1 //===- extra/modularize/Modularize.cpp - Check modularized headers --------===//
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 // Introduction
10 //
11 // This file implements a tool that checks whether a set of headers provides
12 // the consistent definitions required to use modules. It can also check an
13 // existing module map for full coverage of the headers in a directory tree.
14 //
15 // For example, in examining headers, it detects whether the same entity
16 // (say, a NULL macro or size_t typedef) is defined in multiple headers
17 // or whether a header produces different definitions under
18 // different circumstances. These conditions cause modules built from the
19 // headers to behave poorly, and should be fixed before introducing a module
20 // map.
21 //
22 // Modularize takes as input either one or more module maps (by default,
23 // "module.modulemap") or one or more text files containing lists of headers
24 // to check.
25 //
26 // In the case of a module map, the module map must be well-formed in
27 // terms of syntax. Modularize will extract the header file names
28 // from the map. Only normal headers are checked, assuming headers
29 // marked "private", "textual", or "exclude" are not to be checked
30 // as a top-level include, assuming they either are included by
31 // other headers which are checked, or they are not suitable for
32 // modules.
33 //
34 // In the case of a file list, the list is a newline-separated list of headers
35 // to check with respect to each other.
36 // Lines beginning with '#' and empty lines are ignored.
37 // Header file names followed by a colon and other space-separated
38 // file names will include those extra files as dependencies.
39 // The file names can be relative or full paths, but must be on the
40 // same line.
41 //
42 // Modularize also accepts regular clang front-end arguments.
43 //
44 // Usage: modularize [(modularize options)]
45 // [(include-files_list)|(module map)]+ [(front-end-options) ...]
46 //
47 // Options:
48 // -prefix=(optional header path prefix)
49 // Note that unless a "-prefix (header path)" option is specified,
50 // non-absolute file paths in the header list file will be relative
51 // to the header list file directory. Use -prefix to specify a
52 // different directory.
53 // -module-map-path=(module map)
54 // Skip the checks, and instead act as a module.map generation
55 // assistant, generating a module map file based on the header list.
56 // An optional "-root-module=(rootName)" argument can specify a root
57 // module to be created in the generated module.map file. Note that
58 // you will likely need to edit this file to suit the needs of your
59 // headers.
60 // -problem-files-list=(problem files list file name)
61 // For use only with module map assistant. Input list of files that
62 // have problems with respect to modules. These will still be
63 // included in the generated module map, but will be marked as
64 // "excluded" headers.
65 // -root-module=(root module name)
66 // Specifies a root module to be created in the generated module.map
67 // file.
68 // -block-check-header-list-only
69 // Only warn if #include directives are inside extern or namespace
70 // blocks if the included header is in the header list.
71 // -no-coverage-check
72 // Don't do the coverage check.
73 // -coverage-check-only
74 // Only do the coverage check.
75 // -display-file-lists
76 // Display lists of good files (no compile errors), problem files,
77 // and a combined list with problem files preceded by a '#'.
78 // This can be used to quickly determine which files have problems.
79 // The latter combined list might be useful in starting to modularize
80 // a set of headers. You can start with a full list of headers,
81 // use -display-file-lists option, and then use the combined list as
82 // your intermediate list, uncommenting-out headers as you fix them.
83 //
84 // Note that by default, the modularize assumes .h files contain C++ source.
85 // If your .h files in the file list contain another language, you should
86 // append an appropriate -x option to your command line, i.e.: -x c
87 //
88 // Modularization Issue Checks
89 //
90 // In the process of checking headers for modularization issues, modularize
91 // will do normal parsing, reporting normal errors and warnings,
92 // but will also report special error messages like the following:
93 //
94 // error: '(symbol)' defined at multiple locations:
95 // (file):(row):(column)
96 // (file):(row):(column)
97 //
98 // error: header '(file)' has different contents depending on how it was
99 // included
100 //
101 // The latter might be followed by messages like the following:
102 //
103 // note: '(symbol)' in (file) at (row):(column) not always provided
104 //
105 // Checks will also be performed for macro expansions, defined(macro)
106 // expressions, and preprocessor conditional directives that evaluate
107 // inconsistently, and can produce error messages like the following:
108 //
109 // (...)/SubHeader.h:11:5:
110 // #if SYMBOL == 1
111 // ^
112 // error: Macro instance 'SYMBOL' has different values in this header,
113 // depending on how it was included.
114 // 'SYMBOL' expanded to: '1' with respect to these inclusion paths:
115 // (...)/Header1.h
116 // (...)/SubHeader.h
117 // (...)/SubHeader.h:3:9:
118 // #define SYMBOL 1
119 // ^
120 // Macro defined here.
121 // 'SYMBOL' expanded to: '2' with respect to these inclusion paths:
122 // (...)/Header2.h
123 // (...)/SubHeader.h
124 // (...)/SubHeader.h:7:9:
125 // #define SYMBOL 2
126 // ^
127 // Macro defined here.
128 //
129 // Checks will also be performed for '#include' directives that are
130 // nested inside 'extern "C/C++" {}' or 'namespace (name) {}' blocks,
131 // and can produce error message like the following:
132 //
133 // IncludeInExtern.h:2:3
134 // #include "Empty.h"
135 // ^
136 // error: Include directive within extern "C" {}.
137 // IncludeInExtern.h:1:1
138 // extern "C" {
139 // ^
140 // The "extern "C" {}" block is here.
