1 //===- SourceManager.h - Track and cache source files -----------*- C++ -*-===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 /// \file 10 /// Defines the SourceManager interface. 11 /// 12 /// There are three different types of locations in a %file: a spelling 13 /// location, an expansion location, and a presumed location. 14 /// 15 /// Given an example of: 16 /// \code 17 /// #define min(x, y) x < y ? x : y 18 /// \endcode 19 /// 20 /// and then later on a use of min: 21 /// \code 22 /// #line 17 23 /// return min(a, b); 24 /// \endcode 25 /// 26 /// The expansion location is the line in the source code where the macro 27 /// was expanded (the return statement), the spelling location is the 28 /// location in the source where the macro was originally defined, 29 /// and the presumed location is where the line directive states that 30 /// the line is 17, or any other line. 31 // 32 //===----------------------------------------------------------------------===// 33 34 #ifndef LLVM_CLANG_BASIC_SOURCEMANAGER_H 35 #define LLVM_CLANG_BASIC_SOURCEMANAGER_H 36 37 #include "clang/Basic/Diagnostic.h" 38 #include "clang/Basic/FileEntry.h" 39 #include "clang/Basic/FileManager.h" 40 #include "clang/Basic/SourceLocation.h" 41 #include "llvm/ADT/ArrayRef.h" 42 #include "llvm/ADT/BitVector.h" 43 #include "llvm/ADT/DenseMap.h" 44 #include "llvm/ADT/DenseSet.h" 45 #include "llvm/ADT/IntrusiveRefCntPtr.h" 46 #include "llvm/ADT/PagedVector.h" 47 #include "llvm/ADT/PointerIntPair.h" 48 #include "llvm/ADT/SmallVector.h" 49 #include "llvm/ADT/StringRef.h" 50 #include "llvm/Support/Allocator.h" 51 #include "llvm/Support/Compiler.h" 52 #include "llvm/Support/MemoryBuffer.h" 53 #include <cassert> 54 #include <cstddef> 55 #include <map> 56 #include <memory> 57 #include <optional> 58 #include <string> 59 #include <utility> 60 #include <vector> 61 62 namespace clang { 63 64 class ASTReader; 65 class ASTWriter; 66 class FileManager; 67 class LineTableInfo; 68 class SourceManager; 69 70 /// Public enums and private classes that are part of the 71 /// SourceManager implementation. 72 namespace SrcMgr { 73 74 /// Indicates whether a file or directory holds normal user code, 75 /// system code, or system code which is implicitly 'extern "C"' in C++ mode. 76 /// 77 /// Entire directories can be tagged with this (this is maintained by 78 /// DirectoryLookup and friends) as can specific FileInfos when a \#pragma 79 /// system_header is seen or in various other cases. 80 /// 81 enum CharacteristicKind { 82 C_User, 83 C_System, 84 C_ExternCSystem, 85 C_User_ModuleMap, 86 C_System_ModuleMap 87 }; 88 89 /// Determine whether a file / directory characteristic is for system code. 90 inline bool isSystem(CharacteristicKind CK) { 91 return CK != C_User && CK != C_User_ModuleMap; 92 } 93 94 /// Determine whether a file characteristic is for a module map. 95 inline bool isModuleMap(CharacteristicKind CK) { 96 return CK == C_User_ModuleMap || CK == C_System_ModuleMap; 97 } 98 99 /// Mapping of line offsets into a source file. This does not own the storage 100 /// for the line numbers. 101 class LineOffsetMapping { 102 public: 103 explicit operator bool() const { return Storage; } 104 unsigned size() const { 105 assert(Storage); 106 return Storage[0]; 107 } 108 ArrayRef<unsigned> getLines() const { 109 assert(Storage); 110 return ArrayRef<unsigned>(Storage + 1, Storage + 1 + size()); 111 } 112 const unsigned *begin() const { return getLines().begin(); } 113 const unsigned *end() const { return getLines().end(); } 114 const unsigned &operator[](int I) const { return getLines()[I]; } 115 116 static LineOffsetMapping get(llvm::MemoryBufferRef Buffer, 117 llvm::BumpPtrAllocator &Alloc); 118 119 LineOffsetMapping() = default; 120 LineOffsetMapping(ArrayRef<unsigned> LineOffsets, 121 llvm::BumpPtrAllocator &Alloc); 122 123 private: 124 /// First element is the size, followed by elements at off-by-one indexes. 125 unsigned *Storage = nullptr; 126 }; 127 128 /// One instance of this struct is kept for every file loaded or used. 129 /// 130 /// This object owns the MemoryBuffer object. 131 class alignas(8) ContentCache { 132 /// The actual buffer containing the characters from the input 133 /// file. 134 mutable std::unique_ptr<llvm::MemoryBuffer> Buffer; 135 136 public: 137 /// Reference to the file entry representing this ContentCache. 138 /// 139 /// This reference does not own the FileEntry object. 140 /// 141 /// It is possible for this to be NULL if the ContentCache encapsulates 142 /// an imaginary text buffer. 143 /// 144 /// FIXME: Make non-optional using a virtual file as needed, remove \c 145 /// Filename and use \c OrigEntry.getNameAsRequested() instead. 146 OptionalFileEntryRef OrigEntry; 147 148 /// References the file which the contents were actually loaded from. 149 /// 150 /// Can be different from 'Entry' if we overridden the contents of one file 151 /// with the contents of another file. 152 OptionalFileEntryRef ContentsEntry; 153 154 /// The filename that is used to access OrigEntry. 155 /// 156 /// FIXME: Remove this once OrigEntry is a FileEntryRef with a stable name. 157 StringRef Filename; 158 159 /// A bump pointer allocated array of offsets for each source line. 160 /// 161 /// This is lazily computed. The lines are owned by the SourceManager 162 /// BumpPointerAllocator object. 163 mutable LineOffsetMapping SourceLineCache; 164 165 /// Indicates whether the buffer itself was provided to override 166 /// the actual file contents. 167 /// 168 /// When true, the original entry may be a virtual file that does not 169 /// exist. 170 LLVM_PREFERRED_TYPE(bool) 171 unsigned BufferOverridden : 1; 172 173 /// True if this content cache was initially created for a source file 174 /// considered to be volatile (likely to change between stat and open). 175 LLVM_PREFERRED_TYPE(bool) 176 unsigned IsFileVolatile : 1; 177 178 /// True if this file may be transient, that is, if it might not 179 /// exist at some later point in time when this content entry is used, 180 /// after serialization and deserialization. 181 LLVM_PREFERRED_TYPE(bool) 182 unsigned IsTransient : 1; 183 184 LLVM_PREFERRED_TYPE(bool) 185 mutable unsigned IsBufferInvalid : 1; 186 187 ContentCache() 188 : OrigEntry(std::nullopt), ContentsEntry(std::nullopt), 189 BufferOverridden(false), IsFileVolatile(false), IsTransient(false), 190 IsBufferInvalid(false) {} 191 192 ContentCache(FileEntryRef Ent) : ContentCache(Ent, Ent) {} 193 194 ContentCache(FileEntryRef Ent, FileEntryRef contentEnt) 195 : OrigEntry(Ent), ContentsEntry(contentEnt), BufferOverridden(false), 196 IsFileVolatile(false), IsTransient(false), IsBufferInvalid(false) {} 197 198 /// The copy ctor does not allow copies where source object has either 199 /// a non-NULL Buffer or SourceLineCache. Ownership of allocated memory 200 /// is not transferred, so this is a logical error. 201 ContentCache(const ContentCache &RHS) 202 : BufferOverridden(false), IsFileVolatile(false), IsTransient(false), 203 IsBufferInvalid(false) { 204 OrigEntry = RHS.OrigEntry; 205 ContentsEntry = RHS.ContentsEntry; 206 207 assert(!RHS.Buffer && !RHS.SourceLineCache && 208 "Passed ContentCache object cannot own a buffer."); 209 } 210 211 ContentCache &operator=(const ContentCache &RHS) = delete; 212 213 /// Returns the memory buffer for the associated content. 214 /// 215 /// \param Diag Object through which diagnostics will be emitted if the 216 /// buffer cannot be retrieved. 217 /// 218 /// \param Loc If specified, is the location that invalid file diagnostics 219 /// will be emitted at. 220 std::optional<llvm::MemoryBufferRef> 221 getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM, 222 SourceLocation Loc = SourceLocation()) const; 223 224 /// Returns the size of the content encapsulated by this 225 /// ContentCache. 226 /// 227 /// This can be the size of the source file or the size of an 228 /// arbitrary scratch buffer. If the ContentCache encapsulates a source 229 /// file this size is retrieved from the file's FileEntry. 230 unsigned getSize() const; 231 232 /// Returns the number of bytes actually mapped for this 233 /// ContentCache. 234 /// 235 /// This can be 0 if the MemBuffer was not actually expanded. 236 unsigned getSizeBytesMapped() const; 237 238 /// Returns the kind of memory used to back the memory buffer for 239 /// this content cache. This is used for performance analysis. 240 llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const; 241 242 /// Return the buffer, only if it has been loaded. 243 std::optional<llvm::MemoryBufferRef> getBufferIfLoaded() const { 244 if (Buffer) 245 return Buffer->getMemBufferRef(); 246 return std::nullopt; 247 } 248 249 /// Return a StringRef to the source buffer data, only if it has already 250 /// been loaded. 251 std::optional<StringRef> getBufferDataIfLoaded() const { 252 if (Buffer) 253 return Buffer->getBuffer(); 254 return std::nullopt; 255 } 256 257 /// Set the buffer. 258 void setBuffer(std::unique_ptr<llvm::MemoryBuffer> B) { 259 IsBufferInvalid = false; 260 Buffer = std::move(B); 261 } 262 263 /// Set the buffer to one that's not owned (or to nullptr). 264 /// 265 /// \pre Buffer cannot already be set. 266 void setUnownedBuffer(std::optional<llvm::MemoryBufferRef> B) { 267 assert(!Buffer && "Expected to be called right after construction"); 268 if (B) 269 setBuffer(llvm::MemoryBuffer::getMemBuffer(*B)); 270 } 271 272 // If BufStr has an invalid BOM, returns the BOM name; otherwise, returns 273 // nullptr 274 static const char *getInvalidBOM(StringRef BufStr); 275 }; 276 277 // Assert that the \c ContentCache objects will always be 8-byte aligned so 278 // that we can pack 3 bits of integer into pointers to such objects. 279 static_assert(alignof(ContentCache) >= 8, 280 "ContentCache must be 8-byte aligned."); 281 282 /// Information about a FileID, basically just the logical file 283 /// that it represents and include stack information. 