();
207 } else {
208 llvm::sys::fs::file_status S;
209 EC = Iter->status(S);
210 CurrentEntry = Status(S);
211 CurrentEntry.setName(Iter->path());
212 }
213 return EC;
214 }
215 };
216 }
217
dir_begin(const Twine & Dir,std::error_code & EC)218 directory_iterator RealFileSystem::dir_begin(const Twine &Dir,
219 std::error_code &EC) {
220 return directory_iterator(std::make_shared<RealFSDirIter>(Dir, EC));
221 }
222
223 //===-----------------------------------------------------------------------===/
224 // OverlayFileSystem implementation
225 //===-----------------------------------------------------------------------===/
OverlayFileSystem(IntrusiveRefCntPtr<FileSystem> BaseFS)226 OverlayFileSystem::OverlayFileSystem(IntrusiveRefCntPtr<FileSystem> BaseFS) {
227 pushOverlay(BaseFS);
228 }
229
pushOverlay(IntrusiveRefCntPtr<FileSystem> FS)230 void OverlayFileSystem::pushOverlay(IntrusiveRefCntPtr<FileSystem> FS) {
231 FSList.push_back(FS);
232 }
233
status(const Twine & Path)234 ErrorOr<Status> OverlayFileSystem::status(const Twine &Path) {
235 // FIXME: handle symlinks that cross file systems
236 for (iterator I = overlays_begin(), E = overlays_end(); I != E; ++I) {
237 ErrorOr<Status> Status = (*I)->status(Path);
238 if (Status || Status.getError() != llvm::errc::no_such_file_or_directory)
239 return Status;
240 }
241 return make_error_code(llvm::errc::no_such_file_or_directory);
242 }
243
244 ErrorOr<std::unique_ptr<File>>
openFileForRead(const llvm::Twine & Path)245 OverlayFileSystem::openFileForRead(const llvm::Twine &Path) {
246 // FIXME: handle symlinks that cross file systems
247 for (iterator I = overlays_begin(), E = overlays_end(); I != E; ++I) {
248 auto Result = (*I)->openFileForRead(Path);
249 if (Result || Result.getError() != llvm::errc::no_such_file_or_directory)
250 return Result;
251 }
252 return make_error_code(llvm::errc::no_such_file_or_directory);
253 }
254
~DirIterImpl()255 clang::vfs::detail::DirIterImpl::~DirIterImpl() { }
256
257 namespace {
258 class OverlayFSDirIterImpl : public clang::vfs::detail::DirIterImpl {
259 OverlayFileSystem &Overlays;
260 std::string Path;
261 OverlayFileSystem::iterator CurrentFS;
262 directory_iterator CurrentDirIter;
263 llvm::StringSet<> SeenNames;
264
incrementFS()265 std::error_code incrementFS() {
266 assert(CurrentFS != Overlays.overlays_end() && "incrementing past end");
267 ++CurrentFS;
268 for (auto E = Overlays.overlays_end(); CurrentFS != E; ++CurrentFS) {
269 std::error_code EC;
270 CurrentDirIter = (*CurrentFS)->dir_begin(Path, EC);
271 if (EC && EC != errc::no_such_file_or_directory)
272 return EC;
273 if (CurrentDirIter != directory_iterator())
274 break; // found
275 }
276 return std::error_code();
277 }
278
incrementDirIter(bool IsFirstTime)279 std::error_code incrementDirIter(bool IsFirstTime) {
280 assert((IsFirstTime || CurrentDirIter != directory_iterator()) &&
281 "incrementing past end");
282 std::error_code EC;
283 if (!IsFirstTime)
284 CurrentDirIter.increment(EC);
285 if (!EC && CurrentDirIter == directory_iterator())
286 EC = incrementFS();
287 return EC;
288 }
289
incrementImpl(bool IsFirstTime)290 std::error_code incrementImpl(bool IsFirstTime) {
291 while (true) {
292 std::error_code EC = incrementDirIter(IsFirstTime);
293 if (EC || CurrentDirIter == directory_iterator()) {
294 CurrentEntry = Status();
295 return EC;
296 }
297 CurrentEntry = *CurrentDirIter;
298 StringRef Name = llvm::sys::path::filename(CurrentEntry.getName());
299 if (SeenNames.insert(Name).second)
300 return EC; // name not seen before
301 }
302 llvm_unreachable("returned above");
303 }
304
305 public:
OverlayFSDirIterImpl(const Twine & Path,OverlayFileSystem & FS,std::error_code & EC)306 OverlayFSDirIterImpl(const Twine &Path, OverlayFileSystem &FS,
307 std::error_code &EC)
308 : Overlays(FS), Path(Path.str()), CurrentFS(Overlays.overlays_begin()) {
309 CurrentDirIter = (*CurrentFS)->dir_begin(Path, EC);
310 EC = incrementImpl(true);
311 }
312
increment()313 std::error_code increment() override { return incrementImpl(false); }
314 };
315 } // end anonymous namespace
316
dir_begin(const Twine & Dir,std::error_code & EC)317 directory_iterator OverlayFileSystem::dir_begin(const Twine &Dir,
318 std::error_code &EC) {
319 return directory_iterator(
320 std::make_shared<OverlayFSDirIterImpl>(Dir, *this, EC));
321 }
322
323 //===-----------------------------------------------------------------------===/
324 // VFSFromYAML implementation
325 //===-----------------------------------------------------------------------===/
326
327 // Allow DenseMap<StringRef, ...>. This is useful below because we know all the
328 // strings are literals and will outlive the map, and there is no reason to
329 // store them.
