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 "clang/AST/ASTConsumer.h"
146 #include "clang/AST/ASTContext.h"
147 #include "clang/AST/RecursiveASTVisitor.h"
148 #include "clang/Basic/SourceManager.h"
149 #include "clang/Driver/Options.h"
150 #include "clang/Frontend/CompilerInstance.h"
151 #include "clang/Frontend/FrontendActions.h"
152 #include "clang/Lex/Preprocessor.h"
153 #include "clang/Tooling/CompilationDatabase.h"
154 #include "clang/Tooling/Tooling.h"
155 #include "llvm/ADT/OwningPtr.h"
156 #include "llvm/Config/config.h"
157 #include "llvm/Option/Arg.h"
158 #include "llvm/Option/ArgList.h"
159 #include "llvm/Option/OptTable.h"
160 #include "llvm/Option/Option.h"
161 #include "llvm/Support/CommandLine.h"
162 #include "llvm/Support/FileSystem.h"
163 #include "llvm/Support/MemoryBuffer.h"
164 #include "llvm/Support/Path.h"
165 #include <algorithm>
166 #include <fstream>
167 #include <iterator>
168 #include <string>
169 #include <vector>
170 #include "Modularize.h"
171 #include "PreprocessorTracker.h"
172 
173 using namespace clang;
174 using namespace clang::driver;
175 using namespace clang::driver::options;
176 using namespace clang::tooling;
177 using namespace llvm;
178 using namespace llvm::opt;
179 using namespace Modularize;
180 
181 // Option to specify a file name for a list of header files to check.
182 cl::opt<std::string>
183 ListFileName(cl::Positional,
184              cl::desc("<name of file containing list of headers to check>"));
185 
186 // Collect all other arguments, which will be passed to the front end.
187 cl::list<std::string>
188 CC1Arguments(cl::ConsumeAfter,
189              cl::desc("<arguments to be passed to front end>..."));
190 
191 // Option to specify a prefix to be prepended to the header names.
192 cl::opt<std::string> HeaderPrefix(
193     "prefix", cl::init(""),
194     cl::desc(
195         "Prepend header file paths with this prefix."
196         " If not specified,"
197         " the files are considered to be relative to the header list file."));
198 
199 // Option for assistant mode, telling modularize to output a module map
200 // based on the headers list, and where to put it.
201 cl::opt<std::string> ModuleMapPath(
202     "module-map-path", cl::init(""),
203     cl::desc("Turn on module map output and specify output path or file name."
204              " If no path is specified and if prefix option is specified,"
205              " use prefix for file path."));
206 
207 // Option for assistant mode, telling modularize to output a module map
208 // based on the headers list, and where to put it.
209 cl::opt<std::string>
210 RootModule("root-module", cl::init(""),
211            cl::desc("Specify the name of the root module."));
212 
213 // Save the program name for error messages.
214 const char *Argv0;
215 // Save the command line for comments.
216 std::string CommandLine;
217 
218 // Read the header list file and collect the header file names and
219 // optional dependencies.
220 error_code getHeaderFileNames(SmallVectorImpl<std::string> &HeaderFileNames,
221                               DependencyMap &Dependencies,
222                               StringRef ListFileName, StringRef HeaderPrefix) {
223   // By default, use the path component of the list file name.
224   SmallString<256> HeaderDirectory(ListFileName);
225   sys::path::remove_filename(HeaderDirectory);
226   SmallString<256> CurrentDirectory;
227   sys::fs::current_path(CurrentDirectory);
228 
229   // Get the prefix if we have one.
230   if (HeaderPrefix.size() != 0)
231     HeaderDirectory = HeaderPrefix;
232 
233   // Read the header list file into a buffer.
234   OwningPtr<MemoryBuffer> listBuffer;
235   if (error_code ec = MemoryBuffer::getFile(ListFileName, listBuffer)) {
236     return ec;
237   }
238 
239   // Parse the header list into strings.