141 //
142 // See PreprocessorTracker.cpp for additional details.
143 //
144 // Module Map Coverage Check
145 //
146 // The coverage check uses the Clang ModuleMap class to read and parse the
147 // module map file. Starting at the module map file directory, or just the
148 // include paths, if specified, it will collect the names of all the files it
149 // considers headers (no extension, .h, or .inc--if you need more, modify the
150 // isHeader function). It then compares the headers against those referenced
151 // in the module map, either explicitly named, or implicitly named via an
152 // umbrella directory or umbrella file, as parsed by the ModuleMap object.
153 // If headers are found which are not referenced or covered by an umbrella
154 // directory or file, warning messages will be produced, and this program
155 // will return an error code of 1. Other errors result in an error code of 2.
156 // If no problems are found, an error code of 0 is returned.
157 //
158 // Note that in the case of umbrella headers, this tool invokes the compiler
159 // to preprocess the file, and uses a callback to collect the header files
160 // included by the umbrella header or any of its nested includes. If any
161 // front end options are needed for these compiler invocations, these
162 // can be included on the command line after the module map file argument.
163 //
164 // Warning message have the form:
165 //
166 // warning: module.modulemap does not account for file: Level3A.h
167 //
168 // Note that for the case of the module map referencing a file that does
169 // not exist, the module map parser in Clang will (at the time of this
170 // writing) display an error message.
171 //
172 // Module Map Assistant - Module Map Generation
173 //
174 // Modularize also has an option ("-module-map-path=module.modulemap") that will
175 // skip the checks, and instead act as a module.modulemap generation assistant,
176 // generating a module map file based on the header list. An optional
177 // "-root-module=(rootName)" argument can specify a root module to be
178 // created in the generated module.modulemap file. Note that you will likely
179 // need to edit this file to suit the needs of your headers.
180 //
181 // An example command line for generating a module.modulemap file:
182 //
183 // modularize -module-map-path=module.modulemap -root-module=myroot \
184 // headerlist.txt
185 //
186 // Note that if the headers in the header list have partial paths, sub-modules
187 // will be created for the subdirectories involved, assuming that the
188 // subdirectories contain headers to be grouped into a module, but still with
189 // individual modules for the headers in the subdirectory.
190 //
191 // See the ModuleAssistant.cpp file comments for additional details about the
192 // implementation of the assistant mode.
193 //
194 // Future directions:
195 //
196 // Basically, we want to add new checks for whatever we can check with respect
197 // to checking headers for module'ability.
198 //
199 // Some ideas:
200 //
201 // 1. Omit duplicate "not always provided" messages
202 //
203 // 2. Add options to disable any of the checks, in case
204 // there is some problem with them, or the messages get too verbose.
205 //
206 // 3. Try to figure out the preprocessor conditional directives that
207 // contribute to problems and tie them to the inconsistent definitions.
208 //
209 // 4. There are some legitimate uses of preprocessor macros that
210 // modularize will flag as errors, such as repeatedly #include'ing
211 // a file and using interleaving defined/undefined macros
212 // to change declarations in the included file. Is there a way
213 // to address this? Maybe have modularize accept a list of macros
214 // to ignore. Otherwise you can just exclude the file, after checking
215 // for legitimate errors.
216 //
217 // 5. What else?
218 //
219 // General clean-up and refactoring:
220 //
221 // 1. The Location class seems to be something that we might
222 // want to design to be applicable to a wider range of tools, and stick it
223 // somewhere into Tooling/ in mainline
224 //
225 //===----------------------------------------------------------------------===//
226
227 #include "Modularize.h"
228 #include "ModularizeUtilities.h"
229 #include "PreprocessorTracker.h"
230 #include "clang/AST/ASTConsumer.h"
231 #include "clang/AST/ASTContext.h"
232 #include "clang/AST/RecursiveASTVisitor.h"
233 #include "clang/Basic/SourceManager.h"
234 #include "clang/Driver/Options.h"
235 #include "clang/Frontend/CompilerInstance.h"
236 #include "clang/Frontend/FrontendAction.h"
237 #include "clang/Frontend/FrontendActions.h"
238 #include "clang/Lex/Preprocessor.h"
239 #include "clang/Tooling/CompilationDatabase.h"
240 #include "clang/Tooling/Tooling.h"
241 #include "llvm/Option/Arg.h"
242 #include "llvm/Option/ArgList.h"
243 #include "llvm/Option/OptTable.h"
244 #include "llvm/Option/Option.h"
245 #include "llvm/Support/CommandLine.h"
246 #include "llvm/Support/FileSystem.h"
247 #include "llvm/Support/MemoryBuffer.h"
248 #include "llvm/Support/Path.h"
249 #include <algorithm>
250 #include <fstream>
251 #include <iterator>
252 #include <string>
253 #include <vector>
254
255 using namespace clang;
256 using namespace clang::driver;
257 using namespace clang::driver::options;
258 using namespace clang::tooling;
259 using namespace llvm;
260 using namespace llvm::opt;
261 using namespace Modularize;
262
263 // Option to specify a file name for a list of header files to check.
264 static cl::list<std::string>
265 ListFileNames(cl::Positional, cl::value_desc("list"),
266 cl::desc("<list of one or more header list files>"),
267 cl::CommaSeparated);
268
269 // Collect all other arguments, which will be passed to the front end.
