1 //===- extra/modularize/Modularize.cpp - Check modularized headers --------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements a tool that checks whether a set of headers provides
11 // the consistent definitions required to use modules. For example, it detects
12 // whether the same entity (say, a NULL macro or size_t typedef) is defined in
13 // multiple headers or whether a header produces different definitions under
14 // different circumstances. These conditions cause modules built from the
15 // headers to behave poorly, and should be fixed before introducing a module
16 // map.
17 //
18 // Modularize takes as argument a file name for a file containing the
19 // newline-separated list of headers to check with respect to each other.
20 // Lines beginning with '#' and empty lines are ignored.
21 // Header file names followed by a colon and other space-separated
22 // file names will include those extra files as dependencies.
23 // The file names can be relative or full paths, but must be on the
24 // same line.
25 //
26 // Modularize also accepts regular front-end arguments.
27 //
28 // Usage:   modularize [-prefix (optional header path prefix)]
29 //   (include-files_list) [(front-end-options) ...]
30 //
31 // Note that unless a "-prefix (header path)" option is specified,
32 // non-absolute file paths in the header list file will be relative
33 // to the header list file directory.  Use -prefix to specify a different
34 // directory.
35 //
36 // Note that by default, the underlying Clang front end assumes .h files
37 // contain C source.  If your .h files in the file list contain C++ source,
38 // you should append the following to your command lines: -x c++
39 //
40 // Modularize will do normal parsing, reporting normal errors and warnings,
41 // but will also report special error messages like the following:
42 //
43 //   error: '(symbol)' defined at multiple locations:
44 //       (file):(row):(column)
45 //       (file):(row):(column)
46 //
47 //   error: header '(file)' has different contents depending on how it was
48 //     included
49 //
50 // The latter might be followed by messages like the following:
51 //
52 //   note: '(symbol)' in (file) at (row):(column) not always provided
53 //
54 // Checks will also be performed for macro expansions, defined(macro)
55 // expressions, and preprocessor conditional directives that evaluate
56 // inconsistently, and can produce error messages like the following:
57 //
58 //   (...)/SubHeader.h:11:5:
59 //   #if SYMBOL == 1
60 //       ^
61 //   error: Macro instance 'SYMBOL' has different values in this header,
62 //          depending on how it was included.
63 //     'SYMBOL' expanded to: '1' with respect to these inclusion paths:
64 //       (...)/Header1.h
65 //         (...)/SubHeader.h
66 //   (...)/SubHeader.h:3:9:
67 //   #define SYMBOL 1
68 //             ^
69 //   Macro defined here.
70 //     'SYMBOL' expanded to: '2' with respect to these inclusion paths:
71 //       (...)/Header2.h
72 //           (...)/SubHeader.h
73 //   (...)/SubHeader.h:7:9:
74 //   #define SYMBOL 2
75 //             ^
76 //   Macro defined here.
77 //
78 // Checks will also be performed for '#include' directives that are
79 // nested inside 'extern "C/C++" {}' or 'namespace (name) {}' blocks,
80 // and can produce error message like the following:
81 //
82 // IncludeInExtern.h:2:3
83 //   #include "Empty.h"
84 //   ^
85 // error: Include directive within extern "C" {}.
86 // IncludeInExtern.h:1:1
87 // extern "C" {
88 // ^
89 // The "extern "C" {}" block is here.
90 //
91 // See PreprocessorTracker.cpp for additional details.
92 //
93 // Modularize also has an option ("-module-map-path=module.map") that will
94 // skip the checks, and instead act as a module.map generation assistant,
95 // generating a module map file based on the header list.  An optional
96 // "-root-module=(rootName)" argument can specify a root module to be
97 // created in the generated module.map file.  Note that you will likely
98 // need to edit this file to suit the needs of your headers.
99 //
100 // An example command line for generating a module.map file:
101 //
102 //   modularize -module-map-path=module.map -root-module=myroot headerlist.txt
103 //
104 // Note that if the headers in the header list have partial paths, sub-modules
105 // will be created for the subdirectires involved, assuming that the
106 // subdirectories contain headers to be grouped into a module, but still with
107 // individual modules for the headers in the subdirectory.