240   SmallVector<StringRef, 32> Strings;
241   listBuffer->getBuffer().split(Strings, "\n", -1, false);
242 
243   // Collect the header file names from the string list.
244   for (SmallVectorImpl<StringRef>::iterator I = Strings.begin(),
245                                             E = Strings.end();
246        I != E; ++I) {
247     StringRef Line = I->trim();
248     // Ignore comments and empty lines.
249     if (Line.empty() || (Line[0] == '#'))
250       continue;
251     std::pair<StringRef, StringRef> TargetAndDependents = Line.split(':');
252     SmallString<256> HeaderFileName;
253     // Prepend header file name prefix if it's not absolute.
254     if (sys::path::is_absolute(TargetAndDependents.first))
255       llvm::sys::path::native(TargetAndDependents.first, HeaderFileName);
256     else {
257       if (HeaderDirectory.size() != 0)
258         HeaderFileName = HeaderDirectory;
259       else
260         HeaderFileName = CurrentDirectory;
261       sys::path::append(HeaderFileName, TargetAndDependents.first);
262       sys::path::native(HeaderFileName);
263     }
264     // Handle optional dependencies.
265     DependentsVector Dependents;
266     SmallVector<StringRef, 4> DependentsList;
267     TargetAndDependents.second.split(DependentsList, " ", -1, false);
268     int Count = DependentsList.size();
269     for (int Index = 0; Index < Count; ++Index) {
270       SmallString<256> Dependent;
271       if (sys::path::is_absolute(DependentsList[Index]))
272         Dependent = DependentsList[Index];
273       else {
274         if (HeaderDirectory.size() != 0)
275           Dependent = HeaderDirectory;
276         else
277           Dependent = CurrentDirectory;
278         sys::path::append(Dependent, DependentsList[Index]);
279       }
280       sys::path::native(Dependent);
281       Dependents.push_back(Dependent.str());
282     }
283     // Save the resulting header file path and dependencies.
284     HeaderFileNames.push_back(HeaderFileName.str());
285     Dependencies[HeaderFileName.str()] = Dependents;
286   }
287 
288   return error_code::success();
289 }
290 
291 // Helper function for finding the input file in an arguments list.
292 std::string findInputFile(const CommandLineArguments &CLArgs) {
293   OwningPtr<OptTable> Opts(createDriverOptTable());
294   const unsigned IncludedFlagsBitmask = options::CC1Option;
295   unsigned MissingArgIndex, MissingArgCount;
296   SmallVector<const char *, 256> Argv;
297   for (CommandLineArguments::const_iterator I = CLArgs.begin(),
298                                             E = CLArgs.end();
299        I != E; ++I)
300     Argv.push_back(I->c_str());
301   OwningPtr<InputArgList> Args(
302       Opts->ParseArgs(Argv.data(), Argv.data() + Argv.size(), MissingArgIndex,
303                       MissingArgCount, IncludedFlagsBitmask));
304   std::vector<std::string> Inputs = Args->getAllArgValues(OPT_INPUT);
305   return Inputs.back();
306 }
307 
308 // We provide this derivation to add in "-include (file)" arguments for header
309 // dependencies.
310 class AddDependenciesAdjuster : public ArgumentsAdjuster {
311 public:
312   AddDependenciesAdjuster(DependencyMap &Dependencies)
313       : Dependencies(Dependencies) {}
314 
315 private:
316   // Callback for adjusting commandline arguments.