270 static cl::list<std::string>
271 CC1Arguments(cl::ConsumeAfter,
272 cl::desc("<arguments to be passed to front end>..."));
273
274 // Option to specify a prefix to be prepended to the header names.
275 static cl::opt<std::string> HeaderPrefix(
276 "prefix", cl::init(""),
277 cl::desc(
278 "Prepend header file paths with this prefix."
279 " If not specified,"
280 " the files are considered to be relative to the header list file."));
281
282 // Option for assistant mode, telling modularize to output a module map
283 // based on the headers list, and where to put it.
284 static cl::opt<std::string> ModuleMapPath(
285 "module-map-path", cl::init(""),
286 cl::desc("Turn on module map output and specify output path or file name."
287 " If no path is specified and if prefix option is specified,"
288 " use prefix for file path."));
289
290 // Option to specify list of problem files for assistant.
291 // This will cause assistant to exclude these files.
292 static cl::opt<std::string> ProblemFilesList(
293 "problem-files-list", cl::init(""),
294 cl::desc(
295 "List of files with compilation or modularization problems for"
296 " assistant mode. This will be excluded."));
297
298 // Option for assistant mode, telling modularize the name of the root module.
299 static cl::opt<std::string>
300 RootModule("root-module", cl::init(""),
301 cl::desc("Specify the name of the root module."));
302
303 // Option for limiting the #include-inside-extern-or-namespace-block
304 // check to only those headers explicitly listed in the header list.
305 // This is a work-around for private includes that purposefully get
306 // included inside blocks.
307 static cl::opt<bool>
308 BlockCheckHeaderListOnly("block-check-header-list-only", cl::init(false),
309 cl::desc("Only warn if #include directives are inside extern or namespace"
310 " blocks if the included header is in the header list."));
311
312 // Option for include paths for coverage check.
313 static cl::list<std::string>
314 IncludePaths("I", cl::desc("Include path for coverage check."),
315 cl::ZeroOrMore, cl::value_desc("path"));
316
317 // Option for disabling the coverage check.
318 static cl::opt<bool>
319 NoCoverageCheck("no-coverage-check", cl::init(false),
320 cl::desc("Don't do the coverage check."));
321
322 // Option for just doing the coverage check.
323 static cl::opt<bool>
324 CoverageCheckOnly("coverage-check-only", cl::init(false),
325 cl::desc("Only do the coverage check."));
326
327 // Option for displaying lists of good, bad, and mixed files.
328 static cl::opt<bool>
329 DisplayFileLists("display-file-lists", cl::init(false),
330 cl::desc("Display lists of good files (no compile errors), problem files,"
331 " and a combined list with problem files preceded by a '#'."));
332
333 // Save the program name for error messages.
334 const char *Argv0;
335 // Save the command line for comments.
336 std::string CommandLine;
337
338 // Helper function for finding the input file in an arguments list.
findInputFile(const CommandLineArguments & CLArgs)339 static std::string findInputFile(const CommandLineArguments &CLArgs) {
340 const unsigned IncludedFlagsBitmask = options::CC1Option;
341 unsigned MissingArgIndex, MissingArgCount;
342 SmallVector<const char *, 256> Argv;
343 for (auto I = CLArgs.begin(), E = CLArgs.end(); I != E; ++I)
344 Argv.push_back(I->c_str());
345 InputArgList Args = getDriverOptTable().ParseArgs(
346 Argv, MissingArgIndex, MissingArgCount, IncludedFlagsBitmask);
347 std::vector<std::string> Inputs = Args.getAllArgValues(OPT_INPUT);
348 return ModularizeUtilities::getCanonicalPath(Inputs.back());
349 }
350
351 // This arguments adjuster inserts "-include (file)" arguments for header
352 // dependencies. It also inserts a "-w" option and a "-x c++",
353 // if no other "-x" option is present.
354 static ArgumentsAdjuster
getModularizeArgumentsAdjuster(DependencyMap & Dependencies)355 getModularizeArgumentsAdjuster(DependencyMap &Dependencies) {
356 return [&Dependencies](const CommandLineArguments &Args,
357 StringRef /*unused*/) {
358 std::string InputFile = findInputFile(Args);
359 DependentsVector &FileDependents = Dependencies[InputFile];
360 CommandLineArguments NewArgs(Args);
361 if (int Count = FileDependents.size()) {
362 for (int Index = 0; Index < Count; ++Index) {
363 NewArgs.push_back("-include");
364 std::string File(std::string("\"") + FileDependents[Index] +
365 std::string("\""));
366 NewArgs.push_back(FileDependents[Index]);
367 }
368 }
369 // Ignore warnings. (Insert after "clang_tool" at beginning.)
370 NewArgs.insert(NewArgs.begin() + 1, "-w");
371 // Since we are compiling .h files, assume C++ unless given a -x option.