108 //
109 // See the ModuleAssistant.cpp file comments for additional details about the
110 // implementation of the assistant mode.
111 //
112 // Future directions:
113 //
114 // Basically, we want to add new checks for whatever we can check with respect
115 // to checking headers for module'ability.
116 //
117 // Some ideas:
118 //
119 // 1. Omit duplicate "not always provided" messages
120 //
121 // 2. Add options to disable any of the checks, in case
122 // there is some problem with them, or the messages get too verbose.
123 //
124 // 3. Try to figure out the preprocessor conditional directives that
125 // contribute to problems and tie them to the inconsistent definitions.
126 //
127 // 4. There are some legitimate uses of preprocessor macros that
128 // modularize will flag as errors, such as repeatedly #include'ing
129 // a file and using interleaving defined/undefined macros
130 // to change declarations in the included file.  Is there a way
131 // to address this?  Maybe have modularize accept a list of macros
132 // to ignore.  Otherwise you can just exclude the file, after checking
133 // for legitimate errors.
134 //
135 // 5. What else?
136 //
137 // General clean-up and refactoring:
138 //
139 // 1. The Location class seems to be something that we might
140 // want to design to be applicable to a wider range of tools, and stick it
141 // somewhere into Tooling/ in mainline
142 //
143 //===----------------------------------------------------------------------===//
144 
145 #include "Modularize.h"
146 #include "PreprocessorTracker.h"
147 #include "clang/AST/ASTConsumer.h"
148 #include "clang/AST/ASTContext.h"
149 #include "clang/AST/RecursiveASTVisitor.h"
150 #include "clang/Basic/SourceManager.h"
151 #include "clang/Driver/Options.h"
152 #include "clang/Frontend/CompilerInstance.h"
153 #include "clang/Frontend/FrontendActions.h"
154 #include "clang/Lex/Preprocessor.h"
155 #include "clang/Tooling/CompilationDatabase.h"
156 #include "clang/Tooling/Tooling.h"
157 #include "llvm/Config/config.h"
158 #include "llvm/Option/Arg.h"
159 #include "llvm/Option/ArgList.h"
160 #include "llvm/Option/OptTable.h"
161 #include "llvm/Option/Option.h"
162 #include "llvm/Support/CommandLine.h"
163 #include "llvm/Support/FileSystem.h"
164 #include "llvm/Support/MemoryBuffer.h"
165 #include "llvm/Support/Path.h"
166 #include <algorithm>
167 #include <fstream>
168 #include <iterator>
169 #include <string>
170 #include <vector>
171 
172 using namespace clang;
173 using namespace clang::driver;
174 using namespace clang::driver::options;
175 using namespace clang::tooling;
176 using namespace llvm;
177 using namespace llvm::opt;
178 using namespace Modularize;
179 
180 // Option to specify a file name for a list of header files to check.
181 cl::opt<std::string>
182 ListFileName(cl::Positional,
183              cl::desc("<name of file containing list of headers to check>"));
184 
185 // Collect all other arguments, which will be passed to the front end.
186 cl::list<std::string>
187 CC1Arguments(cl::ConsumeAfter,
188              cl::desc("<arguments to be passed to front end>..."));
189 
190 // Option to specify a prefix to be prepended to the header names.
191 cl::opt<std::string> HeaderPrefix(
192     "prefix", cl::init(""),
193     cl::desc(
194         "Prepend header file paths with this prefix."
195         " If not specified,"
196         " the files are considered to be relative to the header list file."));
197 
198 // Option for assistant mode, telling modularize to output a module map
199 // based on the headers list, and where to put it.
200 cl::opt<std::string> ModuleMapPath(
201     "module-map-path", cl::init(""),
202     cl::desc("Turn on module map output and specify output path or file name."
203              " If no path is specified and if prefix option is specified,"
204              " use prefix for file path."));
205 
206 // Option for assistant mode, telling modularize to output a module map
207 // based on the headers list, and where to put it.
