1//===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===//
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/// Implements # directive processing for the Preprocessor.
11///
12//===----------------------------------------------------------------------===//
13
14#include "clang/Basic/AttributeCommonInfo.h"
15#include "clang/Basic/Attributes.h"
16#include "clang/Basic/CharInfo.h"
17#include "clang/Basic/DirectoryEntry.h"
18#include "clang/Basic/FileManager.h"
19#include "clang/Basic/IdentifierTable.h"
20#include "clang/Basic/LangOptions.h"
21#include "clang/Basic/Module.h"
22#include "clang/Basic/SourceLocation.h"
23#include "clang/Basic/SourceManager.h"
24#include "clang/Basic/TargetInfo.h"
25#include "clang/Basic/TokenKinds.h"
26#include "clang/Lex/CodeCompletionHandler.h"
27#include "clang/Lex/HeaderSearch.h"
28#include "clang/Lex/LexDiagnostic.h"
29#include "clang/Lex/LiteralSupport.h"
30#include "clang/Lex/MacroInfo.h"
31#include "clang/Lex/ModuleLoader.h"
32#include "clang/Lex/ModuleMap.h"
33#include "clang/Lex/PPCallbacks.h"
34#include "clang/Lex/Pragma.h"
35#include "clang/Lex/Preprocessor.h"
36#include "clang/Lex/PreprocessorOptions.h"
37#include "clang/Lex/Token.h"
38#include "clang/Lex/VariadicMacroSupport.h"
39#include "llvm/ADT/ArrayRef.h"
40#include "llvm/ADT/STLExtras.h"
41#include "llvm/ADT/ScopeExit.h"
42#include "llvm/ADT/SmallVector.h"
43#include "llvm/ADT/StringExtras.h"
44#include "llvm/ADT/StringRef.h"
45#include "llvm/ADT/StringSwitch.h"
46#include "llvm/Support/ErrorHandling.h"
47#include "llvm/Support/Path.h"
48#include "llvm/Support/SaveAndRestore.h"
49#include <algorithm>
50#include <cassert>
51#include <cstddef>
52#include <cstring>
53#include <optional>
54#include <string>
55#include <utility>
56
57using namespace clang;
58
59//===----------------------------------------------------------------------===//
60// Utility Methods for Preprocessor Directive Handling.
61//===----------------------------------------------------------------------===//
62
63MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
64 static_assert(std::is_trivially_destructible_v<MacroInfo>, "");
65 return new (BP) MacroInfo(L);
66}
67
68DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI,
69 SourceLocation Loc) {
70 return new (BP) DefMacroDirective(MI, Loc);
71}
72
73UndefMacroDirective *
74Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) {
75 return new (BP) UndefMacroDirective(UndefLoc);
76}
77
78VisibilityMacroDirective *
79Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc,
80 bool isPublic) {
81 return new (BP) VisibilityMacroDirective(Loc, isPublic);
82}
83
84/// Read and discard all tokens remaining on the current line until
85/// the tok::eod token is found.
86SourceRange Preprocessor::DiscardUntilEndOfDirective(
87 Token &Tmp, SmallVectorImpl<Token> *DiscardedToks) {
88 SourceRange Res;
89 auto ReadNextTok = [&]() {
90 LexUnexpandedToken(Result&: Tmp);
91 if (DiscardedToks && Tmp.isNot(K: tok::eod))
92 DiscardedToks->push_back(Elt: Tmp);
93 };
94 ReadNextTok();
95 Res.setBegin(Tmp.getLocation());
96 while (Tmp.isNot(K: tok::eod)) {
97 assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
98 ReadNextTok();
99 }
100 Res.setEnd(Tmp.getLocation());
101 return Res;
102}
103
104/// Enumerates possible cases of #define/#undef a reserved identifier.
105enum MacroDiag {
106 MD_NoWarn, //> Not a reserved identifier
107 MD_KeywordDef, //> Macro hides keyword, enabled by default
108 MD_ReservedMacro, //> #define of #undef reserved id, disabled by default
109 MD_ReservedAttributeIdentifier
110};
111
112/// Enumerates possible %select values for the pp_err_elif_after_else and
113/// pp_err_elif_without_if diagnostics.
114enum PPElifDiag {
115 PED_Elif,
116 PED_Elifdef,
117 PED_Elifndef
118};
119
120static bool isFeatureTestMacro(StringRef MacroName) {
121 // list from:
122 // * https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_macros.html
123 // * https://docs.microsoft.com/en-us/cpp/c-runtime-library/security-features-in-the-crt?view=msvc-160
124 // * man 7 feature_test_macros
125 // The list must be sorted for correct binary search.
126 static constexpr StringRef ReservedMacro[] = {
127 "_ATFILE_SOURCE",
128 "_BSD_SOURCE",
129 "_CRT_NONSTDC_NO_WARNINGS",
130 "_CRT_SECURE_CPP_OVERLOAD_STANDARD_NAMES",
131 "_CRT_SECURE_NO_WARNINGS",
132 "_FILE_OFFSET_BITS",
133 "_FORTIFY_SOURCE",
134 "_GLIBCXX_ASSERTIONS",
135 "_GLIBCXX_CONCEPT_CHECKS",
136 "_GLIBCXX_DEBUG",
137 "_GLIBCXX_DEBUG_PEDANTIC",
138 "_GLIBCXX_PARALLEL",
139 "_GLIBCXX_PARALLEL_ASSERTIONS",
140 "_GLIBCXX_SANITIZE_VECTOR",
141 "_GLIBCXX_USE_CXX11_ABI",
142 "_GLIBCXX_USE_DEPRECATED",
143 "_GNU_SOURCE",
144 "_ISOC11_SOURCE",
145 "_ISOC95_SOURCE",
146 "_ISOC99_SOURCE",
147 "_LARGEFILE64_SOURCE",
148 "_POSIX_C_SOURCE",
149 "_REENTRANT",
150 "_SVID_SOURCE",
151 "_THREAD_SAFE",
152 "_XOPEN_SOURCE",
153 "_XOPEN_SOURCE_EXTENDED",
154 "__STDCPP_WANT_MATH_SPEC_FUNCS__",
155 "__STDC_FORMAT_MACROS",
156 };
157 return llvm::binary_search(Range: ReservedMacro, Value&: MacroName);
158}
159
160static bool isLanguageDefinedBuiltin(const SourceManager &SourceMgr,
161 const MacroInfo *MI,
162 const StringRef MacroName) {
163 // If this is a macro with special handling (like __LINE__) then it's language
164 // defined.
165 if (MI->isBuiltinMacro())
166 return true;
167 // Builtin macros are defined in the builtin file
168 if (!SourceMgr.isWrittenInBuiltinFile(Loc: MI->getDefinitionLoc()))
169 return false;
170 // C defines macros starting with __STDC, and C++ defines macros starting with
171 // __STDCPP
172 if (MacroName.starts_with(Prefix: "__STDC"))
173 return true;
174 // C++ defines the __cplusplus macro
175 if (MacroName == "__cplusplus")
176 return true;
177 // C++ defines various feature-test macros starting with __cpp
178 if (MacroName.starts_with(Prefix: "__cpp"))
179 return true;
180 // Anything else isn't language-defined
181 return false;
182}
183
184static bool isReservedCXXAttributeName(Preprocessor &PP, IdentifierInfo *II) {
185 const LangOptions &Lang = PP.getLangOpts();
186 if (Lang.CPlusPlus &&
187 hasAttribute(Syntax: AttributeCommonInfo::AS_CXX11, /* Scope*/ nullptr, Attr: II,
188 Target: PP.getTargetInfo(), LangOpts: Lang, /*CheckPlugins*/ false) > 0) {
189 AttributeCommonInfo::AttrArgsInfo AttrArgsInfo =
190 AttributeCommonInfo::getCXX11AttrArgsInfo(Name: II);
191 if (AttrArgsInfo == AttributeCommonInfo::AttrArgsInfo::Required)
192 return PP.isNextPPTokenOneOf(Ks: tok::l_paren);
193
194 return !PP.isNextPPTokenOneOf(Ks: tok::l_paren) ||
195 AttrArgsInfo == AttributeCommonInfo::AttrArgsInfo::Optional;
196 }
197 return false;
198}
199
200static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) {
201 const LangOptions &Lang = PP.getLangOpts();
202 StringRef Text = II->getName();
203 if (isReservedInAllContexts(Status: II->isReserved(LangOpts: Lang)))
204 return isFeatureTestMacro(MacroName: Text) ? MD_NoWarn : MD_ReservedMacro;
205 if (II->isKeyword(LangOpts: Lang))
206 return MD_KeywordDef;
207 if (Lang.CPlusPlus11 && (Text == "override" || Text == "final"))
208 return MD_KeywordDef;
209 if (isReservedCXXAttributeName(PP, II))
210 return MD_ReservedAttributeIdentifier;
211 return MD_NoWarn;
212}
213
214static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) {
215 const LangOptions &Lang = PP.getLangOpts();
216 // Do not warn on keyword undef. It is generally harmless and widely used.
217 if (isReservedInAllContexts(Status: II->isReserved(LangOpts: Lang)))
218 return MD_ReservedMacro;
219 if (isReservedCXXAttributeName(PP, II))
220 return MD_ReservedAttributeIdentifier;
221 return MD_NoWarn;
222}
223
224// Return true if we want to issue a diagnostic by default if we
225// encounter this name in a #include with the wrong case. For now,
226// this includes the standard C and C++ headers, Posix headers,
227// and Boost headers. Improper case for these #includes is a
228// potential portability issue.
229static bool warnByDefaultOnWrongCase(StringRef Include) {
230 // If the first component of the path is "boost", treat this like a standard header
231 // for the purposes of diagnostics.
232 if (::llvm::sys::path::begin(path: Include)->equals_insensitive(RHS: "boost"))
233 return true;
234
235 // "condition_variable" is the longest standard header name at 18 characters.
236 // If the include file name is longer than that, it can't be a standard header.
237 static const size_t MaxStdHeaderNameLen = 18u;
238 if (Include.size() > MaxStdHeaderNameLen)
239 return false;
240
241 // Lowercase and normalize the search string.
242 SmallString<32> LowerInclude{Include};
243 for (char &Ch : LowerInclude) {
244 // In the ASCII range?
245 if (static_cast<unsigned char>(Ch) > 0x7f)
246 return false; // Can't be a standard header
247 // ASCII lowercase:
248 if (Ch >= 'A' && Ch <= 'Z')
249 Ch += 'a' - 'A';
250 // Normalize path separators for comparison purposes.
251 else if (::llvm::sys::path::is_separator(value: Ch))
252 Ch = '/';
253 }
254
255 // The standard C/C++ and Posix headers
256 return llvm::StringSwitch<bool>(LowerInclude)
257 // C library headers
258 .Cases(CaseStrings: {"assert.h", "complex.h", "ctype.h", "errno.h", "fenv.h"}, Value: true)
259 .Cases(CaseStrings: {"float.h", "inttypes.h", "iso646.h", "limits.h", "locale.h"},
260 Value: true)
261 .Cases(CaseStrings: {"math.h", "setjmp.h", "signal.h", "stdalign.h", "stdarg.h"}, Value: true)
262 .Cases(CaseStrings: {"stdatomic.h", "stdbool.h", "stdckdint.h", "stdcountof.h"}, Value: true)
263 .Cases(CaseStrings: {"stddef.h", "stdint.h", "stdio.h", "stdlib.h", "stdnoreturn.h"},
264 Value: true)
265 .Cases(CaseStrings: {"string.h", "tgmath.h", "threads.h", "time.h", "uchar.h"}, Value: true)
266 .Cases(CaseStrings: {"wchar.h", "wctype.h"}, Value: true)
267
268 // C++ headers for C library facilities
269 .Cases(CaseStrings: {"cassert", "ccomplex", "cctype", "cerrno", "cfenv"}, Value: true)
270 .Cases(CaseStrings: {"cfloat", "cinttypes", "ciso646", "climits", "clocale"}, Value: true)
271 .Cases(CaseStrings: {"cmath", "csetjmp", "csignal", "cstdalign", "cstdarg"}, Value: true)
272 .Cases(CaseStrings: {"cstdbool", "cstddef", "cstdint", "cstdio", "cstdlib"}, Value: true)
273 .Cases(CaseStrings: {"cstring", "ctgmath", "ctime", "cuchar", "cwchar"}, Value: true)
274 .Case(S: "cwctype", Value: true)
275
276 // C++ library headers
277 .Cases(CaseStrings: {"algorithm", "fstream", "list", "regex", "thread"}, Value: true)
278 .Cases(CaseStrings: {"array", "functional", "locale", "scoped_allocator", "tuple"},
279 Value: true)
280 .Cases(CaseStrings: {"atomic", "future", "map", "set", "type_traits"}, Value: true)
281 .Cases(
282 CaseStrings: {"bitset", "initializer_list", "memory", "shared_mutex", "typeindex"},
283 Value: true)
284 .Cases(CaseStrings: {"chrono", "iomanip", "mutex", "sstream", "typeinfo"}, Value: true)
285 .Cases(CaseStrings: {"codecvt", "ios", "new", "stack", "unordered_map"}, Value: true)
286 .Cases(CaseStrings: {"complex", "iosfwd", "numeric", "stdexcept", "unordered_set"},
287 Value: true)
288 .Cases(
289 CaseStrings: {"condition_variable", "iostream", "ostream", "streambuf", "utility"},
290 Value: true)
291 .Cases(CaseStrings: {"deque", "istream", "queue", "string", "valarray"}, Value: true)
292 .Cases(CaseStrings: {"exception", "iterator", "random", "strstream", "vector"}, Value: true)
293 .Cases(CaseStrings: {"forward_list", "limits", "ratio", "system_error"}, Value: true)
294
295 // POSIX headers (which aren't also C headers)
296 .Cases(CaseStrings: {"aio.h", "arpa/inet.h", "cpio.h", "dirent.h", "dlfcn.h"}, Value: true)
297 .Cases(CaseStrings: {"fcntl.h", "fmtmsg.h", "fnmatch.h", "ftw.h", "glob.h"}, Value: true)
298 .Cases(CaseStrings: {"grp.h", "iconv.h", "langinfo.h", "libgen.h", "monetary.h"}, Value: true)
299 .Cases(CaseStrings: {"mqueue.h", "ndbm.h", "net/if.h", "netdb.h", "netinet/in.h"},
300 Value: true)
301 .Cases(CaseStrings: {"netinet/tcp.h", "nl_types.h", "poll.h", "pthread.h", "pwd.h"},
302 Value: true)
303 .Cases(CaseStrings: {"regex.h", "sched.h", "search.h", "semaphore.h", "spawn.h"}, Value: true)
304 .Cases(CaseStrings: {"strings.h", "stropts.h", "sys/ipc.h", "sys/mman.h", "sys/msg.h"},
305 Value: true)
306 .Cases(CaseStrings: {"sys/resource.h", "sys/select.h", "sys/sem.h", "sys/shm.h",
307 "sys/socket.h"},
308 Value: true)
309 .Cases(CaseStrings: {"sys/stat.h", "sys/statvfs.h", "sys/time.h", "sys/times.h",
310 "sys/types.h"},
311 Value: true)
312 .Cases(
313 CaseStrings: {"sys/uio.h", "sys/un.h", "sys/utsname.h", "sys/wait.h", "syslog.h"},
314 Value: true)
315 .Cases(CaseStrings: {"tar.h", "termios.h", "trace.h", "ulimit.h"}, Value: true)
316 .Cases(CaseStrings: {"unistd.h", "utime.h", "utmpx.h", "wordexp.h"}, Value: true)
317 .Default(Value: false);
318}
319
320/// Find a similar string in `Candidates`.
321///
322/// \param LHS a string for a similar string in `Candidates`
323///
324/// \param Candidates the candidates to find a similar string.
325///
326/// \returns a similar string if exists. If no similar string exists,
327/// returns std::nullopt.
328static std::optional<StringRef>
329findSimilarStr(StringRef LHS, const std::vector<StringRef> &Candidates) {
330 // We need to check if `Candidates` has the exact case-insensitive string
331 // because the Levenshtein distance match does not care about it.
332 for (StringRef C : Candidates) {
333 if (LHS.equals_insensitive(RHS: C)) {
334 return C;
335 }
336 }
337
338 // Keep going with the Levenshtein distance match.
339 // If the LHS size is less than 3, use the LHS size minus 1 and if not,
340 // use the LHS size divided by 3.
341 size_t Length = LHS.size();
342 size_t MaxDist = Length < 3 ? Length - 1 : Length / 3;
343
344 std::optional<std::pair<StringRef, size_t>> SimilarStr;
345 for (StringRef C : Candidates) {
346 size_t CurDist = LHS.edit_distance(Other: C, AllowReplacements: true);
347 if (CurDist <= MaxDist) {
348 if (!SimilarStr) {
349 // The first similar string found.
350 SimilarStr = {C, CurDist};
351 } else if (CurDist < SimilarStr->second) {
352 // More similar string found.
353 SimilarStr = {C, CurDist};
354 }
355 }
356 }
357
358 if (SimilarStr) {
359 return SimilarStr->first;
360 } else {
361 return std::nullopt;
362 }
363}
364
365bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
366 bool *ShadowFlag) {
367 // Missing macro name?
368 if (MacroNameTok.is(K: tok::eod))
369 return Diag(Tok: MacroNameTok, DiagID: diag::err_pp_missing_macro_name);
370
371 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
372 if (!II)
373 return Diag(Tok: MacroNameTok, DiagID: diag::err_pp_macro_not_identifier);
374
375 if (II->isCPlusPlusOperatorKeyword()) {
376 // C++ 2.5p2: Alternative tokens behave the same as its primary token
377 // except for their spellings.
378 Diag(Tok: MacroNameTok, DiagID: getLangOpts().MicrosoftExt
379 ? diag::ext_pp_operator_used_as_macro_name
380 : diag::err_pp_operator_used_as_macro_name)
381 << II << MacroNameTok.getKind();
382 // Allow #defining |and| and friends for Microsoft compatibility or
383 // recovery when legacy C headers are included in C++.
384 }
385
386 if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) {
387 // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4.
388 return Diag(Tok: MacroNameTok, DiagID: diag::err_defined_macro_name);
389 }
390
391 // If defining/undefining reserved identifier or a keyword, we need to issue
392 // a warning.
393 SourceLocation MacroNameLoc = MacroNameTok.getLocation();
394 if (ShadowFlag)
395 *ShadowFlag = false;
396 // Macro names with reserved identifiers are accepted if built-in or passed
397 // through the command line (the later may be present if -dD was used to
398 // generate the preprocessed file).
399 // NB: isInPredefinedFile() is relatively expensive, so keep it at the end
400 // of the condition.
401 if (!SourceMgr.isInSystemHeader(Loc: MacroNameLoc) &&
402 !SourceMgr.isInPredefinedFile(Loc: MacroNameLoc)) {
403 MacroDiag D = MD_NoWarn;
404 if (isDefineUndef == MU_Define) {
405 D = shouldWarnOnMacroDef(PP&: *this, II);
406 }
407 else if (isDefineUndef == MU_Undef)
408 D = shouldWarnOnMacroUndef(PP&: *this, II);
409 if (D == MD_KeywordDef) {
410 // We do not want to warn on some patterns widely used in configuration
411 // scripts. This requires analyzing next tokens, so do not issue warnings
412 // now, only inform caller.
413 if (ShadowFlag)
414 *ShadowFlag = true;
415 }
416 if (D == MD_ReservedMacro)
417 Diag(Tok: MacroNameTok, DiagID: diag::warn_pp_macro_is_reserved_id);
418 if (D == MD_ReservedAttributeIdentifier)
419 Diag(Tok: MacroNameTok, DiagID: diag::warn_pp_macro_is_reserved_attribute_id)
420 << II->getName();
421 }
422
423 // Okay, we got a good identifier.
424 return false;
425}
426
427/// Lex and validate a macro name, which occurs after a
428/// \#define or \#undef.
429///
430/// This sets the token kind to eod and discards the rest of the macro line if
431/// the macro name is invalid.
432///
433/// \param MacroNameTok Token that is expected to be a macro name.
434/// \param isDefineUndef Context in which macro is used.
435/// \param ShadowFlag Points to a flag that is set if macro shadows a keyword.
436void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
437 bool *ShadowFlag) {
438 // Read the token, don't allow macro expansion on it.
439 LexUnexpandedToken(Result&: MacroNameTok);
440
441 if (MacroNameTok.is(K: tok::code_completion)) {
442 if (CodeComplete)
443 CodeComplete->CodeCompleteMacroName(IsDefinition: isDefineUndef == MU_Define);
444 setCodeCompletionReached();
445 LexUnexpandedToken(Result&: MacroNameTok);
446 }
447
448 if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag))
449 return;
450
451 // Invalid macro name, read and discard the rest of the line and set the
452 // token kind to tok::eod if necessary.
453 if (MacroNameTok.isNot(K: tok::eod)) {
454 MacroNameTok.setKind(tok::eod);
455 DiscardUntilEndOfDirective();
456 }
457}
458
459/// Ensure that the next token is a tok::eod token.
460///
461/// If not, emit a diagnostic and consume up until the eod. If EnableMacros is
462/// true, then we consider macros that expand to zero tokens as being ok.
463///
464/// Returns the location of the end of the directive.
465SourceLocation
466Preprocessor::CheckEndOfDirective(StringRef DirType, bool EnableMacros,
467 SmallVectorImpl<Token> *ExtraToks) {
468 Token Tmp;
469 auto ReadNextTok = [this, ExtraToks, &Tmp](auto &&LexFn) {
470 std::invoke(LexFn, this, Tmp);
471 if (ExtraToks && Tmp.isNot(K: tok::eod))
472 ExtraToks->push_back(Elt: Tmp);
473 };
474 // Lex unexpanded tokens for most directives: macros might expand to zero
475 // tokens, causing us to miss diagnosing invalid lines. Some directives (like
476 // #line) allow empty macros.
477 if (EnableMacros)
478 ReadNextTok(&Preprocessor::Lex);
479 else
480 ReadNextTok(&Preprocessor::LexUnexpandedToken);
481
482 // There should be no tokens after the directive, but we allow them as an
483 // extension.
484 while (Tmp.is(K: tok::comment)) // Skip comments in -C mode.
485 ReadNextTok(&Preprocessor::LexUnexpandedToken);
486
487 if (Tmp.is(K: tok::eod))
488 return Tmp.getLocation();
489
490 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89,
491 // or if this is a macro-style preprocessing directive, because it is more
492 // trouble than it is worth to insert /**/ and check that there is no /**/
493 // in the range also.
494 FixItHint Hint;
495 if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
496 !CurTokenLexer)
497 Hint = FixItHint::CreateInsertion(InsertionLoc: Tmp.getLocation(),Code: "//");
498
499 unsigned DiagID = diag::ext_pp_extra_tokens_at_eol;
500 // C++20 import or module directive has no '#' prefix.
501 if (getLangOpts().CPlusPlusModules &&
502 (DirType == "import" || DirType == "module"))
503 DiagID = diag::warn_pp_extra_tokens_at_module_directive_eol;
504
505 Diag(Tok: Tmp, DiagID) << DirType << Hint;
506 return DiscardUntilEndOfDirective(DiscardedToks: ExtraToks).getEnd();
507}
508
509void Preprocessor::SuggestTypoedDirective(const Token &Tok,
510 StringRef Directive) const {
511 // If this is a `.S` file, treat unknown # directives as non-preprocessor
512 // directives.
513 if (getLangOpts().AsmPreprocessor) return;
514
515 std::vector<StringRef> Candidates = {
516 "if", "ifdef", "ifndef", "elif", "else", "endif"
517 };
518 if (LangOpts.C23 || LangOpts.CPlusPlus23)
519 Candidates.insert(position: Candidates.end(), l: {"elifdef", "elifndef"});
520
521 if (std::optional<StringRef> Sugg = findSimilarStr(LHS: Directive, Candidates)) {
522 // Directive cannot be coming from macro.
523 assert(Tok.getLocation().isFileID());
524 CharSourceRange DirectiveRange = CharSourceRange::getCharRange(
525 B: Tok.getLocation(),
526 E: Tok.getLocation().getLocWithOffset(Offset: Directive.size()));
527 StringRef SuggValue = *Sugg;
528
529 auto Hint = FixItHint::CreateReplacement(RemoveRange: DirectiveRange, Code: SuggValue);
530 Diag(Tok, DiagID: diag::warn_pp_invalid_directive) << 1 << SuggValue << Hint;
531 }
532}
533
534/// SkipExcludedConditionalBlock - We just read a \#if or related directive and
535/// decided that the subsequent tokens are in the \#if'd out portion of the
536/// file. Lex the rest of the file, until we see an \#endif. If
537/// FoundNonSkipPortion is true, then we have already emitted code for part of
538/// this \#if directive, so \#else/\#elif blocks should never be entered.
539/// If ElseOk is true, then \#else directives are ok, if not, then we have
540/// already seen one so a \#else directive is a duplicate. When this returns,
541/// the caller can lex the first valid token.
542void Preprocessor::SkipExcludedConditionalBlock(SourceLocation HashTokenLoc,
543 SourceLocation IfTokenLoc,
544 bool FoundNonSkipPortion,
545 bool FoundElse,
546 SourceLocation ElseLoc) {
547 // In SkippingRangeStateTy we are depending on SkipExcludedConditionalBlock()
548 // not getting called recursively by storing the RecordedSkippedRanges
549 // DenseMap lookup pointer (field SkipRangePtr). SkippingRangeStateTy expects
550 // that RecordedSkippedRanges won't get modified and SkipRangePtr won't be
551 // invalidated. If this changes and there is a need to call
552 // SkipExcludedConditionalBlock() recursively, SkippingRangeStateTy should
553 // change to do a second lookup in endLexPass function instead of reusing the
554 // lookup pointer.