330 namespace llvm {
331 template<>
332 struct DenseMapInfo<StringRef> {
333 // This assumes that "" will never be a valid key.
getEmptyKeyllvm::DenseMapInfo334 static inline StringRef getEmptyKey() { return StringRef(""); }
getTombstoneKeyllvm::DenseMapInfo335 static inline StringRef getTombstoneKey() { return StringRef(); }
getHashValuellvm::DenseMapInfo336 static unsigned getHashValue(StringRef Val) { return HashString(Val); }
isEqualllvm::DenseMapInfo337 static bool isEqual(StringRef LHS, StringRef RHS) { return LHS == RHS; }
338 };
339 }
340
341 namespace {
342
343 enum EntryKind {
344 EK_Directory,
345 EK_File
346 };
347
348 /// \brief A single file or directory in the VFS.
349 class Entry {
350 EntryKind Kind;
351 std::string Name;
352
353 public:
354 virtual ~Entry();
Entry(EntryKind K,StringRef Name)355 Entry(EntryKind K, StringRef Name) : Kind(K), Name(Name) {}
getName() const356 StringRef getName() const { return Name; }
getKind() const357 EntryKind getKind() const { return Kind; }
358 };
359
360 class DirectoryEntry : public Entry {
361 std::vector<Entry *> Contents;
362 Status S;
363
364 public:
365 ~DirectoryEntry() override;
DirectoryEntry(StringRef Name,std::vector<Entry * > Contents,Status S)366 DirectoryEntry(StringRef Name, std::vector<Entry *> Contents, Status S)
367 : Entry(EK_Directory, Name), Contents(std::move(Contents)),
368 S(std::move(S)) {}
getStatus()369 Status getStatus() { return S; }
370 typedef std::vector<Entry *>::iterator iterator;
contents_begin()371 iterator contents_begin() { return Contents.begin(); }
contents_end()372 iterator contents_end() { return Contents.end(); }
classof(const Entry * E)373 static bool classof(const Entry *E) { return E->getKind() == EK_Directory; }
374 };
375
376 class FileEntry : public Entry {
377 public:
378 enum NameKind {
379 NK_NotSet,
380 NK_External,
381 NK_Virtual
382 };
383 private:
384 std::string ExternalContentsPath;
385 NameKind UseName;
386 public:
FileEntry(StringRef Name,StringRef ExternalContentsPath,NameKind UseName)387 FileEntry(StringRef Name, StringRef ExternalContentsPath, NameKind UseName)
388 : Entry(EK_File, Name), ExternalContentsPath(ExternalContentsPath),
389 UseName(UseName) {}
getExternalContentsPath() const390 StringRef getExternalContentsPath() const { return ExternalContentsPath; }
391 /// \brief whether to use the external path as the name for this file.
useExternalName(bool GlobalUseExternalName) const392 bool useExternalName(bool GlobalUseExternalName) const {
393 return UseName == NK_NotSet ? GlobalUseExternalName
394 : (UseName == NK_External);
395 }
classof(const Entry * E)396 static bool classof(const Entry *E) { return E->getKind() == EK_File; }
397 };
398
399 class VFSFromYAML;
400
401 class VFSFromYamlDirIterImpl : public clang::vfs::detail::DirIterImpl {
402 std::string Dir;
403 VFSFromYAML &FS;
404 DirectoryEntry::iterator Current, End;
405 public:
406 VFSFromYamlDirIterImpl(const Twine &Path, VFSFromYAML &FS,
407 DirectoryEntry::iterator Begin,
408 DirectoryEntry::iterator End, std::error_code &EC);
409 std::error_code increment() override;
410 };
411
412 /// \brief A virtual file system parsed from a YAML file.
413 ///
414 /// Currently, this class allows creating virtual directories and mapping
415 /// virtual file paths to existing external files, available in \c ExternalFS.
416 ///
417 /// The basic structure of the parsed file is:
418 /// \verbatim
419 /// {
420 /// 'version': <version number>,
421 /// <optional configuration>
422 /// 'roots': [
423 /// <directory entries>
424 /// ]
425 /// }
426 /// \endverbatim
427 ///
428 /// All configuration options are optional.
429 /// 'case-sensitive': <boolean, default=true>
430 /// 'use-external-names': <boolean, default=true>
431 ///
432 /// Virtual directories are represented as
433 /// \verbatim
434 /// {
435 /// 'type': 'directory',
436 /// 'name': <string>,
437 /// 'contents': [ <file or directory entries> ]
438 /// }
439 /// \endverbatim
440 ///
441 /// The default attributes for virtual directories are:
442 /// \verbatim
443 /// MTime = now() when created
444 /// Perms = 0777
445 /// User = Group = 0
446 /// Size = 0
447 /// UniqueID = unspecified unique value
448 /// \endverbatim
449 ///
450 /// Re-mapped files are represented as
451 /// \verbatim
452 /// {
453 /// 'type': 'file',
454 /// 'name': <string>,
455 /// 'use-external-name': <boolean> # Optional
456 /// 'external-contents': <path to external file>)
457 /// }
458 /// \endverbatim
459 ///
460 /// and inherit their attributes from the external contents.