284 /// 285 /// Each FileInfo has include stack information, indicating where it came 286 /// from. This information encodes the \#include chain that a token was 287 /// expanded from. The main include file has an invalid IncludeLoc. 288 /// 289 /// FileInfo should not grow larger than ExpansionInfo. Doing so will 290 /// cause memory to bloat in compilations with many unloaded macro 291 /// expansions, since the two data structurs are stored in a union in 292 /// SLocEntry. Extra fields should instead go in "ContentCache *", which 293 /// stores file contents and other bits on the side. 294 /// 295 class FileInfo { 296 friend class clang::SourceManager; 297 friend class clang::ASTWriter; 298 friend class clang::ASTReader; 299 300 /// The location of the \#include that brought in this file. 301 /// 302 /// This is an invalid SLOC for the main file (top of the \#include chain). 303 SourceLocation IncludeLoc; 304 305 /// Number of FileIDs (files and macros) that were created during 306 /// preprocessing of this \#include, including this SLocEntry. 307 /// 308 /// Zero means the preprocessor didn't provide such info for this SLocEntry. 309 unsigned NumCreatedFIDs : 31; 310 311 /// Whether this FileInfo has any \#line directives. 312 LLVM_PREFERRED_TYPE(bool) 313 unsigned HasLineDirectives : 1; 314 315 /// The content cache and the characteristic of the file. 316 llvm::PointerIntPair<const ContentCache *, 3, CharacteristicKind> 317 ContentAndKind; 318 319 public: 320 /// Return a FileInfo object. 321 static FileInfo get(SourceLocation IL, ContentCache &Con, 322 CharacteristicKind FileCharacter, StringRef Filename) { 323 FileInfo X; 324 X.IncludeLoc = IL; 325 X.NumCreatedFIDs = 0; 326 X.HasLineDirectives = false; 327 X.ContentAndKind.setPointer(&Con); 328 X.ContentAndKind.setInt(FileCharacter); 329 Con.Filename = Filename; 330 return X; 331 } 332 333 SourceLocation getIncludeLoc() const { 334 return IncludeLoc; 335 } 336 337 const ContentCache &getContentCache() const { 338 return *ContentAndKind.getPointer(); 339 } 340 341 /// Return whether this is a system header or not. 342 CharacteristicKind getFileCharacteristic() const { 343 return ContentAndKind.getInt(); 344 } 345 346 /// Return true if this FileID has \#line directives in it. 347 bool hasLineDirectives() const { return HasLineDirectives; } 348 349 /// Set the flag that indicates that this FileID has 350 /// line table entries associated with it. 351 void setHasLineDirectives() { HasLineDirectives = true; } 352 353 /// Returns the name of the file that was used when the file was loaded from 354 /// the underlying file system. 355 StringRef getName() const { return getContentCache().Filename; } 356 }; 357 358 /// Each ExpansionInfo encodes the expansion location - where 359 /// the token was ultimately expanded, and the SpellingLoc - where the actual 360 /// character data for the token came from. 361 class ExpansionInfo { 362 // Really these are all SourceLocations. 363 364 /// Where the spelling for the token can be found. 365 SourceLocation SpellingLoc; 366 367 /// In a macro expansion, ExpansionLocStart and ExpansionLocEnd 368 /// indicate the start and end of the expansion. In object-like macros, 369 /// they will be the same. In a function-like macro expansion, the start 370 /// will be the identifier and the end will be the ')'. Finally, in 371 /// macro-argument instantiations, the end will be 'SourceLocation()', an 372 /// invalid location. 373 SourceLocation ExpansionLocStart, ExpansionLocEnd; 374 375 /// Whether the expansion range is a token range. 376 bool ExpansionIsTokenRange; 377 378 public: 379 SourceLocation getSpellingLoc() const { 380 return SpellingLoc.isInvalid() ? getExpansionLocStart() : SpellingLoc; 381 } 382 383 SourceLocation getExpansionLocStart() const { 384 return ExpansionLocStart; 385 } 386 387 SourceLocation getExpansionLocEnd() const { 388 return ExpansionLocEnd.isInvalid() ? getExpansionLocStart() 389 : ExpansionLocEnd; 390 } 391 392 bool isExpansionTokenRange() const { return ExpansionIsTokenRange; } 393 394 CharSourceRange getExpansionLocRange() const { 395 return CharSourceRange( 396 SourceRange(getExpansionLocStart(), getExpansionLocEnd()), 397 isExpansionTokenRange()); 398 } 399 400 bool isMacroArgExpansion() const { 401 // Note that this needs to return false for default constructed objects. 402 return getExpansionLocStart().isValid() && ExpansionLocEnd.isInvalid(); 403 } 404 405 bool isMacroBodyExpansion() const { 406 return getExpansionLocStart().isValid() && ExpansionLocEnd.isValid(); 407 } 408 409 bool isFunctionMacroExpansion() const { 410 return getExpansionLocStart().isValid() && 411 getExpansionLocStart() != getExpansionLocEnd(); 412 } 413 414 /// Return a ExpansionInfo for an expansion. 415 /// 416 /// Start and End specify the expansion range (where the macro is 417 /// expanded), and SpellingLoc specifies the spelling location (where 418 /// the characters from the token come from). All three can refer to 419 /// normal File SLocs or expansion locations. 420 static ExpansionInfo create(SourceLocation SpellingLoc, SourceLocation Start, 421 SourceLocation End, 422 bool ExpansionIsTokenRange = true) { 423 ExpansionInfo X; 424 X.SpellingLoc = SpellingLoc; 425 X.ExpansionLocStart = Start; 426 X.ExpansionLocEnd = End; 427 X.ExpansionIsTokenRange = ExpansionIsTokenRange; 428 return X; 429 } 430 431 /// Return a special ExpansionInfo for the expansion of 432 /// a macro argument into a function-like macro's body. 433 /// 434 /// ExpansionLoc specifies the expansion location (where the macro is 435 /// expanded). This doesn't need to be a range because a macro is always 436 /// expanded at a macro parameter reference, and macro parameters are 437 /// always exactly one token. SpellingLoc specifies the spelling location 438 /// (where the characters from the token come from). ExpansionLoc and 439 /// SpellingLoc can both refer to normal File SLocs or expansion locations. 440 /// 441 /// Given the code: 442 /// \code 443 /// #define F(x) f(x) 444 /// F(42); 445 /// \endcode 446 /// 447 /// When expanding '\c F(42)', the '\c x' would call this with an 448 /// SpellingLoc pointing at '\c 42' and an ExpansionLoc pointing at its 449 /// location in the definition of '\c F'. 450 static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc, 451 SourceLocation ExpansionLoc) { 452 // We store an intentionally invalid source location for the end of the 453 // expansion range to mark that this is a macro argument location rather 454 // than a normal one. 455 return create(SpellingLoc, ExpansionLoc, SourceLocation()); 456 } 457 458 /// Return a special ExpansionInfo representing a token that ends 459 /// prematurely. This is used to model a '>>' token that has been split 460 /// into '>' tokens and similar cases. Unlike for the other forms of 461 /// expansion, the expansion range in this case is a character range, not 462 /// a token range. 463 static ExpansionInfo createForTokenSplit(SourceLocation SpellingLoc, 464 SourceLocation Start, 465 SourceLocation End) { 466 return create(SpellingLoc, Start, End, false); 467 } 468 }; 469 470 // Assert that the \c FileInfo objects are no bigger than \c ExpansionInfo 471 // objects. This controls the size of \c SLocEntry, of which we have one for 472 // each macro expansion. The number of (unloaded) macro expansions can be 473 // very large. Any other fields needed in FileInfo should go in ContentCache. 474 static_assert(sizeof(FileInfo) <= sizeof(ExpansionInfo), 475 "FileInfo must be no larger than ExpansionInfo."); 476 477 /// This is a discriminated union of FileInfo and ExpansionInfo. 478 /// 479 /// SourceManager keeps an array of these objects, and they are uniquely 480 /// identified by the FileID datatype. 481 class SLocEntry { 482 static constexpr int OffsetBits = 8 * sizeof(SourceLocation::UIntTy) - 1; 483 SourceLocation::UIntTy Offset : OffsetBits; 484 LLVM_PREFERRED_TYPE(bool) 485 SourceLocation::UIntTy IsExpansion : 1; 486 union { 487 FileInfo File; 488 ExpansionInfo Expansion; 489 }; 490 491 public: 492 SLocEntry() : Offset(), IsExpansion(), File() {} 493 494 SourceLocation::UIntTy getOffset() const { return Offset; } 495 496 bool isExpansion() const { return IsExpansion; } 497 bool isFile() const { return !isExpansion(); } 498 499 const FileInfo &getFile() const { 500 return const_cast<SLocEntry *>(this)->getFile(); 501 } 502 503 FileInfo &getFile() { 504 assert(isFile() && "Not a file SLocEntry!"); 505 return File; 506 } 507 508 const ExpansionInfo &getExpansion() const { 509 assert(isExpansion() && "Not a macro expansion SLocEntry!"); 510 return Expansion; 511 } 512 513 /// Creates an incomplete SLocEntry that is only able to report its offset. 514 static SLocEntry getOffsetOnly(SourceLocation::UIntTy Offset) { 515 assert(!(Offset & (1ULL << OffsetBits)) && "Offset is too large"); 516 SLocEntry E; 517 E.Offset = Offset; 518 return E; 519 } 520 521 static SLocEntry get(SourceLocation::UIntTy Offset, const FileInfo &FI) { 522 assert(!(Offset & (1ULL << OffsetBits)) && "Offset is too large"); 523 SLocEntry E; 524 E.Offset = Offset; 525 E.IsExpansion = false; 526 E.File = FI; 527 return E; 528 } 529 530 static SLocEntry get(SourceLocation::UIntTy Offset, 531 const ExpansionInfo &Expansion) { 532 assert(!(Offset & (1ULL << OffsetBits)) && "Offset is too large"); 533 SLocEntry E; 534 E.Offset = Offset; 535 E.IsExpansion = true; 536 new (&E.Expansion) ExpansionInfo(Expansion); 537 return E; 538 } 539 }; 540 541 } // namespace SrcMgr 542 543 /// External source of source location entries. 544 class ExternalSLocEntrySource { 545 public: 546 virtual ~ExternalSLocEntrySource(); 547 548 /// Read the source location entry with index ID, which will always be 549 /// less than -1. 550 /// 551 /// \returns true if an error occurred that prevented the source-location 552 /// entry from being loaded. 553 virtual bool ReadSLocEntry(int ID) = 0; 554 555 /// Get the index ID for the loaded SourceLocation offset. 556 /// 557 /// \returns Invalid index ID (0) if an error occurred that prevented the 558 /// SLocEntry from being loaded. 