555 assert(!SkippingExcludedConditionalBlock &&
556 "calling SkipExcludedConditionalBlock recursively");
557 llvm::SaveAndRestore SARSkipping(SkippingExcludedConditionalBlock, true);
558
559 ++NumSkipped;
560 assert(!CurTokenLexer && "Conditional PP block cannot appear in a macro!");
561 assert(CurPPLexer && "Conditional PP block must be in a file!");
562 assert(CurLexer && "Conditional PP block but no current lexer set!");
563
564 if (PreambleConditionalStack.reachedEOFWhileSkipping())
565 PreambleConditionalStack.clearSkipInfo();
566 else
567 CurPPLexer->pushConditionalLevel(DirectiveStart: IfTokenLoc, /*isSkipping*/ WasSkipping: false,
568 FoundNonSkip: FoundNonSkipPortion, FoundElse);
569
570 // Enter raw mode to disable identifier lookup (and thus macro expansion),
571 // disabling warnings, etc.
572 CurPPLexer->LexingRawMode = true;
573 Token Tok;
574 SourceLocation endLoc;
575
576 /// Keeps track and caches skipped ranges and also retrieves a prior skipped
577 /// range if the same block is re-visited.
578 struct SkippingRangeStateTy {
579 Preprocessor &PP;
580
581 const char *BeginPtr = nullptr;
582 unsigned *SkipRangePtr = nullptr;
583
584 SkippingRangeStateTy(Preprocessor &PP) : PP(PP) {}
585
586 void beginLexPass() {
587 if (BeginPtr)
588 return; // continue skipping a block.
589
590 // Initiate a skipping block and adjust the lexer if we already skipped it
591 // before.
592 BeginPtr = PP.CurLexer->getBufferLocation();
593 SkipRangePtr = &PP.RecordedSkippedRanges[BeginPtr];
594 if (*SkipRangePtr) {
595 PP.CurLexer->seek(Offset: PP.CurLexer->getCurrentBufferOffset() + *SkipRangePtr,
596 /*IsAtStartOfLine*/ true);
597 }
598 }
599
600 void endLexPass(const char *Hashptr) {
601 if (!BeginPtr) {
602 // Not doing normal lexing.
603 assert(PP.CurLexer->isDependencyDirectivesLexer());
604 return;
605 }
606
607 // Finished skipping a block, record the range if it's first time visited.
608 if (!*SkipRangePtr) {
609 *SkipRangePtr = Hashptr - BeginPtr;
610 }
611 assert(*SkipRangePtr == unsigned(Hashptr - BeginPtr));
612 BeginPtr = nullptr;
613 SkipRangePtr = nullptr;
614 }
615 } SkippingRangeState(*this);
616
617 while (true) {
618 if (CurLexer->isDependencyDirectivesLexer()) {
619 CurLexer->LexDependencyDirectiveTokenWhileSkipping(Result&: Tok);
620 } else {
621 SkippingRangeState.beginLexPass();
622 while (true) {
623 CurLexer->Lex(Result&: Tok);
624
625 if (Tok.is(K: tok::code_completion)) {
626 setCodeCompletionReached();
627 if (CodeComplete)
628 CodeComplete->CodeCompleteInConditionalExclusion();
629 continue;
630 }
631
632 // There is actually no "skipped block" in the above because the module
633 // directive is not a text-line (https://wg21.link/cpp.pre#2) nor
634 // anything else that is allowed in a group
635 // (https://eel.is/c++draft/cpp.pre#nt:group-part).
636 //
637 // A preprocessor diagnostic (effective with -E) that triggers whenever
638 // a module directive is encountered where a control-line or a text-line
639 // is required.
640 if (getLangOpts().CPlusPlusModules && Tok.isAtStartOfLine() &&
641 Tok.is(K: tok::raw_identifier) &&
642 (Tok.getRawIdentifier() == "export" ||
643 Tok.getRawIdentifier() == "module")) {
644 llvm::SaveAndRestore ModuleDirectiveSkipping(
645 LastTokenWasExportKeyword);
646 LastTokenWasExportKeyword.reset();
647 LookUpIdentifierInfo(Identifier&: Tok);
648 IdentifierInfo *II = Tok.getIdentifierInfo();
649
650 if (II->getName()[0] == 'e') { // export
651 HandleModuleContextualKeyword(Result&: Tok, TokAtPhysicalStartOfLine: Tok.isAtStartOfLine());
652 CurLexer->Lex(Result&: Tok);
653 if (Tok.is(K: tok::raw_identifier)) {
654 LookUpIdentifierInfo(Identifier&: Tok);
655 II = Tok.getIdentifierInfo();
656 }
657 }
658
659 if (II->getName()[0] == 'm') { // module
660 // HandleModuleContextualKeyword changes the lexer state, so we need
661 // to save RawLexingMode
662 llvm::SaveAndRestore RestoreLexingRawMode(CurPPLexer->LexingRawMode,
663 false);
664 if (HandleModuleContextualKeyword(Result&: Tok, TokAtPhysicalStartOfLine: Tok.isAtStartOfLine())) {
665 // We just parsed a # character at the start of a line, so we're
666 // in directive mode. Tell the lexer this so any newlines we see
667 // will be converted into an EOD token (this terminates the
668 // macro).
669 CurPPLexer->ParsingPreprocessorDirective = true;
670 SourceLocation StartLoc = Tok.getLocation();
671 SourceLocation End = DiscardUntilEndOfDirective().getEnd();
672 Diag(Loc: StartLoc, DiagID: diag::err_pp_cond_span_module_decl)
673 << SourceRange(StartLoc, End);
674 CurPPLexer->ParsingPreprocessorDirective = false;
675 // Restore comment saving mode.
676 if (CurLexer)
677 CurLexer->resetExtendedTokenMode();
678 continue;
679 }
680 }
681 }
682
683 // If this is the end of the buffer, we have an error.
684 if (Tok.is(K: tok::eof)) {
685 // We don't emit errors for unterminated conditionals here,
686 // Lexer::LexEndOfFile can do that properly.
687 // Just return and let the caller lex after this #include.
688 if (PreambleConditionalStack.isRecording())
689 PreambleConditionalStack.SkipInfo.emplace(args&: HashTokenLoc, args&: IfTokenLoc,
690 args&: FoundNonSkipPortion,
691 args&: FoundElse, args&: ElseLoc);
692 break;
693 }
694
695 // If this token is not a preprocessor directive, just skip it.
696 if (Tok.isNot(K: tok::hash) || !Tok.isAtStartOfLine())
697 continue;
698
699 break;
700 }
701 }
702 if (Tok.is(K: tok::eof))
703 break;
704
705 // We just parsed a # character at the start of a line, so we're in
706 // directive mode. Tell the lexer this so any newlines we see will be
707 // converted into an EOD token (this terminates the macro).
708 CurPPLexer->ParsingPreprocessorDirective = true;
709 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
710
711 assert(Tok.is(tok::hash));
712 const char *Hashptr = CurLexer->getBufferLocation() - Tok.getLength();
713 assert(CurLexer->getSourceLocation(Hashptr) == Tok.getLocation());
714
715 // Read the next token, the directive flavor.
716 LexUnexpandedToken(Result&: Tok);
717
718 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
719 // something bogus), skip it.
720 if (Tok.isNot(K: tok::raw_identifier)) {
721 CurPPLexer->ParsingPreprocessorDirective = false;
722 // Restore comment saving mode.
723 if (CurLexer) CurLexer->resetExtendedTokenMode();
724 continue;
725 }
726
727 // If the first letter isn't i or e, it isn't intesting to us. We know that
728 // this is safe in the face of spelling differences, because there is no way
729 // to spell an i/e in a strange way that is another letter. Skipping this
730 // allows us to avoid looking up the identifier info for #define/#undef and
731 // other common directives.
732 StringRef RI = Tok.getRawIdentifier();
733
734 char FirstChar = RI[0];
735 if (FirstChar >= 'a' && FirstChar <= 'z' &&
736 FirstChar != 'i' && FirstChar != 'e') {
737 CurPPLexer->ParsingPreprocessorDirective = false;
738 // Restore comment saving mode.
739 if (CurLexer) CurLexer->resetExtendedTokenMode();
740 continue;
741 }
742
743 // Get the identifier name without trigraphs or embedded newlines. Note
744 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
745 // when skipping.
746 char DirectiveBuf[20];
747 StringRef Directive;
748 if (!Tok.needsCleaning() && RI.size() < 20) {
749 Directive = RI;
750 } else {
751 std::string DirectiveStr = getSpelling(Tok);
752 size_t IdLen = DirectiveStr.size();
753 if (IdLen >= 20) {
754 CurPPLexer->ParsingPreprocessorDirective = false;
755 // Restore comment saving mode.
756 if (CurLexer) CurLexer->resetExtendedTokenMode();
757 continue;
758 }
759 memcpy(dest: DirectiveBuf, src: &DirectiveStr[0], n: IdLen);
760 Directive = StringRef(DirectiveBuf, IdLen);
761 }
762
763 if (Directive.starts_with(Prefix: "if")) {
764 StringRef Sub = Directive.substr(Start: 2);
765 if (Sub.empty() || // "if"
766 Sub == "def" || // "ifdef"
767 Sub == "ndef") { // "ifndef"
768 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
769 // bother parsing the condition.
770 DiscardUntilEndOfDirective();
771 CurPPLexer->pushConditionalLevel(DirectiveStart: Tok.getLocation(), /*wasskipping*/WasSkipping: true,
772 /*foundnonskip*/FoundNonSkip: false,
773 /*foundelse*/FoundElse: false);
774 } else {
775 SuggestTypoedDirective(Tok, Directive);
776 }
777 } else if (Directive[0] == 'e') {
778 StringRef Sub = Directive.substr(Start: 1);
779 if (Sub == "ndif") { // "endif"
780 PPConditionalInfo CondInfo;
781 CondInfo.WasSkipping = true; // Silence bogus warning.
782 bool InCond = CurPPLexer->popConditionalLevel(CI&: CondInfo);
783 (void)InCond; // Silence warning in no-asserts mode.
784 assert(!InCond && "Can't be skipping if not in a conditional!");
785
786 // If we popped the outermost skipping block, we're done skipping!
787 if (!CondInfo.WasSkipping) {
788 SkippingRangeState.endLexPass(Hashptr);
789 // Restore the value of LexingRawMode so that trailing comments
790 // are handled correctly, if we've reached the outermost block.
791 CurPPLexer->LexingRawMode = false;
792 endLoc = CheckEndOfDirective(DirType: "endif");
793 CurPPLexer->LexingRawMode = true;
794 if (Callbacks)
795 Callbacks->Endif(Loc: Tok.getLocation(), IfLoc: CondInfo.IfLoc);
796 break;
797 } else {
798 DiscardUntilEndOfDirective();
799 }
800 } else if (Sub == "lse") { // "else".
801 // #else directive in a skipping conditional. If not in some other
802 // skipping conditional, and if #else hasn't already been seen, enter it
803 // as a non-skipping conditional.
804 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
805
806 if (!CondInfo.WasSkipping)
807 SkippingRangeState.endLexPass(Hashptr);
808
809 // If this is a #else with a #else before it, report the error.
810 if (CondInfo.FoundElse)
811 Diag(Tok, DiagID: diag::pp_err_else_after_else);
812
813 // Note that we've seen a #else in this conditional.
814 CondInfo.FoundElse = true;
815
816 // If the conditional is at the top level, and the #if block wasn't
817 // entered, enter the #else block now.
818 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
819 CondInfo.FoundNonSkip = true;
820 // Restore the value of LexingRawMode so that trailing comments
821 // are handled correctly.
822 CurPPLexer->LexingRawMode = false;
823 endLoc = CheckEndOfDirective(DirType: "else");
824 CurPPLexer->LexingRawMode = true;
825 if (Callbacks)
826 Callbacks->Else(Loc: Tok.getLocation(), IfLoc: CondInfo.IfLoc);
827 break;
828 } else {
829 DiscardUntilEndOfDirective(); // C99 6.10p4.
830 }
831 } else if (Sub == "lif") { // "elif".
832 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
833
834 if (!CondInfo.WasSkipping)
835 SkippingRangeState.endLexPass(Hashptr);
836
837 // If this is a #elif with a #else before it, report the error.
838 if (CondInfo.FoundElse)
839 Diag(Tok, DiagID: diag::pp_err_elif_after_else) << PED_Elif;
840
841 // If this is in a skipping block or if we're already handled this #if
842 // block, don't bother parsing the condition.
843 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
844 // FIXME: We should probably do at least some minimal parsing of the
845 // condition to verify that it is well-formed. The current state
846 // allows #elif* directives with completely malformed (or missing)
847 // conditions.
848 DiscardUntilEndOfDirective();
849 } else {
850 // Restore the value of LexingRawMode so that identifiers are
851 // looked up, etc, inside the #elif expression.
852 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
853 CurPPLexer->LexingRawMode = false;
854 IdentifierInfo *IfNDefMacro = nullptr;
855 DirectiveEvalResult DER = EvaluateDirectiveExpression(IfNDefMacro);
856 // Stop if Lexer became invalid after hitting code completion token.
857 if (!CurPPLexer)
858 return;
859 const bool CondValue = DER.Conditional;
860 CurPPLexer->LexingRawMode = true;
861 if (Callbacks) {
862 Callbacks->Elif(
863 Loc: Tok.getLocation(), ConditionRange: DER.ExprRange,
864 ConditionValue: (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False),
865 IfLoc: CondInfo.IfLoc);
866 }
867 // If this condition is true, enter it!
868 if (CondValue) {
869 CondInfo.FoundNonSkip = true;
870 break;
871 }
872 }
873 } else if (Sub == "lifdef" || // "elifdef"
874 Sub == "lifndef") { // "elifndef"
875 bool IsElifDef = Sub == "lifdef";
876 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
877 Token DirectiveToken = Tok;
878
879 if (!CondInfo.WasSkipping)
880 SkippingRangeState.endLexPass(Hashptr);
881
882 // Warn if using `#elifdef` & `#elifndef` in not C23 & C++23 mode even
883 // if this branch is in a skipping block.
884 unsigned DiagID;
885 if (LangOpts.CPlusPlus)
886 DiagID = LangOpts.CPlusPlus23 ? diag::warn_cxx23_compat_pp_directive
887 : diag::ext_cxx23_pp_directive;
888 else
889 DiagID = LangOpts.C23 ? diag::warn_c23_compat_pp_directive
890 : diag::ext_c23_pp_directive;
891 Diag(Tok, DiagID) << (IsElifDef ? PED_Elifdef : PED_Elifndef);
892
893 // If this is a #elif with a #else before it, report the error.
894 if (CondInfo.FoundElse)
895 Diag(Tok, DiagID: diag::pp_err_elif_after_else)
896 << (IsElifDef ? PED_Elifdef : PED_Elifndef);
897
898 // If this is in a skipping block or if we're already handled this #if
899 // block, don't bother parsing the condition.
900 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
901 // FIXME: We should probably do at least some minimal parsing of the
902 // condition to verify that it is well-formed. The current state
903 // allows #elif* directives with completely malformed (or missing)
904 // conditions.
905 DiscardUntilEndOfDirective();
906 } else {
907 // Restore the value of LexingRawMode so that identifiers are
908 // looked up, etc, inside the #elif[n]def expression.
909 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
910 CurPPLexer->LexingRawMode = false;
911 Token MacroNameTok;
912 ReadMacroName(MacroNameTok);
913 CurPPLexer->LexingRawMode = true;
914
915 // If the macro name token is tok::eod, there was an error that was
916 // already reported.
917 if (MacroNameTok.is(K: tok::eod)) {
918 // Skip code until we get to #endif. This helps with recovery by
919 // not emitting an error when the #endif is reached.
920 continue;
921 }
922
923 emitMacroExpansionWarnings(Identifier: MacroNameTok);
924
925 CheckEndOfDirective(DirType: IsElifDef ? "elifdef" : "elifndef");
926
927 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
928 auto MD = getMacroDefinition(II: MII);
929 MacroInfo *MI = MD.getMacroInfo();
930
931 if (Callbacks) {
932 if (IsElifDef) {
933 Callbacks->Elifdef(Loc: DirectiveToken.getLocation(), MacroNameTok,
934 MD);
935 } else {
936 Callbacks->Elifndef(Loc: DirectiveToken.getLocation(), MacroNameTok,
937 MD);
938 }
939 }
940 // If this condition is true, enter it!
941 if (static_cast<bool>(MI) == IsElifDef) {
942 CondInfo.FoundNonSkip = true;
943 break;
944 }
945 }
946 } else {
947 SuggestTypoedDirective(Tok, Directive);
948 }
949 } else {
950 SuggestTypoedDirective(Tok, Directive);
951 }
952
953 CurPPLexer->ParsingPreprocessorDirective = false;
954 // Restore comment saving mode.
955 if (CurLexer) CurLexer->resetExtendedTokenMode();
956 }
957
958 // Finally, if we are out of the conditional (saw an #endif or ran off the end
959 // of the file, just stop skipping and return to lexing whatever came after
960 // the #if block.
961 CurPPLexer->LexingRawMode = false;
962
963 // The last skipped range isn't actually skipped yet if it's truncated
964 // by the end of the preamble; we'll resume parsing after the preamble.
965 if (Callbacks && (Tok.isNot(K: tok::eof) || !isRecordingPreamble()))
966 Callbacks->SourceRangeSkipped(
967 Range: SourceRange(HashTokenLoc, endLoc.isValid()
968 ? endLoc
969 : CurPPLexer->getSourceLocation()),
970 EndifLoc: Tok.getLocation());
971}
972
973Module *Preprocessor::getModuleForLocation(SourceLocation Loc,
974 bool AllowTextual) {
975 if (!SourceMgr.isInMainFile(Loc)) {
976 // Try to determine the module of the include directive.
977 // FIXME: Look into directly passing the FileEntry from LookupFile instead.
978 FileID IDOfIncl = SourceMgr.getFileID(SpellingLoc: SourceMgr.getExpansionLoc(Loc));
979 if (auto EntryOfIncl = SourceMgr.getFileEntryRefForID(FID: IDOfIncl)) {
980 // The include comes from an included file.
981 return HeaderInfo.getModuleMap()
982 .findModuleForHeader(File: *EntryOfIncl, AllowTextual)
983 .getModule();
984 }
985 }
986
987 // This is either in the main file or not in a file at all. It belongs
988 // to the current module, if there is one.
989 return getLangOpts().CurrentModule.empty()
990 ? nullptr
991 : HeaderInfo.lookupModule(ModuleName: getLangOpts().CurrentModule, ImportLoc: Loc);
992}
993
994OptionalFileEntryRef
995Preprocessor::getHeaderToIncludeForDiagnostics(SourceLocation IncLoc,
996 SourceLocation Loc) {
997 Module *IncM = getModuleForLocation(
998 Loc: IncLoc, AllowTextual: LangOpts.ModulesValidateTextualHeaderIncludes);
999
1000 // Walk up through the include stack, looking through textual headers of M
1001 // until we hit a non-textual header that we can #include. (We assume textual
1002 // headers of a module with non-textual headers aren't meant to be used to
1003 // import entities from the module.)
1004 auto &SM = getSourceManager();
1005 while (!Loc.isInvalid() && !SM.isInMainFile(Loc)) {
1006 auto ID = SM.getFileID(SpellingLoc: SM.getExpansionLoc(Loc));
1007 auto FE = SM.getFileEntryRefForID(FID: ID);
1008 if (!FE)
1009 break;
1010
1011 // We want to find all possible modules that might contain this header, so
1012 // search all enclosing directories for module maps and load them.
1013 HeaderInfo.hasModuleMap(Filename: FE->getName(), /*Root*/ nullptr,
1014 IsSystem: SourceMgr.isInSystemHeader(Loc));
1015
1016 bool InPrivateHeader = false;
1017 for (auto Header : HeaderInfo.findAllModulesForHeader(File: *FE)) {
1018 if (!Header.isAccessibleFrom(M: IncM)) {
1019 // It's in a private header; we can't #include it.
1020 // FIXME: If there's a public header in some module that re-exports it,
1021 // then we could suggest including that, but it's not clear that's the
1022 // expected way to make this entity visible.
1023 InPrivateHeader = true;
1024 continue;
1025 }
1026
1027 // Don't suggest explicitly excluded headers.
1028 if (Header.getRole() == ModuleMap::ExcludedHeader)
1029 continue;
1030
1031 // We'll suggest including textual headers below if they're
1032 // include-guarded.
1033 if (Header.getRole() & ModuleMap::TextualHeader)
1034 continue;
1035
1036 // If we have a module import syntax, we shouldn't include a header to
1037 // make a particular module visible. Let the caller know they should
1038 // suggest an import instead.
1039 if (getLangOpts().ObjC || getLangOpts().CPlusPlusModules)
1040 return std::nullopt;
1041
1042 // If this is an accessible, non-textual header of M's top-level module
1043 // that transitively includes the given location and makes the
1044 // corresponding module visible, this is the thing to #include.
1045 return *FE;
1046 }
1047
1048 // FIXME: If we're bailing out due to a private header, we shouldn't suggest
1049 // an import either.
1050 if (InPrivateHeader)
1051 return std::nullopt;
1052
1053 // If the header is includable and has an include guard, assume the
1054 // intended way to expose its contents is by #include, not by importing a
1055 // module that transitively includes it.
1056 if (getHeaderSearchInfo().isFileMultipleIncludeGuarded(File: *FE))
1057 return *FE;
1058
1059 Loc = SM.getIncludeLoc(FID: ID);
1060 }
1061
1062 return std::nullopt;
1063}
1064
1065OptionalFileEntryRef Preprocessor::LookupFile(
1066 SourceLocation FilenameLoc, StringRef Filename, bool isAngled,
1067 ConstSearchDirIterator FromDir, const FileEntry *FromFile,
1068 ConstSearchDirIterator *CurDirArg, SmallVectorImpl<char> *SearchPath,
1069 SmallVectorImpl<char> *RelativePath,
1070 ModuleMap::KnownHeader *SuggestedModule, bool *IsMapped,
1071 bool *IsFrameworkFound, bool SkipCache, bool OpenFile, bool CacheFailures) {
1072 ConstSearchDirIterator CurDirLocal = nullptr;
1073 ConstSearchDirIterator &CurDir = CurDirArg ? *CurDirArg : CurDirLocal;
1074
1075 Module *RequestingModule = getModuleForLocation(
1076 Loc: FilenameLoc, AllowTextual: LangOpts.ModulesValidateTextualHeaderIncludes);
1077
1078 // If the header lookup mechanism may be relative to the current inclusion
1079 // stack, record the parent #includes.
1080 SmallVector<std::pair<OptionalFileEntryRef, DirectoryEntryRef>, 16> Includers;
1081 bool BuildSystemModule = false;
1082 if (!FromDir && !FromFile) {
1083 FileID FID = getCurrentFileLexer()->getFileID();
1084 OptionalFileEntryRef FileEnt = SourceMgr.getFileEntryRefForID(FID);
1085
1086 // If there is no file entry associated with this file, it must be the
1087 // predefines buffer or the module includes buffer. Any other file is not
1088 // lexed with a normal lexer, so it won't be scanned for preprocessor
1089 // directives.
1090 //
1091 // If we have the predefines buffer, resolve #include references (which come
1092 // from the -include command line argument) from the current working
1093 // directory instead of relative to the main file.
1094 //
1095 // If we have the module includes buffer, resolve #include references (which
1096 // come from header declarations in the module map) relative to the module
1097 // map file.
1098 if (!FileEnt) {
1099 if (FID == SourceMgr.getMainFileID() && MainFileDir) {
1100 auto IncludeDir =
1101 HeaderInfo.getModuleMap().shouldImportRelativeToBuiltinIncludeDir(
1102 FileName: Filename, Module: getCurrentModule())
1103 ? HeaderInfo.getModuleMap().getBuiltinDir()
1104 : MainFileDir;
1105 Includers.push_back(Elt: std::make_pair(x: std::nullopt, y&: *IncludeDir));
1106 BuildSystemModule = getCurrentModule()->IsSystem;
1107 } else if ((FileEnt = SourceMgr.getFileEntryRefForID(
1108 FID: SourceMgr.getMainFileID()))) {
1109 auto CWD = FileMgr.getOptionalDirectoryRef(DirName: ".");
1110 Includers.push_back(Elt: std::make_pair(x&: *FileEnt, y&: *CWD));
1111 }
1112 } else {
1113 Includers.push_back(Elt: std::make_pair(x&: *FileEnt, y: FileEnt->getDir()));
1114 }
1115
1116 // MSVC searches the current include stack from top to bottom for
1117 // headers included by quoted include directives.
1118 // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx
1119 if (LangOpts.MSVCCompat && !isAngled) {
1120 for (IncludeStackInfo &ISEntry : llvm::reverse(C&: IncludeMacroStack)) {
1121 if (IsFileLexer(I: ISEntry))
1122 if ((FileEnt = ISEntry.ThePPLexer->getFileEntry()))
1123 Includers.push_back(Elt: std::make_pair(x&: *FileEnt, y: FileEnt->getDir()));
1124 }
1125 }
1126 }
1127
1128 CurDir = CurDirLookup;
1129
1130 if (FromFile) {
1131 // We're supposed to start looking from after a particular file. Search
1132 // the include path until we find that file or run out of files.
1133 ConstSearchDirIterator TmpCurDir = CurDir;
1134 ConstSearchDirIterator TmpFromDir = nullptr;
1135 while (OptionalFileEntryRef FE = HeaderInfo.LookupFile(
1136 Filename, IncludeLoc: FilenameLoc, isAngled, FromDir: TmpFromDir, CurDir: &TmpCurDir,
1137 Includers, SearchPath, RelativePath, RequestingModule,
1138 SuggestedModule, /*IsMapped=*/nullptr,
1139 /*IsFrameworkFound=*/nullptr, SkipCache)) {
1140 // Keep looking as if this file did a #include_next.
1141 TmpFromDir = TmpCurDir;
1142 ++TmpFromDir;
1143 if (&FE->getFileEntry() == FromFile) {
1144 // Found it.
1145 FromDir = TmpFromDir;
1146 CurDir = TmpCurDir;
1147 break;
1148 }
1149 }
1150 }
1151
1152 // Do a standard file entry lookup.