461 ///
462 /// In both cases, the 'name' field may contain multiple path components (e.g.
463 /// /path/to/file). However, any directory that contains more than one child
464 /// must be uniquely represented by a directory entry.
465 class VFSFromYAML : public vfs::FileSystem {
466 std::vector<Entry *> Roots; ///< The root(s) of the virtual file system.
467 /// \brief The file system to use for external references.
468 IntrusiveRefCntPtr<FileSystem> ExternalFS;
469
470 /// @name Configuration
471 /// @{
472
473 /// \brief Whether to perform case-sensitive comparisons.
474 ///
475 /// Currently, case-insensitive matching only works correctly with ASCII.
476 bool CaseSensitive;
477
478 /// \brief Whether to use to use the value of 'external-contents' for the
479 /// names of files. This global value is overridable on a per-file basis.
480 bool UseExternalNames;
481 /// @}
482
483 friend class VFSFromYAMLParser;
484
485 private:
VFSFromYAML(IntrusiveRefCntPtr<FileSystem> ExternalFS)486 VFSFromYAML(IntrusiveRefCntPtr<FileSystem> ExternalFS)
487 : ExternalFS(ExternalFS), CaseSensitive(true), UseExternalNames(true) {}
488
489 /// \brief Looks up \p Path in \c Roots.
490 ErrorOr<Entry *> lookupPath(const Twine &Path);
491
492 /// \brief Looks up the path <tt>[Start, End)</tt> in \p From, possibly
493 /// recursing into the contents of \p From if it is a directory.
494 ErrorOr<Entry *> lookupPath(sys::path::const_iterator Start,
495 sys::path::const_iterator End, Entry *From);
496
497 /// \brief Get the status of a given an \c Entry.
498 ErrorOr<Status> status(const Twine &Path, Entry *E);
499
500 public:
501 ~VFSFromYAML() override;
502
503 /// \brief Parses \p Buffer, which is expected to be in YAML format and
504 /// returns a virtual file system representing its contents.
505 static VFSFromYAML *create(std::unique_ptr<MemoryBuffer> Buffer,
506 SourceMgr::DiagHandlerTy DiagHandler,
507 void *DiagContext,
508 IntrusiveRefCntPtr<FileSystem> ExternalFS);
509
510 ErrorOr<Status> status(const Twine &Path) override;
511 ErrorOr<std::unique_ptr<File>> openFileForRead(const Twine &Path) override;
512
dir_begin(const Twine & Dir,std::error_code & EC)513 directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override{
514 ErrorOr<Entry *> E = lookupPath(Dir);
515 if (!E) {
516 EC = E.getError();
517 return directory_iterator();
518 }
519 ErrorOr<Status> S = status(Dir, *E);
520 if (!S) {
521 EC = S.getError();
522 return directory_iterator();
523 }
524 if (!S->isDirectory()) {
525 EC = std::error_code(static_cast<int>(errc::not_a_directory),
526 std::system_category());
527 return directory_iterator();
528 }
529
530 DirectoryEntry *D = cast<DirectoryEntry>(*E);
531 return directory_iterator(std::make_shared<VFSFromYamlDirIterImpl>(Dir,
532 *this, D->contents_begin(), D->contents_end(), EC));
533 }
534 };
535
536 /// \brief A helper class to hold the common YAML parsing state.