559 virtual int getSLocEntryID(SourceLocation::UIntTy SLocOffset) = 0; 560 561 /// Retrieve the module import location and name for the given ID, if 562 /// in fact it was loaded from a module (rather than, say, a precompiled 563 /// header). 564 virtual std::pair<SourceLocation, StringRef> getModuleImportLoc(int ID) = 0; 565 }; 566 567 /// Holds the cache used by isBeforeInTranslationUnit. 568 /// 569 /// The cache structure is complex enough to be worth breaking out of 570 /// SourceManager. 571 class InBeforeInTUCacheEntry { 572 /// The FileID's of the cached query. 573 /// 574 /// If these match up with a subsequent query, the result can be reused. 575 FileID LQueryFID, RQueryFID; 576 577 /// The relative order of FileIDs that the CommonFID *immediately* includes. 578 /// 579 /// This is used to compare macro expansion locations. 580 bool LChildBeforeRChild; 581 582 /// The file found in common between the two \#include traces, i.e., 583 /// the nearest common ancestor of the \#include tree. 584 FileID CommonFID; 585 586 /// The offset of the previous query in CommonFID. 587 /// 588 /// Usually, this represents the location of the \#include for QueryFID, but 589 /// if LQueryFID is a parent of RQueryFID (or vice versa) then these can be a 590 /// random token in the parent. 591 unsigned LCommonOffset, RCommonOffset; 592 593 public: 594 InBeforeInTUCacheEntry() = default; 595 InBeforeInTUCacheEntry(FileID L, FileID R) : LQueryFID(L), RQueryFID(R) { 596 assert(L != R); 597 } 598 599 /// Return true if the currently cached values match up with 600 /// the specified LHS/RHS query. 601 /// 602 /// If not, we can't use the cache. 603 bool isCacheValid() const { 604 return CommonFID.isValid(); 605 } 606 607 /// If the cache is valid, compute the result given the 608 /// specified offsets in the LHS/RHS FileID's. 609 bool getCachedResult(unsigned LOffset, unsigned ROffset) const { 610 // If one of the query files is the common file, use the offset. Otherwise, 611 // use the #include loc in the common file. 612 if (LQueryFID != CommonFID) LOffset = LCommonOffset; 613 if (RQueryFID != CommonFID) ROffset = RCommonOffset; 614 615 // It is common for multiple macro expansions to be "included" from the same 616 // location (expansion location), in which case use the order of the FileIDs 617 // to determine which came first. This will also take care the case where 618 // one of the locations points at the inclusion/expansion point of the other 619 // in which case its FileID will come before the other. 620 if (LOffset == ROffset) 621 return LChildBeforeRChild; 622 623 return LOffset < ROffset; 624 } 625 626 /// Set up a new query. 627 /// If it matches the old query, we can keep the cached answer. 628 void setQueryFIDs(FileID LHS, FileID RHS) { 629 assert(LHS != RHS); 630 if (LQueryFID != LHS || RQueryFID != RHS) { 631 LQueryFID = LHS; 632 RQueryFID = RHS; 633 CommonFID = FileID(); 634 } 635 } 636 637 void setCommonLoc(FileID commonFID, unsigned lCommonOffset, 638 unsigned rCommonOffset, bool LParentBeforeRParent) { 639 CommonFID = commonFID; 640 LCommonOffset = lCommonOffset; 641 RCommonOffset = rCommonOffset; 642 LChildBeforeRChild = LParentBeforeRParent; 643 } 644 }; 645 646 /// The stack used when building modules on demand, which is used 647 /// to provide a link between the source managers of the different compiler 648 /// instances. 649 using ModuleBuildStack = ArrayRef<std::pair<std::string, FullSourceLoc>>; 650 651 /// This class handles loading and caching of source files into memory. 652 /// 653 /// This object owns the MemoryBuffer objects for all of the loaded 654 /// files and assigns unique FileID's for each unique \#include chain. 655 /// 656 /// The SourceManager can be queried for information about SourceLocation 657 /// objects, turning them into either spelling or expansion locations. Spelling 658 /// locations represent where the bytes corresponding to a token came from and 659 /// expansion locations represent where the location is in the user's view. In 660 /// the case of a macro expansion, for example, the spelling location indicates 661 /// where the expanded token came from and the expansion location specifies 662 /// where it was expanded. 663 class SourceManager : public RefCountedBase<SourceManager> { 664 /// DiagnosticsEngine object. 665 DiagnosticsEngine &Diag; 666 667 FileManager &FileMgr; 668 669 mutable llvm::BumpPtrAllocator ContentCacheAlloc; 670 671 /// Memoized information about all of the files tracked by this 672 /// SourceManager. 673 /// 674 /// This map allows us to merge ContentCache entries based 675 /// on their FileEntry*. All ContentCache objects will thus have unique, 676 /// non-null, FileEntry pointers. 677 llvm::DenseMap<FileEntryRef, SrcMgr::ContentCache*> FileInfos; 678 679 /// True if the ContentCache for files that are overridden by other 680 /// files, should report the original file name. Defaults to true. 681 bool OverridenFilesKeepOriginalName = true; 682 683 /// True if non-system source files should be treated as volatile 684 /// (likely to change while trying to use them). Defaults to false. 685 bool UserFilesAreVolatile; 686 687 /// True if all files read during this compilation should be treated 688 /// as transient (may not be present in later compilations using a module 689 /// file created from this compilation). Defaults to false. 690 bool FilesAreTransient = false; 691 692 struct OverriddenFilesInfoTy { 693 /// Files that have been overridden with the contents from another 694 /// file. 695 llvm::DenseMap<const FileEntry *, FileEntryRef> OverriddenFiles; 696 697 /// Files that were overridden with a memory buffer. 698 llvm::DenseSet<const FileEntry *> OverriddenFilesWithBuffer; 699 }; 700 701 /// Lazily create the object keeping overridden files info, since 702 /// it is uncommonly used. 703 std::unique_ptr<OverriddenFilesInfoTy> OverriddenFilesInfo; 704 705 OverriddenFilesInfoTy &getOverriddenFilesInfo() { 706 if (!OverriddenFilesInfo) 707 OverriddenFilesInfo.reset(new OverriddenFilesInfoTy); 708 return *OverriddenFilesInfo; 709 } 710 711 /// Information about various memory buffers that we have read in. 712 /// 713 /// All FileEntry* within the stored ContentCache objects are NULL, 714 /// as they do not refer to a file. 715 std::vector<SrcMgr::ContentCache*> MemBufferInfos; 716 717 /// The table of SLocEntries that are local to this module. 718 /// 719 /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid 720 /// expansion. 721 SmallVector<SrcMgr::SLocEntry, 0> LocalSLocEntryTable; 722 723 /// The table of SLocEntries that are loaded from other modules. 724 /// 725 /// Negative FileIDs are indexes into this table. To get from ID to an index, 726 /// use (-ID - 2). 727 llvm::PagedVector<SrcMgr::SLocEntry, 32> LoadedSLocEntryTable; 728 729 /// For each allocation in LoadedSLocEntryTable, we keep the first FileID. 730 /// We assume exactly one allocation per AST file, and use that to determine 731 /// whether two FileIDs come from the same AST file. 732 SmallVector<FileID, 0> LoadedSLocEntryAllocBegin; 733 734 /// The starting offset of the next local SLocEntry. 735 /// 736 /// This is LocalSLocEntryTable.back().Offset + the size of that entry. 737 SourceLocation::UIntTy NextLocalOffset; 738 739 /// The starting offset of the latest batch of loaded SLocEntries. 740 /// 741 /// This is LoadedSLocEntryTable.back().Offset, except that that entry might 742 /// not have been loaded, so that value would be unknown. 743 SourceLocation::UIntTy CurrentLoadedOffset; 744 745 /// The highest possible offset is 2^31-1 (2^63-1 for 64-bit source 746 /// locations), so CurrentLoadedOffset starts at 2^31 (2^63 resp.). 747 static const SourceLocation::UIntTy MaxLoadedOffset = 748 1ULL << (8 * sizeof(SourceLocation::UIntTy) - 1); 749 750 /// A bitmap that indicates whether the entries of LoadedSLocEntryTable 751 /// have already been loaded from the external source. 752 /// 753 /// Same indexing as LoadedSLocEntryTable. 754 llvm::BitVector SLocEntryLoaded; 755 756 /// A bitmap that indicates whether the entries of LoadedSLocEntryTable 757 /// have already had their offset loaded from the external source. 758 /// 759 /// Superset of SLocEntryLoaded. Same indexing as SLocEntryLoaded. 760 llvm::BitVector SLocEntryOffsetLoaded; 761 762 /// An external source for source location entries. 763 ExternalSLocEntrySource *ExternalSLocEntries = nullptr; 764 765 /// A one-entry cache to speed up getFileID. 766 /// 767 /// LastFileIDLookup records the last FileID looked up or created, because it 768 /// is very common to look up many tokens from the same file. 769 mutable FileID LastFileIDLookup; 770 771 /// Holds information for \#line directives. 772 /// 773 /// This is referenced by indices from SLocEntryTable. 774 std::unique_ptr<LineTableInfo> LineTable; 775 776 /// These ivars serve as a cache used in the getLineNumber 777 /// method which is used to speedup getLineNumber calls to nearby locations. 778 mutable FileID LastLineNoFileIDQuery; 779 mutable const SrcMgr::ContentCache *LastLineNoContentCache; 780 mutable unsigned LastLineNoFilePos; 781 mutable unsigned LastLineNoResult; 782 783 /// The file ID for the main source file of the translation unit. 784 FileID MainFileID; 785 786 /// The file ID for the precompiled preamble there is one. 787 FileID PreambleFileID; 788 789 // Statistics for -print-stats. 790 mutable unsigned NumLinearScans = 0; 791 mutable unsigned NumBinaryProbes = 0; 792 793 /// Associates a FileID with its "included/expanded in" decomposed 794 /// location. 795 /// 796 /// Used to cache results from and speed-up \c getDecomposedIncludedLoc 797 /// function. 798 mutable llvm::DenseMap<FileID, std::pair<FileID, unsigned>> IncludedLocMap; 799 800 /// The key value into the IsBeforeInTUCache table. 801 using IsBeforeInTUCacheKey = std::pair<FileID, FileID>; 802 803 /// The IsBeforeInTranslationUnitCache is a mapping from FileID pairs 804 /// to cache results. 