317   CommandLineArguments Adjust(const CommandLineArguments &Args) {
318     std::string InputFile = findInputFile(Args);
319     DependentsVector &FileDependents = Dependencies[InputFile];
320     int Count = FileDependents.size();
321     if (Count == 0)
322       return Args;
323     CommandLineArguments NewArgs(Args);
324     for (int Index = 0; Index < Count; ++Index) {
325       NewArgs.push_back("-include");
326       std::string File(std::string("\"") + FileDependents[Index] +
327                        std::string("\""));
328       NewArgs.push_back(FileDependents[Index]);
329     }
330     return NewArgs;
331   }
332   DependencyMap &Dependencies;
333 };
334 
335 // FIXME: The Location class seems to be something that we might
336 // want to design to be applicable to a wider range of tools, and stick it
337 // somewhere into Tooling/ in mainline
338 struct Location {
339   const FileEntry *File;
340   unsigned Line, Column;
341 
342   Location() : File(), Line(), Column() {}
343 
344   Location(SourceManager &SM, SourceLocation Loc) : File(), Line(), Column() {
345     Loc = SM.getExpansionLoc(Loc);
346     if (Loc.isInvalid())
347       return;
348 
349     std::pair<FileID, unsigned> Decomposed = SM.getDecomposedLoc(Loc);
350     File = SM.getFileEntryForID(Decomposed.first);
351     if (!File)
352       return;
353 
354     Line = SM.getLineNumber(Decomposed.first, Decomposed.second);
355     Column = SM.getColumnNumber(Decomposed.first, Decomposed.second);
356   }
357 
358   operator bool() const { return File != 0; }
359 
360   friend bool operator==(const Location &X, const Location &Y) {
361     return X.File == Y.File && X.Line == Y.Line && X.Column == Y.Column;
362   }
363 
364   friend bool operator!=(const Location &X, const Location &Y) {
365     return !(X == Y);
366   }
367 
368   friend bool operator<(const Location &X, const Location &Y) {
369     if (X.File != Y.File)
370       return X.File < Y.File;
371     if (X.Line != Y.Line)
372       return X.Line < Y.Line;
373     return X.Column < Y.Column;
374   }
375   friend bool operator>(const Location &X, const Location &Y) { return Y < X; }
376   friend bool operator<=(const Location &X, const Location &Y) {
377     return !(Y < X);
378   }
379   friend bool operator>=(const Location &X, const Location &Y) {
380     return !(X < Y);
381   }
382 };
383 
384 struct Entry {
385   enum EntryKind {
386     EK_Tag,
387     EK_Value,
388     EK_Macro,
389 
390     EK_NumberOfKinds
391   } Kind;
392 
393   Location Loc;
394 
395   StringRef getKindName() { return getKindName(Kind); }
396   static StringRef getKindName(EntryKind kind);
397 };
398 
399 // Return a string representing the given kind.
400 StringRef Entry::getKindName(Entry::EntryKind kind) {
401   switch (kind) {
402   case EK_Tag:
403     return "tag";
404   case EK_Value:
405     return "value";
406   case EK_Macro:
407     return "macro";
408   case EK_NumberOfKinds:
409     break;
410   }
411   llvm_unreachable("invalid Entry kind");
412 }
413 
414 struct HeaderEntry {
415   std::string Name;
416   Location Loc;
417 
418   friend bool operator==(const HeaderEntry &X, const HeaderEntry &Y) {
419     return X.Loc == Y.Loc && X.Name == Y.Name;
420   }
421   friend bool operator!=(const HeaderEntry &X, const HeaderEntry &Y) {
422     return !(X == Y);
423   }
424   friend bool operator<(const HeaderEntry &X, const HeaderEntry &Y) {
425     return X.Loc < Y.Loc || (X.Loc == Y.Loc && X.Name < Y.Name);
426   }
427   friend bool operator>(const HeaderEntry &X, const HeaderEntry &Y) {
428     return Y < X;
429   }
430   friend bool operator<=(const HeaderEntry &X, const HeaderEntry &Y) {
431     return !(Y < X);
432   }
433   friend bool operator>=(const HeaderEntry &X, const HeaderEntry &Y) {
434     return !(X < Y);
435   }
436 };
437 
438 typedef std::vector<HeaderEntry> HeaderContents;
439 
440 class EntityMap : public StringMap<SmallVector<Entry, 2> > {
441 public:
442   DenseMap<const FileEntry *, HeaderContents> HeaderContentMismatches;
443 
444   void add(const std::string &Name, enum Entry::EntryKind Kind, Location Loc) {
445     // Record this entity in its header.