537 class VFSFromYAMLParser {
538 yaml::Stream &Stream;
539
error(yaml::Node * N,const Twine & Msg)540 void error(yaml::Node *N, const Twine &Msg) {
541 Stream.printError(N, Msg);
542 }
543
544 // false on error
parseScalarString(yaml::Node * N,StringRef & Result,SmallVectorImpl<char> & Storage)545 bool parseScalarString(yaml::Node *N, StringRef &Result,
546 SmallVectorImpl<char> &Storage) {
547 yaml::ScalarNode *S = dyn_cast<yaml::ScalarNode>(N);
548 if (!S) {
549 error(N, "expected string");
550 return false;
551 }
552 Result = S->getValue(Storage);
553 return true;
554 }
555
556 // false on error
parseScalarBool(yaml::Node * N,bool & Result)557 bool parseScalarBool(yaml::Node *N, bool &Result) {
558 SmallString<5> Storage;
559 StringRef Value;
560 if (!parseScalarString(N, Value, Storage))
561 return false;
562
563 if (Value.equals_lower("true") || Value.equals_lower("on") ||
564 Value.equals_lower("yes") || Value == "1") {
565 Result = true;
566 return true;
567 } else if (Value.equals_lower("false") || Value.equals_lower("off") ||
568 Value.equals_lower("no") || Value == "0") {
569 Result = false;
570 return true;
571 }
572
573 error(N, "expected boolean value");
574 return false;
575 }
576
577 struct KeyStatus {
KeyStatus__anon393354d90511::VFSFromYAMLParser::KeyStatus578 KeyStatus(bool Required=false) : Required(Required), Seen(false) {}
579 bool Required;
580 bool Seen;
581 };
582 typedef std::pair<StringRef, KeyStatus> KeyStatusPair;
583
584 // false on error
checkDuplicateOrUnknownKey(yaml::Node * KeyNode,StringRef Key,DenseMap<StringRef,KeyStatus> & Keys)585 bool checkDuplicateOrUnknownKey(yaml::Node *KeyNode, StringRef Key,
586 DenseMap<StringRef, KeyStatus> &Keys) {
587 if (!Keys.count(Key)) {
588 error(KeyNode, "unknown key");
589 return false;
590 }
591 KeyStatus &S = Keys[Key];
592 if (S.Seen) {
593 error(KeyNode, Twine("duplicate key '") + Key + "'");
594 return false;
595 }
596 S.Seen = true;
597 return true;
598 }
599
600 // false on error
checkMissingKeys(yaml::Node * Obj,DenseMap<StringRef,KeyStatus> & Keys)601 bool checkMissingKeys(yaml::Node *Obj, DenseMap<StringRef, KeyStatus> &Keys) {
602 for (DenseMap<StringRef, KeyStatus>::iterator I = Keys.begin(),
603 E = Keys.end();
604 I != E; ++I) {
605 if (I->second.Required && !I->second.Seen) {
606 error(Obj, Twine("missing key '") + I->first + "'");
607 return false;
608 }
609 }
610 return true;
611 }
612
parseEntry(yaml::Node * N)613 Entry *parseEntry(yaml::Node *N) {
614 yaml::MappingNode *M = dyn_cast<yaml::MappingNode>(N);
615 if (!M) {
616 error(N, "expected mapping node for file or directory entry");
617 return nullptr;
618 }
619
620 KeyStatusPair Fields[] = {
621 KeyStatusPair("name", true),
622 KeyStatusPair("type", true),
623 KeyStatusPair("contents", false),
624 KeyStatusPair("external-contents", false),
625 KeyStatusPair("use-external-name", false),
626 };
627
628 DenseMap<StringRef, KeyStatus> Keys(
629 &Fields[0], Fields + sizeof(Fields)/sizeof(Fields[0]));
630
631 bool HasContents = false; // external or otherwise
632 std::vector<Entry *> EntryArrayContents;
633 std::string ExternalContentsPath;
634 std::string Name;
635 FileEntry::NameKind UseExternalName = FileEntry::NK_NotSet;
636 EntryKind Kind;
637
638 for (yaml::MappingNode::iterator I = M->begin(), E = M->end(); I != E;
639 ++I) {
640 StringRef Key;
641 // Reuse the buffer for key and value, since we don't look at key after
642 // parsing value.
643 SmallString<256> Buffer;
644 if (!parseScalarString(I->getKey(), Key, Buffer))
645 return nullptr;
646
647 if (!checkDuplicateOrUnknownKey(I->getKey(), Key, Keys))
648 return nullptr;
649
650 StringRef Value;
651 if (Key == "name") {
652 if (!parseScalarString(I->getValue(), Value, Buffer))
653 return nullptr;
654 Name = Value;
655 } else if (Key == "type") {
656 if (!parseScalarString(I->getValue(), Value, Buffer))
657 return nullptr;
658 if (Value == "file")
659 Kind = EK_File;
660 else if (Value == "directory")
661 Kind = EK_Directory;
662 else {
663 error(I->getValue(), "unknown value for 'type'");
664 return nullptr;
665 }
666 } else if (Key == "contents") {
667 if (HasContents) {
668 error(I->getKey(),
669 "entry already has 'contents' or 'external-contents'");
670 return nullptr;
671 }
672 HasContents = true;
673 yaml::SequenceNode *Contents =
674 dyn_cast<yaml::SequenceNode>(I->getValue());
675 if (!Contents) {
676 // FIXME: this is only for directories, what about files?