805 using InBeforeInTUCache = 806 llvm::DenseMap<IsBeforeInTUCacheKey, InBeforeInTUCacheEntry>; 807 808 /// Cache results for the isBeforeInTranslationUnit method. 809 mutable InBeforeInTUCache IBTUCache; 810 mutable InBeforeInTUCacheEntry IBTUCacheOverflow; 811 812 /// Return the cache entry for comparing the given file IDs 813 /// for isBeforeInTranslationUnit. 814 InBeforeInTUCacheEntry &getInBeforeInTUCache(FileID LFID, FileID RFID) const; 815 816 // Cache for the "fake" buffer used for error-recovery purposes. 817 mutable std::unique_ptr<llvm::MemoryBuffer> FakeBufferForRecovery; 818 819 mutable std::unique_ptr<SrcMgr::ContentCache> FakeContentCacheForRecovery; 820 821 mutable std::unique_ptr<SrcMgr::SLocEntry> FakeSLocEntryForRecovery; 822 823 /// Lazily computed map of macro argument chunks to their expanded 824 /// source location. 825 using MacroArgsMap = std::map<unsigned, SourceLocation>; 826 827 mutable llvm::DenseMap<FileID, std::unique_ptr<MacroArgsMap>> 828 MacroArgsCacheMap; 829 830 /// The stack of modules being built, which is used to detect 831 /// cycles in the module dependency graph as modules are being built, as 832 /// well as to describe why we're rebuilding a particular module. 833 /// 834 /// There is no way to set this value from the command line. If we ever need 835 /// to do so (e.g., if on-demand module construction moves out-of-process), 836 /// we can add a cc1-level option to do so. 837 SmallVector<std::pair<std::string, FullSourceLoc>, 2> StoredModuleBuildStack; 838 839 public: 840 SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr, 841 bool UserFilesAreVolatile = false); 842 explicit SourceManager(const SourceManager &) = delete; 843 SourceManager &operator=(const SourceManager &) = delete; 844 ~SourceManager(); 845 846 void clearIDTables(); 847 848 /// Initialize this source manager suitably to replay the compilation 849 /// described by \p Old. Requires that \p Old outlive \p *this. 850 void initializeForReplay(const SourceManager &Old); 851 852 DiagnosticsEngine &getDiagnostics() const { return Diag; } 853 854 FileManager &getFileManager() const { return FileMgr; } 855 856 /// Set true if the SourceManager should report the original file name 857 /// for contents of files that were overridden by other files. Defaults to 858 /// true. 859 void setOverridenFilesKeepOriginalName(bool value) { 860 OverridenFilesKeepOriginalName = value; 861 } 862 863 /// True if non-system source files should be treated as volatile 864 /// (likely to change while trying to use them). 865 bool userFilesAreVolatile() const { return UserFilesAreVolatile; } 866 867 /// Retrieve the module build stack. 868 ModuleBuildStack getModuleBuildStack() const { 869 return StoredModuleBuildStack; 870 } 871 872 /// Set the module build stack. 873 void setModuleBuildStack(ModuleBuildStack stack) { 874 StoredModuleBuildStack.clear(); 875 StoredModuleBuildStack.append(stack.begin(), stack.end()); 876 } 877 878 /// Push an entry to the module build stack. 879 void pushModuleBuildStack(StringRef moduleName, FullSourceLoc importLoc) { 880 StoredModuleBuildStack.push_back(std::make_pair(moduleName.str(),importLoc)); 881 } 882 883 //===--------------------------------------------------------------------===// 884 // MainFileID creation and querying methods. 885 //===--------------------------------------------------------------------===// 886 887 /// Returns the FileID of the main source file. 888 FileID getMainFileID() const { return MainFileID; } 889 890 /// Set the file ID for the main source file. 891 void setMainFileID(FileID FID) { 892 MainFileID = FID; 893 } 894 895 /// Returns true when the given FileEntry corresponds to the main file. 896 /// 897 /// The main file should be set prior to calling this function. 898 bool isMainFile(const FileEntry &SourceFile); 899 900 /// Set the file ID for the precompiled preamble. 901 void setPreambleFileID(FileID Preamble) { 902 assert(PreambleFileID.isInvalid() && "PreambleFileID already set!"); 903 PreambleFileID = Preamble; 904 } 905 906 /// Get the file ID for the precompiled preamble if there is one. 907 FileID getPreambleFileID() const { return PreambleFileID; } 908 909 //===--------------------------------------------------------------------===// 910 // Methods to create new FileID's and macro expansions. 911 //===--------------------------------------------------------------------===// 912 913 /// Create a new FileID that represents the specified file 914 /// being \#included from the specified IncludePosition. 915 FileID createFileID(FileEntryRef SourceFile, SourceLocation IncludePos, 916 SrcMgr::CharacteristicKind FileCharacter, 917 int LoadedID = 0, 918 SourceLocation::UIntTy LoadedOffset = 0); 919 920 /// Create a new FileID that represents the specified memory buffer. 921 /// 922 /// This does no caching of the buffer and takes ownership of the 923 /// MemoryBuffer, so only pass a MemoryBuffer to this once. 924 FileID createFileID(std::unique_ptr<llvm::MemoryBuffer> Buffer, 925 SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User, 926 int LoadedID = 0, SourceLocation::UIntTy LoadedOffset = 0, 927 SourceLocation IncludeLoc = SourceLocation()); 928 929 /// Create a new FileID that represents the specified memory buffer. 930 /// 931 /// This does not take ownership of the MemoryBuffer. The memory buffer must 932 /// outlive the SourceManager. 933 FileID createFileID(const llvm::MemoryBufferRef &Buffer, 934 SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User, 935 int LoadedID = 0, SourceLocation::UIntTy LoadedOffset = 0, 936 SourceLocation IncludeLoc = SourceLocation()); 937 938 /// Get the FileID for \p SourceFile if it exists. Otherwise, create a 939 /// new FileID for the \p SourceFile. 940 FileID getOrCreateFileID(FileEntryRef SourceFile, 941 SrcMgr::CharacteristicKind FileCharacter); 942 943 /// Creates an expansion SLocEntry for the substitution of an argument into a 944 /// function-like macro's body. Returns the start of the expansion. 945 /// 946 /// The macro argument was written at \p SpellingLoc with length \p Length. 947 /// \p ExpansionLoc is the parameter name in the (expanded) macro body. 948 SourceLocation createMacroArgExpansionLoc(SourceLocation SpellingLoc, 949 SourceLocation ExpansionLoc, 950 unsigned Length); 951 952 /// Creates an expansion SLocEntry for a macro use. Returns its start. 953 /// 954 /// The macro body begins at \p SpellingLoc with length \p Length. 955 /// The macro use spans [ExpansionLocStart, ExpansionLocEnd]. 956 SourceLocation createExpansionLoc(SourceLocation SpellingLoc, 957 SourceLocation ExpansionLocStart, 958 SourceLocation ExpansionLocEnd, 959 unsigned Length, 960 bool ExpansionIsTokenRange = true, 961 int LoadedID = 0, 962 SourceLocation::UIntTy LoadedOffset = 0); 963 964 /// Return a new SourceLocation that encodes that the token starting 965 /// at \p TokenStart ends prematurely at \p TokenEnd. 966 SourceLocation createTokenSplitLoc(SourceLocation SpellingLoc, 967 SourceLocation TokenStart, 968 SourceLocation TokenEnd); 969 970 /// Retrieve the memory buffer associated with the given file. 971 /// 972 /// Returns std::nullopt if the buffer is not valid. 973 std::optional<llvm::MemoryBufferRef> 974 getMemoryBufferForFileOrNone(FileEntryRef File); 975 976 /// Retrieve the memory buffer associated with the given file. 977 /// 978 /// Returns a fake buffer if there isn't a real one. 979 llvm::MemoryBufferRef getMemoryBufferForFileOrFake(FileEntryRef File) { 980 if (auto B = getMemoryBufferForFileOrNone(File)) 981 return *B; 982 return getFakeBufferForRecovery(); 983 } 984 985 /// Override the contents of the given source file by providing an 986 /// already-allocated buffer. 987 /// 988 /// \param SourceFile the source file whose contents will be overridden. 989 /// 990 /// \param Buffer the memory buffer whose contents will be used as the 991 /// data in the given source file. 992 void overrideFileContents(FileEntryRef SourceFile, 993 const llvm::MemoryBufferRef &Buffer) { 994 overrideFileContents(SourceFile, llvm::MemoryBuffer::getMemBuffer(Buffer)); 995 } 996 997 /// Override the contents of the given source file by providing an 998 /// already-allocated buffer. 999 /// 1000 /// \param SourceFile the source file whose contents will be overridden. 1001 /// 1002 /// \param Buffer the memory buffer whose contents will be used as the 1003 /// data in the given source file. 1004 void overrideFileContents(FileEntryRef SourceFile, 1005 std::unique_ptr<llvm::MemoryBuffer> Buffer); 1006 1007 /// Override the given source file with another one. 1008 /// 1009 /// \param SourceFile the source file which will be overridden. 1010 /// 1011 /// \param NewFile the file whose contents will be used as the 1012 /// data instead of the contents of the given source file. 1013 void overrideFileContents(const FileEntry *SourceFile, FileEntryRef NewFile); 1014 1015 /// Returns true if the file contents have been overridden. 1016 bool isFileOverridden(const FileEntry *File) const { 1017 if (OverriddenFilesInfo) { 1018 if (OverriddenFilesInfo->OverriddenFilesWithBuffer.count(File)) 1019 return true; 1020 if (OverriddenFilesInfo->OverriddenFiles.contains(File)) 1021 return true; 1022 } 1023 return false; 1024 } 1025 1026 /// Bypass the overridden contents of a file. This creates a new FileEntry 1027 /// and initializes the content cache for it. Returns std::nullopt if there 1028 /// is no such file in the filesystem. 1029 /// 1030 /// This should be called before parsing has begun. 1031 OptionalFileEntryRef bypassFileContentsOverride(FileEntryRef File); 1032 1033 /// Specify that a file is transient. 1034 void setFileIsTransient(FileEntryRef SourceFile); 1035 1036 /// Specify that all files that are read during this compilation are 1037 /// transient. 1038 void setAllFilesAreTransient(bool Transient) { 1039 FilesAreTransient = Transient; 1040 } 1041 1042 //===--------------------------------------------------------------------===// 1043 // FileID manipulation methods. 1044 //===--------------------------------------------------------------------===// 1045 1046 /// Return the buffer for the specified FileID. 