1153 OptionalFileEntryRef FE = HeaderInfo.LookupFile(
1154 Filename, IncludeLoc: FilenameLoc, isAngled, FromDir, CurDir: &CurDir, Includers, SearchPath,
1155 RelativePath, RequestingModule, SuggestedModule, IsMapped,
1156 IsFrameworkFound, SkipCache, BuildSystemModule, OpenFile, CacheFailures);
1157 if (FE)
1158 return FE;
1159
1160 OptionalFileEntryRef CurFileEnt;
1161 // Otherwise, see if this is a subframework header. If so, this is relative
1162 // to one of the headers on the #include stack. Walk the list of the current
1163 // headers on the #include stack and pass them to HeaderInfo.
1164 if (IsFileLexer()) {
1165 if ((CurFileEnt = CurPPLexer->getFileEntry())) {
1166 if (OptionalFileEntryRef FE = HeaderInfo.LookupSubframeworkHeader(
1167 Filename, ContextFileEnt: *CurFileEnt, SearchPath, RelativePath, RequestingModule,
1168 SuggestedModule)) {
1169 return FE;
1170 }
1171 }
1172 }
1173
1174 for (IncludeStackInfo &ISEntry : llvm::reverse(C&: IncludeMacroStack)) {
1175 if (IsFileLexer(I: ISEntry)) {
1176 if ((CurFileEnt = ISEntry.ThePPLexer->getFileEntry())) {
1177 if (OptionalFileEntryRef FE = HeaderInfo.LookupSubframeworkHeader(
1178 Filename, ContextFileEnt: *CurFileEnt, SearchPath, RelativePath,
1179 RequestingModule, SuggestedModule)) {
1180 return FE;
1181 }
1182 }
1183 }
1184 }
1185
1186 // Otherwise, we really couldn't find the file.
1187 return std::nullopt;
1188}
1189
1190OptionalFileEntryRef Preprocessor::LookupEmbedFile(StringRef Filename,
1191 bool isAngled,
1192 bool OpenFile) {
1193 FileManager &FM = this->getFileManager();
1194 if (llvm::sys::path::is_absolute(path: Filename)) {
1195 // lookup path or immediately fail
1196 llvm::Expected<FileEntryRef> ShouldBeEntry = FM.getFileRef(
1197 Filename, OpenFile, /*CacheFailure=*/true, /*IsText=*/false);
1198 return llvm::expectedToOptional(E: std::move(ShouldBeEntry));
1199 }
1200
1201 auto SeparateComponents = [](SmallVectorImpl<char> &LookupPath,
1202 StringRef StartingFrom, StringRef FileName,
1203 bool RemoveInitialFileComponentFromLookupPath) {
1204 llvm::sys::path::native(path: StartingFrom, result&: LookupPath);
1205 if (RemoveInitialFileComponentFromLookupPath)
1206 llvm::sys::path::remove_filename(path&: LookupPath);
1207 if (!LookupPath.empty() &&
1208 !llvm::sys::path::is_separator(value: LookupPath.back())) {
1209 LookupPath.push_back(Elt: llvm::sys::path::get_separator().front());
1210 }
1211 LookupPath.append(in_start: FileName.begin(), in_end: FileName.end());
1212 };
1213
1214 // Otherwise, it's search time!
1215 SmallString<512> LookupPath;
1216 // Non-angled lookup
1217 if (!isAngled) {
1218 OptionalFileEntryRef LookupFromFile = getCurrentFileLexer()->getFileEntry();
1219 if (LookupFromFile) {
1220 // Use file-based lookup.
1221 SmallString<1024> TmpDir;
1222 TmpDir = LookupFromFile->getDir().getName();
1223 llvm::sys::path::append(path&: TmpDir, a: Filename);
1224 if (!TmpDir.empty()) {
1225 llvm::Expected<FileEntryRef> ShouldBeEntry = FM.getFileRef(
1226 Filename: TmpDir, OpenFile, /*CacheFailure=*/true, /*IsText=*/false);
1227 if (ShouldBeEntry)
1228 return llvm::expectedToOptional(E: std::move(ShouldBeEntry));
1229 llvm::consumeError(Err: ShouldBeEntry.takeError());
1230 }
1231 }
1232
1233 // Otherwise, do working directory lookup.
1234 LookupPath.clear();
1235 auto MaybeWorkingDirEntry = FM.getDirectoryRef(DirName: ".");
1236 if (MaybeWorkingDirEntry) {
1237 DirectoryEntryRef WorkingDirEntry = *MaybeWorkingDirEntry;
1238 StringRef WorkingDir = WorkingDirEntry.getName();
1239 if (!WorkingDir.empty()) {
1240 SeparateComponents(LookupPath, WorkingDir, Filename, false);
1241 llvm::Expected<FileEntryRef> ShouldBeEntry = FM.getFileRef(
1242 Filename: LookupPath, OpenFile, /*CacheFailure=*/true, /*IsText=*/false);
1243 if (ShouldBeEntry)
1244 return llvm::expectedToOptional(E: std::move(ShouldBeEntry));
1245 llvm::consumeError(Err: ShouldBeEntry.takeError());
1246 }
1247 }
1248 }
1249
1250 for (const auto &Entry : PPOpts.EmbedEntries) {
1251 LookupPath.clear();
1252 SeparateComponents(LookupPath, Entry, Filename, false);
1253 llvm::Expected<FileEntryRef> ShouldBeEntry = FM.getFileRef(
1254 Filename: LookupPath, OpenFile, /*CacheFailure=*/true, /*IsText=*/false);
1255 if (ShouldBeEntry)
1256 return llvm::expectedToOptional(E: std::move(ShouldBeEntry));
1257 llvm::consumeError(Err: ShouldBeEntry.takeError());
1258 }
1259 return std::nullopt;
1260}
1261
1262//===----------------------------------------------------------------------===//
1263// Preprocessor Directive Handling.
1264//===----------------------------------------------------------------------===//
1265
1266class Preprocessor::ResetMacroExpansionHelper {
1267public:
1268 ResetMacroExpansionHelper(Preprocessor *pp)
1269 : PP(pp), save(pp->DisableMacroExpansion) {
1270 if (pp->MacroExpansionInDirectivesOverride)
1271 pp->DisableMacroExpansion = false;
1272 }
1273
1274 ~ResetMacroExpansionHelper() {
1275 PP->DisableMacroExpansion = save;
1276 }
1277
1278private:
1279 Preprocessor *PP;
1280 bool save;
1281};
1282
1283/// Process a directive while looking for the through header or a #pragma
1284/// hdrstop. The following directives are handled:
1285/// #include (to check if it is the through header)
1286/// #define (to warn about macros that don't match the PCH)
1287/// #pragma (to check for pragma hdrstop).
1288/// All other directives are completely discarded.
1289void Preprocessor::HandleSkippedDirectiveWhileUsingPCH(Token &Result,
1290 SourceLocation HashLoc) {
1291 if (const IdentifierInfo *II = Result.getIdentifierInfo()) {
1292 if (II->getPPKeywordID() == tok::pp_define) {
1293 return HandleDefineDirective(Tok&: Result,
1294 /*ImmediatelyAfterHeaderGuard=*/false);
1295 }
1296 if (SkippingUntilPCHThroughHeader &&
1297 II->getPPKeywordID() == tok::pp_include) {
1298 return HandleIncludeDirective(HashLoc, Tok&: Result);
1299 }
1300 if (SkippingUntilPragmaHdrStop && II->getPPKeywordID() == tok::pp_pragma) {
1301 Lex(Result);
1302 auto *II = Result.getIdentifierInfo();
1303 if (II && II->getName() == "hdrstop")
1304 return HandlePragmaHdrstop(Tok&: Result);
1305 }
1306 }
1307 DiscardUntilEndOfDirective();
1308}
1309
1310/// HandleDirective - This callback is invoked when the lexer sees a # token
1311/// at the start of a line. This consumes the directive, modifies the
1312/// lexer/preprocessor state, and advances the lexer(s) so that the next token
1313/// read is the correct one.
1314void Preprocessor::HandleDirective(Token &Result) {
1315 // FIXME: Traditional: # with whitespace before it not recognized by K&R?
1316
1317 // We just parsed a # character at the start of a line, so we're in directive
1318 // mode. Tell the lexer this so any newlines we see will be converted into an
1319 // EOD token (which terminates the directive).
1320 CurPPLexer->ParsingPreprocessorDirective = true;
1321 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
1322
1323 bool ImmediatelyAfterTopLevelIfndef =
1324 CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef();
1325 CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef();
1326
1327 ++NumDirectives;
1328
1329 // We are about to read a token. For the multiple-include optimization FA to
1330 // work, we have to remember if we had read any tokens *before* this
1331 // pp-directive.
1332 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
1333
1334 // Save the directive-introducing token('#' and import/module in C++20) in
1335 // case we need to return it later.
1336 Token Introducer = Result;
1337
1338 // Read the next token, the directive flavor. This isn't expanded due to
1339 // C99 6.10.3p8.
1340 if (Introducer.is(K: tok::hash))
1341 LexUnexpandedToken(Result);
1342
1343 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.:
1344 // #define A(x) #x
1345 // A(abc
1346 // #warning blah
1347 // def)
1348 // If so, the user is relying on undefined behavior, emit a diagnostic. Do
1349 // not support this for #include-like directives, since that can result in
1350 // terrible diagnostics, and does not work in GCC.
1351 if (InMacroArgs) {
1352 if (IdentifierInfo *II = Result.getIdentifierInfo()) {
1353 switch (II->getPPKeywordID()) {
1354 case tok::pp_include:
1355 case tok::pp_import:
1356 case tok::pp_include_next:
1357 case tok::pp___include_macros:
1358 case tok::pp_pragma:
1359 case tok::pp_embed:
1360 case tok::pp_module:
1361 case tok::pp___preprocessed_module:
1362 case tok::pp___preprocessed_import:
1363 Diag(Tok: Result, DiagID: diag::err_embedded_directive)
1364 << (getLangOpts().CPlusPlusModules &&
1365 Introducer.isModuleContextualKeyword(
1366 /*AllowExport=*/false))
1367 << II->getName();
1368 Diag(Tok: *ArgMacro, DiagID: diag::note_macro_expansion_here)
1369 << ArgMacro->getIdentifierInfo();
1370 DiscardUntilEndOfDirective();
1371 return;
1372 default:
1373 break;
1374 }
1375 }
1376 Diag(Tok: Result, DiagID: diag::ext_embedded_directive);
1377 }
1378
1379 // Temporarily enable macro expansion if set so
1380 // and reset to previous state when returning from this function.
1381 ResetMacroExpansionHelper helper(this);
1382
1383 if (SkippingUntilPCHThroughHeader || SkippingUntilPragmaHdrStop)
1384 return HandleSkippedDirectiveWhileUsingPCH(Result,
1385 HashLoc: Introducer.getLocation());
1386
1387 switch (Result.getKind()) {
1388 case tok::eod:
1389 // Ignore the null directive with regards to the multiple-include
1390 // optimization, i.e. allow the null directive to appear outside of the
1391 // include guard and still enable the multiple-include optimization.
1392 CurPPLexer->MIOpt.SetReadToken(ReadAnyTokensBeforeDirective);
1393 return; // null directive.
1394 case tok::code_completion:
1395 setCodeCompletionReached();
1396 if (CodeComplete)
1397 CodeComplete->CodeCompleteDirective(
1398 InConditional: CurPPLexer->getConditionalStackDepth() > 0);
1399 return;
1400 case tok::numeric_constant: // # 7 GNU line marker directive.
1401 // In a .S file "# 4" may be a comment so don't treat it as a preprocessor
1402 // directive. However do permit it in the predefines file, as we use line
1403 // markers to mark the builtin macros as being in a system header.
1404 if (getLangOpts().AsmPreprocessor &&
1405 SourceMgr.getFileID(SpellingLoc: Introducer.getLocation()) != getPredefinesFileID())
1406 break;
1407 return HandleDigitDirective(Tok&: Result);
1408 default:
1409 IdentifierInfo *II = Result.getIdentifierInfo();
1410 if (!II) break; // Not an identifier.
1411
1412 // Ask what the preprocessor keyword ID is.
1413 switch (II->getPPKeywordID()) {
1414 default: break;
1415 // C99 6.10.1 - Conditional Inclusion.
1416 case tok::pp_if:
1417 return HandleIfDirective(IfToken&: Result, HashToken: Introducer,
1418 ReadAnyTokensBeforeDirective);
1419 case tok::pp_ifdef:
1420 return HandleIfdefDirective(Result, HashToken: Introducer, isIfndef: false,
1421 ReadAnyTokensBeforeDirective: true /*not valid for miopt*/);
1422 case tok::pp_ifndef:
1423 return HandleIfdefDirective(Result, HashToken: Introducer, isIfndef: true,
1424 ReadAnyTokensBeforeDirective);
1425 case tok::pp_elif:
1426 case tok::pp_elifdef:
1427 case tok::pp_elifndef:
1428 return HandleElifFamilyDirective(ElifToken&: Result, HashToken: Introducer,
1429 Kind: II->getPPKeywordID());
1430
1431 case tok::pp_else:
1432 return HandleElseDirective(Result, HashToken: Introducer);
1433 case tok::pp_endif:
1434 return HandleEndifDirective(EndifToken&: Result);
1435
1436 // C99 6.10.2 - Source File Inclusion.
1437 case tok::pp_include:
1438 // Handle #include.
1439 return HandleIncludeDirective(HashLoc: Introducer.getLocation(), Tok&: Result);
1440 case tok::pp___include_macros:
1441 // Handle -imacros.
1442 return HandleIncludeMacrosDirective(HashLoc: Introducer.getLocation(), Tok&: Result);
1443
1444 // C99 6.10.3 - Macro Replacement.
1445 case tok::pp_define:
1446 return HandleDefineDirective(Tok&: Result, ImmediatelyAfterHeaderGuard: ImmediatelyAfterTopLevelIfndef);
1447 case tok::pp_undef:
1448 return HandleUndefDirective();
1449
1450 // C99 6.10.4 - Line Control.
1451 case tok::pp_line:
1452 return HandleLineDirective();
1453
1454 // C99 6.10.5 - Error Directive.
1455 case tok::pp_error:
1456 return HandleUserDiagnosticDirective(Tok&: Result, isWarning: false);
1457
1458 // C99 6.10.6 - Pragma Directive.
1459 case tok::pp_pragma:
1460 return HandlePragmaDirective(Introducer: {.Kind: PIK_HashPragma, .Loc: Introducer.getLocation()});
1461 case tok::pp_module:
1462 case tok::pp___preprocessed_module:
1463 return HandleCXXModuleDirective(Module: Result);
1464 case tok::pp___preprocessed_import:
1465 return HandleCXXImportDirective(Import: Result);
1466 // GNU Extensions.
1467 case tok::pp_import:
1468 if (getLangOpts().CPlusPlusModules &&
1469 Introducer.isModuleContextualKeyword(
1470 /*AllowExport=*/false))
1471 return HandleCXXImportDirective(Import: Result);
1472 return HandleImportDirective(HashLoc: Introducer.getLocation(), Tok&: Result);
1473 case tok::pp_include_next:
1474 return HandleIncludeNextDirective(HashLoc: Introducer.getLocation(), Tok&: Result);
1475
1476 case tok::pp_warning:
1477 if (LangOpts.CPlusPlus)
1478 Diag(Tok: Result, DiagID: LangOpts.CPlusPlus23
1479 ? diag::warn_cxx23_compat_warning_directive
1480 : diag::ext_pp_warning_directive)
1481 << /*C++23*/ 1;
1482 else
1483 Diag(Tok: Result, DiagID: LangOpts.C23 ? diag::warn_c23_compat_warning_directive
1484 : diag::ext_pp_warning_directive)
1485 << /*C23*/ 0;
1486
1487 return HandleUserDiagnosticDirective(Tok&: Result, isWarning: true);
1488 case tok::pp_ident:
1489 return HandleIdentSCCSDirective(Tok&: Result);
1490 case tok::pp_sccs:
1491 return HandleIdentSCCSDirective(Tok&: Result);
1492 case tok::pp_embed:
1493 return HandleEmbedDirective(HashLoc: Introducer.getLocation(), Tok&: Result);
1494 case tok::pp_assert:
1495 //isExtension = true; // FIXME: implement #assert
1496 break;
1497 case tok::pp_unassert:
1498 //isExtension = true; // FIXME: implement #unassert
1499 break;
1500
1501 case tok::pp___public_macro:
1502 if (getLangOpts().Modules || getLangOpts().ModulesLocalVisibility)
1503 return HandleMacroPublicDirective(Tok&: Result);
1504 break;
1505
1506 case tok::pp___private_macro:
1507 if (getLangOpts().Modules || getLangOpts().ModulesLocalVisibility)
1508 return HandleMacroPrivateDirective();
1509 break;
1510 }
1511 break;
1512 }
1513
1514 // If this is a .S file, treat unknown # directives as non-preprocessor
1515 // directives. This is important because # may be a comment or introduce
1516 // various pseudo-ops. Just return the # token and push back the following
1517 // token to be lexed next time.
1518 if (getLangOpts().AsmPreprocessor) {
1519 auto Toks = std::make_unique<Token[]>(num: 2);
1520 // Return the # and the token after it.
1521 Toks[0] = Introducer;
1522 Toks[1] = Result;
1523
1524 // If the second token is a hashhash token, then we need to translate it to
1525 // unknown so the token lexer doesn't try to perform token pasting.
1526 if (Result.is(K: tok::hashhash))
1527 Toks[1].setKind(tok::unknown);
1528
1529 // Enter this token stream so that we re-lex the tokens. Make sure to
1530 // enable macro expansion, in case the token after the # is an identifier
1531 // that is expanded.
1532 EnterTokenStream(Toks: std::move(Toks), NumToks: 2, DisableMacroExpansion: false, /*IsReinject*/false);
1533 return;
1534 }
1535
1536 // If we reached here, the preprocessing token is not valid!
1537 // Start suggesting if a similar directive found.
1538 Diag(Tok: Result, DiagID: diag::err_pp_invalid_directive) << 0;
1539
1540 // Read the rest of the PP line.
1541 DiscardUntilEndOfDirective();
1542
1543 // Okay, we're done parsing the directive.
1544}
1545
1546/// GetLineValue - Convert a numeric token into an unsigned value, emitting
1547/// Diagnostic DiagID if it is invalid, and returning the value in Val.
1548static bool GetLineValue(Token &DigitTok, unsigned &Val,
1549 unsigned DiagID, Preprocessor &PP,
1550 bool IsGNULineDirective=false) {
1551 if (DigitTok.isNot(K: tok::numeric_constant)) {
1552 PP.Diag(Tok: DigitTok, DiagID);
1553
1554 if (DigitTok.isNot(K: tok::eod))
1555 PP.DiscardUntilEndOfDirective();
1556 return true;
1557 }
1558
1559 SmallString<64> IntegerBuffer;
1560 IntegerBuffer.resize(N: DigitTok.getLength());
1561 const char *DigitTokBegin = &IntegerBuffer[0];
1562 bool Invalid = false;
1563 unsigned ActualLength = PP.getSpelling(Tok: DigitTok, Buffer&: DigitTokBegin, Invalid: &Invalid);
1564 if (Invalid)
1565 return true;
1566
1567 // Verify that we have a simple digit-sequence, and compute the value. This
1568 // is always a simple digit string computed in decimal, so we do this manually
1569 // here.
1570 Val = 0;
1571 for (unsigned i = 0; i != ActualLength; ++i) {
1572 // C++1y [lex.fcon]p1:
1573 // Optional separating single quotes in a digit-sequence are ignored
1574 if (DigitTokBegin[i] == '\'')
1575 continue;
1576
1577 if (!isDigit(c: DigitTokBegin[i])) {
1578 PP.Diag(Loc: PP.AdvanceToTokenCharacter(TokStart: DigitTok.getLocation(), Char: i),
1579 DiagID: diag::err_pp_line_digit_sequence) << IsGNULineDirective;
1580 PP.DiscardUntilEndOfDirective();
1581 return true;
1582 }
1583
1584 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
1585 if (NextVal < Val) { // overflow.
1586 PP.Diag(Tok: DigitTok, DiagID);
1587 PP.DiscardUntilEndOfDirective();
1588 return true;
1589 }
1590 Val = NextVal;
1591 }
1592
1593 if (DigitTokBegin[0] == '0' && Val)
1594 PP.Diag(Loc: DigitTok.getLocation(), DiagID: diag::warn_pp_line_decimal)
1595 << IsGNULineDirective;
1596
1597 return false;
1598}
1599
1600/// Handle a \#line directive: C99 6.10.4.
1601///
1602/// The two acceptable forms are:
1603/// \verbatim
1604/// # line digit-sequence
1605/// # line digit-sequence "s-char-sequence"
1606/// \endverbatim
1607void Preprocessor::HandleLineDirective() {
1608 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are
1609 // expanded.
1610 Token DigitTok;
1611 Lex(Result&: DigitTok);
1612
1613 // Validate the number and convert it to an unsigned.
1614 unsigned LineNo;
1615 if (GetLineValue(DigitTok, Val&: LineNo, DiagID: diag::err_pp_line_requires_integer,PP&: *this))
1616 return;
1617
1618 if (LineNo == 0)
1619 Diag(Tok: DigitTok, DiagID: diag::ext_pp_line_zero);
1620
1621 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
1622 // number greater than 2147483647". C90 requires that the line # be <= 32767.
1623 unsigned LineLimit = 32768U;
1624 if (LangOpts.C99 || LangOpts.CPlusPlus11)
1625 LineLimit = 2147483648U;
1626 if (LineNo >= LineLimit)
1627 Diag(Tok: DigitTok, DiagID: diag::ext_pp_line_too_big) << LineLimit;
1628 else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
1629 Diag(Tok: DigitTok, DiagID: diag::warn_cxx98_compat_pp_line_too_big);
1630
1631 int FilenameID = -1;
1632 Token StrTok;
1633 Lex(Result&: StrTok);
1634
1635 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a
1636 // string followed by eod.
1637 if (StrTok.is(K: tok::eod))
1638 ; // ok
1639 else if (StrTok.isNot(K: tok::string_literal)) {
1640 Diag(Tok: StrTok, DiagID: diag::err_pp_line_invalid_filename);
1641 DiscardUntilEndOfDirective();
1642 return;
1643 } else if (StrTok.hasUDSuffix()) {
1644 Diag(Tok: StrTok, DiagID: diag::err_invalid_string_udl);
1645 DiscardUntilEndOfDirective();
1646 return;
1647 } else {
1648 // Parse and validate the string, converting it into a unique ID.
1649 StringLiteralParser Literal(StrTok, *this);
1650 assert(Literal.isOrdinary() && "Didn't allow wide strings in");
1651 if (Literal.hadError) {
1652 DiscardUntilEndOfDirective();
1653 return;
1654 }
1655 if (Literal.Pascal) {
1656 Diag(Tok: StrTok, DiagID: diag::err_pp_linemarker_invalid_filename);
1657 DiscardUntilEndOfDirective();
1658 return;
1659 }
1660 FilenameID = SourceMgr.getLineTableFilenameID(Str: Literal.GetString());
1661
1662 // Verify that there is nothing after the string, other than EOD. Because
1663 // of C99 6.10.4p5, macros that expand to empty tokens are ok.
1664 CheckEndOfDirective(DirType: "line", EnableMacros: true);
1665 }
1666
1667 // Take the file kind of the file containing the #line directive. #line
1668 // directives are often used for generated sources from the same codebase, so
1669 // the new file should generally be classified the same way as the current
1670 // file. This is visible in GCC's pre-processed output, which rewrites #line
1671 // to GNU line markers.
1672 SrcMgr::CharacteristicKind FileKind =
1673 SourceMgr.getFileCharacteristic(Loc: DigitTok.getLocation());
1674
1675 SourceMgr.AddLineNote(Loc: DigitTok.getLocation(), LineNo, FilenameID, IsFileEntry: false,
1676 IsFileExit: false, FileKind);
1677
1678 if (Callbacks)
1679 Callbacks->FileChanged(Loc: CurPPLexer->getSourceLocation(),
1680 Reason: PPCallbacks::RenameFile, FileType: FileKind);
1681}
1682
1683/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
1684/// marker directive.
1685static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
1686 SrcMgr::CharacteristicKind &FileKind,
1687 Preprocessor &PP) {
1688 unsigned FlagVal;
1689 Token FlagTok;
1690 PP.Lex(Result&: FlagTok);
1691 if (FlagTok.is(K: tok::eod)) return false;
1692 if (GetLineValue(DigitTok&: FlagTok, Val&: FlagVal, DiagID: diag::err_pp_linemarker_invalid_flag, PP))
1693 return true;
1694
1695 if (FlagVal == 1) {
1696 IsFileEntry = true;
1697
1698 PP.Lex(Result&: FlagTok);
1699 if (FlagTok.is(K: tok::eod)) return false;
1700 if (GetLineValue(DigitTok&: FlagTok, Val&: FlagVal, DiagID: diag::err_pp_linemarker_invalid_flag,PP))
1701 return true;
1702 } else if (FlagVal == 2) {
1703 IsFileExit = true;
1704
1705 SourceManager &SM = PP.getSourceManager();
1706 // If we are leaving the current presumed file, check to make sure the
1707 // presumed include stack isn't empty!
1708 FileID CurFileID =
1709 SM.getDecomposedExpansionLoc(Loc: FlagTok.getLocation()).first;
1710 PresumedLoc PLoc = SM.getPresumedLoc(Loc: FlagTok.getLocation());
1711 if (PLoc.isInvalid())
1712 return true;
1713
1714 // If there is no include loc (main file) or if the include loc is in a
1715 // different physical file, then we aren't in a "1" line marker flag region.
1716 SourceLocation IncLoc = PLoc.getIncludeLoc();
1717 if (IncLoc.isInvalid() ||
1718 SM.getDecomposedExpansionLoc(Loc: IncLoc).first != CurFileID) {
1719 PP.Diag(Tok: FlagTok, DiagID: diag::err_pp_linemarker_invalid_pop);
1720 PP.DiscardUntilEndOfDirective();
1721 return true;
1722 }
1723
1724 PP.Lex(Result&: FlagTok);
1725 if (FlagTok.is(K: tok::eod)) return false;
1726 if (GetLineValue(DigitTok&: FlagTok, Val&: FlagVal, DiagID: diag::err_pp_linemarker_invalid_flag,PP))
1727 return true;
1728 }
1729
1730 // We must have 3 if there are still flags.