208 cl::opt<std::string>
209 RootModule("root-module", cl::init(""),
210            cl::desc("Specify the name of the root module."));
211 
212 // Save the program name for error messages.
213 const char *Argv0;
214 // Save the command line for comments.
215 std::string CommandLine;
216 
217 // Read the header list file and collect the header file names and
218 // optional dependencies.
219 error_code getHeaderFileNames(SmallVectorImpl<std::string> &HeaderFileNames,
220                               DependencyMap &Dependencies,
221                               StringRef ListFileName, StringRef HeaderPrefix) {
222   // By default, use the path component of the list file name.
223   SmallString<256> HeaderDirectory(ListFileName);
224   sys::path::remove_filename(HeaderDirectory);
225   SmallString<256> CurrentDirectory;
226   sys::fs::current_path(CurrentDirectory);
227 
228   // Get the prefix if we have one.
229   if (HeaderPrefix.size() != 0)
230     HeaderDirectory = HeaderPrefix;
231 
232   // Read the header list file into a buffer.
233   std::unique_ptr<MemoryBuffer> listBuffer;
234   if (error_code ec = MemoryBuffer::getFile(ListFileName, listBuffer)) {
235     return ec;
236   }
237 
238   // Parse the header list into strings.
239   SmallVector<StringRef, 32> Strings;
240   listBuffer->getBuffer().split(Strings, "\n", -1, false);
241 
242   // Collect the header file names from the string list.
243   for (SmallVectorImpl<StringRef>::iterator I = Strings.begin(),
244                                             E = Strings.end();
245        I != E; ++I) {
246     StringRef Line = I->trim();
247     // Ignore comments and empty lines.
248     if (Line.empty() || (Line[0] == '#'))
249       continue;
250     std::pair<StringRef, StringRef> TargetAndDependents = Line.split(':');
251     SmallString<256> HeaderFileName;
252     // Prepend header file name prefix if it's not absolute.
253     if (sys::path::is_absolute(TargetAndDependents.first))
254       llvm::sys::path::native(TargetAndDependents.first, HeaderFileName);
255     else {
256       if (HeaderDirectory.size() != 0)
257         HeaderFileName = HeaderDirectory;
258       else
259         HeaderFileName = CurrentDirectory;
260       sys::path::append(HeaderFileName, TargetAndDependents.first);
261       sys::path::native(HeaderFileName);
262     }
263     // Handle optional dependencies.
264     DependentsVector Dependents;
265     SmallVector<StringRef, 4> DependentsList;
266     TargetAndDependents.second.split(DependentsList, " ", -1, false);
267     int Count = DependentsList.size();
268     for (int Index = 0; Index < Count; ++Index) {
269       SmallString<256> Dependent;
270       if (sys::path::is_absolute(DependentsList[Index]))
271         Dependent = DependentsList[Index];
272       else {
273         if (HeaderDirectory.size() != 0)
274           Dependent = HeaderDirectory;
275         else
276           Dependent = CurrentDirectory;
277         sys::path::append(Dependent, DependentsList[Index]);
278       }
279       sys::path::native(Dependent);
280       Dependents.push_back(Dependent.str());
281     }
282     // Save the resulting header file path and dependencies.
283     HeaderFileNames.push_back(HeaderFileName.str());
284     Dependencies[HeaderFileName.str()] = Dependents;
285   }
286 
287   return error_code::success();
288 }
289 
290 // Helper function for finding the input file in an arguments list.
291 std::string findInputFile(const CommandLineArguments &CLArgs) {
292   std::unique_ptr<OptTable> Opts(createDriverOptTable());
293   const unsigned IncludedFlagsBitmask = options::CC1Option;
294   unsigned MissingArgIndex, MissingArgCount;
295   SmallVector<const char *, 256> Argv;
296   for (CommandLineArguments::const_iterator I = CLArgs.begin(),
297                                             E = CLArgs.end();
298        I != E; ++I)
299     Argv.push_back(I->c_str());
300   std::unique_ptr<InputArgList> Args(
301       Opts->ParseArgs(Argv.data(), Argv.data() + Argv.size(), MissingArgIndex,
302                       MissingArgCount, IncludedFlagsBitmask));
303   std::vector<std::string> Inputs = Args->getAllArgValues(OPT_INPUT);
304   return Inputs.back();
305 }
306 
307 // We provide this derivation to add in "-include (file)" arguments for header
308 // dependencies.