446     HeaderEntry HE = { Name, Loc };
447     CurHeaderContents[Loc.File].push_back(HE);
448 
449     // Check whether we've seen this entry before.
450     SmallVector<Entry, 2> &Entries = (*this)[Name];
451     for (unsigned I = 0, N = Entries.size(); I != N; ++I) {
452       if (Entries[I].Kind == Kind && Entries[I].Loc == Loc)
453         return;
454     }
455 
456     // We have not seen this entry before; record it.
457     Entry E = { Kind, Loc };
458     Entries.push_back(E);
459   }
460 
461   void mergeCurHeaderContents() {
462     for (DenseMap<const FileEntry *, HeaderContents>::iterator
463              H = CurHeaderContents.begin(),
464              HEnd = CurHeaderContents.end();
465          H != HEnd; ++H) {
466       // Sort contents.
467       std::sort(H->second.begin(), H->second.end());
468 
469       // Check whether we've seen this header before.
470       DenseMap<const FileEntry *, HeaderContents>::iterator KnownH =
471           AllHeaderContents.find(H->first);
472       if (KnownH == AllHeaderContents.end()) {
473         // We haven't seen this header before; record its contents.
474         AllHeaderContents.insert(*H);
475         continue;
476       }
477 
478       // If the header contents are the same, we're done.
479       if (H->second == KnownH->second)
480         continue;
481 
482       // Determine what changed.
483       std::set_symmetric_difference(
484           H->second.begin(), H->second.end(), KnownH->second.begin(),
485           KnownH->second.end(),
486           std::back_inserter(HeaderContentMismatches[H->first]));
487     }
488 
489     CurHeaderContents.clear();
490   }
491 
492 private:
493   DenseMap<const FileEntry *, HeaderContents> CurHeaderContents;
494   DenseMap<const FileEntry *, HeaderContents> AllHeaderContents;
495 };
496 
497 class CollectEntitiesVisitor
498     : public RecursiveASTVisitor<CollectEntitiesVisitor> {
499 public:
500   CollectEntitiesVisitor(SourceManager &SM, EntityMap &Entities,
501                          Preprocessor &PP, PreprocessorTracker &PPTracker,
502                          int &HadErrors)
503       : SM(SM), Entities(Entities), PP(PP), PPTracker(PPTracker),
504         HadErrors(HadErrors) {}
505 
506   bool TraverseStmt(Stmt *S) { return true; }
507   bool TraverseType(QualType T) { return true; }
508   bool TraverseTypeLoc(TypeLoc TL) { return true; }
509   bool TraverseNestedNameSpecifier(NestedNameSpecifier *NNS) { return true; }
510   bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS) {
511     return true;
512   }
513   bool TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo) {
514     return true;
515   }
516   bool TraverseTemplateName(TemplateName Template) { return true; }
517   bool TraverseTemplateArgument(const TemplateArgument &Arg) { return true; }
518   bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &ArgLoc) {
519     return true;
520   }
521   bool TraverseTemplateArguments(const TemplateArgument *Args,
522                                  unsigned NumArgs) {
523     return true;
524   }
525   bool TraverseConstructorInitializer(CXXCtorInitializer *Init) { return true; }
526   bool TraverseLambdaCapture(LambdaExpr::Capture C) { return true; }
527 
528   // Check 'extern "*" {}' block for #include directives.
529   bool VisitLinkageSpecDecl(LinkageSpecDecl *D) {
530     // Bail if not a block.