677 error(I->getValue(), "expected array");
678 return nullptr;
679 }
680
681 for (yaml::SequenceNode::iterator I = Contents->begin(),
682 E = Contents->end();
683 I != E; ++I) {
684 if (Entry *E = parseEntry(&*I))
685 EntryArrayContents.push_back(E);
686 else
687 return nullptr;
688 }
689 } else if (Key == "external-contents") {
690 if (HasContents) {
691 error(I->getKey(),
692 "entry already has 'contents' or 'external-contents'");
693 return nullptr;
694 }
695 HasContents = true;
696 if (!parseScalarString(I->getValue(), Value, Buffer))
697 return nullptr;
698 ExternalContentsPath = Value;
699 } else if (Key == "use-external-name") {
700 bool Val;
701 if (!parseScalarBool(I->getValue(), Val))
702 return nullptr;
703 UseExternalName = Val ? FileEntry::NK_External : FileEntry::NK_Virtual;
704 } else {
705 llvm_unreachable("key missing from Keys");
706 }
707 }
708
709 if (Stream.failed())
710 return nullptr;
711
712 // check for missing keys
713 if (!HasContents) {
714 error(N, "missing key 'contents' or 'external-contents'");
715 return nullptr;
716 }
717 if (!checkMissingKeys(N, Keys))
718 return nullptr;
719
720 // check invalid configuration
721 if (Kind == EK_Directory && UseExternalName != FileEntry::NK_NotSet) {
722 error(N, "'use-external-name' is not supported for directories");
723 return nullptr;
724 }
725
726 // Remove trailing slash(es), being careful not to remove the root path
727 StringRef Trimmed(Name);
728 size_t RootPathLen = sys::path::root_path(Trimmed).size();
729 while (Trimmed.size() > RootPathLen &&
730 sys::path::is_separator(Trimmed.back()))
731 Trimmed = Trimmed.slice(0, Trimmed.size()-1);
732 // Get the last component
733 StringRef LastComponent = sys::path::filename(Trimmed);
734
735 Entry *Result = nullptr;
736 switch (Kind) {
737 case EK_File:
738 Result = new FileEntry(LastComponent, std::move(ExternalContentsPath),
739 UseExternalName);
740 break;
741 case EK_Directory:
742 Result = new DirectoryEntry(LastComponent, std::move(EntryArrayContents),
743 Status("", "", getNextVirtualUniqueID(), sys::TimeValue::now(), 0, 0,
744 0, file_type::directory_file, sys::fs::all_all));
745 break;
746 }
747
748 StringRef Parent = sys::path::parent_path(Trimmed);
749 if (Parent.empty())
750 return Result;
751
752 // if 'name' contains multiple components, create implicit directory entries
753 for (sys::path::reverse_iterator I = sys::path::rbegin(Parent),
754 E = sys::path::rend(Parent);
755 I != E; ++I) {
756 Result = new DirectoryEntry(*I, llvm::makeArrayRef(Result),
757 Status("", "", getNextVirtualUniqueID(), sys::TimeValue::now(), 0, 0,
758 0, file_type::directory_file, sys::fs::all_all));
759 }
760 return Result;
761 }
762
763 public:
VFSFromYAMLParser(yaml::Stream & S)764 VFSFromYAMLParser(yaml::Stream &S) : Stream(S) {}
765
766 // false on error
parse(yaml::Node * Root,VFSFromYAML * FS)767 bool parse(yaml::Node *Root, VFSFromYAML *FS) {
768 yaml::MappingNode *Top = dyn_cast<yaml::MappingNode>(Root);
769 if (!Top) {
770 error(Root, "expected mapping node");
771 return false;
772 }
773
774 KeyStatusPair Fields[] = {
775 KeyStatusPair("version", true),
776 KeyStatusPair("case-sensitive", false),
777 KeyStatusPair("use-external-names", false),
778 KeyStatusPair("roots", true),
779 };
780
781 DenseMap<StringRef, KeyStatus> Keys(
782 &Fields[0], Fields + sizeof(Fields)/sizeof(Fields[0]));
783
784 // Parse configuration and 'roots'
785 for (yaml::MappingNode::iterator I = Top->begin(), E = Top->end(); I != E;
786 ++I) {
787 SmallString<10> KeyBuffer;
788 StringRef Key;
789 if (!parseScalarString(I->getKey(), Key, KeyBuffer))
790 return false;
791
792 if (!checkDuplicateOrUnknownKey(I->getKey(), Key, Keys))
793 return false;
794
795 if (Key == "roots") {
796 yaml::SequenceNode *Roots = dyn_cast<yaml::SequenceNode>(I->getValue());
797 if (!Roots) {
798 error(I->getValue(), "expected array");
799 return false;
800 }
801
802 for (yaml::SequenceNode::iterator I = Roots->begin(), E = Roots->end();
803 I != E; ++I) {
804 if (Entry *E = parseEntry(&*I))
805 FS->Roots.push_back(E);
806 else
807 return false;
808 }
809 } else if (Key == "version") {
810 StringRef VersionString;
811 SmallString<4> Storage;
812 if (!parseScalarString(I->getValue(), VersionString, Storage))
813 return false;
814 int Version;
815 if (VersionString.getAsInteger<int>(10, Version)) {
816 error(I->getValue(), "expected integer");
817 return false;
818 }
819 if (Version < 0) {
820 error(I->getValue(), "invalid version number");
821 return false;
822 }
823 if (Version != 0) {
824 error(I->getValue(), "version mismatch, expected 0");
825 return false;