1047 /// 1048 /// If there is an error opening this buffer the first time, return 1049 /// std::nullopt. 1050 std::optional<llvm::MemoryBufferRef> 1051 getBufferOrNone(FileID FID, SourceLocation Loc = SourceLocation()) const { 1052 if (auto *Entry = getSLocEntryForFile(FID)) 1053 return Entry->getFile().getContentCache().getBufferOrNone( 1054 Diag, getFileManager(), Loc); 1055 return std::nullopt; 1056 } 1057 1058 /// Return the buffer for the specified FileID. 1059 /// 1060 /// If there is an error opening this buffer the first time, this 1061 /// manufactures a temporary buffer and returns it. 1062 llvm::MemoryBufferRef 1063 getBufferOrFake(FileID FID, SourceLocation Loc = SourceLocation()) const { 1064 if (auto B = getBufferOrNone(FID, Loc)) 1065 return *B; 1066 return getFakeBufferForRecovery(); 1067 } 1068 1069 /// Returns the FileEntry record for the provided FileID. 1070 const FileEntry *getFileEntryForID(FileID FID) const { 1071 if (auto FE = getFileEntryRefForID(FID)) 1072 return *FE; 1073 return nullptr; 1074 } 1075 1076 /// Returns the FileEntryRef for the provided FileID. 1077 OptionalFileEntryRef getFileEntryRefForID(FileID FID) const { 1078 if (auto *Entry = getSLocEntryForFile(FID)) 1079 return Entry->getFile().getContentCache().OrigEntry; 1080 return std::nullopt; 1081 } 1082 1083 /// Returns the filename for the provided FileID, unless it's a built-in 1084 /// buffer that's not represented by a filename. 1085 /// 1086 /// Returns std::nullopt for non-files and built-in files. 1087 std::optional<StringRef> getNonBuiltinFilenameForID(FileID FID) const; 1088 1089 /// Returns the FileEntry record for the provided SLocEntry. 1090 const FileEntry * 1091 getFileEntryForSLocEntry(const SrcMgr::SLocEntry &SLocEntry) const { 1092 if (auto FE = SLocEntry.getFile().getContentCache().OrigEntry) 1093 return *FE; 1094 return nullptr; 1095 } 1096 1097 /// Return a StringRef to the source buffer data for the 1098 /// specified FileID. 1099 /// 1100 /// \param FID The file ID whose contents will be returned. 1101 /// \param Invalid If non-NULL, will be set true if an error occurred. 1102 StringRef getBufferData(FileID FID, bool *Invalid = nullptr) const; 1103 1104 /// Return a StringRef to the source buffer data for the 1105 /// specified FileID, returning std::nullopt if invalid. 1106 /// 1107 /// \param FID The file ID whose contents will be returned. 1108 std::optional<StringRef> getBufferDataOrNone(FileID FID) const; 1109 1110 /// Return a StringRef to the source buffer data for the 1111 /// specified FileID, returning std::nullopt if it's not yet loaded. 1112 /// 1113 /// \param FID The file ID whose contents will be returned. 1114 std::optional<StringRef> getBufferDataIfLoaded(FileID FID) const; 1115 1116 /// Get the number of FileIDs (files and macros) that were created 1117 /// during preprocessing of \p FID, including it. 1118 unsigned getNumCreatedFIDsForFileID(FileID FID) const { 1119 if (auto *Entry = getSLocEntryForFile(FID)) 1120 return Entry->getFile().NumCreatedFIDs; 1121 return 0; 1122 } 1123 1124 /// Set the number of FileIDs (files and macros) that were created 1125 /// during preprocessing of \p FID, including it. 1126 void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs, 1127 bool Force = false) { 1128 auto *Entry = getSLocEntryForFile(FID); 1129 if (!Entry) 1130 return; 1131 assert((Force || Entry->getFile().NumCreatedFIDs == 0) && "Already set!"); 1132 Entry->getFile().NumCreatedFIDs = NumFIDs; 1133 } 1134 1135 //===--------------------------------------------------------------------===// 1136 // SourceLocation manipulation methods. 1137 //===--------------------------------------------------------------------===// 1138 1139 /// Return the FileID for a SourceLocation. 1140 /// 1141 /// This is a very hot method that is used for all SourceManager queries 1142 /// that start with a SourceLocation object. It is responsible for finding 1143 /// the entry in SLocEntryTable which contains the specified location. 1144 /// 1145 FileID getFileID(SourceLocation SpellingLoc) const { 1146 return getFileID(SpellingLoc.getOffset()); 1147 } 1148 1149 /// Return the filename of the file containing a SourceLocation. 1150 StringRef getFilename(SourceLocation SpellingLoc) const; 1151 1152 /// Return the source location corresponding to the first byte of 1153 /// the specified file. 1154 SourceLocation getLocForStartOfFile(FileID FID) const { 1155 if (auto *Entry = getSLocEntryForFile(FID)) 1156 return SourceLocation::getFileLoc(Entry->getOffset()); 1157 return SourceLocation(); 1158 } 1159 1160 /// Return the source location corresponding to the last byte of the 1161 /// specified file. 1162 SourceLocation getLocForEndOfFile(FileID FID) const { 1163 if (auto *Entry = getSLocEntryForFile(FID)) 1164 return SourceLocation::getFileLoc(Entry->getOffset() + 1165 getFileIDSize(FID)); 1166 return SourceLocation(); 1167 } 1168 1169 /// Returns the include location if \p FID is a \#include'd file 1170 /// otherwise it returns an invalid location. 1171 SourceLocation getIncludeLoc(FileID FID) const { 1172 if (auto *Entry = getSLocEntryForFile(FID)) 1173 return Entry->getFile().getIncludeLoc(); 1174 return SourceLocation(); 1175 } 1176 1177 // Returns the import location if the given source location is 1178 // located within a module, or an invalid location if the source location 1179 // is within the current translation unit. 1180 std::pair<SourceLocation, StringRef> 1181 getModuleImportLoc(SourceLocation Loc) const { 1182 FileID FID = getFileID(Loc); 1183 1184 // Positive file IDs are in the current translation unit, and -1 is a 1185 // placeholder. 1186 if (FID.ID >= -1) 1187 return std::make_pair(SourceLocation(), ""); 1188 1189 return ExternalSLocEntries->getModuleImportLoc(FID.ID); 1190 } 1191 1192 /// Given a SourceLocation object \p Loc, return the expansion 1193 /// location referenced by the ID. 1194 SourceLocation getExpansionLoc(SourceLocation Loc) const { 1195 // Handle the non-mapped case inline, defer to out of line code to handle 1196 // expansions. 1197 if (Loc.isFileID()) return Loc; 1198 return getExpansionLocSlowCase(Loc); 1199 } 1200 1201 /// Given \p Loc, if it is a macro location return the expansion 1202 /// location or the spelling location, depending on if it comes from a 1203 /// macro argument or not. 1204 SourceLocation getFileLoc(SourceLocation Loc) const { 1205 if (Loc.isFileID()) return Loc; 1206 return getFileLocSlowCase(Loc); 1207 } 1208 1209 /// Return the start/end of the expansion information for an 1210 /// expansion location. 1211 /// 1212 /// \pre \p Loc is required to be an expansion location. 1213 CharSourceRange getImmediateExpansionRange(SourceLocation Loc) const; 1214 1215 /// Given a SourceLocation object, return the range of 1216 /// tokens covered by the expansion in the ultimate file. 1217 CharSourceRange getExpansionRange(SourceLocation Loc) const; 1218 1219 /// Given a SourceRange object, return the range of 1220 /// tokens or characters covered by the expansion in the ultimate file. 1221 CharSourceRange getExpansionRange(SourceRange Range) const { 1222 SourceLocation Begin = getExpansionRange(Range.getBegin()).getBegin(); 1223 CharSourceRange End = getExpansionRange(Range.getEnd()); 1224 return CharSourceRange(SourceRange(Begin, End.getEnd()), 1225 End.isTokenRange()); 1226 } 1227 1228 /// Given a CharSourceRange object, return the range of 1229 /// tokens or characters covered by the expansion in the ultimate file. 1230 CharSourceRange getExpansionRange(CharSourceRange Range) const { 1231 CharSourceRange Expansion = getExpansionRange(Range.getAsRange()); 1232 if (Expansion.getEnd() == Range.getEnd()) 1233 Expansion.setTokenRange(Range.isTokenRange()); 1234 return Expansion; 1235 } 1236 1237 /// Given a SourceLocation object, return the spelling 1238 /// location referenced by the ID. 1239 /// 1240 /// This is the place where the characters that make up the lexed token 1241 /// can be found. 1242 SourceLocation getSpellingLoc(SourceLocation Loc) const { 1243 // Handle the non-mapped case inline, defer to out of line code to handle 1244 // expansions. 1245 if (Loc.isFileID()) return Loc; 1246 return getSpellingLocSlowCase(Loc); 1247 } 1248 1249 /// Given a SourceLocation object, return the spelling location 1250 /// referenced by the ID. 1251 /// 1252 /// This is the first level down towards the place where the characters 1253 /// that make up the lexed token can be found. This should not generally 1254 /// be used by clients. 1255 SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const; 1256 1257 /// Form a SourceLocation from a FileID and Offset pair. 1258 SourceLocation getComposedLoc(FileID FID, unsigned Offset) const { 1259 auto *Entry = getSLocEntryOrNull(FID); 1260 if (!Entry) 1261 return SourceLocation(); 1262 1263 SourceLocation::UIntTy GlobalOffset = Entry->getOffset() + Offset; 1264 return Entry->isFile() ? SourceLocation::getFileLoc(GlobalOffset) 1265 : SourceLocation::getMacroLoc(GlobalOffset); 1266 } 1267 1268 /// Decompose the specified location into a raw FileID + Offset pair. 1269 /// 1270 /// The first element is the FileID, the second is the offset from the 1271 /// start of the buffer of the location. 1272 std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const { 1273 FileID FID = getFileID(Loc); 1274 auto *Entry = getSLocEntryOrNull(FID); 1275 if (!Entry) 1276 return std::make_pair(FileID(), 0); 1277 return std::make_pair(FID, Loc.getOffset() - Entry->getOffset()); 1278 } 1279 1280 /// Decompose the specified location into a raw FileID + Offset pair. 1281 /// 1282 /// If the location is an expansion record, walk through it until we find 1283 /// the final location expanded. 