309 class AddDependenciesAdjuster : public ArgumentsAdjuster {
310 public:
311   AddDependenciesAdjuster(DependencyMap &Dependencies)
312       : Dependencies(Dependencies) {}
313 
314 private:
315   // Callback for adjusting commandline arguments.
316   CommandLineArguments Adjust(const CommandLineArguments &Args) {
317     std::string InputFile = findInputFile(Args);
318     DependentsVector &FileDependents = Dependencies[InputFile];
319     int Count = FileDependents.size();
320     if (Count == 0)
321       return Args;
322     CommandLineArguments NewArgs(Args);
323     for (int Index = 0; Index < Count; ++Index) {
324       NewArgs.push_back("-include");
325       std::string File(std::string("\"") + FileDependents[Index] +
326                        std::string("\""));
327       NewArgs.push_back(FileDependents[Index]);
328     }
329     return NewArgs;
330   }
331   DependencyMap &Dependencies;
332 };
333 
334 // FIXME: The Location class seems to be something that we might
335 // want to design to be applicable to a wider range of tools, and stick it
336 // somewhere into Tooling/ in mainline
337 struct Location {
338   const FileEntry *File;
339   unsigned Line, Column;
340 
341   Location() : File(), Line(), Column() {}
342 
343   Location(SourceManager &SM, SourceLocation Loc) : File(), Line(), Column() {
344     Loc = SM.getExpansionLoc(Loc);
345     if (Loc.isInvalid())
346       return;
347 
348     std::pair<FileID, unsigned> Decomposed = SM.getDecomposedLoc(Loc);
349     File = SM.getFileEntryForID(Decomposed.first);
350     if (!File)
351       return;
352 
353     Line = SM.getLineNumber(Decomposed.first, Decomposed.second);
354     Column = SM.getColumnNumber(Decomposed.first, Decomposed.second);
355   }
356 
357   operator bool() const { return File != 0; }
358 
359   friend bool operator==(const Location &X, const Location &Y) {
360     return X.File == Y.File && X.Line == Y.Line && X.Column == Y.Column;
361   }
362 
363   friend bool operator!=(const Location &X, const Location &Y) {
364     return !(X == Y);
365   }
366 
367   friend bool operator<(const Location &X, const Location &Y) {
368     if (X.File != Y.File)
369       return X.File < Y.File;
370     if (X.Line != Y.Line)
371       return X.Line < Y.Line;
372     return X.Column < Y.Column;
373   }
374   friend bool operator>(const Location &X, const Location &Y) { return Y < X; }
375   friend bool operator<=(const Location &X, const Location &Y) {
376     return !(Y < X);
377   }
378   friend bool operator>=(const Location &X, const Location &Y) {
379     return !(X < Y);
380   }
381 };
382 
383 struct Entry {
384   enum EntryKind {
385     EK_Tag,
386     EK_Value,
387     EK_Macro,
388 
389     EK_NumberOfKinds
390   } Kind;
391 
392   Location Loc;
393 
394   StringRef getKindName() { return getKindName(Kind); }
395   static StringRef getKindName(EntryKind kind);
396 };
397 
398 // Return a string representing the given kind.