372 if (!llvm::is_contained(NewArgs, "-x")) {
373 NewArgs.insert(NewArgs.begin() + 2, "-x");
374 NewArgs.insert(NewArgs.begin() + 3, "c++");
375 }
376 return NewArgs;
377 };
378 }
379
380 // FIXME: The Location class seems to be something that we might
381 // want to design to be applicable to a wider range of tools, and stick it
382 // somewhere into Tooling/ in mainline
383 struct Location {
384 const FileEntry *File;
385 unsigned Line, Column;
386
LocationLocation387 Location() : File(), Line(), Column() {}
388
LocationLocation389 Location(SourceManager &SM, SourceLocation Loc) : File(), Line(), Column() {
390 Loc = SM.getExpansionLoc(Loc);
391 if (Loc.isInvalid())
392 return;
393
394 std::pair<FileID, unsigned> Decomposed = SM.getDecomposedLoc(Loc);
395 File = SM.getFileEntryForID(Decomposed.first);
396 if (!File)
397 return;
398
399 Line = SM.getLineNumber(Decomposed.first, Decomposed.second);
400 Column = SM.getColumnNumber(Decomposed.first, Decomposed.second);
401 }
402
operator boolLocation403 operator bool() const { return File != nullptr; }
404
operator ==(const Location & X,const Location & Y)405 friend bool operator==(const Location &X, const Location &Y) {
406 return X.File == Y.File && X.Line == Y.Line && X.Column == Y.Column;
407 }
408
operator !=(const Location & X,const Location & Y)409 friend bool operator!=(const Location &X, const Location &Y) {
410 return !(X == Y);
411 }
412
operator <(const Location & X,const Location & Y)413 friend bool operator<(const Location &X, const Location &Y) {
414 if (X.File != Y.File)
415 return X.File < Y.File;
416 if (X.Line != Y.Line)
417 return X.Line < Y.Line;
418 return X.Column < Y.Column;
419 }
operator >(const Location & X,const Location & Y)420 friend bool operator>(const Location &X, const Location &Y) { return Y < X; }
operator <=(const Location & X,const Location & Y)421 friend bool operator<=(const Location &X, const Location &Y) {
422 return !(Y < X);
423 }
operator >=(const Location & X,const Location & Y)424 friend bool operator>=(const Location &X, const Location &Y) {
425 return !(X < Y);
426 }
427 };
428
429 struct Entry {
430 enum EntryKind {
431 EK_Tag,
432 EK_Value,
433 EK_Macro,
434
435 EK_NumberOfKinds
436 } Kind;
437
438 Location Loc;
439
getKindNameEntry440 StringRef getKindName() { return getKindName(Kind); }
441 static StringRef getKindName(EntryKind kind);
442 };
443
444 // Return a string representing the given kind.
getKindName(Entry::EntryKind kind)445 StringRef Entry::getKindName(Entry::EntryKind kind) {
446 switch (kind) {
447 case EK_Tag:
448 return "tag";
449 case EK_Value:
450 return "value";
451 case EK_Macro:
452 return "macro";
453 case EK_NumberOfKinds:
454 break;
455 }
456 llvm_unreachable("invalid Entry kind");
457 }
458
459 struct HeaderEntry {
460 std::string Name;
461 Location Loc;
462
operator ==(const HeaderEntry & X,const HeaderEntry & Y)463 friend bool operator==(const HeaderEntry &X, const HeaderEntry &Y) {
464 return X.Loc == Y.Loc && X.Name == Y.Name;
465 }
operator !=(const HeaderEntry & X,const HeaderEntry & Y)466 friend bool operator!=(const HeaderEntry &X, const HeaderEntry &Y) {
467 return !(X == Y);
468 }
operator <(const HeaderEntry & X,const HeaderEntry & Y)469 friend bool operator<(const HeaderEntry &X, const HeaderEntry &Y) {
470 return X.Loc < Y.Loc || (X.Loc == Y.Loc && X.Name < Y.Name);
471 }
operator >(const HeaderEntry & X,const HeaderEntry & Y)472 friend bool operator>(const HeaderEntry &X, const HeaderEntry &Y) {
473 return Y < X;
474 }
operator <=(const HeaderEntry & X,const HeaderEntry & Y)475 friend bool operator<=(const HeaderEntry &X, const HeaderEntry &Y) {
476 return !(Y < X);
477 }
operator >=(const HeaderEntry & X,const HeaderEntry & Y)478 friend bool operator>=(const HeaderEntry &X, const HeaderEntry &Y) {
479 return !(X < Y);
480 }
481 };
482
483 typedef std::vector<HeaderEntry> HeaderContents;
484
485 class EntityMap : public StringMap<SmallVector<Entry, 2> > {
486 public:
487 DenseMap<const FileEntry *, HeaderContents> HeaderContentMismatches;
488
add(const std::string & Name,enum Entry::EntryKind Kind,Location Loc)489 void add(const std::string &Name, enum Entry::EntryKind Kind, Location Loc) {
490 // Record this entity in its header.
491 HeaderEntry HE = { Name, Loc };
492 CurHeaderContents[Loc.File].push_back(HE);
493
494 // Check whether we've seen this entry before.
495 SmallVector<Entry, 2> &Entries = (*this)[Name];
496 for (unsigned I = 0, N = Entries.size(); I != N; ++I) {
497 if (Entries[I].Kind == Kind && Entries[I].Loc == Loc)
498 return;
499 }
500
501 // We have not seen this entry before; record it.
502 Entry E = { Kind, Loc };
503 Entries.push_back(E);
504 }
505
mergeCurHeaderContents()506 void mergeCurHeaderContents() {
507 for (DenseMap<const FileEntry *, HeaderContents>::iterator
508 H = CurHeaderContents.begin(),
509 HEnd = CurHeaderContents.end();
510 H != HEnd; ++H) {
511 // Sort contents.
512 llvm::sort(H->second);
513
514 // Check whether we've seen this header before.
515 DenseMap<const FileEntry *, HeaderContents>::iterator KnownH =
516 AllHeaderContents.find(H->first);
517 if (KnownH == AllHeaderContents.end()) {
518 // We haven't seen this header before; record its contents.