531     if (!D->hasBraces())
532       return true;
533     SourceRange BlockRange = D->getSourceRange();
534     const char *LinkageLabel;
535     switch (D->getLanguage()) {
536     case LinkageSpecDecl::lang_c:
537       LinkageLabel = "extern \"C\" {}";
538       break;
539     case LinkageSpecDecl::lang_cxx:
540       LinkageLabel = "extern \"C++\" {}";
541       break;
542     }
543     if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, LinkageLabel,
544                                            errs()))
545       HadErrors = 1;
546     return true;
547   }
548 
549   // Check 'namespace (name) {}' block for #include directives.
550   bool VisitNamespaceDecl(const NamespaceDecl *D) {
551     SourceRange BlockRange = D->getSourceRange();
552     std::string Label("namespace ");
553     Label += D->getName();
554     Label += " {}";
555     if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, Label.c_str(),
556                                            errs()))
557       HadErrors = 1;
558     return true;
559   }
560 
561   // Collect definition entities.
562   bool VisitNamedDecl(NamedDecl *ND) {
563     // We only care about file-context variables.
564     if (!ND->getDeclContext()->isFileContext())
565       return true;
566 
567     // Skip declarations that tend to be properly multiply-declared.
568     if (isa<NamespaceDecl>(ND) || isa<UsingDirectiveDecl>(ND) ||
569         isa<NamespaceAliasDecl>(ND) ||
570         isa<ClassTemplateSpecializationDecl>(ND) || isa<UsingDecl>(ND) ||
571         isa<ClassTemplateDecl>(ND) || isa<TemplateTypeParmDecl>(ND) ||
572         isa<TypeAliasTemplateDecl>(ND) || isa<UsingShadowDecl>(ND) ||
573         isa<FunctionDecl>(ND) || isa<FunctionTemplateDecl>(ND) ||
574         (isa<TagDecl>(ND) &&
575          !cast<TagDecl>(ND)->isThisDeclarationADefinition()))
576       return true;
577 
578     // Skip anonymous declarations.
579     if (!ND->getDeclName())
580       return true;
581 
582     // Get the qualified name.
583     std::string Name;
584     llvm::raw_string_ostream OS(Name);
585     ND->printQualifiedName(OS);
586     OS.flush();
587     if (Name.empty())
588       return true;
589 
590     Location Loc(SM, ND->getLocation());
591     if (!Loc)
592       return true;
593 
594     Entities.add(Name, isa<TagDecl>(ND) ? Entry::EK_Tag : Entry::EK_Value, Loc);
595     return true;
596   }
597 
598 private:
599   SourceManager &SM;
600   EntityMap &Entities;
601   Preprocessor &PP;
602   PreprocessorTracker &PPTracker;
603   int &HadErrors;
604 };
605 
606 class CollectEntitiesConsumer : public ASTConsumer {
607 public:
608   CollectEntitiesConsumer(EntityMap &Entities,
609                           PreprocessorTracker &preprocessorTracker,
610                           Preprocessor &PP, StringRef InFile, int &HadErrors)
611       : Entities(Entities), PPTracker(preprocessorTracker), PP(PP),
612         HadErrors(HadErrors) {
613     PPTracker.handlePreprocessorEntry(PP, InFile);
614   }
615 
616   ~CollectEntitiesConsumer() { PPTracker.handlePreprocessorExit(); }
617 
618   virtual void HandleTranslationUnit(ASTContext &Ctx) {
619     SourceManager &SM = Ctx.getSourceManager();
620 
621     // Collect declared entities.
622     CollectEntitiesVisitor(SM, Entities, PP, PPTracker, HadErrors)
623         .TraverseDecl(Ctx.getTranslationUnitDecl());
624 
625     // Collect macro definitions.
626     for (Preprocessor::macro_iterator M = PP.macro_begin(),
627                                       MEnd = PP.macro_end();
628          M != MEnd; ++M) {
629       Location Loc(SM, M->second->getLocation());
630       if (!Loc)
631         continue;
632 
633       Entities.add(M->first->getName().str(), Entry::EK_Macro, Loc);
634     }
635 
636     // Merge header contents.