399 StringRef Entry::getKindName(Entry::EntryKind kind) {
400   switch (kind) {
401   case EK_Tag:
402     return "tag";
403   case EK_Value:
404     return "value";
405   case EK_Macro:
406     return "macro";
407   case EK_NumberOfKinds:
408     break;
409   }
410   llvm_unreachable("invalid Entry kind");
411 }
412 
413 struct HeaderEntry {
414   std::string Name;
415   Location Loc;
416 
417   friend bool operator==(const HeaderEntry &X, const HeaderEntry &Y) {
418     return X.Loc == Y.Loc && X.Name == Y.Name;
419   }
420   friend bool operator!=(const HeaderEntry &X, const HeaderEntry &Y) {
421     return !(X == Y);
422   }
423   friend bool operator<(const HeaderEntry &X, const HeaderEntry &Y) {
424     return X.Loc < Y.Loc || (X.Loc == Y.Loc && X.Name < Y.Name);
425   }
426   friend bool operator>(const HeaderEntry &X, const HeaderEntry &Y) {
427     return Y < X;
428   }
429   friend bool operator<=(const HeaderEntry &X, const HeaderEntry &Y) {
430     return !(Y < X);
431   }
432   friend bool operator>=(const HeaderEntry &X, const HeaderEntry &Y) {
433     return !(X < Y);
434   }
435 };
436 
437 typedef std::vector<HeaderEntry> HeaderContents;
438 
439 class EntityMap : public StringMap<SmallVector<Entry, 2> > {
440 public:
441   DenseMap<const FileEntry *, HeaderContents> HeaderContentMismatches;
442 
443   void add(const std::string &Name, enum Entry::EntryKind Kind, Location Loc) {
444     // Record this entity in its header.
445     HeaderEntry HE = { Name, Loc };
446     CurHeaderContents[Loc.File].push_back(HE);
447 
448     // Check whether we've seen this entry before.
449     SmallVector<Entry, 2> &Entries = (*this)[Name];
450     for (unsigned I = 0, N = Entries.size(); I != N; ++I) {
451       if (Entries[I].Kind == Kind && Entries[I].Loc == Loc)
452         return;
453     }
454 
455     // We have not seen this entry before; record it.
456     Entry E = { Kind, Loc };
457     Entries.push_back(E);
458   }
459 
460   void mergeCurHeaderContents() {
461     for (DenseMap<const FileEntry *, HeaderContents>::iterator
462              H = CurHeaderContents.begin(),
463              HEnd = CurHeaderContents.end();
464          H != HEnd; ++H) {
465       // Sort contents.
466       std::sort(H->second.begin(), H->second.end());
467 
468       // Check whether we've seen this header before.
469       DenseMap<const FileEntry *, HeaderContents>::iterator KnownH =
470           AllHeaderContents.find(H->first);
471       if (KnownH == AllHeaderContents.end()) {
472         // We haven't seen this header before; record its contents.
473         AllHeaderContents.insert(*H);
474         continue;
475       }
476 
477       // If the header contents are the same, we're done.
478       if (H->second == KnownH->second)
479         continue;
480 
481       // Determine what changed.
482       std::set_symmetric_difference(
483           H->second.begin(), H->second.end(), KnownH->second.begin(),
484           KnownH->second.end(),
485           std::back_inserter(HeaderContentMismatches[H->first]));
486     }
487 
488     CurHeaderContents.clear();
489   }
490 
491 private:
492   DenseMap<const FileEntry *, HeaderContents> CurHeaderContents;
493   DenseMap<const FileEntry *, HeaderContents> AllHeaderContents;
494 };
495 
496 class CollectEntitiesVisitor
497     : public RecursiveASTVisitor<CollectEntitiesVisitor> {
498 public:
499   CollectEntitiesVisitor(SourceManager &SM, EntityMap &Entities,
500                          Preprocessor &PP, PreprocessorTracker &PPTracker,
501                          int &HadErrors)
502       : SM(SM), Entities(Entities), PP(PP), PPTracker(PPTracker),
503         HadErrors(HadErrors) {}
504 
505   bool TraverseStmt(Stmt *S) { return true; }
506   bool TraverseType(QualType T) { return true; }
507   bool TraverseTypeLoc(TypeLoc TL) { return true; }
508   bool TraverseNestedNameSpecifier(NestedNameSpecifier *NNS) { return true; }
509   bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS) {
510     return true;
511   }
512   bool TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo) {
513     return true;
514   }
515   bool TraverseTemplateName(TemplateName Template) { return true; }
516   bool TraverseTemplateArgument(const TemplateArgument &Arg) { return true; }
517   bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &ArgLoc) {
518     return true;
519   }
520   bool TraverseTemplateArguments(const TemplateArgument *Args,
521                                  unsigned NumArgs) {
522     return true;
523   }
524   bool TraverseConstructorInitializer(CXXCtorInitializer *Init) { return true; }
525   bool TraverseLambdaCapture(LambdaCapture C) { return true; }
526 
527   // Check 'extern "*" {}' block for #include directives.