826 }
827 } else if (Key == "case-sensitive") {
828 if (!parseScalarBool(I->getValue(), FS->CaseSensitive))
829 return false;
830 } else if (Key == "use-external-names") {
831 if (!parseScalarBool(I->getValue(), FS->UseExternalNames))
832 return false;
833 } else {
834 llvm_unreachable("key missing from Keys");
835 }
836 }
837
838 if (Stream.failed())
839 return false;
840
841 if (!checkMissingKeys(Top, Keys))
842 return false;
843 return true;
844 }
845 };
846 } // end of anonymous namespace
847
~Entry()848 Entry::~Entry() {}
~DirectoryEntry()849 DirectoryEntry::~DirectoryEntry() { llvm::DeleteContainerPointers(Contents); }
850
~VFSFromYAML()851 VFSFromYAML::~VFSFromYAML() { llvm::DeleteContainerPointers(Roots); }
852
create(std::unique_ptr<MemoryBuffer> Buffer,SourceMgr::DiagHandlerTy DiagHandler,void * DiagContext,IntrusiveRefCntPtr<FileSystem> ExternalFS)853 VFSFromYAML *VFSFromYAML::create(std::unique_ptr<MemoryBuffer> Buffer,
854 SourceMgr::DiagHandlerTy DiagHandler,
855 void *DiagContext,
856 IntrusiveRefCntPtr<FileSystem> ExternalFS) {
857
858 SourceMgr SM;
859 yaml::Stream Stream(Buffer->getMemBufferRef(), SM);
860
861 SM.setDiagHandler(DiagHandler, DiagContext);
862 yaml::document_iterator DI = Stream.begin();
863 yaml::Node *Root = DI->getRoot();
864 if (DI == Stream.end() || !Root) {
865 SM.PrintMessage(SMLoc(), SourceMgr::DK_Error, "expected root node");
866 return nullptr;
867 }
868
869 VFSFromYAMLParser P(Stream);
870
871 std::unique_ptr<VFSFromYAML> FS(new VFSFromYAML(ExternalFS));
872 if (!P.parse(Root, FS.get()))
873 return nullptr;
874
875 return FS.release();
876 }
877
lookupPath(const Twine & Path_)878 ErrorOr<Entry *> VFSFromYAML::lookupPath(const Twine &Path_) {
879 SmallString<256> Path;
880 Path_.toVector(Path);
881
882 // Handle relative paths
883 if (std::error_code EC = sys::fs::make_absolute(Path))
884 return EC;
885
886 if (Path.empty())
887 return make_error_code(llvm::errc::invalid_argument);
888
889 sys::path::const_iterator Start = sys::path::begin(Path);
890 sys::path::const_iterator End = sys::path::end(Path);
891 for (std::vector<Entry *>::iterator I = Roots.begin(), E = Roots.end();
892 I != E; ++I) {
893 ErrorOr<Entry *> Result = lookupPath(Start, End, *I);
894 if (Result || Result.getError() != llvm::errc::no_such_file_or_directory)
895 return Result;
896 }
897 return make_error_code(llvm::errc::no_such_file_or_directory);
898 }
899
lookupPath(sys::path::const_iterator Start,sys::path::const_iterator End,Entry * From)900 ErrorOr<Entry *> VFSFromYAML::lookupPath(sys::path::const_iterator Start,
901 sys::path::const_iterator End,
902 Entry *From) {
903 if (Start->equals("."))
904 ++Start;
905
906 // FIXME: handle ..
907 if (CaseSensitive ? !Start->equals(From->getName())
908 : !Start->equals_lower(From->getName()))
909 // failure to match
910 return make_error_code(llvm::errc::no_such_file_or_directory);
911
912 ++Start;
913
914 if (Start == End) {
915 // Match!
916 return From;
917 }
918
919 DirectoryEntry *DE = dyn_cast<DirectoryEntry>(From);
920 if (!DE)
921 return make_error_code(llvm::errc::not_a_directory);
922
923 for (DirectoryEntry::iterator I = DE->contents_begin(),
924 E = DE->contents_end();
925 I != E; ++I) {
926 ErrorOr<Entry *> Result = lookupPath(Start, End, *I);
927 if (Result || Result.getError() != llvm::errc::no_such_file_or_directory)
928 return Result;
929 }
930 return make_error_code(llvm::errc::no_such_file_or_directory);
931 }
932
status(const Twine & Path,Entry * E)933 ErrorOr<Status> VFSFromYAML::status(const Twine &Path, Entry *E) {
934 assert(E != nullptr);
935 std::string PathStr(Path.str());
936 if (FileEntry *F = dyn_cast<FileEntry>(E)) {
937 ErrorOr<Status> S = ExternalFS->status(F->getExternalContentsPath());
938 assert(!S || S->getName() == F->getExternalContentsPath());
939 if (S && !F->useExternalName(UseExternalNames))
940 S->setName(PathStr);
941 if (S)
942 S->IsVFSMapped = true;
943 return S;
944 } else { // directory
945 DirectoryEntry *DE = cast<DirectoryEntry>(E);
946 Status S = DE->getStatus();
947 S.setName(PathStr);
948 return S;
949 }
950 }
951
status(const Twine & Path)952 ErrorOr<Status> VFSFromYAML::status(const Twine &Path) {
953 ErrorOr<Entry *> Result = lookupPath(Path);
954 if (!Result)
955 return Result.getError();
956 return status(Path, *Result);
957 }
958
openFileForRead(const Twine & Path)959 ErrorOr<std::unique_ptr<File>> VFSFromYAML::openFileForRead(const Twine &Path) {
960 ErrorOr<Entry *> E = lookupPath(Path);
961 if (!E)
962 return E.getError();
963
964 FileEntry *F = dyn_cast<FileEntry>(*E);
965 if (!F) // FIXME: errc::not_a_file?