1731 if (FlagVal != 3) {
1732 PP.Diag(Tok: FlagTok, DiagID: diag::err_pp_linemarker_invalid_flag);
1733 PP.DiscardUntilEndOfDirective();
1734 return true;
1735 }
1736
1737 FileKind = SrcMgr::C_System;
1738
1739 PP.Lex(Result&: FlagTok);
1740 if (FlagTok.is(K: tok::eod)) return false;
1741 if (GetLineValue(DigitTok&: FlagTok, Val&: FlagVal, DiagID: diag::err_pp_linemarker_invalid_flag, PP))
1742 return true;
1743
1744 // We must have 4 if there is yet another flag.
1745 if (FlagVal != 4) {
1746 PP.Diag(Tok: FlagTok, DiagID: diag::err_pp_linemarker_invalid_flag);
1747 PP.DiscardUntilEndOfDirective();
1748 return true;
1749 }
1750
1751 FileKind = SrcMgr::C_ExternCSystem;
1752
1753 PP.Lex(Result&: FlagTok);
1754 if (FlagTok.is(K: tok::eod)) return false;
1755
1756 // There are no more valid flags here.
1757 PP.Diag(Tok: FlagTok, DiagID: diag::err_pp_linemarker_invalid_flag);
1758 PP.DiscardUntilEndOfDirective();
1759 return true;
1760}
1761
1762/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
1763/// one of the following forms:
1764///
1765/// # 42
1766/// # 42 "file" ('1' | '2')?
1767/// # 42 "file" ('1' | '2')? '3' '4'?
1768///
1769void Preprocessor::HandleDigitDirective(Token &DigitTok) {
1770 // Validate the number and convert it to an unsigned. GNU does not have a
1771 // line # limit other than it fit in 32-bits.
1772 unsigned LineNo;
1773 if (GetLineValue(DigitTok, Val&: LineNo, DiagID: diag::err_pp_linemarker_requires_integer,
1774 PP&: *this, IsGNULineDirective: true))
1775 return;
1776
1777 Token StrTok;
1778 Lex(Result&: StrTok);
1779
1780 bool IsFileEntry = false, IsFileExit = false;
1781 int FilenameID = -1;
1782 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1783
1784 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a
1785 // string followed by eod.
1786 if (StrTok.is(K: tok::eod)) {
1787 Diag(Tok: StrTok, DiagID: diag::ext_pp_gnu_line_directive);
1788 // Treat this like "#line NN", which doesn't change file characteristics.
1789 FileKind = SourceMgr.getFileCharacteristic(Loc: DigitTok.getLocation());
1790 } else if (StrTok.isNot(K: tok::string_literal)) {
1791 Diag(Tok: StrTok, DiagID: diag::err_pp_linemarker_invalid_filename);
1792 DiscardUntilEndOfDirective();
1793 return;
1794 } else if (StrTok.hasUDSuffix()) {
1795 Diag(Tok: StrTok, DiagID: diag::err_invalid_string_udl);
1796 DiscardUntilEndOfDirective();
1797 return;
1798 } else {
1799 // Parse and validate the string, converting it into a unique ID.
1800 StringLiteralParser Literal(StrTok, *this);
1801 assert(Literal.isOrdinary() && "Didn't allow wide strings in");
1802 if (Literal.hadError) {
1803 DiscardUntilEndOfDirective();
1804 return;
1805 }
1806 if (Literal.Pascal) {
1807 Diag(Tok: StrTok, DiagID: diag::err_pp_linemarker_invalid_filename);
1808 DiscardUntilEndOfDirective();
1809 return;
1810 }
1811
1812 // If a filename was present, read any flags that are present.
1813 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, FileKind, PP&: *this))
1814 return;
1815 if (!SourceMgr.isInPredefinedFile(Loc: DigitTok.getLocation()))
1816 Diag(Tok: StrTok, DiagID: diag::ext_pp_gnu_line_directive);
1817
1818 // Exiting to an empty string means pop to the including file, so leave
1819 // FilenameID as -1 in that case.
1820 if (!(IsFileExit && Literal.GetString().empty()))
1821 FilenameID = SourceMgr.getLineTableFilenameID(Str: Literal.GetString());
1822 }
1823
1824 // Create a line note with this information.
1825 SourceMgr.AddLineNote(Loc: DigitTok.getLocation(), LineNo, FilenameID, IsFileEntry,
1826 IsFileExit, FileKind);
1827
1828 // If the preprocessor has callbacks installed, notify them of the #line
1829 // change. This is used so that the line marker comes out in -E mode for
1830 // example.
1831 if (Callbacks) {
1832 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1833 if (IsFileEntry)
1834 Reason = PPCallbacks::EnterFile;
1835 else if (IsFileExit)
1836 Reason = PPCallbacks::ExitFile;
1837
1838 Callbacks->FileChanged(Loc: CurPPLexer->getSourceLocation(), Reason, FileType: FileKind);
1839 }
1840}
1841
1842/// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1843///
1844void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1845 bool isWarning) {
1846 // Read the rest of the line raw. We do this because we don't want macros
1847 // to be expanded and we don't require that the tokens be valid preprocessing
1848 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does
1849 // collapse multiple consecutive white space between tokens, but this isn't
1850 // specified by the standard.
1851 SmallString<128> Message;
1852 CurLexer->ReadToEndOfLine(Result: &Message);
1853
1854 // Find the first non-whitespace character, so that we can make the
1855 // diagnostic more succinct.
1856 StringRef Msg = Message.str().ltrim(Char: ' ');
1857
1858 if (isWarning)
1859 Diag(Tok, DiagID: diag::pp_hash_warning) << Msg;
1860 else
1861 Diag(Tok, DiagID: diag::err_pp_hash_error) << Msg;
1862}
1863
1864/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1865///
1866void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1867 // Yes, this directive is an extension.
1868 Diag(Tok, DiagID: diag::ext_pp_ident_directive);
1869
1870 // Read the string argument.
1871 Token StrTok;
1872 Lex(Result&: StrTok);
1873
1874 // If the token kind isn't a string, it's a malformed directive.
1875 if (StrTok.isNot(K: tok::string_literal) &&
1876 StrTok.isNot(K: tok::wide_string_literal)) {
1877 Diag(Tok: StrTok, DiagID: diag::err_pp_malformed_ident);
1878 if (StrTok.isNot(K: tok::eod))
1879 DiscardUntilEndOfDirective();
1880 return;
1881 }
1882
1883 if (StrTok.hasUDSuffix()) {
1884 Diag(Tok: StrTok, DiagID: diag::err_invalid_string_udl);
1885 DiscardUntilEndOfDirective();
1886 return;
1887 }
1888
1889 // Verify that there is nothing after the string, other than EOD.
1890 CheckEndOfDirective(DirType: "ident");
1891
1892 if (Callbacks) {
1893 bool Invalid = false;
1894 std::string Str = getSpelling(Tok: StrTok, Invalid: &Invalid);
1895 if (!Invalid)
1896 Callbacks->Ident(Loc: Tok.getLocation(), str: Str);
1897 }
1898}
1899
1900/// Handle a #public directive.
1901void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1902 Token MacroNameTok;
1903 ReadMacroName(MacroNameTok, isDefineUndef: MU_Undef);
1904
1905 // Error reading macro name? If so, diagnostic already issued.
1906 if (MacroNameTok.is(K: tok::eod))
1907 return;
1908
1909 // Check to see if this is the last token on the #__public_macro line.
1910 CheckEndOfDirective(DirType: "__public_macro");
1911
1912 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1913 // Okay, we finally have a valid identifier to undef.
1914 MacroDirective *MD = getLocalMacroDirective(II);
1915
1916 // If the macro is not defined, this is an error.
1917 if (!MD) {
1918 Diag(Tok: MacroNameTok, DiagID: diag::err_pp_visibility_non_macro) << II;
1919 return;
1920 }
1921
1922 // Note that this macro has now been exported.
1923 appendMacroDirective(II, MD: AllocateVisibilityMacroDirective(
1924 Loc: MacroNameTok.getLocation(), /*isPublic=*/true));
1925}
1926
1927/// Handle a #private directive.
1928void Preprocessor::HandleMacroPrivateDirective() {
1929 Token MacroNameTok;
1930 ReadMacroName(MacroNameTok, isDefineUndef: MU_Undef);
1931
1932 // Error reading macro name? If so, diagnostic already issued.
1933 if (MacroNameTok.is(K: tok::eod))
1934 return;
1935
1936 // Check to see if this is the last token on the #__private_macro line.
1937 CheckEndOfDirective(DirType: "__private_macro");
1938
1939 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1940 // Okay, we finally have a valid identifier to undef.
1941 MacroDirective *MD = getLocalMacroDirective(II);
1942
1943 // If the macro is not defined, this is an error.
1944 if (!MD) {
1945 Diag(Tok: MacroNameTok, DiagID: diag::err_pp_visibility_non_macro) << II;
1946 return;
1947 }
1948
1949 // Note that this macro has now been marked private.
1950 appendMacroDirective(II, MD: AllocateVisibilityMacroDirective(
1951 Loc: MacroNameTok.getLocation(), /*isPublic=*/false));
1952}
1953
1954//===----------------------------------------------------------------------===//
1955// Preprocessor Include Directive Handling.
1956//===----------------------------------------------------------------------===//
1957
1958/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1959/// checked and spelled filename, e.g. as an operand of \#include. This returns
1960/// true if the input filename was in <>'s or false if it were in ""'s. The
1961/// caller is expected to provide a buffer that is large enough to hold the
1962/// spelling of the filename, but is also expected to handle the case when
1963/// this method decides to use a different buffer.
1964bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1965 StringRef &Buffer) {
1966 // Get the text form of the filename.
1967 assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1968
1969 // FIXME: Consider warning on some of the cases described in C11 6.4.7/3 and
1970 // C++20 [lex.header]/2:
1971 //
1972 // If `"`, `'`, `\`, `/*`, or `//` appears in a header-name, then
1973 // in C: behavior is undefined
1974 // in C++: program is conditionally-supported with implementation-defined
1975 // semantics
1976
1977 // Make sure the filename is <x> or "x".
1978 bool isAngled;
1979 if (Buffer[0] == '<') {
1980 if (Buffer.back() != '>') {
1981 Diag(Loc, DiagID: diag::err_pp_expects_filename);
1982 Buffer = StringRef();
1983 return true;
1984 }
1985 isAngled = true;
1986 } else if (Buffer[0] == '"') {
1987 if (Buffer.back() != '"') {
1988 Diag(Loc, DiagID: diag::err_pp_expects_filename);
1989 Buffer = StringRef();
1990 return true;
1991 }
1992 isAngled = false;
1993 } else {
1994 Diag(Loc, DiagID: diag::err_pp_expects_filename);
1995 Buffer = StringRef();
1996 return true;
1997 }
1998
1999 // Diagnose #include "" as invalid.
2000 if (Buffer.size() <= 2) {
2001 Diag(Loc, DiagID: diag::err_pp_empty_filename);
2002 Buffer = StringRef();
2003 return true;
2004 }
2005
2006 // Skip the brackets.
2007 Buffer = Buffer.substr(Start: 1, N: Buffer.size()-2);
2008 return isAngled;
2009}
2010
2011/// Push a token onto the token stream containing an annotation.
2012void Preprocessor::EnterAnnotationToken(SourceRange Range,
2013 tok::TokenKind Kind,
2014 void *AnnotationVal) {
2015 // FIXME: Produce this as the current token directly, rather than
2016 // allocating a new token for it.
2017 auto Tok = std::make_unique<Token[]>(num: 1);
2018 Tok[0].startToken();
2019 Tok[0].setKind(Kind);
2020 Tok[0].setLocation(Range.getBegin());
2021 Tok[0].setAnnotationEndLoc(Range.getEnd());
2022 Tok[0].setAnnotationValue(AnnotationVal);
2023 EnterTokenStream(Toks: std::move(Tok), NumToks: 1, DisableMacroExpansion: true, /*IsReinject*/ false);
2024}
2025
2026/// Produce a diagnostic informing the user that a #include or similar
2027/// was implicitly treated as a module import.
2028static void diagnoseAutoModuleImport(Preprocessor &PP, SourceLocation HashLoc,
2029 Token &IncludeTok,
2030 ArrayRef<IdentifierLoc> Path,
2031 SourceLocation PathEnd) {
2032 SmallString<128> PathString;
2033 for (size_t I = 0, N = Path.size(); I != N; ++I) {
2034 if (I)
2035 PathString += '.';
2036 PathString += Path[I].getIdentifierInfo()->getName();
2037 }
2038
2039 int IncludeKind = 0;
2040 switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
2041 case tok::pp_include:
2042 IncludeKind = 0;
2043 break;
2044
2045 case tok::pp_import:
2046 IncludeKind = 1;
2047 break;
2048
2049 case tok::pp_include_next:
2050 IncludeKind = 2;
2051 break;
2052
2053 case tok::pp___include_macros:
2054 IncludeKind = 3;
2055 break;
2056
2057 default:
2058 llvm_unreachable("unknown include directive kind");
2059 }
2060
2061 PP.Diag(Loc: HashLoc, DiagID: diag::remark_pp_include_directive_modular_translation)
2062 << IncludeKind << PathString;
2063}
2064
2065// Given a vector of path components and a string containing the real
2066// path to the file, build a properly-cased replacement in the vector,
2067// and return true if the replacement should be suggested.
2068static bool trySimplifyPath(SmallVectorImpl<StringRef> &Components,
2069 StringRef RealPathName,
2070 llvm::sys::path::Style Separator) {
2071 auto RealPathComponentIter = llvm::sys::path::rbegin(path: RealPathName);
2072 auto RealPathComponentEnd = llvm::sys::path::rend(path: RealPathName);
2073 int Cnt = 0;
2074 bool SuggestReplacement = false;
2075
2076 auto IsSep = [Separator](StringRef Component) {
2077 return Component.size() == 1 &&
2078 llvm::sys::path::is_separator(value: Component[0], style: Separator);
2079 };
2080
2081 // Below is a best-effort to handle ".." in paths. It is admittedly
2082 // not 100% correct in the presence of symlinks.
2083 for (auto &Component : llvm::reverse(C&: Components)) {
2084 if ("." == Component) {
2085 } else if (".." == Component) {
2086 ++Cnt;
2087 } else if (Cnt) {
2088 --Cnt;
2089 } else if (RealPathComponentIter != RealPathComponentEnd) {
2090 if (!IsSep(Component) && !IsSep(*RealPathComponentIter) &&
2091 Component != *RealPathComponentIter) {
2092 // If these non-separator path components differ by more than just case,
2093 // then we may be looking at symlinked paths. Bail on this diagnostic to
2094 // avoid noisy false positives.
2095 SuggestReplacement =
2096 RealPathComponentIter->equals_insensitive(RHS: Component);
2097 if (!SuggestReplacement)
2098 break;
2099 Component = *RealPathComponentIter;
2100 }
2101 ++RealPathComponentIter;
2102 }
2103 }
2104 return SuggestReplacement;
2105}
2106
2107bool Preprocessor::checkModuleIsAvailable(const LangOptions &LangOpts,
2108 const TargetInfo &TargetInfo,
2109 const Module &M,
2110 DiagnosticsEngine &Diags) {
2111 Module::Requirement Requirement;
2112 Module::UnresolvedHeaderDirective MissingHeader;
2113 Module *ShadowingModule = nullptr;
2114 if (M.isAvailable(LangOpts, Target: TargetInfo, Req&: Requirement, MissingHeader,
2115 ShadowingModule))
2116 return false;
2117
2118 if (MissingHeader.FileNameLoc.isValid()) {
2119 Diags.Report(Loc: MissingHeader.FileNameLoc, DiagID: diag::err_module_header_missing)
2120 << MissingHeader.IsUmbrella << MissingHeader.FileName;
2121 } else if (ShadowingModule) {
2122 Diags.Report(Loc: M.DefinitionLoc, DiagID: diag::err_module_shadowed) << M.Name;
2123 Diags.Report(Loc: ShadowingModule->DefinitionLoc,
2124 DiagID: diag::note_previous_definition);
2125 } else {
2126 // FIXME: Track the location at which the requirement was specified, and
2127 // use it here.
2128 Diags.Report(Loc: M.DefinitionLoc, DiagID: diag::err_module_unavailable)
2129 << M.getFullModuleName() << Requirement.RequiredState
2130 << Requirement.FeatureName;
2131 }
2132 return true;
2133}
2134
2135std::pair<ConstSearchDirIterator, const FileEntry *>
2136Preprocessor::getIncludeNextStart(const Token &IncludeNextTok) const {
2137 // #include_next is like #include, except that we start searching after
2138 // the current found directory. If we can't do this, issue a
2139 // diagnostic.
2140 ConstSearchDirIterator Lookup = CurDirLookup;
2141 const FileEntry *LookupFromFile = nullptr;
2142
2143 if (isInPrimaryFile() && LangOpts.IsHeaderFile) {
2144 // If the main file is a header, then it's either for PCH/AST generation,
2145 // or libclang opened it. Either way, handle it as a normal include below
2146 // and do not complain about include_next.
2147 } else if (isInPrimaryFile()) {
2148 Lookup = nullptr;
2149 Diag(Tok: IncludeNextTok, DiagID: diag::pp_include_next_in_primary);
2150 } else if (CurLexerSubmodule) {
2151 // Start looking up in the directory *after* the one in which the current
2152 // file would be found, if any.
2153 assert(CurPPLexer && "#include_next directive in macro?");
2154 if (auto FE = CurPPLexer->getFileEntry())
2155 LookupFromFile = *FE;
2156 Lookup = nullptr;
2157 } else if (!Lookup) {
2158 // The current file was not found by walking the include path. Either it
2159 // is the primary file (handled above), or it was found by absolute path,
2160 // or it was found relative to such a file.
2161 // FIXME: Track enough information so we know which case we're in.
2162 Diag(Tok: IncludeNextTok, DiagID: diag::pp_include_next_absolute_path);
2163 } else {
2164 // Start looking up in the next directory.
2165 ++Lookup;
2166 }
2167
2168 return {Lookup, LookupFromFile};
2169}
2170
2171/// HandleIncludeDirective - The "\#include" tokens have just been read, read
2172/// the file to be included from the lexer, then include it! This is a common
2173/// routine with functionality shared between \#include, \#include_next and
2174/// \#import. LookupFrom is set when this is a \#include_next directive, it
2175/// specifies the file to start searching from.
2176void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
2177 Token &IncludeTok,
2178 ConstSearchDirIterator LookupFrom,
2179 const FileEntry *LookupFromFile) {
2180 Token FilenameTok;
2181 if (LexHeaderName(Result&: FilenameTok))
2182 return;
2183
2184 if (FilenameTok.isNot(K: tok::header_name)) {
2185 if (FilenameTok.is(K: tok::identifier) &&
2186 (PPOpts.SingleFileParseMode || PPOpts.SingleModuleParseMode)) {
2187 // If we saw #include IDENTIFIER and lexing didn't turn in into a header
2188 // name, it was undefined. In 'single-{file,module}-parse' mode, just skip
2189 // the directive without emitting diagnostics - the identifier might be
2190 // normally defined in previously-skipped include directive.
2191 DiscardUntilEndOfDirective();
2192 return;
2193 }
2194
2195 Diag(Loc: FilenameTok.getLocation(), DiagID: diag::err_pp_expects_filename);
2196 if (FilenameTok.isNot(K: tok::eod))
2197 DiscardUntilEndOfDirective();
2198 return;
2199 }
2200
2201 // Verify that there is nothing after the filename, other than EOD. Note
2202 // that we allow macros that expand to nothing after the filename, because
2203 // this falls into the category of "#include pp-tokens new-line" specified
2204 // in C99 6.10.2p4.
2205 SourceLocation EndLoc =
2206 CheckEndOfDirective(DirType: IncludeTok.getIdentifierInfo()->getNameStart(), EnableMacros: true);
2207
2208 auto Action = HandleHeaderIncludeOrImport(HashLoc, IncludeTok, FilenameTok,
2209 EndLoc, LookupFrom, LookupFromFile);
2210 switch (Action.Kind) {
2211 case ImportAction::None:
2212 case ImportAction::SkippedModuleImport:
2213 break;
2214 case ImportAction::ModuleBegin:
2215 EnterAnnotationToken(Range: SourceRange(HashLoc, EndLoc),
2216 Kind: tok::annot_module_begin, AnnotationVal: Action.ModuleForHeader);
2217 break;
2218 case ImportAction::HeaderUnitImport:
2219 EnterAnnotationToken(Range: SourceRange(HashLoc, EndLoc), Kind: tok::annot_header_unit,
2220 AnnotationVal: Action.ModuleForHeader);
2221 break;
2222 case ImportAction::ModuleImport:
2223 EnterAnnotationToken(Range: SourceRange(HashLoc, EndLoc),
2224 Kind: tok::annot_module_include, AnnotationVal: Action.ModuleForHeader);
2225 break;
2226 case ImportAction::Failure:
2227 assert(TheModuleLoader.HadFatalFailure &&
2228 "This should be an early exit only to a fatal error");
2229 TheModuleLoader.HadFatalFailure = true;
2230 IncludeTok.setKind(tok::eof);
2231 CurLexer->cutOffLexing();
2232 return;
2233 }
2234}
2235
2236OptionalFileEntryRef Preprocessor::LookupHeaderIncludeOrImport(
2237 ConstSearchDirIterator *CurDir, StringRef &Filename,
2238 SourceLocation FilenameLoc, CharSourceRange FilenameRange,
2239 const Token &FilenameTok, bool &IsFrameworkFound, bool IsImportDecl,
2240 bool &IsMapped, ConstSearchDirIterator LookupFrom,
2241 const FileEntry *LookupFromFile, StringRef &LookupFilename,
2242 SmallVectorImpl<char> &RelativePath, SmallVectorImpl<char> &SearchPath,
2243 ModuleMap::KnownHeader &SuggestedModule, bool isAngled) {
2244 auto DiagnoseHeaderInclusion = [&](FileEntryRef FE) {
2245 if (LangOpts.AsmPreprocessor)
2246 return;
2247
2248 Module *RequestingModule = getModuleForLocation(
2249 Loc: FilenameLoc, AllowTextual: LangOpts.ModulesValidateTextualHeaderIncludes);
2250 bool RequestingModuleIsModuleInterface =
2251 !SourceMgr.isInMainFile(Loc: FilenameLoc);
2252
2253 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
2254 RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
2255 Filename, File: FE);
2256 };
2257
2258 OptionalFileEntryRef File = LookupFile(
2259 FilenameLoc, Filename: LookupFilename, isAngled, FromDir: LookupFrom, FromFile: LookupFromFile, CurDirArg: CurDir,
2260 SearchPath: Callbacks ? &SearchPath : nullptr, RelativePath: Callbacks ? &RelativePath : nullptr,
2261 SuggestedModule: &SuggestedModule, IsMapped: &IsMapped, IsFrameworkFound: &IsFrameworkFound);
2262 if (File) {
2263 DiagnoseHeaderInclusion(*File);
2264 return File;
2265 }
2266
2267 // Give the clients a chance to silently skip this include.
2268 if (Callbacks && Callbacks->FileNotFound(FileName: Filename))
2269 return std::nullopt;
2270
2271 if (SuppressIncludeNotFoundError)
2272 return std::nullopt;
2273
2274 // If the file could not be located and it was included via angle
2275 // brackets, we can attempt a lookup as though it were a quoted path to
2276 // provide the user with a possible fixit.
2277 if (isAngled) {
2278 OptionalFileEntryRef File = LookupFile(
2279 FilenameLoc, Filename: LookupFilename, isAngled: false, FromDir: LookupFrom, FromFile: LookupFromFile, CurDirArg: CurDir,
2280 SearchPath: Callbacks ? &SearchPath : nullptr, RelativePath: Callbacks ? &RelativePath : nullptr,
2281 SuggestedModule: &SuggestedModule, IsMapped: &IsMapped,
2282 /*IsFrameworkFound=*/nullptr);
2283 if (File) {
2284 DiagnoseHeaderInclusion(*File);
2285 Diag(Tok: FilenameTok, DiagID: diag::err_pp_file_not_found_angled_include_not_fatal)
2286 << Filename << IsImportDecl
2287 << FixItHint::CreateReplacement(RemoveRange: FilenameRange,
2288 Code: "\"" + Filename.str() + "\"");
2289 return File;
2290 }
2291 }
2292
2293 // Check for likely typos due to leading or trailing non-isAlphanumeric
2294 // characters
2295 StringRef OriginalFilename = Filename;
2296 if (LangOpts.SpellChecking) {
2297 // A heuristic to correct a typo file name by removing leading and
2298 // trailing non-isAlphanumeric characters.
2299 auto CorrectTypoFilename = [](llvm::StringRef Filename) {
2300 Filename = Filename.drop_until(F: isAlphanumeric);
2301 while (!Filename.empty() && !isAlphanumeric(c: Filename.back())) {
2302 Filename = Filename.drop_back();
2303 }
2304 return Filename;
2305 };
2306 StringRef TypoCorrectionName = CorrectTypoFilename(Filename);
2307 StringRef TypoCorrectionLookupName = CorrectTypoFilename(LookupFilename);
2308
2309 OptionalFileEntryRef File = LookupFile(
2310 FilenameLoc, Filename: TypoCorrectionLookupName, isAngled, FromDir: LookupFrom,
2311 FromFile: LookupFromFile, CurDirArg: CurDir, SearchPath: Callbacks ? &SearchPath : nullptr,
2312 RelativePath: Callbacks ? &RelativePath : nullptr, SuggestedModule: &SuggestedModule, IsMapped: &IsMapped,
2313 /*IsFrameworkFound=*/nullptr);
2314 if (File) {
2315 DiagnoseHeaderInclusion(*File);
2316 auto Hint =
2317 isAngled ? FixItHint::CreateReplacement(
2318 RemoveRange: FilenameRange, Code: "<" + TypoCorrectionName.str() + ">")
2319 : FixItHint::CreateReplacement(
2320 RemoveRange: FilenameRange, Code: "\"" + TypoCorrectionName.str() + "\"");
2321 Diag(Tok: FilenameTok, DiagID: diag::err_pp_file_not_found_typo_not_fatal)
2322 << OriginalFilename << TypoCorrectionName << Hint;
2323 // We found the file, so set the Filename to the name after typo
2324 // correction.