519 AllHeaderContents.insert(*H);
520 continue;
521 }
522
523 // If the header contents are the same, we're done.
524 if (H->second == KnownH->second)
525 continue;
526
527 // Determine what changed.
528 std::set_symmetric_difference(
529 H->second.begin(), H->second.end(), KnownH->second.begin(),
530 KnownH->second.end(),
531 std::back_inserter(HeaderContentMismatches[H->first]));
532 }
533
534 CurHeaderContents.clear();
535 }
536
537 private:
538 DenseMap<const FileEntry *, HeaderContents> CurHeaderContents;
539 DenseMap<const FileEntry *, HeaderContents> AllHeaderContents;
540 };
541
542 class CollectEntitiesVisitor
543 : public RecursiveASTVisitor<CollectEntitiesVisitor> {
544 public:
CollectEntitiesVisitor(SourceManager & SM,EntityMap & Entities,Preprocessor & PP,PreprocessorTracker & PPTracker,int & HadErrors)545 CollectEntitiesVisitor(SourceManager &SM, EntityMap &Entities,
546 Preprocessor &PP, PreprocessorTracker &PPTracker,
547 int &HadErrors)
548 : SM(SM), Entities(Entities), PP(PP), PPTracker(PPTracker),
549 HadErrors(HadErrors) {}
550
TraverseStmt(Stmt * S)551 bool TraverseStmt(Stmt *S) { return true; }
TraverseType(QualType T)552 bool TraverseType(QualType T) { return true; }
TraverseTypeLoc(TypeLoc TL)553 bool TraverseTypeLoc(TypeLoc TL) { return true; }
TraverseNestedNameSpecifier(NestedNameSpecifier * NNS)554 bool TraverseNestedNameSpecifier(NestedNameSpecifier *NNS) { return true; }
TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS)555 bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS) {
556 return true;
557 }
TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo)558 bool TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo) {
559 return true;
560 }
TraverseTemplateName(TemplateName Template)561 bool TraverseTemplateName(TemplateName Template) { return true; }
TraverseTemplateArgument(const TemplateArgument & Arg)562 bool TraverseTemplateArgument(const TemplateArgument &Arg) { return true; }
TraverseTemplateArgumentLoc(const TemplateArgumentLoc & ArgLoc)563 bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &ArgLoc) {
564 return true;
565 }
TraverseTemplateArguments(const TemplateArgument * Args,unsigned NumArgs)566 bool TraverseTemplateArguments(const TemplateArgument *Args,
567 unsigned NumArgs) {
568 return true;
569 }
TraverseConstructorInitializer(CXXCtorInitializer * Init)570 bool TraverseConstructorInitializer(CXXCtorInitializer *Init) { return true; }
TraverseLambdaCapture(LambdaExpr * LE,const LambdaCapture * C,Expr * Init)571 bool TraverseLambdaCapture(LambdaExpr *LE, const LambdaCapture *C,
572 Expr *Init) {
573 return true;
574 }
575
576 // Check 'extern "*" {}' block for #include directives.
VisitLinkageSpecDecl(LinkageSpecDecl * D)577 bool VisitLinkageSpecDecl(LinkageSpecDecl *D) {
578 // Bail if not a block.
579 if (!D->hasBraces())
580 return true;
581 SourceRange BlockRange = D->getSourceRange();
582 const char *LinkageLabel;
583 switch (D->getLanguage()) {
584 case LinkageSpecDecl::lang_c:
585 LinkageLabel = "extern \"C\" {}";
586 break;
587 case LinkageSpecDecl::lang_cxx:
588 LinkageLabel = "extern \"C++\" {}";
589 break;
590 }
591 if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, LinkageLabel,
592 errs()))
593 HadErrors = 1;
594 return true;
595 }
596
597 // Check 'namespace (name) {}' block for #include directives.
VisitNamespaceDecl(const NamespaceDecl * D)598 bool VisitNamespaceDecl(const NamespaceDecl *D) {
599 SourceRange BlockRange = D->getSourceRange();
600 std::string Label("namespace ");
601 Label += D->getName();
602 Label += " {}";
603 if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, Label.c_str(),
604 errs()))
605 HadErrors = 1;
606 return true;
607 }
608
609 // Collect definition entities.
VisitNamedDecl(NamedDecl * ND)610 bool VisitNamedDecl(NamedDecl *ND) {
611 // We only care about file-context variables.
612 if (!ND->getDeclContext()->isFileContext())
613 return true;
614
615 // Skip declarations that tend to be properly multiply-declared.
616 if (isa<NamespaceDecl>(ND) || isa<UsingDirectiveDecl>(ND) ||
617 isa<NamespaceAliasDecl>(ND) ||
618 isa<ClassTemplateSpecializationDecl>(ND) || isa<UsingDecl>(ND) ||
619 isa<ClassTemplateDecl>(ND) || isa<TemplateTypeParmDecl>(ND) ||
620 isa<TypeAliasTemplateDecl>(ND) || isa<UsingShadowDecl>(ND) ||
621 isa<FunctionDecl>(ND) || isa<FunctionTemplateDecl>(ND) ||
622 (isa<TagDecl>(ND) &&
623 !cast<TagDecl>(ND)->isThisDeclarationADefinition()))
624 return true;
625
626 // Skip anonymous declarations.
627 if (!ND->getDeclName())
628 return true;
629
630 // Get the qualified name.