966 return make_error_code(llvm::errc::invalid_argument);
967
968 auto Result = ExternalFS->openFileForRead(F->getExternalContentsPath());
969 if (!Result)
970 return Result;
971
972 if (!F->useExternalName(UseExternalNames))
973 (*Result)->setName(Path.str());
974
975 return Result;
976 }
977
978 IntrusiveRefCntPtr<FileSystem>
getVFSFromYAML(std::unique_ptr<MemoryBuffer> Buffer,SourceMgr::DiagHandlerTy DiagHandler,void * DiagContext,IntrusiveRefCntPtr<FileSystem> ExternalFS)979 vfs::getVFSFromYAML(std::unique_ptr<MemoryBuffer> Buffer,
980 SourceMgr::DiagHandlerTy DiagHandler, void *DiagContext,
981 IntrusiveRefCntPtr<FileSystem> ExternalFS) {
982 return VFSFromYAML::create(std::move(Buffer), DiagHandler, DiagContext,
983 ExternalFS);
984 }
985
getNextVirtualUniqueID()986 UniqueID vfs::getNextVirtualUniqueID() {
987 static std::atomic<unsigned> UID;
988 unsigned ID = ++UID;
989 // The following assumes that uint64_t max will never collide with a real
990 // dev_t value from the OS.
991 return UniqueID(std::numeric_limits<uint64_t>::max(), ID);
992 }
993
994 #ifndef NDEBUG
pathHasTraversal(StringRef Path)995 static bool pathHasTraversal(StringRef Path) {
996 using namespace llvm::sys;
997 for (StringRef Comp : llvm::make_range(path::begin(Path), path::end(Path)))
998 if (Comp == "." || Comp == "..")
999 return true;
1000 return false;
1001 }
1002 #endif
1003
addFileMapping(StringRef VirtualPath,StringRef RealPath)1004 void YAMLVFSWriter::addFileMapping(StringRef VirtualPath, StringRef RealPath) {
1005 assert(sys::path::is_absolute(VirtualPath) && "virtual path not absolute");
1006 assert(sys::path::is_absolute(RealPath) && "real path not absolute");
1007 assert(!pathHasTraversal(VirtualPath) && "path traversal is not supported");
1008 Mappings.emplace_back(VirtualPath, RealPath);
1009 }
1010
1011 namespace {
1012 class JSONWriter {
1013 llvm::raw_ostream &OS;
1014 SmallVector<StringRef, 16> DirStack;
getDirIndent()1015 inline unsigned getDirIndent() { return 4 * DirStack.size(); }
getFileIndent()1016 inline unsigned getFileIndent() { return 4 * (DirStack.size() + 1); }
1017 bool containedIn(StringRef Parent, StringRef Path);
1018 StringRef containedPart(StringRef Parent, StringRef Path);
1019 void startDirectory(StringRef Path);
1020 void endDirectory();
1021 void writeEntry(StringRef VPath, StringRef RPath);
1022
1023 public:
JSONWriter(llvm::raw_ostream & OS)1024 JSONWriter(llvm::raw_ostream &OS) : OS(OS) {}
1025 void write(ArrayRef<YAMLVFSEntry> Entries, Optional<bool> IsCaseSensitive);
1026 };
1027 }
1028
containedIn(StringRef Parent,StringRef Path)1029 bool JSONWriter::containedIn(StringRef Parent, StringRef Path) {
1030 using namespace llvm::sys;
1031 // Compare each path component.
1032 auto IParent = path::begin(Parent), EParent = path::end(Parent);
1033 for (auto IChild = path::begin(Path), EChild = path::end(Path);
1034 IParent != EParent && IChild != EChild; ++IParent, ++IChild) {
1035 if (*IParent != *IChild)
1036 return false;
1037 }
1038 // Have we exhausted the parent path?
1039 return IParent == EParent;
1040 }
1041
containedPart(StringRef Parent,StringRef Path)1042 StringRef JSONWriter::containedPart(StringRef Parent, StringRef Path) {
1043 assert(!Parent.empty());
1044 assert(containedIn(Parent, Path));
1045 return Path.slice(Parent.size() + 1, StringRef::npos);
1046 }
1047
startDirectory(StringRef Path)1048 void JSONWriter::startDirectory(StringRef Path) {
1049 StringRef Name =
1050 DirStack.empty() ? Path : containedPart(DirStack.back(), Path);
1051 DirStack.push_back(Path);
1052 unsigned Indent = getDirIndent();
1053 OS.indent(Indent) << "{\n";
1054 OS.indent(Indent + 2) << "'type': 'directory',\n";
1055 OS.indent(Indent + 2) << "'name': \"" << llvm::yaml::escape(Name) << "\",\n";
1056 OS.indent(Indent + 2) << "'contents': [\n";
1057 }
1058
endDirectory()1059 void JSONWriter::endDirectory() {
1060 unsigned Indent = getDirIndent();
1061 OS.indent(Indent + 2) << "]\n";
1062 OS.indent(Indent) << "}";
1063
1064 DirStack.pop_back();
1065 }
1066
writeEntry(StringRef VPath,StringRef RPath)1067 void JSONWriter::writeEntry(StringRef VPath, StringRef RPath) {