2325 Filename = TypoCorrectionName;
2326 LookupFilename = TypoCorrectionLookupName;
2327 return File;
2328 }
2329 }
2330
2331 // If the file is still not found, just go with the vanilla diagnostic
2332 assert(!File && "expected missing file");
2333 Diag(Tok: FilenameTok, DiagID: diag::err_pp_file_not_found)
2334 << OriginalFilename << FilenameRange;
2335 if (IsFrameworkFound) {
2336 size_t SlashPos = OriginalFilename.find(C: '/');
2337 assert(SlashPos != StringRef::npos &&
2338 "Include with framework name should have '/' in the filename");
2339 StringRef FrameworkName = OriginalFilename.substr(Start: 0, N: SlashPos);
2340 FrameworkCacheEntry &CacheEntry =
2341 HeaderInfo.LookupFrameworkCache(FWName: FrameworkName);
2342 assert(CacheEntry.Directory && "Found framework should be in cache");
2343 Diag(Tok: FilenameTok, DiagID: diag::note_pp_framework_without_header)
2344 << OriginalFilename.substr(Start: SlashPos + 1) << FrameworkName
2345 << CacheEntry.Directory->getName();
2346 }
2347
2348 return std::nullopt;
2349}
2350
2351/// Handle either a #include-like directive or an import declaration that names
2352/// a header file.
2353///
2354/// \param HashLoc The location of the '#' token for an include, or
2355/// SourceLocation() for an import declaration.
2356/// \param IncludeTok The include / include_next / import token.
2357/// \param FilenameTok The header-name token.
2358/// \param EndLoc The location at which any imported macros become visible.
2359/// \param LookupFrom For #include_next, the starting directory for the
2360/// directory lookup.
2361/// \param LookupFromFile For #include_next, the starting file for the directory
2362/// lookup.
2363Preprocessor::ImportAction Preprocessor::HandleHeaderIncludeOrImport(
2364 SourceLocation HashLoc, Token &IncludeTok, Token &FilenameTok,
2365 SourceLocation EndLoc, ConstSearchDirIterator LookupFrom,
2366 const FileEntry *LookupFromFile) {
2367 SmallString<128> FilenameBuffer;
2368 StringRef Filename = getSpelling(Tok: FilenameTok, Buffer&: FilenameBuffer);
2369 SourceLocation CharEnd = FilenameTok.getEndLoc();
2370
2371 CharSourceRange FilenameRange
2372 = CharSourceRange::getCharRange(B: FilenameTok.getLocation(), E: CharEnd);
2373 StringRef OriginalFilename = Filename;
2374 bool isAngled =
2375 GetIncludeFilenameSpelling(Loc: FilenameTok.getLocation(), Buffer&: Filename);
2376
2377 // If GetIncludeFilenameSpelling set the start ptr to null, there was an
2378 // error.
2379 if (Filename.empty())
2380 return {ImportAction::None};
2381
2382 bool IsImportDecl = HashLoc.isInvalid();
2383 SourceLocation StartLoc = IsImportDecl ? IncludeTok.getLocation() : HashLoc;
2384
2385 // Complain about attempts to #include files in an audit pragma.
2386 if (PragmaARCCFCodeAuditedInfo.getLoc().isValid()) {
2387 Diag(Loc: StartLoc, DiagID: diag::err_pp_include_in_arc_cf_code_audited) << IsImportDecl;
2388 Diag(Loc: PragmaARCCFCodeAuditedInfo.getLoc(), DiagID: diag::note_pragma_entered_here);
2389
2390 // Immediately leave the pragma.
2391 PragmaARCCFCodeAuditedInfo = IdentifierLoc();
2392 }
2393
2394 // Complain about attempts to #include files in an assume-nonnull pragma.
2395 if (PragmaAssumeNonNullLoc.isValid()) {
2396 Diag(Loc: StartLoc, DiagID: diag::err_pp_include_in_assume_nonnull) << IsImportDecl;
2397 Diag(Loc: PragmaAssumeNonNullLoc, DiagID: diag::note_pragma_entered_here);
2398
2399 // Immediately leave the pragma.
2400 PragmaAssumeNonNullLoc = SourceLocation();
2401 }
2402
2403 if (HeaderInfo.HasIncludeAliasMap()) {
2404 // Map the filename with the brackets still attached. If the name doesn't
2405 // map to anything, fall back on the filename we've already gotten the
2406 // spelling for.
2407 StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(Source: OriginalFilename);
2408 if (!NewName.empty())
2409 Filename = NewName;
2410 }
2411
2412 // Search include directories.
2413 bool IsMapped = false;
2414 bool IsFrameworkFound = false;
2415 ConstSearchDirIterator CurDir = nullptr;
2416 SmallString<1024> SearchPath;
2417 SmallString<1024> RelativePath;
2418 // We get the raw path only if we have 'Callbacks' to which we later pass
2419 // the path.
2420 ModuleMap::KnownHeader SuggestedModule;
2421 SourceLocation FilenameLoc = FilenameTok.getLocation();
2422 StringRef LookupFilename = Filename;
2423
2424 // Normalize slashes when compiling with -fms-extensions on non-Windows. This
2425 // is unnecessary on Windows since the filesystem there handles backslashes.
2426 SmallString<128> NormalizedPath;
2427 llvm::sys::path::Style BackslashStyle = llvm::sys::path::Style::native;
2428 if (is_style_posix(S: BackslashStyle) && LangOpts.MicrosoftExt) {
2429 NormalizedPath = Filename.str();
2430 llvm::sys::path::native(path&: NormalizedPath);
2431 LookupFilename = NormalizedPath;
2432 BackslashStyle = llvm::sys::path::Style::windows;
2433 }
2434
2435 OptionalFileEntryRef File = LookupHeaderIncludeOrImport(
2436 CurDir: &CurDir, Filename, FilenameLoc, FilenameRange, FilenameTok,
2437 IsFrameworkFound, IsImportDecl, IsMapped, LookupFrom, LookupFromFile,
2438 LookupFilename, RelativePath, SearchPath, SuggestedModule, isAngled);
2439
2440 if (usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) {
2441 if (File && isPCHThroughHeader(FE: &File->getFileEntry()))
2442 SkippingUntilPCHThroughHeader = false;
2443 return {ImportAction::None};
2444 }
2445
2446 // Should we enter the source file? Set to Skip if either the source file is
2447 // known to have no effect beyond its effect on module visibility -- that is,
2448 // if it's got an include guard that is already defined, set to Import if it
2449 // is a modular header we've already built and should import.
2450
2451 // For C++20 Modules
2452 // [cpp.include]/7 If the header identified by the header-name denotes an
2453 // importable header, it is implementation-defined whether the #include
2454 // preprocessing directive is instead replaced by an import directive.
2455 // For this implementation, the translation is permitted when we are parsing
2456 // the Global Module Fragment, and not otherwise (the cases where it would be
2457 // valid to replace an include with an import are highly constrained once in
2458 // named module purview; this choice avoids considerable complexity in
2459 // determining valid cases).
2460
2461 enum { Enter, Import, Skip, IncludeLimitReached } Action = Enter;
2462
2463 if (PPOpts.SingleFileParseMode)
2464 Action = IncludeLimitReached;
2465
2466 // If we've reached the max allowed include depth, it is usually due to an
2467 // include cycle. Don't enter already processed files again as it can lead to
2468 // reaching the max allowed include depth again.
2469 if (Action == Enter && HasReachedMaxIncludeDepth && File &&
2470 alreadyIncluded(File: *File))
2471 Action = IncludeLimitReached;
2472
2473 // FIXME: We do not have a good way to disambiguate C++ clang modules from
2474 // C++ standard modules (other than use/non-use of Header Units).
2475
2476 Module *ModuleToImport = SuggestedModule.getModule();
2477
2478 bool MaybeTranslateInclude = Action == Enter && File && ModuleToImport &&
2479 !ModuleToImport->isForBuilding(LangOpts: getLangOpts());
2480
2481 // Maybe a usable Header Unit
2482 bool UsableHeaderUnit = false;
2483 if (getLangOpts().CPlusPlusModules && ModuleToImport &&
2484 ModuleToImport->isHeaderUnit()) {
2485 if (TrackGMFState.inGMF() || IsImportDecl)
2486 UsableHeaderUnit = true;
2487 else if (!IsImportDecl) {
2488 // This is a Header Unit that we do not include-translate
2489 ModuleToImport = nullptr;
2490 }
2491 }
2492 // Maybe a usable clang header module.
2493 bool UsableClangHeaderModule =
2494 (getLangOpts().CPlusPlusModules || getLangOpts().Modules) &&
2495 ModuleToImport && !ModuleToImport->isHeaderUnit();
2496
2497 if (MaybeTranslateInclude && (UsableHeaderUnit || UsableClangHeaderModule) &&
2498 PPOpts.SingleModuleParseMode) {
2499 Action = IncludeLimitReached;
2500 }
2501 // Determine whether we should try to import the module for this #include, if
2502 // there is one. Don't do so if precompiled module support is disabled or we
2503 // are processing this module textually (because we're building the module).
2504 else if (MaybeTranslateInclude &&
2505 (UsableHeaderUnit || UsableClangHeaderModule)) {
2506 // If this include corresponds to a module but that module is
2507 // unavailable, diagnose the situation and bail out.
2508 // FIXME: Remove this; loadModule does the same check (but produces
2509 // slightly worse diagnostics).
2510 if (checkModuleIsAvailable(LangOpts: getLangOpts(), TargetInfo: getTargetInfo(), M: *ModuleToImport,
2511 Diags&: getDiagnostics())) {
2512 Diag(Loc: FilenameTok.getLocation(),
2513 DiagID: diag::note_implicit_top_level_module_import_here)
2514 << ModuleToImport->getTopLevelModuleName();
2515 return {ImportAction::None};
2516 }
2517
2518 // Compute the module access path corresponding to this module.
2519 // FIXME: Should we have a second loadModule() overload to avoid this
2520 // extra lookup step?
2521 SmallVector<IdentifierLoc, 2> Path;
2522 for (Module *Mod = ModuleToImport; Mod; Mod = Mod->Parent)
2523 Path.emplace_back(Args: FilenameTok.getLocation(),
2524 Args: getIdentifierInfo(Name: Mod->Name));
2525 std::reverse(first: Path.begin(), last: Path.end());
2526
2527 // Warn that we're replacing the include/import with a module import.
2528 if (!IsImportDecl)
2529 diagnoseAutoModuleImport(PP&: *this, HashLoc: StartLoc, IncludeTok, Path, PathEnd: CharEnd);
2530
2531 // Load the module to import its macros. We'll make the declarations
2532 // visible when the parser gets here.
2533 // FIXME: Pass ModuleToImport in here rather than converting it to a path
2534 // and making the module loader convert it back again.
2535 ModuleLoadResult Imported = TheModuleLoader.loadModule(
2536 ImportLoc: IncludeTok.getLocation(), Path, Visibility: Module::Hidden,
2537 /*IsInclusionDirective=*/true);
2538 assert((Imported == nullptr || Imported == ModuleToImport) &&
2539 "the imported module is different than the suggested one");
2540
2541 if (Imported) {
2542 Action = Import;
2543 } else if (Imported.isMissingExpected()) {
2544 markClangModuleAsAffecting(
2545 M: static_cast<Module *>(Imported)->getTopLevelModule());
2546 // We failed to find a submodule that we assumed would exist (because it
2547 // was in the directory of an umbrella header, for instance), but no
2548 // actual module containing it exists (because the umbrella header is
2549 // incomplete). Treat this as a textual inclusion.
2550 ModuleToImport = nullptr;
2551 } else if (Imported.isConfigMismatch()) {
2552 // On a configuration mismatch, enter the header textually. We still know
2553 // that it's part of the corresponding module.
2554 } else {
2555 // We hit an error processing the import. Bail out.
2556 if (hadModuleLoaderFatalFailure()) {
2557 // With a fatal failure in the module loader, we abort parsing.
2558 Token &Result = IncludeTok;
2559 assert(CurLexer && "#include but no current lexer set!");
2560 Result.startToken();
2561 CurLexer->FormTokenWithChars(Result, TokEnd: CurLexer->BufferEnd, Kind: tok::eof);
2562 CurLexer->cutOffLexing();
2563 }
2564 return {ImportAction::None};
2565 }
2566 }
2567
2568 // The #included file will be considered to be a system header if either it is
2569 // in a system include directory, or if the #includer is a system include
2570 // header.
2571 SrcMgr::CharacteristicKind FileCharacter =
2572 SourceMgr.getFileCharacteristic(Loc: FilenameTok.getLocation());
2573 if (File)
2574 FileCharacter = std::max(a: HeaderInfo.getFileDirFlavor(File: *File), b: FileCharacter);
2575
2576 // If this is a '#import' or an import-declaration, don't re-enter the file.
2577 //
2578 // FIXME: If we have a suggested module for a '#include', and we've already
2579 // visited this file, don't bother entering it again. We know it has no
2580 // further effect.
2581 bool EnterOnce =
2582 IsImportDecl ||
2583 IncludeTok.getIdentifierInfo()->getPPKeywordID() == tok::pp_import;
2584
2585 bool IsFirstIncludeOfFile = false;
2586
2587 // Ask HeaderInfo if we should enter this #include file. If not, #including
2588 // this file will have no effect.
2589 if (Action == Enter && File &&
2590 !HeaderInfo.ShouldEnterIncludeFile(PP&: *this, File: *File, isImport: EnterOnce,
2591 ModulesEnabled: getLangOpts().Modules, M: ModuleToImport,
2592 IsFirstIncludeOfFile)) {
2593 // C++ standard modules:
2594 // If we are not in the GMF, then we textually include only
2595 // clang modules:
2596 // Even if we've already preprocessed this header once and know that we
2597 // don't need to see its contents again, we still need to import it if it's
2598 // modular because we might not have imported it from this submodule before.
2599 //
2600 // FIXME: We don't do this when compiling a PCH because the AST
2601 // serialization layer can't cope with it. This means we get local
2602 // submodule visibility semantics wrong in that case.
2603 if (UsableHeaderUnit && !getLangOpts().CompilingPCH)
2604 Action = TrackGMFState.inGMF() ? Import : Skip;
2605 else
2606 Action = (ModuleToImport && !getLangOpts().CompilingPCH) ? Import : Skip;
2607 }
2608
2609 // Check for circular inclusion of the main file.
2610 // We can't generate a consistent preamble with regard to the conditional
2611 // stack if the main file is included again as due to the preamble bounds
2612 // some directives (e.g. #endif of a header guard) will never be seen.
2613 // Since this will lead to confusing errors, avoid the inclusion.
2614 if (Action == Enter && File && PreambleConditionalStack.isRecording() &&
2615 SourceMgr.isMainFile(SourceFile: File->getFileEntry())) {
2616 Diag(Loc: FilenameTok.getLocation(),
2617 DiagID: diag::err_pp_including_mainfile_in_preamble);
2618 return {ImportAction::None};
2619 }
2620
2621 if (Callbacks && !IsImportDecl) {
2622 // Notify the callback object that we've seen an inclusion directive.
2623 // FIXME: Use a different callback for a pp-import?
2624 Callbacks->InclusionDirective(HashLoc, IncludeTok, FileName: LookupFilename, IsAngled: isAngled,
2625 FilenameRange, File, SearchPath, RelativePath,
2626 SuggestedModule: SuggestedModule.getModule(), ModuleImported: Action == Import,
2627 FileType: FileCharacter);
2628 if (Action == Skip && File)
2629 Callbacks->FileSkipped(SkippedFile: *File, FilenameTok, FileType: FileCharacter);
2630 }
2631
2632 if (!File)
2633 return {ImportAction::None};
2634
2635 // If this is a C++20 pp-import declaration, diagnose if we didn't find any
2636 // module corresponding to the named header.
2637 if (IsImportDecl && !ModuleToImport) {
2638 Diag(Tok: FilenameTok, DiagID: diag::err_header_import_not_header_unit)
2639 << OriginalFilename << File->getName();
2640 return {ImportAction::None};
2641 }
2642
2643 // Issue a diagnostic if the name of the file on disk has a different case
2644 // than the one we're about to open.
2645 const bool CheckIncludePathPortability =
2646 !IsMapped && !File->getFileEntry().tryGetRealPathName().empty();
2647
2648 if (CheckIncludePathPortability) {
2649 StringRef Name = LookupFilename;
2650 StringRef NameWithoriginalSlashes = Filename;
2651#if defined(_WIN32)
2652 // Skip UNC prefix if present. (tryGetRealPathName() always
2653 // returns a path with the prefix skipped.)
2654 bool NameWasUNC = Name.consume_front("\\\\?\\");
2655 NameWithoriginalSlashes.consume_front("\\\\?\\");
2656#endif
2657 StringRef RealPathName = File->getFileEntry().tryGetRealPathName();
2658 SmallVector<StringRef, 16> Components(llvm::sys::path::begin(path: Name),
2659 llvm::sys::path::end(path: Name));
2660#if defined(_WIN32)
2661 // -Wnonportable-include-path is designed to diagnose includes using
2662 // case even on systems with a case-insensitive file system.
2663 // On Windows, RealPathName always starts with an upper-case drive
2664 // letter for absolute paths, but Name might start with either
2665 // case depending on if `cd c:\foo` or `cd C:\foo` was used in the shell.
2666 // ("foo" will always have on-disk case, no matter which case was
2667 // used in the cd command). To not emit this warning solely for
2668 // the drive letter, whose case is dependent on if `cd` is used
2669 // with upper- or lower-case drive letters, always consider the
2670 // given drive letter case as correct for the purpose of this warning.
2671 SmallString<128> FixedDriveRealPath;
2672 if (llvm::sys::path::is_absolute(Name) &&
2673 llvm::sys::path::is_absolute(RealPathName) &&
2674 toLowercase(Name[0]) == toLowercase(RealPathName[0]) &&
2675 isLowercase(Name[0]) != isLowercase(RealPathName[0])) {
2676 assert(Components.size() >= 3 && "should have drive, backslash, name");
2677 assert(Components[0].size() == 2 && "should start with drive");
2678 assert(Components[0][1] == ':' && "should have colon");
2679 FixedDriveRealPath = (Name.substr(0, 1) + RealPathName.substr(1)).str();
2680 RealPathName = FixedDriveRealPath;
2681 }
2682#endif
2683
2684 if (trySimplifyPath(Components, RealPathName, Separator: BackslashStyle)) {
2685 SmallString<128> Path;
2686 Path.reserve(N: Name.size()+2);
2687 Path.push_back(Elt: isAngled ? '<' : '"');
2688
2689 const auto IsSep = [BackslashStyle](char c) {
2690 return llvm::sys::path::is_separator(value: c, style: BackslashStyle);
2691 };
2692
2693 for (auto Component : Components) {
2694 // On POSIX, Components will contain a single '/' as first element
2695 // exactly if Name is an absolute path.
2696 // On Windows, it will contain "C:" followed by '\' for absolute paths.
2697 // The drive letter is optional for absolute paths on Windows, but
2698 // clang currently cannot process absolute paths in #include lines that
2699 // don't have a drive.
2700 // If the first entry in Components is a directory separator,
2701 // then the code at the bottom of this loop that keeps the original
2702 // directory separator style copies it. If the second entry is
2703 // a directory separator (the C:\ case), then that separator already
2704 // got copied when the C: was processed and we want to skip that entry.
2705 if (!(Component.size() == 1 && IsSep(Component[0])))
2706 Path.append(RHS: Component);
2707 else if (Path.size() != 1)
2708 continue;
2709
2710 // Append the separator(s) the user used, or the close quote
2711 if (Path.size() > NameWithoriginalSlashes.size()) {
2712 Path.push_back(Elt: isAngled ? '>' : '"');
2713 continue;
2714 }
2715 assert(IsSep(NameWithoriginalSlashes[Path.size()-1]));
2716 do
2717 Path.push_back(Elt: NameWithoriginalSlashes[Path.size()-1]);
2718 while (Path.size() <= NameWithoriginalSlashes.size() &&
2719 IsSep(NameWithoriginalSlashes[Path.size()-1]));
2720 }
2721
2722#if defined(_WIN32)
2723 // Restore UNC prefix if it was there.
2724 if (NameWasUNC)
2725 Path = (Path.substr(0, 1) + "\\\\?\\" + Path.substr(1)).str();
2726#endif
2727
2728 // For user files and known standard headers, issue a diagnostic.
2729 // For other system headers, don't. They can be controlled separately.
2730 auto DiagId =
2731 (FileCharacter == SrcMgr::C_User || warnByDefaultOnWrongCase(Include: Name))
2732 ? diag::pp_nonportable_path
2733 : diag::pp_nonportable_system_path;
2734 Diag(Tok: FilenameTok, DiagID: DiagId) << Path <<
2735 FixItHint::CreateReplacement(RemoveRange: FilenameRange, Code: Path);
2736 }
2737 }
2738
2739 switch (Action) {
2740 case Skip:
2741 // If we don't need to enter the file, stop now.
2742 if (ModuleToImport)
2743 return {ImportAction::SkippedModuleImport, ModuleToImport};
2744 return {ImportAction::None};
2745
2746 case IncludeLimitReached:
2747 // If we reached our include limit and don't want to enter any more files,
2748 // don't go any further.
2749 return {ImportAction::None};
2750
2751 case Import: {
2752 // If this is a module import, make it visible if needed.
2753 assert(ModuleToImport && "no module to import");
2754
2755 makeModuleVisible(M: ModuleToImport, Loc: EndLoc);
2756
2757 if (IncludeTok.getIdentifierInfo()->getPPKeywordID() ==
2758 tok::pp___include_macros)
2759 return {ImportAction::None};
2760
2761 return {ImportAction::ModuleImport, ModuleToImport};
2762 }
2763
2764 case Enter:
2765 break;
2766 }
2767
2768 // Check that we don't have infinite #include recursion.
2769 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
2770 Diag(Tok: FilenameTok, DiagID: diag::err_pp_include_too_deep);
2771 HasReachedMaxIncludeDepth = true;
2772 return {ImportAction::None};
2773 }
2774
2775 if (isAngled && isInNamedModule())
2776 Diag(Tok: FilenameTok, DiagID: diag::warn_pp_include_angled_in_module_purview)
2777 << getNamedModuleName();
2778
2779 // Look up the file, create a File ID for it.
2780 SourceLocation IncludePos = FilenameTok.getLocation();
2781 // If the filename string was the result of macro expansions, set the include
2782 // position on the file where it will be included and after the expansions.
2783 if (IncludePos.isMacroID())
2784 IncludePos = SourceMgr.getExpansionRange(Loc: IncludePos).getEnd();
2785 FileID FID = SourceMgr.createFileID(SourceFile: *File, IncludePos, FileCharacter);
2786 if (!FID.isValid()) {
2787 TheModuleLoader.HadFatalFailure = true;
2788 return ImportAction::Failure;
2789 }
2790
2791 // If all is good, enter the new file!
2792 if (EnterSourceFile(FID, Dir: CurDir, Loc: FilenameTok.getLocation(),
2793 IsFirstIncludeOfFile))
2794 return {ImportAction::None};
2795
2796 // Determine if we're switching to building a new submodule, and which one.
2797 // This does not apply for C++20 modules header units.
2798 if (ModuleToImport && !ModuleToImport->isHeaderUnit()) {
2799 if (ModuleToImport->getTopLevelModule()->ShadowingModule) {
2800 // We are building a submodule that belongs to a shadowed module. This
2801 // means we find header files in the shadowed module.
2802 Diag(Loc: ModuleToImport->DefinitionLoc,
2803 DiagID: diag::err_module_build_shadowed_submodule)
2804 << ModuleToImport->getFullModuleName();
2805 Diag(Loc: ModuleToImport->getTopLevelModule()->ShadowingModule->DefinitionLoc,
2806 DiagID: diag::note_previous_definition);
2807 return {ImportAction::None};
2808 }
2809 // When building a pch, -fmodule-name tells the compiler to textually
2810 // include headers in the specified module. We are not building the
2811 // specified module.
2812 //
2813 // FIXME: This is the wrong way to handle this. We should produce a PCH
2814 // that behaves the same as the header would behave in a compilation using
2815 // that PCH, which means we should enter the submodule. We need to teach
2816 // the AST serialization layer to deal with the resulting AST.
2817 if (getLangOpts().CompilingPCH &&
2818 ModuleToImport->isForBuilding(LangOpts: getLangOpts()))
2819 return {ImportAction::None};
2820
2821 assert(!CurLexerSubmodule && "should not have marked this as a module yet");
2822 CurLexerSubmodule = ModuleToImport;
2823
2824 // Let the macro handling code know that any future macros are within
2825 // the new submodule.
2826 EnterSubmodule(M: ModuleToImport, ImportLoc: EndLoc, /*ForPragma*/ false);
2827
2828 // Let the parser know that any future declarations are within the new
2829 // submodule.
2830 // FIXME: There's no point doing this if we're handling a #__include_macros
2831 // directive.
2832 return {ImportAction::ModuleBegin, ModuleToImport};
2833 }
2834
2835 assert(!IsImportDecl && "failed to diagnose missing module for import decl");
2836 return {ImportAction::None};
2837}
2838
2839/// HandleIncludeNextDirective - Implements \#include_next.