637     Entities.mergeCurHeaderContents();
638   }
639 
640 private:
641   EntityMap &Entities;
642   PreprocessorTracker &PPTracker;
643   Preprocessor &PP;
644   int &HadErrors;
645 };
646 
647 class CollectEntitiesAction : public SyntaxOnlyAction {
648 public:
649   CollectEntitiesAction(EntityMap &Entities,
650                         PreprocessorTracker &preprocessorTracker,
651                         int &HadErrors)
652       : Entities(Entities), PPTracker(preprocessorTracker),
653         HadErrors(HadErrors) {}
654 
655 protected:
656   virtual clang::ASTConsumer *CreateASTConsumer(CompilerInstance &CI,
657                                                 StringRef InFile) {
658     return new CollectEntitiesConsumer(Entities, PPTracker,
659                                        CI.getPreprocessor(), InFile, HadErrors);
660   }
661 
662 private:
663   EntityMap &Entities;
664   PreprocessorTracker &PPTracker;
665   int &HadErrors;
666 };
667 
668 class ModularizeFrontendActionFactory : public FrontendActionFactory {
669 public:
670   ModularizeFrontendActionFactory(EntityMap &Entities,
671                                   PreprocessorTracker &preprocessorTracker,
672                                   int &HadErrors)
673       : Entities(Entities), PPTracker(preprocessorTracker),
674         HadErrors(HadErrors) {}
675 
676   virtual CollectEntitiesAction *create() {
677     return new CollectEntitiesAction(Entities, PPTracker, HadErrors);
678   }
679 
680 private:
681   EntityMap &Entities;
682   PreprocessorTracker &PPTracker;
683   int &HadErrors;
684 };
685 
686 int main(int Argc, const char **Argv) {
687 
688   // Save program name for error messages.
689   Argv0 = Argv[0];
690 
691   // Save program arguments for use in module.map comment.
692   CommandLine = sys::path::stem(sys::path::filename(Argv0));
693   for (int ArgIndex = 1; ArgIndex < Argc; ArgIndex++) {
694     CommandLine.append(" ");
695     CommandLine.append(Argv[ArgIndex]);
696   }
697 
698   // This causes options to be parsed.
699   cl::ParseCommandLineOptions(Argc, Argv, "modularize.\n");
700 
701   // No go if we have no header list file.
702   if (ListFileName.size() == 0) {
703     cl::PrintHelpMessage();
704     return 1;
705   }
706 
707   // Get header file names and dependencies.
708   SmallVector<std::string, 32> Headers;
709   DependencyMap Dependencies;
710   if (error_code EC = getHeaderFileNames(Headers, Dependencies, ListFileName,
711                                          HeaderPrefix)) {
712     errs() << Argv[0] << ": error: Unable to get header list '" << ListFileName
713            << "': " << EC.message() << '\n';
714     return 1;
715   }
716 
717   // If we are in assistant mode, output the module map and quit.
718   if (ModuleMapPath[0]) {
719     if (!createModuleMap(ModuleMapPath, Headers, Dependencies, HeaderPrefix,
720                          RootModule))
721       return 1; // Failed.
722     return 0;   // Success - Skip checks in assistant mode.
723   }
724 
725   // Create the compilation database.
726   SmallString<256> PathBuf;
727   sys::fs::current_path(PathBuf);
728   OwningPtr<CompilationDatabase> Compilations;
729   Compilations.reset(
730       new FixedCompilationDatabase(Twine(PathBuf), CC1Arguments));
731 
732   // Create preprocessor tracker, to watch for macro and conditional problems.
733   OwningPtr<PreprocessorTracker> PPTracker(PreprocessorTracker::create());
734 
735   // Parse all of the headers, detecting duplicates.