1284 std::pair<FileID, unsigned> 1285 getDecomposedExpansionLoc(SourceLocation Loc) const { 1286 FileID FID = getFileID(Loc); 1287 auto *E = getSLocEntryOrNull(FID); 1288 if (!E) 1289 return std::make_pair(FileID(), 0); 1290 1291 unsigned Offset = Loc.getOffset()-E->getOffset(); 1292 if (Loc.isFileID()) 1293 return std::make_pair(FID, Offset); 1294 1295 return getDecomposedExpansionLocSlowCase(E); 1296 } 1297 1298 /// Decompose the specified location into a raw FileID + Offset pair. 1299 /// 1300 /// If the location is an expansion record, walk through it until we find 1301 /// its spelling record. 1302 std::pair<FileID, unsigned> 1303 getDecomposedSpellingLoc(SourceLocation Loc) const { 1304 FileID FID = getFileID(Loc); 1305 auto *E = getSLocEntryOrNull(FID); 1306 if (!E) 1307 return std::make_pair(FileID(), 0); 1308 1309 unsigned Offset = Loc.getOffset()-E->getOffset(); 1310 if (Loc.isFileID()) 1311 return std::make_pair(FID, Offset); 1312 return getDecomposedSpellingLocSlowCase(E, Offset); 1313 } 1314 1315 /// Returns the "included/expanded in" decomposed location of the given 1316 /// FileID. 1317 std::pair<FileID, unsigned> getDecomposedIncludedLoc(FileID FID) const; 1318 1319 /// Returns the offset from the start of the file that the 1320 /// specified SourceLocation represents. 1321 /// 1322 /// This is not very meaningful for a macro ID. 1323 unsigned getFileOffset(SourceLocation SpellingLoc) const { 1324 return getDecomposedLoc(SpellingLoc).second; 1325 } 1326 1327 /// Tests whether the given source location represents a macro 1328 /// argument's expansion into the function-like macro definition. 1329 /// 1330 /// \param StartLoc If non-null and function returns true, it is set to the 1331 /// start location of the macro argument expansion. 1332 /// 1333 /// Such source locations only appear inside of the expansion 1334 /// locations representing where a particular function-like macro was 1335 /// expanded. 1336 bool isMacroArgExpansion(SourceLocation Loc, 1337 SourceLocation *StartLoc = nullptr) const; 1338 1339 /// Tests whether the given source location represents the expansion of 1340 /// a macro body. 1341 /// 1342 /// This is equivalent to testing whether the location is part of a macro 1343 /// expansion but not the expansion of an argument to a function-like macro. 1344 bool isMacroBodyExpansion(SourceLocation Loc) const; 1345 1346 /// Returns true if the given MacroID location points at the beginning 1347 /// of the immediate macro expansion. 1348 /// 1349 /// \param MacroBegin If non-null and function returns true, it is set to the 1350 /// begin location of the immediate macro expansion. 1351 bool isAtStartOfImmediateMacroExpansion(SourceLocation Loc, 1352 SourceLocation *MacroBegin = nullptr) const; 1353 1354 /// Returns true if the given MacroID location points at the character 1355 /// end of the immediate macro expansion. 1356 /// 1357 /// \param MacroEnd If non-null and function returns true, it is set to the 1358 /// character end location of the immediate macro expansion. 1359 bool 1360 isAtEndOfImmediateMacroExpansion(SourceLocation Loc, 1361 SourceLocation *MacroEnd = nullptr) const; 1362 1363 /// Returns true if \p Loc is inside the [\p Start, +\p Length) 1364 /// chunk of the source location address space. 1365 /// 1366 /// If it's true and \p RelativeOffset is non-null, it will be set to the 1367 /// relative offset of \p Loc inside the chunk. 1368 bool 1369 isInSLocAddrSpace(SourceLocation Loc, SourceLocation Start, unsigned Length, 1370 SourceLocation::UIntTy *RelativeOffset = nullptr) const { 1371 assert(((Start.getOffset() < NextLocalOffset && 1372 Start.getOffset()+Length <= NextLocalOffset) || 1373 (Start.getOffset() >= CurrentLoadedOffset && 1374 Start.getOffset()+Length < MaxLoadedOffset)) && 1375 "Chunk is not valid SLoc address space"); 1376 SourceLocation::UIntTy LocOffs = Loc.getOffset(); 1377 SourceLocation::UIntTy BeginOffs = Start.getOffset(); 1378 SourceLocation::UIntTy EndOffs = BeginOffs + Length; 1379 if (LocOffs >= BeginOffs && LocOffs < EndOffs) { 1380 if (RelativeOffset) 1381 *RelativeOffset = LocOffs - BeginOffs; 1382 return true; 1383 } 1384 1385 return false; 1386 } 1387 1388 /// Return true if both \p LHS and \p RHS are in the local source 1389 /// location address space or the loaded one. 1390 /// 1391 /// If it's true and \p RelativeOffset is non-null, it will be set to the 1392 /// offset of \p RHS relative to \p LHS. 1393 bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS, 1394 SourceLocation::IntTy *RelativeOffset) const { 1395 SourceLocation::UIntTy LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset(); 1396 bool LHSLoaded = LHSOffs >= CurrentLoadedOffset; 1397 bool RHSLoaded = RHSOffs >= CurrentLoadedOffset; 1398 1399 if (LHSLoaded == RHSLoaded) { 1400 if (RelativeOffset) 1401 *RelativeOffset = RHSOffs - LHSOffs; 1402 return true; 1403 } 1404 1405 return false; 1406 } 1407 1408 //===--------------------------------------------------------------------===// 1409 // Queries about the code at a SourceLocation. 1410 //===--------------------------------------------------------------------===// 1411 1412 /// Return a pointer to the start of the specified location 1413 /// in the appropriate spelling MemoryBuffer. 1414 /// 1415 /// \param Invalid If non-NULL, will be set \c true if an error occurs. 1416 const char *getCharacterData(SourceLocation SL, 1417 bool *Invalid = nullptr) const; 1418 1419 /// Return the column # for the specified file position. 1420 /// 1421 /// This is significantly cheaper to compute than the line number. This 1422 /// returns zero if the column number isn't known. This may only be called 1423 /// on a file sloc, so you must choose a spelling or expansion location 1424 /// before calling this method. 1425 unsigned getColumnNumber(FileID FID, unsigned FilePos, 1426 bool *Invalid = nullptr) const; 1427 unsigned getSpellingColumnNumber(SourceLocation Loc, 1428 bool *Invalid = nullptr) const; 1429 unsigned getExpansionColumnNumber(SourceLocation Loc, 1430 bool *Invalid = nullptr) const; 1431 unsigned getPresumedColumnNumber(SourceLocation Loc, 1432 bool *Invalid = nullptr) const; 1433 1434 /// Given a SourceLocation, return the spelling line number 1435 /// for the position indicated. 1436 /// 1437 /// This requires building and caching a table of line offsets for the 1438 /// MemoryBuffer, so this is not cheap: use only when about to emit a 1439 /// diagnostic. 1440 unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = nullptr) const; 1441 unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const; 1442 unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const; 1443 unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const; 1444 1445 /// Return the filename or buffer identifier of the buffer the 1446 /// location is in. 1447 /// 1448 /// Note that this name does not respect \#line directives. Use 1449 /// getPresumedLoc for normal clients. 1450 StringRef getBufferName(SourceLocation Loc, bool *Invalid = nullptr) const; 1451 1452 /// Return the file characteristic of the specified source 1453 /// location, indicating whether this is a normal file, a system 1454 /// header, or an "implicit extern C" system header. 1455 /// 1456 /// This state can be modified with flags on GNU linemarker directives like: 1457 /// \code 1458 /// # 4 "foo.h" 3 1459 /// \endcode 1460 /// which changes all source locations in the current file after that to be 1461 /// considered to be from a system header. 1462 SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const; 1463 1464 /// Returns the "presumed" location of a SourceLocation specifies. 1465 /// 1466 /// A "presumed location" can be modified by \#line or GNU line marker 1467 /// directives. This provides a view on the data that a user should see 1468 /// in diagnostics, for example. 1469 /// 1470 /// Note that a presumed location is always given as the expansion point of 1471 /// an expansion location, not at the spelling location. 1472 /// 1473 /// \returns The presumed location of the specified SourceLocation. If the 1474 /// presumed location cannot be calculated (e.g., because \p Loc is invalid 1475 /// or the file containing \p Loc has changed on disk), returns an invalid 1476 /// presumed location. 1477 PresumedLoc getPresumedLoc(SourceLocation Loc, 1478 bool UseLineDirectives = true) const; 1479 1480 /// Returns whether the PresumedLoc for a given SourceLocation is 1481 /// in the main file. 1482 /// 1483 /// This computes the "presumed" location for a SourceLocation, then checks 1484 /// whether it came from a file other than the main file. This is different 1485 /// from isWrittenInMainFile() because it takes line marker directives into 1486 /// account. 1487 bool isInMainFile(SourceLocation Loc) const; 1488 1489 /// Returns true if the spelling locations for both SourceLocations 1490 /// are part of the same file buffer. 1491 /// 1492 /// This check ignores line marker directives. 1493 bool isWrittenInSameFile(SourceLocation Loc1, SourceLocation Loc2) const { 1494 return getFileID(Loc1) == getFileID(Loc2); 1495 } 1496 1497 /// Returns true if the spelling location for the given location 1498 /// is in the main file buffer. 1499 /// 1500 /// This check ignores line marker directives. 1501 bool isWrittenInMainFile(SourceLocation Loc) const { 1502 return getFileID(Loc) == getMainFileID(); 1503 } 1504 1505 /// Returns whether \p Loc is located in a <built-in> file. 1506 bool isWrittenInBuiltinFile(SourceLocation Loc) const { 1507 PresumedLoc Presumed = getPresumedLoc(Loc); 1508 if (Presumed.isInvalid()) 1509 return false; 1510 StringRef Filename(Presumed.getFilename()); 1511 return Filename == "<built-in>"; 1512 } 1513 1514 /// Returns whether \p Loc is located in a <command line> file. 1515 bool isWrittenInCommandLineFile(SourceLocation Loc) const { 1516 PresumedLoc Presumed = getPresumedLoc(Loc); 1517 if (Presumed.isInvalid()) 1518 return false; 1519 StringRef Filename(Presumed.getFilename()); 1520 return Filename == "<command line>"; 1521 } 1522 1523 /// Returns whether \p Loc is located in a <scratch space> file. 1524 bool isWrittenInScratchSpace(SourceLocation Loc) const { 1525 PresumedLoc Presumed = getPresumedLoc(Loc); 1526 if (Presumed.isInvalid()) 1527 return false; 1528 StringRef Filename(Presumed.getFilename()); 1529 return Filename == "<scratch space>"; 1530 } 1531 1532 /// Returns if a SourceLocation is in a system header. 