2840///
2841void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
2842 Token &IncludeNextTok) {
2843 Diag(Tok: IncludeNextTok, DiagID: diag::ext_pp_include_next_directive);
2844
2845 ConstSearchDirIterator Lookup = nullptr;
2846 const FileEntry *LookupFromFile;
2847 std::tie(args&: Lookup, args&: LookupFromFile) = getIncludeNextStart(IncludeNextTok);
2848
2849 return HandleIncludeDirective(HashLoc, IncludeTok&: IncludeNextTok, LookupFrom: Lookup,
2850 LookupFromFile);
2851}
2852
2853/// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
2854void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
2855 // The Microsoft #import directive takes a type library and generates header
2856 // files from it, and includes those. This is beyond the scope of what clang
2857 // does, so we ignore it and error out. However, #import can optionally have
2858 // trailing attributes that span multiple lines. We're going to eat those
2859 // so we can continue processing from there.
2860 Diag(Tok, DiagID: diag::err_pp_import_directive_ms );
2861
2862 // Read tokens until we get to the end of the directive. Note that the
2863 // directive can be split over multiple lines using the backslash character.
2864 DiscardUntilEndOfDirective();
2865}
2866
2867/// HandleImportDirective - Implements \#import.
2868///
2869void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
2870 Token &ImportTok) {
2871 if (!LangOpts.ObjC) { // #import is standard for ObjC.
2872 if (LangOpts.MSVCCompat)
2873 return HandleMicrosoftImportDirective(Tok&: ImportTok);
2874 Diag(Tok: ImportTok, DiagID: diag::ext_pp_import_directive);
2875 }
2876 return HandleIncludeDirective(HashLoc, IncludeTok&: ImportTok);
2877}
2878
2879/// HandleIncludeMacrosDirective - The -imacros command line option turns into a
2880/// pseudo directive in the predefines buffer. This handles it by sucking all
2881/// tokens through the preprocessor and discarding them (only keeping the side
2882/// effects on the preprocessor).
2883void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
2884 Token &IncludeMacrosTok) {
2885 // This directive should only occur in the predefines buffer. If not, emit an
2886 // error and reject it.
2887 SourceLocation Loc = IncludeMacrosTok.getLocation();
2888 if (SourceMgr.getBufferName(Loc) != "<built-in>") {
2889 Diag(Loc: IncludeMacrosTok.getLocation(),
2890 DiagID: diag::pp_include_macros_out_of_predefines);
2891 DiscardUntilEndOfDirective();
2892 return;
2893 }
2894
2895 // Treat this as a normal #include for checking purposes. If this is
2896 // successful, it will push a new lexer onto the include stack.
2897 HandleIncludeDirective(HashLoc, IncludeTok&: IncludeMacrosTok);
2898
2899 Token TmpTok;
2900 do {
2901 Lex(Result&: TmpTok);
2902 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
2903 } while (TmpTok.isNot(K: tok::hashhash));
2904}
2905
2906//===----------------------------------------------------------------------===//
2907// Preprocessor Macro Directive Handling.
2908//===----------------------------------------------------------------------===//
2909
2910/// ReadMacroParameterList - The ( starting a parameter list of a macro
2911/// definition has just been read. Lex the rest of the parameters and the
2912/// closing ), updating MI with what we learn. Return true if an error occurs
2913/// parsing the param list.
2914bool Preprocessor::ReadMacroParameterList(MacroInfo *MI, Token &Tok) {
2915 SmallVector<IdentifierInfo*, 32> Parameters;
2916
2917 while (true) {
2918 LexUnexpandedNonComment(Result&: Tok);
2919 switch (Tok.getKind()) {
2920 case tok::r_paren:
2921 // Found the end of the parameter list.
2922 if (Parameters.empty()) // #define FOO()
2923 return false;
2924 // Otherwise we have #define FOO(A,)
2925 Diag(Tok, DiagID: diag::err_pp_expected_ident_in_arg_list);
2926 return true;
2927 case tok::ellipsis: // #define X(... -> C99 varargs
2928 if (!LangOpts.C99)
2929 Diag(Tok, DiagID: LangOpts.CPlusPlus11 ?
2930 diag::warn_cxx98_compat_variadic_macro :
2931 diag::ext_variadic_macro);
2932
2933 // OpenCL v1.2 s6.9.e: variadic macros are not supported.
2934 if (LangOpts.OpenCL && !LangOpts.OpenCLCPlusPlus) {
2935 Diag(Tok, DiagID: diag::ext_pp_opencl_variadic_macros);
2936 }
2937
2938 // Lex the token after the identifier.
2939 LexUnexpandedNonComment(Result&: Tok);
2940 if (Tok.isNot(K: tok::r_paren)) {
2941 Diag(Tok, DiagID: diag::err_pp_missing_rparen_in_macro_def);
2942 return true;
2943 }
2944 // Add the __VA_ARGS__ identifier as a parameter.
2945 Parameters.push_back(Elt: Ident__VA_ARGS__);
2946 MI->setIsC99Varargs();
2947 MI->setParameterList(List: Parameters, PPAllocator&: BP);
2948 return false;
2949 case tok::eod: // #define X(
2950 Diag(Tok, DiagID: diag::err_pp_missing_rparen_in_macro_def);
2951 return true;
2952 default:
2953 // Handle keywords and identifiers here to accept things like
2954 // #define Foo(for) for.
2955 IdentifierInfo *II = Tok.getIdentifierInfo();
2956 if (!II) {
2957 // #define X(1
2958 Diag(Tok, DiagID: diag::err_pp_invalid_tok_in_arg_list);
2959 return true;
2960 }
2961
2962 // If this is already used as a parameter, it is used multiple times (e.g.
2963 // #define X(A,A.
2964 if (llvm::is_contained(Range&: Parameters, Element: II)) { // C99 6.10.3p6
2965 Diag(Tok, DiagID: diag::err_pp_duplicate_name_in_arg_list) << II;
2966 return true;
2967 }
2968
2969 // Add the parameter to the macro info.
2970 Parameters.push_back(Elt: II);
2971
2972 // Lex the token after the identifier.
2973 LexUnexpandedNonComment(Result&: Tok);
2974
2975 switch (Tok.getKind()) {
2976 default: // #define X(A B
2977 Diag(Tok, DiagID: diag::err_pp_expected_comma_in_arg_list);
2978 return true;
2979 case tok::r_paren: // #define X(A)
2980 MI->setParameterList(List: Parameters, PPAllocator&: BP);
2981 return false;
2982 case tok::comma: // #define X(A,
2983 break;
2984 case tok::ellipsis: // #define X(A... -> GCC extension
2985 // Diagnose extension.
2986 Diag(Tok, DiagID: diag::ext_named_variadic_macro);
2987
2988 // Lex the token after the identifier.
2989 LexUnexpandedNonComment(Result&: Tok);
2990 if (Tok.isNot(K: tok::r_paren)) {
2991 Diag(Tok, DiagID: diag::err_pp_missing_rparen_in_macro_def);
2992 return true;
2993 }
2994
2995 MI->setIsGNUVarargs();
2996 MI->setParameterList(List: Parameters, PPAllocator&: BP);
2997 return false;
2998 }
2999 }
3000 }
3001}
3002
3003static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI,
3004 const LangOptions &LOptions) {
3005 if (MI->getNumTokens() == 1) {
3006 const Token &Value = MI->getReplacementToken(Tok: 0);
3007
3008 // Macro that is identity, like '#define inline inline' is a valid pattern.
3009 if (MacroName.getKind() == Value.getKind())
3010 return true;
3011
3012 // Macro that maps a keyword to the same keyword decorated with leading/
3013 // trailing underscores is a valid pattern:
3014 // #define inline __inline
3015 // #define inline __inline__
3016 // #define inline _inline (in MS compatibility mode)
3017 StringRef MacroText = MacroName.getIdentifierInfo()->getName();
3018 if (IdentifierInfo *II = Value.getIdentifierInfo()) {
3019 if (!II->isKeyword(LangOpts: LOptions))
3020 return false;
3021 StringRef ValueText = II->getName();
3022 StringRef TrimmedValue = ValueText;
3023 if (!ValueText.starts_with(Prefix: "__")) {
3024 if (ValueText.starts_with(Prefix: "_"))
3025 TrimmedValue = TrimmedValue.drop_front(N: 1);
3026 else
3027 return false;
3028 } else {
3029 TrimmedValue = TrimmedValue.drop_front(N: 2);
3030 if (TrimmedValue.ends_with(Suffix: "__"))
3031 TrimmedValue = TrimmedValue.drop_back(N: 2);
3032 }
3033 return TrimmedValue == MacroText;
3034 } else {
3035 return false;
3036 }
3037 }
3038
3039 // #define inline
3040 return MacroName.isOneOf(Ks: tok::kw_extern, Ks: tok::kw_inline, Ks: tok::kw_static,
3041 Ks: tok::kw_const) &&
3042 MI->getNumTokens() == 0;
3043}
3044
3045// ReadOptionalMacroParameterListAndBody - This consumes all (i.e. the
3046// entire line) of the macro's tokens and adds them to MacroInfo, and while
3047// doing so performs certain validity checks including (but not limited to):
3048// - # (stringization) is followed by a macro parameter
3049//
3050// Returns a nullptr if an invalid sequence of tokens is encountered or returns
3051// a pointer to a MacroInfo object.
3052
3053MacroInfo *Preprocessor::ReadOptionalMacroParameterListAndBody(
3054 const Token &MacroNameTok, const bool ImmediatelyAfterHeaderGuard) {
3055
3056 Token LastTok = MacroNameTok;
3057 // Create the new macro.
3058 MacroInfo *const MI = AllocateMacroInfo(L: MacroNameTok.getLocation());
3059
3060 Token Tok;
3061 LexUnexpandedToken(Result&: Tok);
3062
3063 // Ensure we consume the rest of the macro body if errors occur.
3064 llvm::scope_exit _([&]() {
3065 // The flag indicates if we are still waiting for 'eod'.
3066 if (CurLexer->ParsingPreprocessorDirective)
3067 DiscardUntilEndOfDirective();
3068 });
3069
3070 // Used to un-poison and then re-poison identifiers of the __VA_ARGS__ ilk
3071 // within their appropriate context.
3072 VariadicMacroScopeGuard VariadicMacroScopeGuard(*this);
3073
3074 // If this is a function-like macro definition, parse the argument list,
3075 // marking each of the identifiers as being used as macro arguments. Also,
3076 // check other constraints on the first token of the macro body.
3077 if (Tok.is(K: tok::eod)) {
3078 if (ImmediatelyAfterHeaderGuard) {
3079 // Save this macro information since it may part of a header guard.
3080 CurPPLexer->MIOpt.SetDefinedMacro(M: MacroNameTok.getIdentifierInfo(),
3081 Loc: MacroNameTok.getLocation());
3082 }
3083 // If there is no body to this macro, we have no special handling here.
3084 } else if (Tok.hasLeadingSpace()) {
3085 // This is a normal token with leading space. Clear the leading space
3086 // marker on the first token to get proper expansion.
3087 Tok.clearFlag(Flag: Token::LeadingSpace);
3088 } else if (Tok.is(K: tok::l_paren)) {
3089 // This is a function-like macro definition. Read the argument list.
3090 MI->setIsFunctionLike();
3091 if (ReadMacroParameterList(MI, Tok&: LastTok))
3092 return nullptr;
3093
3094 // If this is a definition of an ISO C/C++ variadic function-like macro (not
3095 // using the GNU named varargs extension) inform our variadic scope guard
3096 // which un-poisons and re-poisons certain identifiers (e.g. __VA_ARGS__)
3097 // allowed only within the definition of a variadic macro.
3098
3099 if (MI->isC99Varargs()) {
3100 VariadicMacroScopeGuard.enterScope();
3101 }
3102
3103 // Read the first token after the arg list for down below.
3104 LexUnexpandedToken(Result&: Tok);
3105 } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
3106 // C99 requires whitespace between the macro definition and the body. Emit
3107 // a diagnostic for something like "#define X+".
3108 Diag(Tok, DiagID: diag::ext_c99_whitespace_required_after_macro_name);
3109 } else {
3110 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
3111 // first character of a replacement list is not a character required by
3112 // subclause 5.2.1, then there shall be white-space separation between the
3113 // identifier and the replacement list.". 5.2.1 lists this set:
3114 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
3115 // is irrelevant here.
3116 bool isInvalid = false;
3117 if (Tok.is(K: tok::at)) // @ is not in the list above.
3118 isInvalid = true;
3119 else if (Tok.is(K: tok::unknown)) {
3120 // If we have an unknown token, it is something strange like "`". Since
3121 // all of valid characters would have lexed into a single character
3122 // token of some sort, we know this is not a valid case.
3123 isInvalid = true;
3124 }
3125 if (isInvalid)
3126 Diag(Tok, DiagID: diag::ext_missing_whitespace_after_macro_name);
3127 else
3128 Diag(Tok, DiagID: diag::warn_missing_whitespace_after_macro_name);
3129 }
3130
3131 if (!Tok.is(K: tok::eod))
3132 LastTok = Tok;
3133
3134 SmallVector<Token, 16> Tokens;
3135
3136 // Read the rest of the macro body.
3137 if (MI->isObjectLike()) {
3138 // Object-like macros are very simple, just read their body.
3139 while (Tok.isNot(K: tok::eod)) {
3140 LastTok = Tok;
3141 Tokens.push_back(Elt: Tok);
3142 // Get the next token of the macro.
3143 LexUnexpandedToken(Result&: Tok);
3144 }
3145 } else {
3146 // Otherwise, read the body of a function-like macro. While we are at it,
3147 // check C99 6.10.3.2p1: ensure that # operators are followed by macro
3148 // parameters in function-like macro expansions.
3149
3150 VAOptDefinitionContext VAOCtx(*this);
3151
3152 while (Tok.isNot(K: tok::eod)) {
3153 LastTok = Tok;
3154
3155 if (!Tok.isOneOf(Ks: tok::hash, Ks: tok::hashat, Ks: tok::hashhash)) {
3156 Tokens.push_back(Elt: Tok);
3157
3158 if (VAOCtx.isVAOptToken(T: Tok)) {
3159 // If we're already within a VAOPT, emit an error.
3160 if (VAOCtx.isInVAOpt()) {
3161 Diag(Tok, DiagID: diag::err_pp_vaopt_nested_use);
3162 return nullptr;
3163 }
3164 // Ensure VAOPT is followed by a '(' .
3165 LexUnexpandedToken(Result&: Tok);
3166 if (Tok.isNot(K: tok::l_paren)) {
3167 Diag(Tok, DiagID: diag::err_pp_missing_lparen_in_vaopt_use);
3168 return nullptr;
3169 }
3170 Tokens.push_back(Elt: Tok);
3171 VAOCtx.sawVAOptFollowedByOpeningParens(LParenLoc: Tok.getLocation());
3172 LexUnexpandedToken(Result&: Tok);
3173 if (Tok.is(K: tok::hashhash)) {
3174 Diag(Tok, DiagID: diag::err_vaopt_paste_at_start);
3175 return nullptr;
3176 }
3177 continue;
3178 } else if (VAOCtx.isInVAOpt()) {
3179 if (Tok.is(K: tok::r_paren)) {
3180 if (VAOCtx.sawClosingParen()) {
3181 assert(Tokens.size() >= 3 &&
3182 "Must have seen at least __VA_OPT__( "
3183 "and a subsequent tok::r_paren");
3184 if (Tokens[Tokens.size() - 2].is(K: tok::hashhash)) {
3185 Diag(Tok, DiagID: diag::err_vaopt_paste_at_end);
3186 return nullptr;
3187 }
3188 }
3189 } else if (Tok.is(K: tok::l_paren)) {
3190 VAOCtx.sawOpeningParen(LParenLoc: Tok.getLocation());
3191 }
3192 }
3193 // Get the next token of the macro.
3194 LexUnexpandedToken(Result&: Tok);
3195 continue;
3196 }
3197
3198 // If we're in -traditional mode, then we should ignore stringification
3199 // and token pasting. Mark the tokens as unknown so as not to confuse
3200 // things.
3201 if (getLangOpts().TraditionalCPP) {
3202 Tok.setKind(tok::unknown);
3203 Tokens.push_back(Elt: Tok);
3204
3205 // Get the next token of the macro.
3206 LexUnexpandedToken(Result&: Tok);
3207 continue;
3208 }
3209
3210 if (Tok.is(K: tok::hashhash)) {
3211 // If we see token pasting, check if it looks like the gcc comma
3212 // pasting extension. We'll use this information to suppress
3213 // diagnostics later on.
3214
3215 // Get the next token of the macro.
3216 LexUnexpandedToken(Result&: Tok);
3217
3218 if (Tok.is(K: tok::eod)) {
3219 Tokens.push_back(Elt: LastTok);
3220 break;
3221 }
3222
3223 if (!Tokens.empty() && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
3224 Tokens[Tokens.size() - 1].is(K: tok::comma))
3225 MI->setHasCommaPasting();
3226
3227 // Things look ok, add the '##' token to the macro.
3228 Tokens.push_back(Elt: LastTok);
3229 continue;
3230 }
3231
3232 // Our Token is a stringization operator.
3233 // Get the next token of the macro.
3234 LexUnexpandedToken(Result&: Tok);
3235
3236 // Check for a valid macro arg identifier or __VA_OPT__.
3237 if (!VAOCtx.isVAOptToken(T: Tok) &&
3238 (Tok.getIdentifierInfo() == nullptr ||
3239 MI->getParameterNum(Arg: Tok.getIdentifierInfo()) == -1)) {
3240
3241 // If this is assembler-with-cpp mode, we accept random gibberish after
3242 // the '#' because '#' is often a comment character. However, change
3243 // the kind of the token to tok::unknown so that the preprocessor isn't
3244 // confused.
3245 if (getLangOpts().AsmPreprocessor && Tok.isNot(K: tok::eod)) {
3246 LastTok.setKind(tok::unknown);
3247 Tokens.push_back(Elt: LastTok);
3248 continue;
3249 } else {
3250 Diag(Tok, DiagID: diag::err_pp_stringize_not_parameter)
3251 << LastTok.is(K: tok::hashat);
3252 return nullptr;
3253 }
3254 }
3255
3256 // Things look ok, add the '#' and param name tokens to the macro.
3257 Tokens.push_back(Elt: LastTok);
3258
3259 // If the token following '#' is VAOPT, let the next iteration handle it
3260 // and check it for correctness, otherwise add the token and prime the
3261 // loop with the next one.
3262 if (!VAOCtx.isVAOptToken(T: Tok)) {
3263 Tokens.push_back(Elt: Tok);
3264 LastTok = Tok;
3265
3266 // Get the next token of the macro.
3267 LexUnexpandedToken(Result&: Tok);
3268 }
3269 }
3270 if (VAOCtx.isInVAOpt()) {
3271 assert(Tok.is(tok::eod) && "Must be at End Of preprocessing Directive");
3272 Diag(Tok, DiagID: diag::err_pp_expected_after)
3273 << LastTok.getKind() << tok::r_paren;
3274 Diag(Loc: VAOCtx.getUnmatchedOpeningParenLoc(), DiagID: diag::note_matching) << tok::l_paren;
3275 return nullptr;
3276 }
3277 }
3278 MI->setDefinitionEndLoc(LastTok.getLocation());
3279
3280 MI->setTokens(Tokens, PPAllocator&: BP);
3281 return MI;
3282}
3283
3284static bool isObjCProtectedMacro(const IdentifierInfo *II) {
3285 return II->isStr(Str: "__strong") || II->isStr(Str: "__weak") ||
3286 II->isStr(Str: "__unsafe_unretained") || II->isStr(Str: "__autoreleasing");
3287}
3288
3289/// HandleDefineDirective - Implements \#define. This consumes the entire macro
3290/// line then lets the caller lex the next real token.
3291void Preprocessor::HandleDefineDirective(
3292 Token &DefineTok, const bool ImmediatelyAfterHeaderGuard) {
3293 ++NumDefined;
3294
3295 Token MacroNameTok;
3296 bool MacroShadowsKeyword;
3297 ReadMacroName(MacroNameTok, isDefineUndef: MU_Define, ShadowFlag: &MacroShadowsKeyword);
3298
3299 // Error reading macro name? If so, diagnostic already issued.
3300 if (MacroNameTok.is(K: tok::eod))
3301 return;
3302
3303 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
3304 // Issue a final pragma warning if we're defining a macro that was has been
3305 // undefined and is being redefined.
3306 if (!II->hasMacroDefinition() && II->hadMacroDefinition() && II->isFinal())
3307 emitFinalMacroWarning(Identifier: MacroNameTok, /*IsUndef=*/false);
3308
3309 // If we are supposed to keep comments in #defines, reenable comment saving
3310 // mode.
3311 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
3312
3313 MacroInfo *const MI = ReadOptionalMacroParameterListAndBody(
3314 MacroNameTok, ImmediatelyAfterHeaderGuard);
3315
3316 if (!MI) return;
3317
3318 if (MacroShadowsKeyword &&
3319 !isConfigurationPattern(MacroName&: MacroNameTok, MI, LOptions: getLangOpts())) {
3320 Diag(Tok: MacroNameTok, DiagID: diag::warn_pp_macro_hides_keyword);
3321 }
3322 // Check that there is no paste (##) operator at the beginning or end of the
3323 // replacement list.
3324 unsigned NumTokens = MI->getNumTokens();
3325 if (NumTokens != 0) {
3326 if (MI->getReplacementToken(Tok: 0).is(K: tok::hashhash)) {
3327 Diag(Tok: MI->getReplacementToken(Tok: 0), DiagID: diag::err_paste_at_start);
3328 return;
3329 }
3330 if (MI->getReplacementToken(Tok: NumTokens-1).is(K: tok::hashhash)) {
3331 Diag(Tok: MI->getReplacementToken(Tok: NumTokens-1), DiagID: diag::err_paste_at_end);
3332 return;
3333 }
3334 }
3335
3336 // When skipping just warn about macros that do not match.
3337 if (SkippingUntilPCHThroughHeader) {
3338 const MacroInfo *OtherMI = getMacroInfo(II: MacroNameTok.getIdentifierInfo());
3339 if (!OtherMI || !MI->isIdenticalTo(Other: *OtherMI, PP&: *this,
3340 /*Syntactic=*/Syntactically: LangOpts.MicrosoftExt))
3341 Diag(Loc: MI->getDefinitionLoc(), DiagID: diag::warn_pp_macro_def_mismatch_with_pch)
3342 << MacroNameTok.getIdentifierInfo();
3343 // Issue the diagnostic but allow the change if msvc extensions are enabled
3344 if (!LangOpts.MicrosoftExt)
3345 return;
3346 }
3347
3348 // Finally, if this identifier already had a macro defined for it, verify that
3349 // the macro bodies are identical, and issue diagnostics if they are not.
3350 if (const MacroInfo *OtherMI=getMacroInfo(II: MacroNameTok.getIdentifierInfo())) {
3351 // Final macros are hard-mode: they always warn. Even if the bodies are
3352 // identical. Even if they are in system headers. Even if they are things we
3353 // would silently allow in the past.
3354 if (MacroNameTok.getIdentifierInfo()->isFinal())
3355 emitFinalMacroWarning(Identifier: MacroNameTok, /*IsUndef=*/false);
3356
3357 // In Objective-C, ignore attempts to directly redefine the builtin
3358 // definitions of the ownership qualifiers. It's still possible to
3359 // #undef them.
3360 if (getLangOpts().ObjC &&
3361 SourceMgr.getFileID(SpellingLoc: OtherMI->getDefinitionLoc()) ==
3362 getPredefinesFileID() &&
3363 isObjCProtectedMacro(II: MacroNameTok.getIdentifierInfo())) {
3364 // Warn if it changes the tokens.
3365 if ((!getDiagnostics().getSuppressSystemWarnings() ||
3366 !SourceMgr.isInSystemHeader(Loc: DefineTok.getLocation())) &&
3367 !MI->isIdenticalTo(Other: *OtherMI, PP&: *this,
3368 /*Syntactic=*/Syntactically: LangOpts.MicrosoftExt)) {
3369 Diag(Loc: MI->getDefinitionLoc(), DiagID: diag::warn_pp_objc_macro_redef_ignored);
3370 }
3371 assert(!OtherMI->isWarnIfUnused());
3372 return;
3373 }
3374
3375 // It is very common for system headers to have tons of macro redefinitions
3376 // and for warnings to be disabled in system headers. If this is the case,
3377 // then don't bother calling MacroInfo::isIdenticalTo.
3378 if (!getDiagnostics().getSuppressSystemWarnings() ||
3379 !SourceMgr.isInSystemHeader(Loc: DefineTok.getLocation())) {
3380
3381 if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
3382 Diag(Loc: OtherMI->getDefinitionLoc(), DiagID: diag::pp_macro_not_used);
3383
3384 // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
3385 // C++ [cpp.predefined]p4, but allow it as an extension.
3386 if (isLanguageDefinedBuiltin(SourceMgr, MI: OtherMI, MacroName: II->getName()))
3387 Diag(Tok: MacroNameTok, DiagID: diag::ext_pp_redef_builtin_macro);
3388 // Macros must be identical. This means all tokens and whitespace
3389 // separation must be the same. C99 6.10.3p2.
3390 else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
3391 !MI->isIdenticalTo(Other: *OtherMI, PP&: *this, /*Syntactic=*/Syntactically: LangOpts.MicrosoftExt)) {
3392 Diag(Loc: MI->getDefinitionLoc(), DiagID: diag::ext_pp_macro_redef)
3393 << MacroNameTok.getIdentifierInfo();
3394 Diag(Loc: OtherMI->getDefinitionLoc(), DiagID: diag::note_previous_definition);
3395 }
3396 }
3397 if (OtherMI->isWarnIfUnused())
3398 WarnUnusedMacroLocs.erase(V: OtherMI->getDefinitionLoc());
3399 }
3400
3401 DefMacroDirective *MD =
3402 appendDefMacroDirective(II: MacroNameTok.getIdentifierInfo(), MI);
3403
3404 assert(!MI->isUsed());
3405 // If we need warning for not using the macro, add its location in the
3406 // warn-because-unused-macro set. If it gets used it will be removed from set.
3407 if (getSourceManager().isInMainFile(Loc: MI->getDefinitionLoc()) &&
3408 !Diags->isIgnored(DiagID: diag::pp_macro_not_used, Loc: MI->getDefinitionLoc()) &&
3409 !MacroExpansionInDirectivesOverride &&
3410 getSourceManager().getFileID(SpellingLoc: MI->getDefinitionLoc()) !=
3411 getPredefinesFileID()) {
3412 MI->setIsWarnIfUnused(true);
3413 WarnUnusedMacroLocs.insert(V: MI->getDefinitionLoc());
3414 }
3415
3416 // If the callbacks want to know, tell them about the macro definition.