528   bool VisitLinkageSpecDecl(LinkageSpecDecl *D) {
529     // Bail if not a block.
530     if (!D->hasBraces())
531       return true;
532     SourceRange BlockRange = D->getSourceRange();
533     const char *LinkageLabel;
534     switch (D->getLanguage()) {
535     case LinkageSpecDecl::lang_c:
536       LinkageLabel = "extern \"C\" {}";
537       break;
538     case LinkageSpecDecl::lang_cxx:
539       LinkageLabel = "extern \"C++\" {}";
540       break;
541     }
542     if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, LinkageLabel,
543                                            errs()))
544       HadErrors = 1;
545     return true;
546   }
547 
548   // Check 'namespace (name) {}' block for #include directives.
549   bool VisitNamespaceDecl(const NamespaceDecl *D) {
550     SourceRange BlockRange = D->getSourceRange();
551     std::string Label("namespace ");
552     Label += D->getName();
553     Label += " {}";
554     if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, Label.c_str(),
555                                            errs()))
556       HadErrors = 1;
557     return true;
558   }
559 
560   // Collect definition entities.
561   bool VisitNamedDecl(NamedDecl *ND) {
562     // We only care about file-context variables.
563     if (!ND->getDeclContext()->isFileContext())
564       return true;
565 
566     // Skip declarations that tend to be properly multiply-declared.
567     if (isa<NamespaceDecl>(ND) || isa<UsingDirectiveDecl>(ND) ||
568         isa<NamespaceAliasDecl>(ND) ||
569         isa<ClassTemplateSpecializationDecl>(ND) || isa<UsingDecl>(ND) ||
570         isa<ClassTemplateDecl>(ND) || isa<TemplateTypeParmDecl>(ND) ||
571         isa<TypeAliasTemplateDecl>(ND) || isa<UsingShadowDecl>(ND) ||
572         isa<FunctionDecl>(ND) || isa<FunctionTemplateDecl>(ND) ||
573         (isa<TagDecl>(ND) &&
574          !cast<TagDecl>(ND)->isThisDeclarationADefinition()))
575       return true;
576 
577     // Skip anonymous declarations.
578     if (!ND->getDeclName())
579       return true;
580 
581     // Get the qualified name.
582     std::string Name;
583     llvm::raw_string_ostream OS(Name);
584     ND->printQualifiedName(OS);
585     OS.flush();
586     if (Name.empty())
587       return true;
588 
589     Location Loc(SM, ND->getLocation());
590     if (!Loc)
591       return true;
592 
593     Entities.add(Name, isa<TagDecl>(ND) ? Entry::EK_Tag : Entry::EK_Value, Loc);
594     return true;
595   }
596 
597 private:
598   SourceManager &SM;
599   EntityMap &Entities;
600   Preprocessor &PP;
601   PreprocessorTracker &PPTracker;
602   int &HadErrors;
603 };
604 
605 class CollectEntitiesConsumer : public ASTConsumer {
606 public:
607   CollectEntitiesConsumer(EntityMap &Entities,
608                           PreprocessorTracker &preprocessorTracker,
609                           Preprocessor &PP, StringRef InFile, int &HadErrors)
610       : Entities(Entities), PPTracker(preprocessorTracker), PP(PP),
611         HadErrors(HadErrors) {
612     PPTracker.handlePreprocessorEntry(PP, InFile);
613   }
614 
615   ~CollectEntitiesConsumer() { PPTracker.handlePreprocessorExit(); }
616 
617   virtual void HandleTranslationUnit(ASTContext &Ctx) {
618     SourceManager &SM = Ctx.getSourceManager();
619 
620     // Collect declared entities.