631 std::string Name;
632 llvm::raw_string_ostream OS(Name);
633 ND->printQualifiedName(OS);
634 OS.flush();
635 if (Name.empty())
636 return true;
637
638 Location Loc(SM, ND->getLocation());
639 if (!Loc)
640 return true;
641
642 Entities.add(Name, isa<TagDecl>(ND) ? Entry::EK_Tag : Entry::EK_Value, Loc);
643 return true;
644 }
645
646 private:
647 SourceManager &SM;
648 EntityMap &Entities;
649 Preprocessor &PP;
650 PreprocessorTracker &PPTracker;
651 int &HadErrors;
652 };
653
654 class CollectEntitiesConsumer : public ASTConsumer {
655 public:
CollectEntitiesConsumer(EntityMap & Entities,PreprocessorTracker & preprocessorTracker,Preprocessor & PP,StringRef InFile,int & HadErrors)656 CollectEntitiesConsumer(EntityMap &Entities,
657 PreprocessorTracker &preprocessorTracker,
658 Preprocessor &PP, StringRef InFile, int &HadErrors)
659 : Entities(Entities), PPTracker(preprocessorTracker), PP(PP),
660 HadErrors(HadErrors) {
661 PPTracker.handlePreprocessorEntry(PP, InFile);
662 }
663
~CollectEntitiesConsumer()664 ~CollectEntitiesConsumer() override { PPTracker.handlePreprocessorExit(); }
665
HandleTranslationUnit(ASTContext & Ctx)666 void HandleTranslationUnit(ASTContext &Ctx) override {
667 SourceManager &SM = Ctx.getSourceManager();
668
669 // Collect declared entities.
670 CollectEntitiesVisitor(SM, Entities, PP, PPTracker, HadErrors)
671 .TraverseDecl(Ctx.getTranslationUnitDecl());
672
673 // Collect macro definitions.
674 for (Preprocessor::macro_iterator M = PP.macro_begin(),
675 MEnd = PP.macro_end();
676 M != MEnd; ++M) {
677 Location Loc(SM, M->second.getLatest()->getLocation());
678 if (!Loc)
679 continue;
680
681 Entities.add(M->first->getName().str(), Entry::EK_Macro, Loc);
682 }
683
684 // Merge header contents.
685 Entities.mergeCurHeaderContents();
686 }
687
688 private:
689 EntityMap &Entities;
690 PreprocessorTracker &PPTracker;
691 Preprocessor &PP;
692 int &HadErrors;
693 };
694
695 class CollectEntitiesAction : public SyntaxOnlyAction {
696 public:
CollectEntitiesAction(EntityMap & Entities,PreprocessorTracker & preprocessorTracker,int & HadErrors)697 CollectEntitiesAction(EntityMap &Entities,
698 PreprocessorTracker &preprocessorTracker,
699 int &HadErrors)
700 : Entities(Entities), PPTracker(preprocessorTracker),
701 HadErrors(HadErrors) {}
702
703 protected:
704 std::unique_ptr<clang::ASTConsumer>
CreateASTConsumer(CompilerInstance & CI,StringRef InFile)705 CreateASTConsumer(CompilerInstance &CI, StringRef InFile) override {
706 return std::make_unique<CollectEntitiesConsumer>(
707 Entities, PPTracker, CI.getPreprocessor(), InFile, HadErrors);
708 }
709
710 private:
711 EntityMap &Entities;
712 PreprocessorTracker &PPTracker;
713 int &HadErrors;
714 };
715
716 class ModularizeFrontendActionFactory : public FrontendActionFactory {
717 public:
ModularizeFrontendActionFactory(EntityMap & Entities,PreprocessorTracker & preprocessorTracker,int & HadErrors)718 ModularizeFrontendActionFactory(EntityMap &Entities,
719 PreprocessorTracker &preprocessorTracker,
720 int &HadErrors)
721 : Entities(Entities), PPTracker(preprocessorTracker),
722 HadErrors(HadErrors) {}
723
create()724 std::unique_ptr<FrontendAction> create() override {
725 return std::make_unique<CollectEntitiesAction>(Entities, PPTracker,
726 HadErrors);
727 }
728
729 private:
730 EntityMap &Entities;
731 PreprocessorTracker &PPTracker;
732 int &HadErrors;
733 };
734
735 class CompileCheckVisitor
736 : public RecursiveASTVisitor<CompileCheckVisitor> {
737 public:
CompileCheckVisitor()738 CompileCheckVisitor() {}
739
TraverseStmt(Stmt * S)740 bool TraverseStmt(Stmt *S) { return true; }
TraverseType(QualType T)741 bool TraverseType(QualType T) { return true; }
TraverseTypeLoc(TypeLoc TL)742 bool TraverseTypeLoc(TypeLoc TL) { return true; }
TraverseNestedNameSpecifier(NestedNameSpecifier * NNS)743 bool TraverseNestedNameSpecifier(NestedNameSpecifier *NNS) { return true; }
TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS)744 bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS) {
745 return true;
746 }
TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo)747 bool TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo) {
748 return true;
749 }
TraverseTemplateName(TemplateName Template)750 bool TraverseTemplateName(TemplateName Template) { return true; }
TraverseTemplateArgument(const TemplateArgument & Arg)751 bool TraverseTemplateArgument(const TemplateArgument &Arg) { return true; }
TraverseTemplateArgumentLoc(const TemplateArgumentLoc & ArgLoc)752 bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &ArgLoc) {
753 return true;
754 }
TraverseTemplateArguments(const TemplateArgument * Args,unsigned NumArgs)755 bool TraverseTemplateArguments(const TemplateArgument *Args,
756 unsigned NumArgs) {
757 return true;
758 }
TraverseConstructorInitializer(CXXCtorInitializer * Init)759 bool TraverseConstructorInitializer(CXXCtorInitializer *Init) { return true; }
TraverseLambdaCapture(LambdaExpr * LE,const LambdaCapture * C,Expr * Init)760 bool TraverseLambdaCapture(LambdaExpr *LE, const LambdaCapture *C,
761 Expr *Init) {
762 return true;
763 }
764
765 // Check 'extern "*" {}' block for #include directives.