736   EntityMap Entities;
737   ClangTool Tool(*Compilations, Headers);
738   Tool.appendArgumentsAdjuster(new AddDependenciesAdjuster(Dependencies));
739   int HadErrors = 0;
740   HadErrors |= Tool.run(
741       new ModularizeFrontendActionFactory(Entities, *PPTracker, HadErrors));
742 
743   // Create a place to save duplicate entity locations, separate bins per kind.
744   typedef SmallVector<Location, 8> LocationArray;
745   typedef SmallVector<LocationArray, Entry::EK_NumberOfKinds> EntryBinArray;
746   EntryBinArray EntryBins;
747   int KindIndex;
748   for (KindIndex = 0; KindIndex < Entry::EK_NumberOfKinds; ++KindIndex) {
749     LocationArray Array;
750     EntryBins.push_back(Array);
751   }
752 
753   // Check for the same entity being defined in multiple places.
754   for (EntityMap::iterator E = Entities.begin(), EEnd = Entities.end();
755        E != EEnd; ++E) {
756     // If only one occurance, exit early.
757     if (E->second.size() == 1)
758       continue;
759     // Clear entity locations.
760     for (EntryBinArray::iterator CI = EntryBins.begin(), CE = EntryBins.end();
761          CI != CE; ++CI) {
762       CI->clear();
763     }
764     // Walk the entities of a single name, collecting the locations,
765     // separated into separate bins.
766     for (unsigned I = 0, N = E->second.size(); I != N; ++I) {
767       EntryBins[E->second[I].Kind].push_back(E->second[I].Loc);
768     }
769     // Report any duplicate entity definition errors.
770     int KindIndex = 0;
771     for (EntryBinArray::iterator DI = EntryBins.begin(), DE = EntryBins.end();
772          DI != DE; ++DI, ++KindIndex) {
773       int ECount = DI->size();
774       // If only 1 occurance, skip;
775       if (ECount <= 1)
776         continue;
777       LocationArray::iterator FI = DI->begin();
778       StringRef kindName = Entry::getKindName((Entry::EntryKind)KindIndex);
779       errs() << "error: " << kindName << " '" << E->first()
780              << "' defined at multiple locations:\n";
781       for (LocationArray::iterator FE = DI->end(); FI != FE; ++FI) {
782         errs() << "    " << FI->File->getName() << ":" << FI->Line << ":"
783                << FI->Column << "\n";
784       }
785       HadErrors = 1;
786     }
787   }
788 
789   // Complain about macro instance in header files that differ based on how
790   // they are included.
791   if (PPTracker->reportInconsistentMacros(errs()))
792     HadErrors = 1;
793 
794   // Complain about preprocessor conditional directives in header files that
795   // differ based on how they are included.
796   if (PPTracker->reportInconsistentConditionals(errs()))
797     HadErrors = 1;
798 
799   // Complain about any headers that have contents that differ based on how
800   // they are included.
801   // FIXME: Could we provide information about which preprocessor conditionals
802   // are involved?
803   for (DenseMap<const FileEntry *, HeaderContents>::iterator
804            H = Entities.HeaderContentMismatches.begin(),
805            HEnd = Entities.HeaderContentMismatches.end();
806        H != HEnd; ++H) {
807     if (H->second.empty()) {
808       errs() << "internal error: phantom header content mismatch\n";
809       continue;
810     }
811 
812     HadErrors = 1;
813     errs() << "error: header '" << H->first->getName()
814            << "' has different contents depending on how it was included.\n";
815     for (unsigned I = 0, N = H->second.size(); I != N; ++I) {
816       errs() << "note: '" << H->second[I].Name << "' in "
817              << H->second[I].Loc.File->getName() << " at "
818              << H->second[I].Loc.Line << ":" << H->second[I].Loc.Column
819              << " not always provided\n";
820     }
821   }
822 
823   return HadErrors;
824 }
825