621     CollectEntitiesVisitor(SM, Entities, PP, PPTracker, HadErrors)
622         .TraverseDecl(Ctx.getTranslationUnitDecl());
623 
624     // Collect macro definitions.
625     for (Preprocessor::macro_iterator M = PP.macro_begin(),
626                                       MEnd = PP.macro_end();
627          M != MEnd; ++M) {
628       Location Loc(SM, M->second->getLocation());
629       if (!Loc)
630         continue;
631 
632       Entities.add(M->first->getName().str(), Entry::EK_Macro, Loc);
633     }
634 
635     // Merge header contents.
636     Entities.mergeCurHeaderContents();
637   }
638 
639 private:
640   EntityMap &Entities;
641   PreprocessorTracker &PPTracker;
642   Preprocessor &PP;
643   int &HadErrors;
644 };
645 
646 class CollectEntitiesAction : public SyntaxOnlyAction {
647 public:
648   CollectEntitiesAction(EntityMap &Entities,
649                         PreprocessorTracker &preprocessorTracker,
650                         int &HadErrors)
651       : Entities(Entities), PPTracker(preprocessorTracker),
652         HadErrors(HadErrors) {}
653 
654 protected:
655   virtual clang::ASTConsumer *CreateASTConsumer(CompilerInstance &CI,
656                                                 StringRef InFile) {
657     return new CollectEntitiesConsumer(Entities, PPTracker,
658                                        CI.getPreprocessor(), InFile, HadErrors);
659   }
660 
661 private:
662   EntityMap &Entities;
663   PreprocessorTracker &PPTracker;
664   int &HadErrors;
665 };
666 
667 class ModularizeFrontendActionFactory : public FrontendActionFactory {
668 public:
669   ModularizeFrontendActionFactory(EntityMap &Entities,
670                                   PreprocessorTracker &preprocessorTracker,
671                                   int &HadErrors)
672       : Entities(Entities), PPTracker(preprocessorTracker),
673         HadErrors(HadErrors) {}
674 
675   virtual CollectEntitiesAction *create() {
676     return new CollectEntitiesAction(Entities, PPTracker, HadErrors);
677   }
678 
679 private:
680   EntityMap &Entities;
681   PreprocessorTracker &PPTracker;
682   int &HadErrors;
683 };
684 
685 int main(int Argc, const char **Argv) {
686 
687   // Save program name for error messages.
688   Argv0 = Argv[0];
689 
690   // Save program arguments for use in module.map comment.
691   CommandLine = sys::path::stem(sys::path::filename(Argv0));
692   for (int ArgIndex = 1; ArgIndex < Argc; ArgIndex++) {
693     CommandLine.append(" ");
694     CommandLine.append(Argv[ArgIndex]);
695   }
696 
697   // This causes options to be parsed.
698   cl::ParseCommandLineOptions(Argc, Argv, "modularize.\n");
699 
700   // No go if we have no header list file.
701   if (ListFileName.size() == 0) {
702     cl::PrintHelpMessage();
703     return 1;
704   }
705 
706   // Get header file names and dependencies.
707   SmallVector<std::string, 32> Headers;
708   DependencyMap Dependencies;
709   if (error_code EC = getHeaderFileNames(Headers, Dependencies, ListFileName,
710                                          HeaderPrefix)) {
711     errs() << Argv[0] << ": error: Unable to get header list '" << ListFileName
712            << "': " << EC.message() << '\n';
713     return 1;
714   }
715 
716   // If we are in assistant mode, output the module map and quit.
717   if (ModuleMapPath.length() != 0) {
718     if (!createModuleMap(ModuleMapPath, Headers, Dependencies, HeaderPrefix,
719                          RootModule))
720       return 1; // Failed.
721     return 0;   // Success - Skip checks in assistant mode.
722   }
723 
724   // Create the compilation database.
725   SmallString<256> PathBuf;
726   sys::fs::current_path(PathBuf);
727   std::unique_ptr<CompilationDatabase> Compilations;
728   Compilations.reset(
729       new FixedCompilationDatabase(Twine(PathBuf), CC1Arguments));
730 
731   // Create preprocessor tracker, to watch for macro and conditional problems.