1068 unsigned Indent = getFileIndent();
1069 OS.indent(Indent) << "{\n";
1070 OS.indent(Indent + 2) << "'type': 'file',\n";
1071 OS.indent(Indent + 2) << "'name': \"" << llvm::yaml::escape(VPath) << "\",\n";
1072 OS.indent(Indent + 2) << "'external-contents': \""
1073 << llvm::yaml::escape(RPath) << "\"\n";
1074 OS.indent(Indent) << "}";
1075 }
1076
write(ArrayRef<YAMLVFSEntry> Entries,Optional<bool> IsCaseSensitive)1077 void JSONWriter::write(ArrayRef<YAMLVFSEntry> Entries,
1078 Optional<bool> IsCaseSensitive) {
1079 using namespace llvm::sys;
1080
1081 OS << "{\n"
1082 " 'version': 0,\n";
1083 if (IsCaseSensitive.hasValue())
1084 OS << " 'case-sensitive': '"
1085 << (IsCaseSensitive.getValue() ? "true" : "false") << "',\n";
1086 OS << " 'roots': [\n";
1087
1088 if (!Entries.empty()) {
1089 const YAMLVFSEntry &Entry = Entries.front();
1090 startDirectory(path::parent_path(Entry.VPath));
1091 writeEntry(path::filename(Entry.VPath), Entry.RPath);
1092
1093 for (const auto &Entry : Entries.slice(1)) {
1094 StringRef Dir = path::parent_path(Entry.VPath);
1095 if (Dir == DirStack.back())
1096 OS << ",\n";
1097 else {
1098 while (!DirStack.empty() && !containedIn(DirStack.back(), Dir)) {
1099 OS << "\n";
1100 endDirectory();
1101 }
1102 OS << ",\n";
1103 startDirectory(Dir);
1104 }
1105 writeEntry(path::filename(Entry.VPath), Entry.RPath);
1106 }
1107
1108 while (!DirStack.empty()) {
1109 OS << "\n";
1110 endDirectory();
1111 }
1112 OS << "\n";
1113 }
1114
1115 OS << " ]\n"
1116 << "}\n";
1117 }
1118
write(llvm::raw_ostream & OS)1119 void YAMLVFSWriter::write(llvm::raw_ostream &OS) {
1120 std::sort(Mappings.begin(), Mappings.end(),
1121 [](const YAMLVFSEntry &LHS, const YAMLVFSEntry &RHS) {
1122 return LHS.VPath < RHS.VPath;
1123 });
1124
1125 JSONWriter(OS).write(Mappings, IsCaseSensitive);
1126 }
1127
VFSFromYamlDirIterImpl(const Twine & _Path,VFSFromYAML & FS,DirectoryEntry::iterator Begin,DirectoryEntry::iterator End,std::error_code & EC)1128 VFSFromYamlDirIterImpl::VFSFromYamlDirIterImpl(const Twine &_Path,
1129 VFSFromYAML &FS,
1130 DirectoryEntry::iterator Begin,
1131 DirectoryEntry::iterator End,
1132 std::error_code &EC)
1133 : Dir(_Path.str()), FS(FS), Current(Begin), End(End) {
1134 if (Current != End) {
1135 SmallString<128> PathStr(Dir);
1136 llvm::sys::path::append(PathStr, (*Current)->getName());
1137 llvm::ErrorOr<vfs::Status> S = FS.status(PathStr);
1138 if (S)
1139 CurrentEntry = *S;
1140 else
1141 EC = S.getError();
1142 }
1143 }
1144
increment()1145 std::error_code VFSFromYamlDirIterImpl::increment() {
1146 assert(Current != End && "cannot iterate past end");
1147 if (++Current != End) {
1148 SmallString<128> PathStr(Dir);
1149 llvm::sys::path::append(PathStr, (*Current)->getName());
1150 llvm::ErrorOr<vfs::Status> S = FS.status(PathStr);
1151 if (!S)
1152 return S.getError();
1153 CurrentEntry = *S;
1154 } else {
1155 CurrentEntry = Status();
1156 }
1157 return std::error_code();
1158 }
1159
recursive_directory_iterator(FileSystem & FS_,const Twine & Path,std::error_code & EC)1160 vfs::recursive_directory_iterator::recursive_directory_iterator(FileSystem &FS_,
1161 const Twine &Path,
1162 std::error_code &EC)
1163 : FS(&FS_) {
1164 directory_iterator I = FS->dir_begin(Path, EC);
1165 if (!EC && I != directory_iterator()) {
1166 State = std::make_shared<IterState>();
1167 State->push(I);
1168 }
1169 }
1170
1171 vfs::recursive_directory_iterator &
increment(std::error_code & EC)1172 recursive_directory_iterator::increment(std::error_code &EC) {
1173 assert(FS && State && !State->empty() && "incrementing past end");
1174 assert(State->top()->isStatusKnown() && "non-canonical end iterator");
1175 vfs::directory_iterator End;
1176 if (State->top()->isDirectory()) {
1177 vfs::directory_iterator I = FS->dir_begin(State->top()->getName(), EC);
1178 if (EC)
1179 return *this;
1180 if (I != End) {
1181 State->push(I);
1182 return *this;
1183 }
1184 }
1185
1186 while (!State->empty() && State->top().increment(EC) == End)
1187 State->pop();
1188
1189 if (State->empty())
1190 State.reset(); // end iterator
1191
1192 return *this;
1193 }
1194