3417 if (Callbacks)
3418 Callbacks->MacroDefined(MacroNameTok, MD);
3419}
3420
3421/// HandleUndefDirective - Implements \#undef.
3422///
3423void Preprocessor::HandleUndefDirective() {
3424 ++NumUndefined;
3425
3426 Token MacroNameTok;
3427 ReadMacroName(MacroNameTok, isDefineUndef: MU_Undef);
3428
3429 // Error reading macro name? If so, diagnostic already issued.
3430 if (MacroNameTok.is(K: tok::eod))
3431 return;
3432
3433 // Check to see if this is the last token on the #undef line.
3434 CheckEndOfDirective(DirType: "undef");
3435
3436 // Okay, we have a valid identifier to undef.
3437 auto *II = MacroNameTok.getIdentifierInfo();
3438 auto MD = getMacroDefinition(II);
3439 UndefMacroDirective *Undef = nullptr;
3440
3441 if (II->isFinal())
3442 emitFinalMacroWarning(Identifier: MacroNameTok, /*IsUndef=*/true);
3443
3444 // If the macro is not defined, this is a noop undef.
3445 if (const MacroInfo *MI = MD.getMacroInfo()) {
3446 if (!MI->isUsed() && MI->isWarnIfUnused())
3447 Diag(Loc: MI->getDefinitionLoc(), DiagID: diag::pp_macro_not_used);
3448
3449 // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4 and
3450 // C++ [cpp.predefined]p4, but allow it as an extension.
3451 if (isLanguageDefinedBuiltin(SourceMgr, MI, MacroName: II->getName()))
3452 Diag(Tok: MacroNameTok, DiagID: diag::ext_pp_undef_builtin_macro);
3453
3454 if (MI->isWarnIfUnused())
3455 WarnUnusedMacroLocs.erase(V: MI->getDefinitionLoc());
3456
3457 Undef = AllocateUndefMacroDirective(UndefLoc: MacroNameTok.getLocation());
3458 }
3459
3460 // If the callbacks want to know, tell them about the macro #undef.
3461 // Note: no matter if the macro was defined or not.
3462 if (Callbacks)
3463 Callbacks->MacroUndefined(MacroNameTok, MD, Undef);
3464
3465 if (Undef)
3466 appendMacroDirective(II, MD: Undef);
3467}
3468
3469//===----------------------------------------------------------------------===//
3470// Preprocessor Conditional Directive Handling.
3471//===----------------------------------------------------------------------===//
3472
3473/// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive. isIfndef
3474/// is true when this is a \#ifndef directive. ReadAnyTokensBeforeDirective is
3475/// true if any tokens have been returned or pp-directives activated before this
3476/// \#ifndef has been lexed.
3477///
3478void Preprocessor::HandleIfdefDirective(Token &Result,
3479 const Token &HashToken,
3480 bool isIfndef,
3481 bool ReadAnyTokensBeforeDirective) {
3482 ++NumIf;
3483 Token DirectiveTok = Result;
3484
3485 Token MacroNameTok;
3486 ReadMacroName(MacroNameTok);
3487
3488 // Error reading macro name? If so, diagnostic already issued.
3489 if (MacroNameTok.is(K: tok::eod)) {
3490 // Skip code until we get to #endif. This helps with recovery by not
3491 // emitting an error when the #endif is reached.
3492 SkipExcludedConditionalBlock(HashTokenLoc: HashToken.getLocation(),
3493 IfTokenLoc: DirectiveTok.getLocation(),
3494 /*Foundnonskip*/ FoundNonSkipPortion: false, /*FoundElse*/ false);
3495 return;
3496 }
3497
3498 emitMacroExpansionWarnings(Identifier: MacroNameTok, /*IsIfnDef=*/true);
3499
3500 // Check to see if this is the last token on the #if[n]def line.
3501 CheckEndOfDirective(DirType: isIfndef ? "ifndef" : "ifdef");
3502
3503 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
3504 auto MD = getMacroDefinition(II: MII);
3505 MacroInfo *MI = MD.getMacroInfo();
3506
3507 if (CurPPLexer->getConditionalStackDepth() == 0) {
3508 // If the start of a top-level #ifdef and if the macro is not defined,
3509 // inform MIOpt that this might be the start of a proper include guard.
3510 // Otherwise it is some other form of unknown conditional which we can't
3511 // handle.
3512 if (!ReadAnyTokensBeforeDirective && !MI) {
3513 assert(isIfndef && "#ifdef shouldn't reach here");
3514 CurPPLexer->MIOpt.EnterTopLevelIfndef(M: MII, Loc: MacroNameTok.getLocation());
3515 } else
3516 CurPPLexer->MIOpt.EnterTopLevelConditional();
3517 }
3518
3519 // If there is a macro, process it.
3520 if (MI) // Mark it used.
3521 markMacroAsUsed(MI);
3522
3523 if (Callbacks) {
3524 if (isIfndef)
3525 Callbacks->Ifndef(Loc: DirectiveTok.getLocation(), MacroNameTok, MD);
3526 else
3527 Callbacks->Ifdef(Loc: DirectiveTok.getLocation(), MacroNameTok, MD);
3528 }
3529
3530 bool RetainExcludedCB = PPOpts.RetainExcludedConditionalBlocks &&
3531 getSourceManager().isInMainFile(Loc: DirectiveTok.getLocation());
3532
3533 // Should we include the stuff contained by this directive?
3534 if (PPOpts.SingleFileParseMode && !MI) {
3535 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
3536 // the directive blocks.
3537 CurPPLexer->pushConditionalLevel(DirectiveStart: DirectiveTok.getLocation(),
3538 /*wasskip*/WasSkipping: false, /*foundnonskip*/FoundNonSkip: false,
3539 /*foundelse*/FoundElse: false);
3540 } else if (PPOpts.SingleModuleParseMode && !MI) {
3541 // In 'single-module-parse mode' undefined identifiers trigger skipping of
3542 // all the directive blocks. We lie here and set FoundNonSkipPortion so that
3543 // even any \#else blocks get skipped.
3544 SkipExcludedConditionalBlock(
3545 HashTokenLoc: HashToken.getLocation(), IfTokenLoc: DirectiveTok.getLocation(),
3546 /*FoundNonSkipPortion=*/true, /*FoundElse=*/false);
3547 } else if (!MI == isIfndef || RetainExcludedCB) {
3548 // Yes, remember that we are inside a conditional, then lex the next token.
3549 CurPPLexer->pushConditionalLevel(DirectiveStart: DirectiveTok.getLocation(),
3550 /*wasskip*/WasSkipping: false, /*foundnonskip*/FoundNonSkip: true,
3551 /*foundelse*/FoundElse: false);
3552 } else {
3553 // No, skip the contents of this block.
3554 SkipExcludedConditionalBlock(HashTokenLoc: HashToken.getLocation(),
3555 IfTokenLoc: DirectiveTok.getLocation(),
3556 /*Foundnonskip*/ FoundNonSkipPortion: false,
3557 /*FoundElse*/ false);
3558 }
3559}
3560
3561/// HandleIfDirective - Implements the \#if directive.
3562///
3563void Preprocessor::HandleIfDirective(Token &IfToken,
3564 const Token &HashToken,
3565 bool ReadAnyTokensBeforeDirective) {
3566 ++NumIf;
3567
3568 // Parse and evaluate the conditional expression.
3569 IdentifierInfo *IfNDefMacro = nullptr;
3570 const DirectiveEvalResult DER = EvaluateDirectiveExpression(IfNDefMacro);
3571 const bool ConditionalTrue = DER.Conditional;
3572 // Lexer might become invalid if we hit code completion point while evaluating
3573 // expression.
3574 if (!CurPPLexer)
3575 return;
3576
3577 // If this condition is equivalent to #ifndef X, and if this is the first
3578 // directive seen, handle it for the multiple-include optimization.
3579 if (CurPPLexer->getConditionalStackDepth() == 0) {
3580 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
3581 // FIXME: Pass in the location of the macro name, not the 'if' token.
3582 CurPPLexer->MIOpt.EnterTopLevelIfndef(M: IfNDefMacro, Loc: IfToken.getLocation());
3583 else
3584 CurPPLexer->MIOpt.EnterTopLevelConditional();
3585 }
3586
3587 if (Callbacks)
3588 Callbacks->If(
3589 Loc: IfToken.getLocation(), ConditionRange: DER.ExprRange,
3590 ConditionValue: (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False));
3591
3592 bool RetainExcludedCB = PPOpts.RetainExcludedConditionalBlocks &&
3593 getSourceManager().isInMainFile(Loc: IfToken.getLocation());
3594
3595 // Should we include the stuff contained by this directive?
3596 if (PPOpts.SingleFileParseMode && DER.IncludedUndefinedIds) {
3597 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
3598 // the directive blocks.
3599 CurPPLexer->pushConditionalLevel(DirectiveStart: IfToken.getLocation(), /*wasskip*/WasSkipping: false,
3600 /*foundnonskip*/FoundNonSkip: false, /*foundelse*/FoundElse: false);
3601 } else if (PPOpts.SingleModuleParseMode && DER.IncludedUndefinedIds) {
3602 // In 'single-module-parse mode' undefined identifiers trigger skipping of
3603 // all the directive blocks. We lie here and set FoundNonSkipPortion so that
3604 // even any \#else blocks get skipped.
3605 SkipExcludedConditionalBlock(HashTokenLoc: HashToken.getLocation(), IfTokenLoc: IfToken.getLocation(),
3606 /*FoundNonSkipPortion=*/true,
3607 /*FoundElse=*/false);
3608 } else if (ConditionalTrue || RetainExcludedCB) {
3609 // Yes, remember that we are inside a conditional, then lex the next token.
3610 CurPPLexer->pushConditionalLevel(DirectiveStart: IfToken.getLocation(), /*wasskip*/WasSkipping: false,
3611 /*foundnonskip*/FoundNonSkip: true, /*foundelse*/FoundElse: false);
3612 } else {
3613 // No, skip the contents of this block.
3614 SkipExcludedConditionalBlock(HashTokenLoc: HashToken.getLocation(), IfTokenLoc: IfToken.getLocation(),
3615 /*Foundnonskip*/ FoundNonSkipPortion: false,
3616 /*FoundElse*/ false);
3617 }
3618}
3619
3620/// HandleEndifDirective - Implements the \#endif directive.
3621///
3622void Preprocessor::HandleEndifDirective(Token &EndifToken) {
3623 ++NumEndif;
3624
3625 // Check that this is the whole directive.
3626 CheckEndOfDirective(DirType: "endif");
3627
3628 PPConditionalInfo CondInfo;
3629 if (CurPPLexer->popConditionalLevel(CI&: CondInfo)) {
3630 // No conditionals on the stack: this is an #endif without an #if.
3631 Diag(Tok: EndifToken, DiagID: diag::err_pp_endif_without_if);
3632 return;
3633 }
3634
3635 // If this the end of a top-level #endif, inform MIOpt.
3636 if (CurPPLexer->getConditionalStackDepth() == 0)
3637 CurPPLexer->MIOpt.ExitTopLevelConditional();
3638
3639 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
3640 "This code should only be reachable in the non-skipping case!");
3641
3642 if (Callbacks)
3643 Callbacks->Endif(Loc: EndifToken.getLocation(), IfLoc: CondInfo.IfLoc);
3644}
3645
3646/// HandleElseDirective - Implements the \#else directive.
3647///
3648void Preprocessor::HandleElseDirective(Token &Result, const Token &HashToken) {
3649 ++NumElse;
3650
3651 // #else directive in a non-skipping conditional... start skipping.
3652 CheckEndOfDirective(DirType: "else");
3653
3654 PPConditionalInfo CI;
3655 if (CurPPLexer->popConditionalLevel(CI)) {
3656 Diag(Tok: Result, DiagID: diag::pp_err_else_without_if);
3657 return;
3658 }
3659
3660 // If this is a top-level #else, inform the MIOpt.
3661 if (CurPPLexer->getConditionalStackDepth() == 0)
3662 CurPPLexer->MIOpt.EnterTopLevelConditional();
3663
3664 // If this is a #else with a #else before it, report the error.
3665 if (CI.FoundElse) Diag(Tok: Result, DiagID: diag::pp_err_else_after_else);
3666
3667 if (Callbacks)
3668 Callbacks->Else(Loc: Result.getLocation(), IfLoc: CI.IfLoc);
3669
3670 bool RetainExcludedCB = PPOpts.RetainExcludedConditionalBlocks &&
3671 getSourceManager().isInMainFile(Loc: Result.getLocation());
3672
3673 if ((PPOpts.SingleFileParseMode && !CI.FoundNonSkip) || RetainExcludedCB) {
3674 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
3675 // the directive blocks.
3676 CurPPLexer->pushConditionalLevel(DirectiveStart: CI.IfLoc, /*wasskip*/WasSkipping: false,
3677 /*foundnonskip*/FoundNonSkip: false, /*foundelse*/FoundElse: true);
3678 return;
3679 }
3680
3681 // Finally, skip the rest of the contents of this block.
3682 SkipExcludedConditionalBlock(HashTokenLoc: HashToken.getLocation(), IfTokenLoc: CI.IfLoc,
3683 /*Foundnonskip*/ FoundNonSkipPortion: true,
3684 /*FoundElse*/ true, ElseLoc: Result.getLocation());
3685}
3686
3687/// Implements the \#elif, \#elifdef, and \#elifndef directives.
3688void Preprocessor::HandleElifFamilyDirective(Token &ElifToken,
3689 const Token &HashToken,
3690 tok::PPKeywordKind Kind) {
3691 PPElifDiag DirKind = Kind == tok::pp_elif ? PED_Elif
3692 : Kind == tok::pp_elifdef ? PED_Elifdef
3693 : PED_Elifndef;
3694 ++NumElse;
3695
3696 // Warn if using `#elifdef` & `#elifndef` in not C23 & C++23 mode.
3697 switch (DirKind) {
3698 case PED_Elifdef:
3699 case PED_Elifndef:
3700 unsigned DiagID;
3701 if (LangOpts.CPlusPlus)
3702 DiagID = LangOpts.CPlusPlus23 ? diag::warn_cxx23_compat_pp_directive
3703 : diag::ext_cxx23_pp_directive;
3704 else
3705 DiagID = LangOpts.C23 ? diag::warn_c23_compat_pp_directive
3706 : diag::ext_c23_pp_directive;
3707 Diag(Tok: ElifToken, DiagID) << DirKind;
3708 break;
3709 default:
3710 break;
3711 }
3712
3713 // #elif directive in a non-skipping conditional... start skipping.
3714 // We don't care what the condition is, because we will always skip it (since
3715 // the block immediately before it was included).
3716 SourceRange ConditionRange = DiscardUntilEndOfDirective();
3717
3718 PPConditionalInfo CI;
3719 if (CurPPLexer->popConditionalLevel(CI)) {
3720 Diag(Tok: ElifToken, DiagID: diag::pp_err_elif_without_if) << DirKind;
3721 return;
3722 }
3723
3724 // If this is a top-level #elif, inform the MIOpt.
3725 if (CurPPLexer->getConditionalStackDepth() == 0)
3726 CurPPLexer->MIOpt.EnterTopLevelConditional();
3727
3728 // If this is a #elif with a #else before it, report the error.
3729 if (CI.FoundElse)
3730 Diag(Tok: ElifToken, DiagID: diag::pp_err_elif_after_else) << DirKind;
3731
3732 if (Callbacks) {
3733 switch (Kind) {
3734 case tok::pp_elif:
3735 Callbacks->Elif(Loc: ElifToken.getLocation(), ConditionRange,
3736 ConditionValue: PPCallbacks::CVK_NotEvaluated, IfLoc: CI.IfLoc);
3737 break;
3738 case tok::pp_elifdef:
3739 Callbacks->Elifdef(Loc: ElifToken.getLocation(), ConditionRange, IfLoc: CI.IfLoc);
3740 break;
3741 case tok::pp_elifndef:
3742 Callbacks->Elifndef(Loc: ElifToken.getLocation(), ConditionRange, IfLoc: CI.IfLoc);
3743 break;
3744 default:
3745 assert(false && "unexpected directive kind");
3746 break;
3747 }
3748 }
3749
3750 bool RetainExcludedCB = PPOpts.RetainExcludedConditionalBlocks &&
3751 getSourceManager().isInMainFile(Loc: ElifToken.getLocation());
3752
3753 if ((PPOpts.SingleFileParseMode && !CI.FoundNonSkip) || RetainExcludedCB) {
3754 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
3755 // the directive blocks.
3756 CurPPLexer->pushConditionalLevel(DirectiveStart: ElifToken.getLocation(), /*wasskip*/WasSkipping: false,
3757 /*foundnonskip*/FoundNonSkip: false, /*foundelse*/FoundElse: false);
3758 return;
3759 }
3760
3761 // Finally, skip the rest of the contents of this block.
3762 SkipExcludedConditionalBlock(
3763 HashTokenLoc: HashToken.getLocation(), IfTokenLoc: CI.IfLoc, /*Foundnonskip*/ FoundNonSkipPortion: true,
3764 /*FoundElse*/ CI.FoundElse, ElseLoc: ElifToken.getLocation());
3765}
3766
3767std::optional<LexEmbedParametersResult>
3768Preprocessor::LexEmbedParameters(Token &CurTok, bool ForHasEmbed) {
3769 LexEmbedParametersResult Result{};
3770 tok::TokenKind EndTokenKind = ForHasEmbed ? tok::r_paren : tok::eod;
3771
3772 auto DiagMismatchedBracesAndSkipToEOD =
3773 [&](tok::TokenKind Expected,
3774 std::pair<tok::TokenKind, SourceLocation> Matches) {
3775 Diag(Tok: CurTok, DiagID: diag::err_expected) << Expected;
3776 Diag(Loc: Matches.second, DiagID: diag::note_matching) << Matches.first;
3777 if (CurTok.isNot(K: tok::eod))
3778 DiscardUntilEndOfDirective(Tmp&: CurTok);
3779 };
3780
3781 auto ExpectOrDiagAndSkipToEOD = [&](tok::TokenKind Kind) {
3782 if (CurTok.isNot(K: Kind)) {
3783 Diag(Tok: CurTok, DiagID: diag::err_expected) << Kind;
3784 if (CurTok.isNot(K: tok::eod))
3785 DiscardUntilEndOfDirective(Tmp&: CurTok);
3786 return false;
3787 }
3788 return true;
3789 };
3790
3791 // C23 6.10:
3792 // pp-parameter-name:
3793 // pp-standard-parameter
3794 // pp-prefixed-parameter
3795 //
3796 // pp-standard-parameter:
3797 // identifier
3798 //
3799 // pp-prefixed-parameter:
3800 // identifier :: identifier
3801 auto LexPPParameterName = [&]() -> std::optional<std::string> {
3802 // We expect the current token to be an identifier; if it's not, things
3803 // have gone wrong.
3804 if (!ExpectOrDiagAndSkipToEOD(tok::identifier))
3805 return std::nullopt;
3806
3807 const IdentifierInfo *Prefix = CurTok.getIdentifierInfo();
3808
3809 // Lex another token; it is either a :: or we're done with the parameter
3810 // name.
3811 LexNonComment(Result&: CurTok);
3812 if (CurTok.is(K: tok::coloncolon)) {
3813 // We found a ::, so lex another identifier token.
3814 LexNonComment(Result&: CurTok);
3815 if (!ExpectOrDiagAndSkipToEOD(tok::identifier))
3816 return std::nullopt;
3817
3818 const IdentifierInfo *Suffix = CurTok.getIdentifierInfo();
3819
3820 // Lex another token so we're past the name.
3821 LexNonComment(Result&: CurTok);
3822 return (llvm::Twine(Prefix->getName()) + "::" + Suffix->getName()).str();
3823 }
3824 return Prefix->getName().str();
3825 };
3826
3827 // C23 6.10p5: In all aspects, a preprocessor standard parameter specified by
3828 // this document as an identifier pp_param and an identifier of the form
3829 // __pp_param__ shall behave the same when used as a preprocessor parameter,
3830 // except for the spelling.
3831 auto NormalizeParameterName = [](StringRef Name) {
3832 if (Name.size() > 4 && Name.starts_with(Prefix: "__") && Name.ends_with(Suffix: "__"))
3833 return Name.substr(Start: 2, N: Name.size() - 4);
3834 return Name;
3835 };
3836
3837 auto LexParenthesizedIntegerExpr = [&]() -> std::optional<size_t> {
3838 // we have a limit parameter and its internals are processed using
3839 // evaluation rules from #if.
3840 if (!ExpectOrDiagAndSkipToEOD(tok::l_paren))
3841 return std::nullopt;
3842
3843 // We do not consume the ( because EvaluateDirectiveExpression will lex
3844 // the next token for us.
3845 IdentifierInfo *ParameterIfNDef = nullptr;
3846 bool EvaluatedDefined;
3847 DirectiveEvalResult LimitEvalResult = EvaluateDirectiveExpression(
3848 IfNDefMacro&: ParameterIfNDef, Tok&: CurTok, EvaluatedDefined, /*CheckForEOD=*/CheckForEoD: false);
3849
3850 if (!LimitEvalResult.Value) {
3851 // If there was an error evaluating the directive expression, we expect
3852 // to be at the end of directive token.
3853 assert(CurTok.is(tok::eod) && "expect to be at the end of directive");
3854 return std::nullopt;
3855 }
3856
3857 if (!ExpectOrDiagAndSkipToEOD(tok::r_paren))
3858 return std::nullopt;
3859
3860 // Eat the ).
3861 LexNonComment(Result&: CurTok);
3862
3863 // C23 6.10.3.2p2: The token defined shall not appear within the constant
3864 // expression.
3865 if (EvaluatedDefined) {
3866 Diag(Tok: CurTok, DiagID: diag::err_defined_in_pp_embed);
3867 return std::nullopt;
3868 }
3869
3870 if (LimitEvalResult.Value) {
3871 const llvm::APSInt &Result = *LimitEvalResult.Value;
3872 if (Result.isNegative()) {
3873 Diag(Tok: CurTok, DiagID: diag::err_requires_positive_value)
3874 << toString(I: Result, Radix: 10) << /*positive*/ 0;
3875 if (CurTok.isNot(K: EndTokenKind))
3876 DiscardUntilEndOfDirective(Tmp&: CurTok);
3877 return std::nullopt;
3878 }
3879 return Result.getLimitedValue();
3880 }
3881 return std::nullopt;
3882 };
3883
3884 auto GetMatchingCloseBracket = [](tok::TokenKind Kind) {
3885 switch (Kind) {
3886 case tok::l_paren:
3887 return tok::r_paren;
3888 case tok::l_brace:
3889 return tok::r_brace;
3890 case tok::l_square:
3891 return tok::r_square;
3892 default:
3893 llvm_unreachable("should not get here");
3894 }
3895 };
3896
3897 auto LexParenthesizedBalancedTokenSoup =
3898 [&](llvm::SmallVectorImpl<Token> &Tokens) {
3899 std::vector<std::pair<tok::TokenKind, SourceLocation>> BracketStack;
3900
3901 // We expect the current token to be a left paren.
3902 if (!ExpectOrDiagAndSkipToEOD(tok::l_paren))
3903 return false;
3904 LexNonComment(Result&: CurTok); // Eat the (
3905
3906 bool WaitingForInnerCloseParen = false;
3907 while (CurTok.isNot(K: tok::eod) &&
3908 (WaitingForInnerCloseParen || CurTok.isNot(K: tok::r_paren))) {
3909 switch (CurTok.getKind()) {
3910 default: // Shutting up diagnostics about not fully-covered switch.
3911 break;
3912 case tok::l_paren:
3913 WaitingForInnerCloseParen = true;
3914 [[fallthrough]];
3915 case tok::l_brace:
3916 case tok::l_square:
3917 BracketStack.push_back(x: {CurTok.getKind(), CurTok.getLocation()});
3918 break;
3919 case tok::r_paren:
3920 WaitingForInnerCloseParen = false;
3921 [[fallthrough]];
3922 case tok::r_brace:
3923 case tok::r_square: {
3924 if (BracketStack.empty()) {
3925 ExpectOrDiagAndSkipToEOD(tok::r_paren);
3926 return false;
3927 }
3928 tok::TokenKind Matching =
3929 GetMatchingCloseBracket(BracketStack.back().first);
3930 if (CurTok.getKind() != Matching) {
3931 DiagMismatchedBracesAndSkipToEOD(Matching, BracketStack.back());
3932 return false;
3933 }
3934 BracketStack.pop_back();
3935 } break;
3936 }
3937 Tokens.push_back(Elt: CurTok);
3938 LexNonComment(Result&: CurTok);
3939 }
3940
3941 // When we're done, we want to eat the closing paren.
3942 if (!ExpectOrDiagAndSkipToEOD(tok::r_paren))
3943 return false;
3944
3945 LexNonComment(Result&: CurTok); // Eat the )
3946 return true;
3947 };
3948
3949 LexNonComment(Result&: CurTok); // Prime the pump.
3950 while (!CurTok.isOneOf(Ks: EndTokenKind, Ks: tok::eod)) {
3951 SourceLocation ParamStartLoc = CurTok.getLocation();
3952 std::optional<std::string> ParamName = LexPPParameterName();
3953 if (!ParamName)
3954 return std::nullopt;
3955 StringRef Parameter = NormalizeParameterName(*ParamName);
3956
3957 // Lex the parameters (dependent on the parameter type we want!).
3958 //
3959 // C23 6.10.3.Xp1: The X standard embed parameter may appear zero times or
3960 // one time in the embed parameter sequence.