732   std::unique_ptr<PreprocessorTracker> PPTracker(PreprocessorTracker::create());
733 
734   // Parse all of the headers, detecting duplicates.
735   EntityMap Entities;
736   ClangTool Tool(*Compilations, Headers);
737   Tool.appendArgumentsAdjuster(new AddDependenciesAdjuster(Dependencies));
738   int HadErrors = 0;
739   HadErrors |= Tool.run(
740       new ModularizeFrontendActionFactory(Entities, *PPTracker, HadErrors));
741 
742   // Create a place to save duplicate entity locations, separate bins per kind.
743   typedef SmallVector<Location, 8> LocationArray;
744   typedef SmallVector<LocationArray, Entry::EK_NumberOfKinds> EntryBinArray;
745   EntryBinArray EntryBins;
746   int KindIndex;
747   for (KindIndex = 0; KindIndex < Entry::EK_NumberOfKinds; ++KindIndex) {
748     LocationArray Array;
749     EntryBins.push_back(Array);
750   }
751 
752   // Check for the same entity being defined in multiple places.
753   for (EntityMap::iterator E = Entities.begin(), EEnd = Entities.end();
754        E != EEnd; ++E) {
755     // If only one occurrence, exit early.
756     if (E->second.size() == 1)
757       continue;
758     // Clear entity locations.
759     for (EntryBinArray::iterator CI = EntryBins.begin(), CE = EntryBins.end();
760          CI != CE; ++CI) {
761       CI->clear();
762     }
763     // Walk the entities of a single name, collecting the locations,
764     // separated into separate bins.
765     for (unsigned I = 0, N = E->second.size(); I != N; ++I) {
766       EntryBins[E->second[I].Kind].push_back(E->second[I].Loc);
767     }
768     // Report any duplicate entity definition errors.
769     int KindIndex = 0;
770     for (EntryBinArray::iterator DI = EntryBins.begin(), DE = EntryBins.end();
771          DI != DE; ++DI, ++KindIndex) {
772       int ECount = DI->size();
773       // If only 1 occurrence of this entity, skip it, as we only report duplicates.
774       if (ECount <= 1)
775         continue;
776       LocationArray::iterator FI = DI->begin();
777       StringRef kindName = Entry::getKindName((Entry::EntryKind)KindIndex);
778       errs() << "error: " << kindName << " '" << E->first()
779              << "' defined at multiple locations:\n";
780       for (LocationArray::iterator FE = DI->end(); FI != FE; ++FI) {
781         errs() << "    " << FI->File->getName() << ":" << FI->Line << ":"
782                << FI->Column << "\n";
783       }
784       HadErrors = 1;
785     }
786   }
787 
788   // Complain about macro instance in header files that differ based on how
789   // they are included.
790   if (PPTracker->reportInconsistentMacros(errs()))
791     HadErrors = 1;
792 
793   // Complain about preprocessor conditional directives in header files that
794   // differ based on how they are included.
795   if (PPTracker->reportInconsistentConditionals(errs()))
796     HadErrors = 1;
797 
798   // Complain about any headers that have contents that differ based on how
799   // they are included.
800   // FIXME: Could we provide information about which preprocessor conditionals
801   // are involved?
802   for (DenseMap<const FileEntry *, HeaderContents>::iterator
803            H = Entities.HeaderContentMismatches.begin(),
804            HEnd = Entities.HeaderContentMismatches.end();
805        H != HEnd; ++H) {
806     if (H->second.empty()) {
807       errs() << "internal error: phantom header content mismatch\n";
808       continue;
809     }
810 
811     HadErrors = 1;
812     errs() << "error: header '" << H->first->getName()
813            << "' has different contents depending on how it was included.\n";
814     for (unsigned I = 0, N = H->second.size(); I != N; ++I) {
815       errs() << "note: '" << H->second[I].Name << "' in "
816              << H->second[I].Loc.File->getName() << " at "
817              << H->second[I].Loc.Line << ":" << H->second[I].Loc.Column
818              << " not always provided\n";
819     }
820   }
821 
822   return HadErrors;
823 }
824