1533 bool isInSystemHeader(SourceLocation Loc) const { 1534 if (Loc.isInvalid()) 1535 return false; 1536 return isSystem(getFileCharacteristic(Loc)); 1537 } 1538 1539 /// Returns if a SourceLocation is in an "extern C" system header. 1540 bool isInExternCSystemHeader(SourceLocation Loc) const { 1541 return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem; 1542 } 1543 1544 /// Returns whether \p Loc is expanded from a macro in a system header. 1545 bool isInSystemMacro(SourceLocation loc) const { 1546 if (!loc.isMacroID()) 1547 return false; 1548 1549 // This happens when the macro is the result of a paste, in that case 1550 // its spelling is the scratch memory, so we take the parent context. 1551 // There can be several level of token pasting. 1552 if (isWrittenInScratchSpace(getSpellingLoc(loc))) { 1553 do { 1554 loc = getImmediateMacroCallerLoc(loc); 1555 } while (isWrittenInScratchSpace(getSpellingLoc(loc))); 1556 return isInSystemMacro(loc); 1557 } 1558 1559 return isInSystemHeader(getSpellingLoc(loc)); 1560 } 1561 1562 /// The size of the SLocEntry that \p FID represents. 1563 unsigned getFileIDSize(FileID FID) const; 1564 1565 /// Given a specific FileID, returns true if \p Loc is inside that 1566 /// FileID chunk and sets relative offset (offset of \p Loc from beginning 1567 /// of FileID) to \p relativeOffset. 1568 bool isInFileID(SourceLocation Loc, FileID FID, 1569 unsigned *RelativeOffset = nullptr) const { 1570 SourceLocation::UIntTy Offs = Loc.getOffset(); 1571 if (isOffsetInFileID(FID, Offs)) { 1572 if (RelativeOffset) 1573 *RelativeOffset = Offs - getSLocEntry(FID).getOffset(); 1574 return true; 1575 } 1576 1577 return false; 1578 } 1579 1580 //===--------------------------------------------------------------------===// 1581 // Line Table Manipulation Routines 1582 //===--------------------------------------------------------------------===// 1583 1584 /// Return the uniqued ID for the specified filename. 1585 unsigned getLineTableFilenameID(StringRef Str); 1586 1587 /// Add a line note to the line table for the FileID and offset 1588 /// specified by Loc. 1589 /// 1590 /// If FilenameID is -1, it is considered to be unspecified. 1591 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID, 1592 bool IsFileEntry, bool IsFileExit, 1593 SrcMgr::CharacteristicKind FileKind); 1594 1595 /// Determine if the source manager has a line table. 1596 bool hasLineTable() const { return LineTable != nullptr; } 1597 1598 /// Retrieve the stored line table. 1599 LineTableInfo &getLineTable(); 1600 1601 //===--------------------------------------------------------------------===// 1602 // Queries for performance analysis. 1603 //===--------------------------------------------------------------------===// 1604 1605 /// Return the total amount of physical memory allocated by the 1606 /// ContentCache allocator. 1607 size_t getContentCacheSize() const { 1608 return ContentCacheAlloc.getTotalMemory(); 1609 } 1610 1611 struct MemoryBufferSizes { 1612 const size_t malloc_bytes; 1613 const size_t mmap_bytes; 1614 1615 MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes) 1616 : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {} 1617 }; 1618 1619 /// Return the amount of memory used by memory buffers, breaking down 1620 /// by heap-backed versus mmap'ed memory. 1621 MemoryBufferSizes getMemoryBufferSizes() const; 1622 1623 /// Return the amount of memory used for various side tables and 1624 /// data structures in the SourceManager. 1625 size_t getDataStructureSizes() const; 1626 1627 //===--------------------------------------------------------------------===// 1628 // Other miscellaneous methods. 1629 //===--------------------------------------------------------------------===// 1630 1631 /// Get the source location for the given file:line:col triplet. 1632 /// 1633 /// If the source file is included multiple times, the source location will 1634 /// be based upon the first inclusion. 1635 SourceLocation translateFileLineCol(const FileEntry *SourceFile, 1636 unsigned Line, unsigned Col) const; 1637 1638 /// Get the FileID for the given file. 1639 /// 1640 /// If the source file is included multiple times, the FileID will be the 1641 /// first inclusion. 1642 FileID translateFile(const FileEntry *SourceFile) const; 1643 FileID translateFile(FileEntryRef SourceFile) const { 1644 return translateFile(&SourceFile.getFileEntry()); 1645 } 1646 1647 /// Get the source location in \p FID for the given line:col. 1648 /// Returns null location if \p FID is not a file SLocEntry. 1649 SourceLocation translateLineCol(FileID FID, 1650 unsigned Line, unsigned Col) const; 1651 1652 /// If \p Loc points inside a function macro argument, the returned 1653 /// location will be the macro location in which the argument was expanded. 1654 /// If a macro argument is used multiple times, the expanded location will 1655 /// be at the first expansion of the argument. 1656 /// e.g. 1657 /// MY_MACRO(foo); 1658 /// ^ 1659 /// Passing a file location pointing at 'foo', will yield a macro location 1660 /// where 'foo' was expanded into. 1661 SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const; 1662 1663 /// Determines the order of 2 source locations in the translation unit. 1664 /// 1665 /// \returns true if LHS source location comes before RHS, false otherwise. 1666 bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const; 1667 1668 /// Determines whether the two decomposed source location is in the 1669 /// same translation unit. As a byproduct, it also calculates the order 1670 /// of the source locations in case they are in the same TU. 1671 /// 1672 /// \returns Pair of bools the first component is true if the two locations 1673 /// are in the same TU. The second bool is true if the first is true 1674 /// and \p LOffs is before \p ROffs. 1675 std::pair<bool, bool> 1676 isInTheSameTranslationUnit(std::pair<FileID, unsigned> &LOffs, 1677 std::pair<FileID, unsigned> &ROffs) const; 1678 1679 /// \param Loc a source location in a loaded AST (of a PCH/Module file). 1680 /// \returns a FileID uniquely identifies the AST of a loaded 1681 /// module/PCH where `Loc` is at. 1682 FileID getUniqueLoadedASTFileID(SourceLocation Loc) const; 1683 1684 /// Determines whether the two decomposed source location is in the same TU. 1685 bool isInTheSameTranslationUnitImpl( 1686 const std::pair<FileID, unsigned> &LOffs, 1687 const std::pair<FileID, unsigned> &ROffs) const; 1688 1689 /// Determines the order of 2 source locations in the "source location 1690 /// address space". 1691 bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const { 1692 return isBeforeInSLocAddrSpace(LHS, RHS.getOffset()); 1693 } 1694 1695 /// Determines the order of a source location and a source location 1696 /// offset in the "source location address space". 1697 /// 1698 /// Note that we always consider source locations loaded from 1699 bool isBeforeInSLocAddrSpace(SourceLocation LHS, 1700 SourceLocation::UIntTy RHS) const { 1701 SourceLocation::UIntTy LHSOffset = LHS.getOffset(); 1702 bool LHSLoaded = LHSOffset >= CurrentLoadedOffset; 1703 bool RHSLoaded = RHS >= CurrentLoadedOffset; 1704 if (LHSLoaded == RHSLoaded) 1705 return LHSOffset < RHS; 1706 1707 return LHSLoaded; 1708 } 1709 1710 /// Return true if the Point is within Start and End. 1711 bool isPointWithin(SourceLocation Location, SourceLocation Start, 1712 SourceLocation End) const { 1713 return Location == Start || Location == End || 1714 (isBeforeInTranslationUnit(Start, Location) && 1715 isBeforeInTranslationUnit(Location, End)); 1716 } 1717 1718 // Iterators over FileInfos. 1719 using fileinfo_iterator = 1720 llvm::DenseMap<FileEntryRef, SrcMgr::ContentCache *>::const_iterator; 1721 1722 fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); } 1723 fileinfo_iterator fileinfo_end() const { return FileInfos.end(); } 1724 bool hasFileInfo(const FileEntry *File) const { 1725 return FileInfos.find_as(File) != FileInfos.end(); 1726 } 1727 1728 /// Print statistics to stderr. 1729 void PrintStats() const; 1730 1731 void dump() const; 1732 1733 // Produce notes describing the current source location address space usage. 1734 void noteSLocAddressSpaceUsage(DiagnosticsEngine &Diag, 1735 std::optional<unsigned> MaxNotes = 32) const; 1736 1737 /// Get the number of local SLocEntries we have. 1738 unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); } 1739 1740 /// Get a local SLocEntry. This is exposed for indexing. 1741 const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index) const { 1742 return const_cast<SourceManager *>(this)->getLocalSLocEntry(Index); 1743 } 1744 1745 /// Get a local SLocEntry. This is exposed for indexing. 1746 SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index) { 1747 assert(Index < LocalSLocEntryTable.size() && "Invalid index"); 1748 return LocalSLocEntryTable[Index]; 1749 } 1750 1751 /// Get the number of loaded SLocEntries we have. 1752 unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();} 1753 1754 /// Get a loaded SLocEntry. This is exposed for indexing. 1755 const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index, 1756 bool *Invalid = nullptr) const { 1757 return const_cast<SourceManager *>(this)->getLoadedSLocEntry(Index, 1758 Invalid); 1759 } 1760 1761 /// Get a loaded SLocEntry. This is exposed for indexing. 