3961 if (Parameter == "limit") {
3962 if (Result.MaybeLimitParam)
3963 Diag(Tok: CurTok, DiagID: diag::err_pp_embed_dup_params) << Parameter;
3964
3965 std::optional<size_t> Limit = LexParenthesizedIntegerExpr();
3966 if (!Limit)
3967 return std::nullopt;
3968 Result.MaybeLimitParam =
3969 PPEmbedParameterLimit{*Limit, {ParamStartLoc, CurTok.getLocation()}};
3970 } else if (Parameter == "clang::offset") {
3971 if (Result.MaybeOffsetParam)
3972 Diag(Tok: CurTok, DiagID: diag::err_pp_embed_dup_params) << Parameter;
3973
3974 std::optional<size_t> Offset = LexParenthesizedIntegerExpr();
3975 if (!Offset)
3976 return std::nullopt;
3977 Result.MaybeOffsetParam = PPEmbedParameterOffset{
3978 *Offset, {ParamStartLoc, CurTok.getLocation()}};
3979 } else if (Parameter == "prefix") {
3980 if (Result.MaybePrefixParam)
3981 Diag(Tok: CurTok, DiagID: diag::err_pp_embed_dup_params) << Parameter;
3982
3983 SmallVector<Token, 4> Soup;
3984 if (!LexParenthesizedBalancedTokenSoup(Soup))
3985 return std::nullopt;
3986 Result.MaybePrefixParam = PPEmbedParameterPrefix{
3987 std::move(Soup), {ParamStartLoc, CurTok.getLocation()}};
3988 } else if (Parameter == "suffix") {
3989 if (Result.MaybeSuffixParam)
3990 Diag(Tok: CurTok, DiagID: diag::err_pp_embed_dup_params) << Parameter;
3991
3992 SmallVector<Token, 4> Soup;
3993 if (!LexParenthesizedBalancedTokenSoup(Soup))
3994 return std::nullopt;
3995 Result.MaybeSuffixParam = PPEmbedParameterSuffix{
3996 std::move(Soup), {ParamStartLoc, CurTok.getLocation()}};
3997 } else if (Parameter == "if_empty") {
3998 if (Result.MaybeIfEmptyParam)
3999 Diag(Tok: CurTok, DiagID: diag::err_pp_embed_dup_params) << Parameter;
4000
4001 SmallVector<Token, 4> Soup;
4002 if (!LexParenthesizedBalancedTokenSoup(Soup))
4003 return std::nullopt;
4004 Result.MaybeIfEmptyParam = PPEmbedParameterIfEmpty{
4005 std::move(Soup), {ParamStartLoc, CurTok.getLocation()}};
4006 } else {
4007 ++Result.UnrecognizedParams;
4008
4009 // If there's a left paren, we need to parse a balanced token sequence
4010 // and just eat those tokens.
4011 if (CurTok.is(K: tok::l_paren)) {
4012 SmallVector<Token, 4> Soup;
4013 if (!LexParenthesizedBalancedTokenSoup(Soup))
4014 return std::nullopt;
4015 }
4016 if (!ForHasEmbed) {
4017 Diag(Loc: ParamStartLoc, DiagID: diag::err_pp_unknown_parameter) << 1 << Parameter;
4018 if (CurTok.isNot(K: EndTokenKind))
4019 DiscardUntilEndOfDirective(Tmp&: CurTok);
4020 return std::nullopt;
4021 }
4022 }
4023 }
4024 return Result;
4025}
4026
4027void Preprocessor::HandleEmbedDirectiveImpl(
4028 SourceLocation HashLoc, const LexEmbedParametersResult &Params,
4029 StringRef BinaryContents, StringRef FileName) {
4030 if (BinaryContents.empty()) {
4031 // If we have no binary contents, the only thing we need to emit are the
4032 // if_empty tokens, if any.
4033 // FIXME: this loses AST fidelity; nothing in the compiler will see that
4034 // these tokens came from #embed. We have to hack around this when printing
4035 // preprocessed output. The same is true for prefix and suffix tokens.
4036 if (Params.MaybeIfEmptyParam) {
4037 ArrayRef<Token> Toks = Params.MaybeIfEmptyParam->Tokens;
4038 size_t TokCount = Toks.size();
4039 auto NewToks = std::make_unique<Token[]>(num: TokCount);
4040 llvm::copy(Range&: Toks, Out: NewToks.get());
4041 EnterTokenStream(Toks: std::move(NewToks), NumToks: TokCount, DisableMacroExpansion: true, IsReinject: true);
4042 }
4043 return;
4044 }
4045
4046 size_t NumPrefixToks = Params.PrefixTokenCount(),
4047 NumSuffixToks = Params.SuffixTokenCount();
4048 size_t TotalNumToks = 1 + NumPrefixToks + NumSuffixToks;
4049 size_t CurIdx = 0;
4050 auto Toks = std::make_unique<Token[]>(num: TotalNumToks);
4051
4052 // Add the prefix tokens, if any.
4053 if (Params.MaybePrefixParam) {
4054 llvm::copy(Range: Params.MaybePrefixParam->Tokens, Out: &Toks[CurIdx]);
4055 CurIdx += NumPrefixToks;
4056 }
4057
4058 EmbedAnnotationData *Data = new (BP) EmbedAnnotationData;
4059 Data->BinaryData = BinaryContents;
4060 Data->FileName = FileName;
4061
4062 Toks[CurIdx].startToken();
4063 Toks[CurIdx].setKind(tok::annot_embed);
4064 Toks[CurIdx].setAnnotationRange(HashLoc);
4065 Toks[CurIdx++].setAnnotationValue(Data);
4066
4067 // Now add the suffix tokens, if any.
4068 if (Params.MaybeSuffixParam) {
4069 llvm::copy(Range: Params.MaybeSuffixParam->Tokens, Out: &Toks[CurIdx]);
4070 CurIdx += NumSuffixToks;
4071 }
4072
4073 assert(CurIdx == TotalNumToks && "Calculated the incorrect number of tokens");
4074 EnterTokenStream(Toks: std::move(Toks), NumToks: TotalNumToks, DisableMacroExpansion: true, IsReinject: true);
4075}
4076
4077void Preprocessor::HandleEmbedDirective(SourceLocation HashLoc,
4078 Token &EmbedTok) {
4079 // Give the usual extension/compatibility warnings.
4080 if (LangOpts.C23)
4081 Diag(Tok: EmbedTok, DiagID: diag::warn_compat_pp_embed_directive);
4082 else
4083 Diag(Tok: EmbedTok, DiagID: diag::ext_pp_embed_directive)
4084 << (LangOpts.CPlusPlus ? /*Clang*/ 1 : /*C23*/ 0);
4085
4086 // Parse the filename header
4087 Token FilenameTok;
4088 if (LexHeaderName(Result&: FilenameTok))
4089 return;
4090
4091 if (FilenameTok.isNot(K: tok::header_name)) {
4092 Diag(Loc: FilenameTok.getLocation(), DiagID: diag::err_pp_expects_filename);
4093 if (FilenameTok.isNot(K: tok::eod))
4094 DiscardUntilEndOfDirective();
4095 return;
4096 }
4097
4098 // Parse the optional sequence of
4099 // directive-parameters:
4100 // identifier parameter-name-list[opt] directive-argument-list[opt]
4101 // directive-argument-list:
4102 // '(' balanced-token-sequence ')'
4103 // parameter-name-list:
4104 // '::' identifier parameter-name-list[opt]
4105 Token CurTok;
4106 std::optional<LexEmbedParametersResult> Params =
4107 LexEmbedParameters(CurTok, /*ForHasEmbed=*/false);
4108
4109 assert((Params || CurTok.is(tok::eod)) &&
4110 "expected success or to be at the end of the directive");
4111 if (!Params)
4112 return;
4113
4114 // Now, splat the data out!
4115 SmallString<128> FilenameBuffer;
4116 StringRef Filename = getSpelling(Tok: FilenameTok, Buffer&: FilenameBuffer);
4117 StringRef OriginalFilename = Filename;
4118 bool isAngled =
4119 GetIncludeFilenameSpelling(Loc: FilenameTok.getLocation(), Buffer&: Filename);
4120
4121 // If GetIncludeFilenameSpelling set the start ptr to null, there was an
4122 // error.
4123 if (Filename.empty())
4124 return;
4125
4126 OptionalFileEntryRef MaybeFileRef =
4127 this->LookupEmbedFile(Filename, isAngled, /*OpenFile=*/true);
4128 if (!MaybeFileRef) {
4129 // could not find file
4130 if (Callbacks && Callbacks->EmbedFileNotFound(FileName: Filename)) {
4131 return;
4132 }
4133 Diag(Tok: FilenameTok, DiagID: diag::err_pp_file_not_found) << Filename;
4134 return;
4135 }
4136
4137 if (MaybeFileRef->isDeviceFile()) {
4138 Diag(Tok: FilenameTok, DiagID: diag::err_pp_embed_device_file) << Filename;
4139 return;
4140 }
4141
4142 std::optional<llvm::MemoryBufferRef> MaybeFile =
4143 getSourceManager().getMemoryBufferForFileOrNone(File: *MaybeFileRef);
4144 if (!MaybeFile) {
4145 // could not find file
4146 Diag(Tok: FilenameTok, DiagID: diag::err_cannot_open_file)
4147 << Filename << "a buffer to the contents could not be created";
4148 return;
4149 }
4150 StringRef BinaryContents = MaybeFile->getBuffer();
4151
4152 // The order is important between 'offset' and 'limit'; we want to offset
4153 // first and then limit second; otherwise we may reduce the notional resource
4154 // size to something too small to offset into.
4155 if (Params->MaybeOffsetParam) {
4156 // FIXME: just like with the limit() and if_empty() parameters, this loses
4157 // source fidelity in the AST; it has no idea that there was an offset
4158 // involved.
4159 // offsets all the way to the end of the file make for an empty file.
4160 BinaryContents = BinaryContents.substr(Start: Params->MaybeOffsetParam->Offset);
4161 }
4162
4163 if (Params->MaybeLimitParam) {
4164 // FIXME: just like with the clang::offset() and if_empty() parameters,
4165 // this loses source fidelity in the AST; it has no idea there was a limit
4166 // involved.
4167 BinaryContents = BinaryContents.substr(Start: 0, N: Params->MaybeLimitParam->Limit);
4168 }
4169
4170 if (Callbacks)
4171 Callbacks->EmbedDirective(HashLoc, FileName: Filename, IsAngled: isAngled, File: MaybeFileRef,
4172 Params: *Params);
4173 // getSpelling() may return a buffer from the token itself or it may use the
4174 // SmallString buffer we provided. getSpelling() may also return a string that
4175 // is actually longer than FilenameTok.getLength(), so we first pass a
4176 // locally created buffer to getSpelling() to get the string of real length
4177 // and then we allocate a long living buffer because the buffer we used
4178 // previously will only live till the end of this function and we need
4179 // filename info to live longer.
4180 void *Mem = BP.Allocate(Size: OriginalFilename.size(), Alignment: alignof(char *));
4181 memcpy(dest: Mem, src: OriginalFilename.data(), n: OriginalFilename.size());
4182 StringRef FilenameToGo =
4183 StringRef(static_cast<char *>(Mem), OriginalFilename.size());
4184 HandleEmbedDirectiveImpl(HashLoc, Params: *Params, BinaryContents, FileName: FilenameToGo);
4185}
4186
4187/// HandleCXXImportDirective - Handle the C++ modules import directives
4188///
4189/// pp-import:
4190/// export[opt] import header-name pp-tokens[opt] ; new-line
4191/// export[opt] import header-name-tokens pp-tokens[opt] ; new-line
4192/// export[opt] import pp-tokens ; new-line
4193///
4194/// The header importing are replaced by annot_header_unit token, and the
4195/// lexed module name are replaced by annot_module_name token.
4196void Preprocessor::HandleCXXImportDirective(Token ImportTok) {
4197 assert(getLangOpts().CPlusPlusModules && ImportTok.is(tok::kw_import));
4198 llvm::SaveAndRestore<bool> SaveImportingCXXModules(
4199 this->ImportingCXXNamedModules, true);
4200
4201 if (LastTokenWasExportKeyword.isValid())
4202 LastTokenWasExportKeyword.reset();
4203
4204 Token Tok;
4205 if (LexHeaderName(Result&: Tok)) {
4206 if (Tok.isNot(K: tok::eod))
4207 CheckEndOfDirective(DirType: ImportTok.getIdentifierInfo()->getName());
4208 return;
4209 }
4210
4211 SourceLocation UseLoc = ImportTok.getLocation();
4212 SmallVector<Token, 4> DirToks{ImportTok};
4213 SmallVector<IdentifierLoc, 2> Path;
4214 bool ImportingHeader = false;
4215 bool IsPartition = false;
4216 std::string FlatName;
4217 switch (Tok.getKind()) {
4218 case tok::header_name:
4219 ImportingHeader = true;
4220 DirToks.push_back(Elt: Tok);
4221 Lex(Result&: DirToks.emplace_back());
4222 break;
4223 case tok::colon:
4224 IsPartition = true;
4225 DirToks.push_back(Elt: Tok);
4226 UseLoc = Tok.getLocation();
4227 Lex(Result&: Tok);
4228 [[fallthrough]];
4229 case tok::identifier: {
4230 bool LeadingSpace = Tok.hasLeadingSpace();
4231 unsigned NumToksInDirective = DirToks.size();
4232 if (LexModuleNameContinue(Tok, UseLoc, Suffix&: DirToks, Path)) {
4233 if (Tok.isNot(K: tok::eod))
4234 CheckEndOfDirective(DirType: ImportTok.getIdentifierInfo()->getName(),
4235 /*EnableMacros=*/false, ExtraToks: &DirToks);
4236 EnterModuleSuffixTokenStream(Toks: DirToks);
4237 return;
4238 }
4239
4240 // Clean the module-name tokens and replace these tokens with
4241 // annot_module_name.
4242 DirToks.resize(N: NumToksInDirective);
4243 ModuleNameLoc *NameLoc = ModuleNameLoc::Create(PP&: *this, Path);
4244 DirToks.emplace_back();
4245 DirToks.back().setKind(tok::annot_module_name);
4246 DirToks.back().setAnnotationRange(NameLoc->getRange());
4247 DirToks.back().setAnnotationValue(static_cast<void *>(NameLoc));
4248 DirToks.back().setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace);
4249 DirToks.push_back(Elt: Tok);
4250
4251 bool IsValid =
4252 (IsPartition && ModuleDeclState.isNamedModule()) || !IsPartition;
4253 if (Callbacks && IsValid) {
4254 if (IsPartition && ModuleDeclState.isNamedModule()) {
4255 FlatName += ModuleDeclState.getPrimaryName();
4256 FlatName += ":";
4257 }
4258
4259 FlatName += ModuleLoader::getFlatNameFromPath(Path);
4260 SourceLocation StartLoc = IsPartition ? UseLoc : Path[0].getLoc();
4261 IdentifierLoc FlatNameLoc(StartLoc, getIdentifierInfo(Name: FlatName));
4262
4263 // We don't/shouldn't load the standard c++20 modules when preprocessing.
4264 // so the imported module is nullptr.
4265 Callbacks->moduleImport(ImportLoc: ImportTok.getLocation(),
4266 Path: ModuleIdPath(FlatNameLoc),
4267 /*Imported=*/nullptr);
4268 }
4269 break;
4270 }
4271 default:
4272 DirToks.push_back(Elt: Tok);
4273 break;
4274 }
4275
4276 // Consume the pp-import-suffix and expand any macros in it now, if we're not
4277 // at the semicolon already.
4278 if (!DirToks.back().isOneOf(Ks: tok::semi, Ks: tok::eod))
4279 CollectPPImportSuffix(Toks&: DirToks);
4280
4281 if (DirToks.back().isNot(K: tok::eod))
4282 CheckEndOfDirective(DirType: ImportTok.getIdentifierInfo()->getName());
4283 else
4284 DirToks.pop_back();
4285
4286 // This is not a pp-import after all.
4287 if (DirToks.back().isNot(K: tok::semi)) {
4288 EnterModuleSuffixTokenStream(Toks: DirToks);
4289 return;
4290 }
4291
4292 if (ImportingHeader) {
4293 // C++2a [cpp.module]p1:
4294 // The ';' preprocessing-token terminating a pp-import shall not have
4295 // been produced by macro replacement.
4296 SourceLocation SemiLoc = DirToks.back().getLocation();
4297 if (SemiLoc.isMacroID())
4298 Diag(Loc: SemiLoc, DiagID: diag::err_header_import_semi_in_macro);
4299
4300 auto Action = HandleHeaderIncludeOrImport(
4301 /*HashLoc*/ SourceLocation(), IncludeTok&: ImportTok, FilenameTok&: Tok, EndLoc: SemiLoc);
4302 switch (Action.Kind) {
4303 case ImportAction::None:
4304 break;
4305
4306 case ImportAction::ModuleBegin:
4307 // Let the parser know we're textually entering the module.
4308 DirToks.emplace_back();
4309 DirToks.back().startToken();
4310 DirToks.back().setKind(tok::annot_module_begin);
4311 DirToks.back().setLocation(SemiLoc);
4312 DirToks.back().setAnnotationEndLoc(SemiLoc);
4313 DirToks.back().setAnnotationValue(Action.ModuleForHeader);
4314 [[fallthrough]];
4315
4316 case ImportAction::ModuleImport:
4317 case ImportAction::HeaderUnitImport:
4318 case ImportAction::SkippedModuleImport:
4319 // We chose to import (or textually enter) the file. Convert the
4320 // header-name token into a header unit annotation token.
4321 DirToks[1].setKind(tok::annot_header_unit);
4322 DirToks[1].setAnnotationEndLoc(DirToks[0].getLocation());
4323 DirToks[1].setAnnotationValue(Action.ModuleForHeader);
4324 // FIXME: Call the moduleImport callback?
4325 break;
4326 case ImportAction::Failure:
4327 assert(TheModuleLoader.HadFatalFailure &&
4328 "This should be an early exit only to a fatal error");
4329 CurLexer->cutOffLexing();
4330 return;
4331 }
4332 }
4333
4334 EnterModuleSuffixTokenStream(Toks: DirToks);
4335}
4336
4337/// HandleCXXModuleDirective - Handle C++ module declaration directives.
4338///
4339/// pp-module:
4340/// export[opt] module pp-tokens[opt] ; new-line
4341///
4342/// pp-module-name:
4343/// pp-module-name-qualifier[opt] identifier
4344/// pp-module-partition:
4345/// : pp-module-name-qualifier[opt] identifier
4346/// pp-module-name-qualifier:
4347/// identifier .
4348/// pp-module-name-qualifier identifier .
4349///
4350/// global-module-fragment:
4351/// module-keyword ; declaration-seq[opt]
4352///
4353/// private-module-fragment:
4354/// module-keyword : private ; declaration-seq[opt]
4355///
4356/// The lexed module name are replaced by annot_module_name token.
4357void Preprocessor::HandleCXXModuleDirective(Token ModuleTok) {
4358 assert(getLangOpts().CPlusPlusModules && ModuleTok.is(tok::kw_module));
4359 Token Introducer = ModuleTok;
4360 if (LastTokenWasExportKeyword.isValid()) {
4361 Introducer = LastTokenWasExportKeyword.getExportTok();
4362 LastTokenWasExportKeyword.reset();
4363 }
4364
4365 SourceLocation StartLoc = Introducer.getLocation();
4366
4367 Token Tok;
4368 SourceLocation UseLoc = ModuleTok.getLocation();
4369 SmallVector<Token, 4> DirToks{ModuleTok};
4370 SmallVector<IdentifierLoc, 2> Path, Partition;
4371 LexUnexpandedToken(Result&: Tok);
4372
4373 switch (Tok.getKind()) {
4374 // Global Module Fragment.
4375 case tok::semi:
4376 DirToks.push_back(Elt: Tok);
4377 break;
4378 case tok::colon:
4379 DirToks.push_back(Elt: Tok);
4380 LexUnexpandedToken(Result&: Tok);
4381 if (Tok.isNot(K: tok::kw_private)) {
4382 if (Tok.isNot(K: tok::eod))
4383 CheckEndOfDirective(DirType: ModuleTok.getIdentifierInfo()->getName(),
4384 /*EnableMacros=*/false, ExtraToks: &DirToks);
4385 EnterModuleSuffixTokenStream(Toks: DirToks);
4386 return;
4387 }
4388 DirToks.push_back(Elt: Tok);
4389 break;
4390 case tok::identifier: {
4391 bool LeadingSpace = Tok.hasLeadingSpace();
4392 unsigned NumToksInDirective = DirToks.size();
4393
4394 // C++ [cpp.module]p3: Any preprocessing tokens after the module
4395 // preprocessing token in the module directive are processed just as in
4396 // normal text.
4397 //
4398 // P3034R1 Module Declarations Shouldn’t be Macros.
4399 if (LexModuleNameContinue(Tok, UseLoc, Suffix&: DirToks, Path,
4400 /*AllowMacroExpansion=*/false)) {
4401 if (Tok.isNot(K: tok::eod))
4402 CheckEndOfDirective(DirType: ModuleTok.getIdentifierInfo()->getName(),
4403 /*EnableMacros=*/false, ExtraToks: &DirToks);
4404 EnterModuleSuffixTokenStream(Toks: DirToks);
4405 return;
4406 }
4407
4408 ModuleNameLoc *NameLoc = ModuleNameLoc::Create(PP&: *this, Path);
4409 DirToks.resize(N: NumToksInDirective);
4410 DirToks.emplace_back();
4411 DirToks.back().setKind(tok::annot_module_name);
4412 DirToks.back().setAnnotationRange(NameLoc->getRange());
4413 DirToks.back().setAnnotationValue(static_cast<void *>(NameLoc));
4414 DirToks.back().setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace);
4415 DirToks.push_back(Elt: Tok);
4416
4417 // C++20 [cpp.module]p
4418 // The pp-tokens, if any, of a pp-module shall be of the form:
4419 // pp-module-name pp-module-partition[opt] pp-tokens[opt]
4420 if (Tok.is(K: tok::colon)) {
4421 NumToksInDirective = DirToks.size();
4422 LexUnexpandedToken(Result&: Tok);
4423 LeadingSpace = Tok.hasLeadingSpace();
4424 if (LexModuleNameContinue(Tok, UseLoc, Suffix&: DirToks, Path&: Partition,
4425 /*AllowMacroExpansion=*/false,
4426 /*IsPartition=*/true)) {
4427 if (Tok.isNot(K: tok::eod))
4428 CheckEndOfDirective(DirType: ModuleTok.getIdentifierInfo()->getName(),
4429 /*EnableMacros=*/false, ExtraToks: &DirToks);
4430 EnterModuleSuffixTokenStream(Toks: DirToks);
4431 return;
4432 }
4433
4434 ModuleNameLoc *PartitionLoc = ModuleNameLoc::Create(PP&: *this, Path: Partition);
4435 DirToks.resize(N: NumToksInDirective);
4436 DirToks.emplace_back();
4437 DirToks.back().setKind(tok::annot_module_name);
4438 DirToks.back().setAnnotationRange(NameLoc->getRange());
4439 DirToks.back().setAnnotationValue(static_cast<void *>(PartitionLoc));
4440 DirToks.back().setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace);
4441 DirToks.push_back(Elt: Tok);
4442 }
4443
4444 // If the current token is a macro definition, put it back to token stream
4445 // and expand any macros in it later.
4446 //
4447 // export module M ATTR(some_attr); // -D'ATTR(x)=[[x]]'
4448 //
4449 // Current token is `ATTR`.
4450 if (Tok.is(K: tok::identifier) &&
4451 getMacroDefinition(II: Tok.getIdentifierInfo())) {
4452 std::unique_ptr<Token[]> TokCopy = std::make_unique<Token[]>(num: 1);
4453 TokCopy[0] = Tok;
4454 EnterTokenStream(Toks: std::move(TokCopy), /*NumToks=*/1,
4455 /*DisableMacroExpansion=*/false, /*IsReinject=*/false);
4456 Lex(Result&: Tok);
4457 DirToks.back() = Tok;
4458 }
4459 break;
4460 }
4461 default:
4462 DirToks.push_back(Elt: Tok);
4463 break;
4464 }
4465
4466 // Consume the pp-import-suffix and expand any macros in it now, if we're not
4467 // at the semicolon already.
4468 SourceLocation End = DirToks.back().getLocation();
4469 std::optional<Token> NextPPTok = DirToks.back();
4470 if (DirToks.back().is(K: tok::eod)) {
4471 NextPPTok = peekNextPPToken();
4472 if (NextPPTok && NextPPTok->is(K: tok::raw_identifier))
4473 LookUpIdentifierInfo(Identifier&: *NextPPTok);
4474 }
4475
4476 // Only ';' and '[' are allowed after module name.
4477 // We also check 'private' because the previous is not a module name.
4478 if (!NextPPTok->isOneOf(Ks: tok::semi, Ks: tok::eod, Ks: tok::l_square, Ks: tok::kw_private))
4479 Diag(Tok: *NextPPTok, DiagID: diag::err_pp_unexpected_tok_after_module_name)
4480 << getSpelling(Tok: *NextPPTok);
4481
4482 if (!DirToks.back().isOneOf(Ks: tok::semi, Ks: tok::eod)) {
4483 // Consume the pp-import-suffix and expand any macros in it now. We'll add
4484 // it back into the token stream later.
4485 CollectPPImportSuffix(Toks&: DirToks);
4486 End = DirToks.back().getLocation();
4487 }
4488
4489 if (DirToks.back().isNot(K: tok::eod))
4490 End = CheckEndOfDirective(DirType: ModuleTok.getIdentifierInfo()->getName(),
4491 /*EnableMacros=*/false, ExtraToks: &DirToks);
4492 else
4493 End = DirToks.pop_back_val().getLocation();
4494
4495 if (!IncludeMacroStack.empty()) {
4496 Diag(Loc: StartLoc, DiagID: diag::err_pp_module_decl_in_header)
4497 << SourceRange(StartLoc, End);
4498 }
4499
4500 if (CurPPLexer->getConditionalStackDepth() != 0) {
4501 Diag(Loc: StartLoc, DiagID: diag::err_pp_cond_span_module_decl)
4502 << SourceRange(StartLoc, End);
4503 }
4504 EnterModuleSuffixTokenStream(Toks: DirToks);
4505}
4506