VisitLinkageSpecDecl(LinkageSpecDecl * D)766 bool VisitLinkageSpecDecl(LinkageSpecDecl *D) {
767 return true;
768 }
769
770 // Check 'namespace (name) {}' block for #include directives.
VisitNamespaceDecl(const NamespaceDecl * D)771 bool VisitNamespaceDecl(const NamespaceDecl *D) {
772 return true;
773 }
774
775 // Collect definition entities.
VisitNamedDecl(NamedDecl * ND)776 bool VisitNamedDecl(NamedDecl *ND) {
777 return true;
778 }
779 };
780
781 class CompileCheckConsumer : public ASTConsumer {
782 public:
CompileCheckConsumer()783 CompileCheckConsumer() {}
784
HandleTranslationUnit(ASTContext & Ctx)785 void HandleTranslationUnit(ASTContext &Ctx) override {
786 CompileCheckVisitor().TraverseDecl(Ctx.getTranslationUnitDecl());
787 }
788 };
789
790 class CompileCheckAction : public SyntaxOnlyAction {
791 public:
CompileCheckAction()792 CompileCheckAction() {}
793
794 protected:
795 std::unique_ptr<clang::ASTConsumer>
CreateASTConsumer(CompilerInstance & CI,StringRef InFile)796 CreateASTConsumer(CompilerInstance &CI, StringRef InFile) override {
797 return std::make_unique<CompileCheckConsumer>();
798 }
799 };
800
801 class CompileCheckFrontendActionFactory : public FrontendActionFactory {
802 public:
CompileCheckFrontendActionFactory()803 CompileCheckFrontendActionFactory() {}
804
create()805 std::unique_ptr<FrontendAction> create() override {
806 return std::make_unique<CompileCheckAction>();
807 }
808 };
809
main(int Argc,const char ** Argv)810 int main(int Argc, const char **Argv) {
811
812 // Save program name for error messages.
813 Argv0 = Argv[0];
814
815 // Save program arguments for use in module.modulemap comment.
816 CommandLine = std::string(sys::path::stem(sys::path::filename(Argv0)));
817 for (int ArgIndex = 1; ArgIndex < Argc; ArgIndex++) {
818 CommandLine.append(" ");
819 CommandLine.append(Argv[ArgIndex]);
820 }
821
822 // This causes options to be parsed.
823 cl::ParseCommandLineOptions(Argc, Argv, "modularize.\n");
824
825 // No go if we have no header list file.
826 if (ListFileNames.size() == 0) {
827 cl::PrintHelpMessage();
828 return 1;
829 }
830
831 std::unique_ptr<ModularizeUtilities> ModUtil;
832 int HadErrors = 0;
833
834 ModUtil.reset(
835 ModularizeUtilities::createModularizeUtilities(
836 ListFileNames, HeaderPrefix, ProblemFilesList));
837
838 // Get header file names and dependencies.
839 if (ModUtil->loadAllHeaderListsAndDependencies())
840 HadErrors = 1;
841
842 // If we are in assistant mode, output the module map and quit.
843 if (ModuleMapPath.length() != 0) {
844 if (!createModuleMap(ModuleMapPath, ModUtil->HeaderFileNames,
845 ModUtil->ProblemFileNames,
846 ModUtil->Dependencies, HeaderPrefix, RootModule))
847 return 1; // Failed.
848 return 0; // Success - Skip checks in assistant mode.
849 }
850
851 // If we're doing module maps.
852 if (!NoCoverageCheck && ModUtil->HasModuleMap) {
853 // Do coverage check.
854 if (ModUtil->doCoverageCheck(IncludePaths, CommandLine))
855 HadErrors = 1;
856 }
857
858 // Bail early if only doing the coverage check.
859 if (CoverageCheckOnly)
860 return HadErrors;
861
862 // Create the compilation database.
863 SmallString<256> PathBuf;
864 sys::fs::current_path(PathBuf);
865 std::unique_ptr<CompilationDatabase> Compilations;
866 Compilations.reset(
867 new FixedCompilationDatabase(Twine(PathBuf), CC1Arguments));
868
869 // Create preprocessor tracker, to watch for macro and conditional problems.
870 std::unique_ptr<PreprocessorTracker> PPTracker(
871 PreprocessorTracker::create(ModUtil->HeaderFileNames,
872 BlockCheckHeaderListOnly));
873
874 // Coolect entities here.
875 EntityMap Entities;
876
877 // Because we can't easily determine which files failed
878 // during the tool run, if we're collecting the file lists
879 // for display, we do a first compile pass on individual
880 // files to find which ones don't compile stand-alone.
881 if (DisplayFileLists) {
882 // First, make a pass to just get compile errors.