1762 SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index, 1763 bool *Invalid = nullptr) { 1764 assert(Index < LoadedSLocEntryTable.size() && "Invalid index"); 1765 if (SLocEntryLoaded[Index]) 1766 return LoadedSLocEntryTable[Index]; 1767 return loadSLocEntry(Index, Invalid); 1768 } 1769 1770 const SrcMgr::SLocEntry &getSLocEntry(FileID FID, 1771 bool *Invalid = nullptr) const { 1772 return const_cast<SourceManager *>(this)->getSLocEntry(FID, Invalid); 1773 } 1774 1775 SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = nullptr) { 1776 if (FID.ID == 0 || FID.ID == -1) { 1777 if (Invalid) *Invalid = true; 1778 return LocalSLocEntryTable[0]; 1779 } 1780 return getSLocEntryByID(FID.ID, Invalid); 1781 } 1782 1783 SourceLocation::UIntTy getNextLocalOffset() const { return NextLocalOffset; } 1784 1785 void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) { 1786 assert(LoadedSLocEntryTable.empty() && 1787 "Invalidating existing loaded entries"); 1788 ExternalSLocEntries = Source; 1789 } 1790 1791 /// Allocate a number of loaded SLocEntries, which will be actually 1792 /// loaded on demand from the external source. 1793 /// 1794 /// NumSLocEntries will be allocated, which occupy a total of TotalSize space 1795 /// in the global source view. The lowest ID and the base offset of the 1796 /// entries will be returned. 1797 std::pair<int, SourceLocation::UIntTy> 1798 AllocateLoadedSLocEntries(unsigned NumSLocEntries, 1799 SourceLocation::UIntTy TotalSize); 1800 1801 /// Returns true if \p Loc came from a PCH/Module. 1802 bool isLoadedSourceLocation(SourceLocation Loc) const { 1803 return isLoadedOffset(Loc.getOffset()); 1804 } 1805 1806 /// Returns true if \p Loc did not come from a PCH/Module. 1807 bool isLocalSourceLocation(SourceLocation Loc) const { 1808 return isLocalOffset(Loc.getOffset()); 1809 } 1810 1811 /// Returns true if \p FID came from a PCH/Module. 1812 bool isLoadedFileID(FileID FID) const { 1813 assert(FID.ID != -1 && "Using FileID sentinel value"); 1814 return FID.ID < 0; 1815 } 1816 1817 /// Returns true if \p FID did not come from a PCH/Module. 1818 bool isLocalFileID(FileID FID) const { 1819 return !isLoadedFileID(FID); 1820 } 1821 1822 /// Gets the location of the immediate macro caller, one level up the stack 1823 /// toward the initial macro typed into the source. 1824 SourceLocation getImmediateMacroCallerLoc(SourceLocation Loc) const { 1825 if (!Loc.isMacroID()) return Loc; 1826 1827 // When we have the location of (part of) an expanded parameter, its 1828 // spelling location points to the argument as expanded in the macro call, 1829 // and therefore is used to locate the macro caller. 1830 if (isMacroArgExpansion(Loc)) 1831 return getImmediateSpellingLoc(Loc); 1832 1833 // Otherwise, the caller of the macro is located where this macro is 1834 // expanded (while the spelling is part of the macro definition). 1835 return getImmediateExpansionRange(Loc).getBegin(); 1836 } 1837 1838 /// \return Location of the top-level macro caller. 1839 SourceLocation getTopMacroCallerLoc(SourceLocation Loc) const; 1840 1841 private: 1842 friend class ASTReader; 1843 friend class ASTWriter; 1844 1845 llvm::MemoryBufferRef getFakeBufferForRecovery() const; 1846 SrcMgr::ContentCache &getFakeContentCacheForRecovery() const; 1847 1848 const SrcMgr::SLocEntry &loadSLocEntry(unsigned Index, bool *Invalid) const; 1849 SrcMgr::SLocEntry &loadSLocEntry(unsigned Index, bool *Invalid); 1850 1851 const SrcMgr::SLocEntry *getSLocEntryOrNull(FileID FID) const { 1852 return const_cast<SourceManager *>(this)->getSLocEntryOrNull(FID); 1853 } 1854 1855 SrcMgr::SLocEntry *getSLocEntryOrNull(FileID FID) { 1856 bool Invalid = false; 1857 SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid); 1858 return Invalid ? nullptr : &Entry; 1859 } 1860 1861 const SrcMgr::SLocEntry *getSLocEntryForFile(FileID FID) const { 1862 return const_cast<SourceManager *>(this)->getSLocEntryForFile(FID); 1863 } 1864 1865 SrcMgr::SLocEntry *getSLocEntryForFile(FileID FID) { 1866 if (auto *Entry = getSLocEntryOrNull(FID)) 1867 if (Entry->isFile()) 1868 return Entry; 1869 return nullptr; 1870 } 1871 1872 /// Get the entry with the given unwrapped FileID. 1873 /// Invalid will not be modified for Local IDs. 1874 const SrcMgr::SLocEntry &getSLocEntryByID(int ID, 1875 bool *Invalid = nullptr) const { 1876 return const_cast<SourceManager *>(this)->getSLocEntryByID(ID, Invalid); 1877 } 1878 1879 SrcMgr::SLocEntry &getSLocEntryByID(int ID, bool *Invalid = nullptr) { 1880 assert(ID != -1 && "Using FileID sentinel value"); 1881 if (ID < 0) 1882 return getLoadedSLocEntryByID(ID, Invalid); 1883 return getLocalSLocEntry(static_cast<unsigned>(ID)); 1884 } 1885 1886 const SrcMgr::SLocEntry & 1887 getLoadedSLocEntryByID(int ID, bool *Invalid = nullptr) const { 1888 return const_cast<SourceManager *>(this)->getLoadedSLocEntryByID(ID, 1889 Invalid); 1890 } 1891 1892 SrcMgr::SLocEntry &getLoadedSLocEntryByID(int ID, bool *Invalid = nullptr) { 1893 return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2), Invalid); 1894 } 1895 1896 FileID getFileID(SourceLocation::UIntTy SLocOffset) const { 1897 // If our one-entry cache covers this offset, just return it. 1898 if (isOffsetInFileID(LastFileIDLookup, SLocOffset)) 1899 return LastFileIDLookup; 1900 1901 return getFileIDSlow(SLocOffset); 1902 } 1903 1904 bool isLocalOffset(SourceLocation::UIntTy SLocOffset) const { 1905 return SLocOffset < CurrentLoadedOffset; 1906 } 1907 1908 bool isLoadedOffset(SourceLocation::UIntTy SLocOffset) const { 1909 return SLocOffset >= CurrentLoadedOffset; 1910 } 1911 1912 /// Implements the common elements of storing an expansion info struct into 1913 /// the SLocEntry table and producing a source location that refers to it. 1914 SourceLocation 1915 createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion, 1916 unsigned Length, int LoadedID = 0, 1917 SourceLocation::UIntTy LoadedOffset = 0); 1918 1919 /// Return true if the specified FileID contains the 1920 /// specified SourceLocation offset. This is a very hot method. 1921 inline bool isOffsetInFileID(FileID FID, 1922 SourceLocation::UIntTy SLocOffset) const { 1923 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID); 1924 // If the entry is after the offset, it can't contain it. 1925 if (SLocOffset < Entry.getOffset()) return false; 1926 1927 // If this is the very last entry then it does. 1928 if (FID.ID == -2) 1929 return true; 1930 1931 // If it is the last local entry, then it does if the location is local. 1932 if (FID.ID+1 == static_cast<int>(LocalSLocEntryTable.size())) 1933 return SLocOffset < NextLocalOffset; 1934 1935 // Otherwise, the entry after it has to not include it. This works for both 1936 // local and loaded entries. 1937 return SLocOffset < getSLocEntryByID(FID.ID+1).getOffset(); 1938 } 1939 1940 /// Returns the previous in-order FileID or an invalid FileID if there 1941 /// is no previous one. 1942 FileID getPreviousFileID(FileID FID) const; 1943 1944 /// Returns the next in-order FileID or an invalid FileID if there is 1945 /// no next one. 1946 FileID getNextFileID(FileID FID) const; 1947 1948 /// Create a new fileID for the specified ContentCache and 1949 /// include position. 1950 /// 1951 /// This works regardless of whether the ContentCache corresponds to a 1952 /// file or some other input source. 1953 FileID createFileIDImpl(SrcMgr::ContentCache &File, StringRef Filename, 1954 SourceLocation IncludePos, 1955 SrcMgr::CharacteristicKind DirCharacter, int LoadedID, 1956 SourceLocation::UIntTy LoadedOffset); 1957 1958 SrcMgr::ContentCache &getOrCreateContentCache(FileEntryRef SourceFile, 1959 bool isSystemFile = false); 1960 1961 /// Create a new ContentCache for the specified memory buffer. 1962 SrcMgr::ContentCache & 1963 createMemBufferContentCache(std::unique_ptr<llvm::MemoryBuffer> Buf); 1964 1965 FileID getFileIDSlow(SourceLocation::UIntTy SLocOffset) const; 1966 FileID getFileIDLocal(SourceLocation::UIntTy SLocOffset) const; 1967 FileID getFileIDLoaded(SourceLocation::UIntTy SLocOffset) const; 1968 1969 SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const; 1970 SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const; 1971 SourceLocation getFileLocSlowCase(SourceLocation Loc) const; 1972 1973 std::pair<FileID, unsigned> 1974 getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const; 1975 std::pair<FileID, unsigned> 1976 getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E, 1977 unsigned Offset) const; 1978 void computeMacroArgsCache(MacroArgsMap &MacroArgsCache, FileID FID) const; 1979 void associateFileChunkWithMacroArgExp(MacroArgsMap &MacroArgsCache, 1980 FileID FID, 1981 SourceLocation SpellLoc, 1982 SourceLocation ExpansionLoc, 1983 unsigned ExpansionLength) const; 1984 void updateSlocUsageStats() const; 1985 }; 1986 1987 /// Comparison function object. 1988 template<typename T> 1989 class BeforeThanCompare; 1990 1991 /// Compare two source locations. 1992 template<> 1993 class BeforeThanCompare<SourceLocation> { 1994 SourceManager &SM; 1995 1996 public: 1997 explicit BeforeThanCompare(SourceManager &SM) : SM(SM) {} 1998 1999 bool operator()(SourceLocation LHS, SourceLocation RHS) const { 2000 return SM.isBeforeInTranslationUnit(LHS, RHS); 2001 } 2002 }; 2003 2004 /// Compare two non-overlapping source ranges. 2005 template<> 2006 class BeforeThanCompare<SourceRange> { 2007 SourceManager &SM; 2008 2009 public: 2010 explicit BeforeThanCompare(SourceManager &SM) : SM(SM) {} 2011 2012 bool operator()(SourceRange LHS, SourceRange RHS) const { 2013 return SM.isBeforeInTranslationUnit(LHS.getBegin(), RHS.getBegin()); 2014 } 2015 }; 2016 2017 /// SourceManager and necessary dependencies (e.g. VFS, FileManager) for a 2018 /// single in-memorty file. 2019 class SourceManagerForFile { 2020 public: 2021 /// Creates SourceManager and necessary dependencies (e.g. VFS, FileManager). 2022 /// The main file in the SourceManager will be \p FileName with \p Content. 2023 SourceManagerForFile(StringRef FileName, StringRef Content); 2024 2025 SourceManager &get() { 2026 assert(SourceMgr); 2027 return *SourceMgr; 2028 } 2029 2030 private: 2031 // The order of these fields are important - they should be in the same order 2032 // as they are created in `createSourceManagerForFile` so that they can be 2033 // deleted in the reverse order as they are created. 2034 std::unique_ptr<FileManager> FileMgr; 2035 std::unique_ptr<DiagnosticsEngine> Diagnostics; 2036 std::unique_ptr<SourceManager> SourceMgr; 2037 }; 2038 2039 } // namespace clang 2040 2041 #endif // LLVM_CLANG_BASIC_SOURCEMANAGER_H 2042