883 for (auto &CompileCheckFile : ModUtil->HeaderFileNames) {
884 llvm::SmallVector<std::string, 32> CompileCheckFileArray;
885 CompileCheckFileArray.push_back(CompileCheckFile);
886 ClangTool CompileCheckTool(*Compilations, CompileCheckFileArray);
887 CompileCheckTool.appendArgumentsAdjuster(
888 getModularizeArgumentsAdjuster(ModUtil->Dependencies));
889 int CompileCheckFileErrors = 0;
890 // FIXME: use newFrontendActionFactory.
891 CompileCheckFrontendActionFactory CompileCheckFactory;
892 CompileCheckFileErrors |= CompileCheckTool.run(&CompileCheckFactory);
893 if (CompileCheckFileErrors != 0) {
894 ModUtil->addUniqueProblemFile(CompileCheckFile); // Save problem file.
895 HadErrors |= 1;
896 }
897 else
898 ModUtil->addNoCompileErrorsFile(CompileCheckFile); // Save good file.
899 }
900 }
901
902 // Then we make another pass on the good files to do the rest of the work.
903 ClangTool Tool(*Compilations,
904 (DisplayFileLists ? ModUtil->GoodFileNames : ModUtil->HeaderFileNames));
905 Tool.appendArgumentsAdjuster(
906 getModularizeArgumentsAdjuster(ModUtil->Dependencies));
907 ModularizeFrontendActionFactory Factory(Entities, *PPTracker, HadErrors);
908 HadErrors |= Tool.run(&Factory);
909
910 // Create a place to save duplicate entity locations, separate bins per kind.
911 typedef SmallVector<Location, 8> LocationArray;
912 typedef SmallVector<LocationArray, Entry::EK_NumberOfKinds> EntryBinArray;
913 EntryBinArray EntryBins;
914 int KindIndex;
915 for (KindIndex = 0; KindIndex < Entry::EK_NumberOfKinds; ++KindIndex) {
916 LocationArray Array;
917 EntryBins.push_back(Array);
918 }
919
920 // Check for the same entity being defined in multiple places.
921 for (EntityMap::iterator E = Entities.begin(), EEnd = Entities.end();
922 E != EEnd; ++E) {
923 // If only one occurrence, exit early.
924 if (E->second.size() == 1)
925 continue;
926 // Clear entity locations.
927 for (EntryBinArray::iterator CI = EntryBins.begin(), CE = EntryBins.end();
928 CI != CE; ++CI) {
929 CI->clear();
930 }
931 // Walk the entities of a single name, collecting the locations,
932 // separated into separate bins.
933 for (unsigned I = 0, N = E->second.size(); I != N; ++I) {
934 EntryBins[E->second[I].Kind].push_back(E->second[I].Loc);
935 }
936 // Report any duplicate entity definition errors.
937 int KindIndex = 0;
938 for (EntryBinArray::iterator DI = EntryBins.begin(), DE = EntryBins.end();
939 DI != DE; ++DI, ++KindIndex) {
940 int ECount = DI->size();
941 // If only 1 occurrence of this entity, skip it, we only report duplicates.
942 if (ECount <= 1)
943 continue;
944 LocationArray::iterator FI = DI->begin();
945 StringRef kindName = Entry::getKindName((Entry::EntryKind)KindIndex);
946 errs() << "error: " << kindName << " '" << E->first()
947 << "' defined at multiple locations:\n";
948 for (LocationArray::iterator FE = DI->end(); FI != FE; ++FI) {
949 errs() << " " << FI->File->getName() << ":" << FI->Line << ":"
950 << FI->Column << "\n";
951 ModUtil->addUniqueProblemFile(std::string(FI->File->getName()));
952 }
953 HadErrors = 1;
954 }
955 }
956
957 // Complain about macro instance in header files that differ based on how
958 // they are included.
959 if (PPTracker->reportInconsistentMacros(errs()))
960 HadErrors = 1;
961
962 // Complain about preprocessor conditional directives in header files that
963 // differ based on how they are included.
964 if (PPTracker->reportInconsistentConditionals(errs()))
965 HadErrors = 1;
966
967 // Complain about any headers that have contents that differ based on how
968 // they are included.
969 // FIXME: Could we provide information about which preprocessor conditionals
970 // are involved?
971 for (DenseMap<const FileEntry *, HeaderContents>::iterator
972 H = Entities.HeaderContentMismatches.begin(),
973 HEnd = Entities.HeaderContentMismatches.end();
974 H != HEnd; ++H) {
975 if (H->second.empty()) {
976 errs() << "internal error: phantom header content mismatch\n";
977 continue;
978 }
979
980 HadErrors = 1;
981 ModUtil->addUniqueProblemFile(std::string(H->first->getName()));
982 errs() << "error: header '" << H->first->getName()
983 << "' has different contents depending on how it was included.\n";
984 for (unsigned I = 0, N = H->second.size(); I != N; ++I) {
985 errs() << "note: '" << H->second[I].Name << "' in "
986 << H->second[I].Loc.File->getName() << " at "
987 << H->second[I].Loc.Line << ":" << H->second[I].Loc.Column
988 << " not always provided\n";
989 }
990 }
991
992 if (DisplayFileLists) {
993 ModUtil->displayProblemFiles();
994 ModUtil->displayGoodFiles();
995 ModUtil->displayCombinedFiles();
996 }
997
998 return HadErrors;
999 }
1000