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 <iterator>
251#include <map>
252#include <string>
253#include <vector>
254
255using namespace clang;
256using namespace clang::driver;
257using namespace clang::driver::options;
258using namespace clang::tooling;
259using namespace llvm;
260using namespace llvm::opt;
261using namespace Modularize;
262
263// Option to specify a file name for a list of header files to check.
264static 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.
270static 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.
275static 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.
284static 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.
292static 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.
299static cl::opt<std::string>
300RootModule("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.
307static cl::opt<bool>
308BlockCheckHeaderListOnly("block-check-header-list-only", cl::init(false),
309cl::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.
313static cl::list<std::string>
314 IncludePaths("I", cl::desc("Include path for coverage check."),
315 cl::value_desc("path"));
316
317// Option for disabling the coverage check.
318static cl::opt<bool> NoCoverageCheck("no-coverage-check",
319 cl::desc("Don't do the coverage check."));
320
321// Option for just doing the coverage check.
322static cl::opt<bool>
323CoverageCheckOnly("coverage-check-only", cl::init(false),
324cl::desc("Only do the coverage check."));
325
326// Option for displaying lists of good, bad, and mixed files.
327static cl::opt<bool>
328DisplayFileLists("display-file-lists", cl::init(false),
329cl::desc("Display lists of good files (no compile errors), problem files,"
330 " and a combined list with problem files preceded by a '#'."));
331
332// Save the program name for error messages.
333const char *Argv0;
334// Save the command line for comments.
335std::string CommandLine;
336
337// Helper function for finding the input file in an arguments list.
338static std::string findInputFile(const CommandLineArguments &CLArgs) {
339 const unsigned IncludedFlagsBitmask = options::CC1Option;
340 unsigned MissingArgIndex, MissingArgCount;
341 SmallVector<const char *, 256> Argv;
342 for (auto I = CLArgs.begin(), E = CLArgs.end(); I != E; ++I)
343 Argv.push_back(I->c_str());
344 InputArgList Args = getDriverOptTable().ParseArgs(
345 Argv, MissingArgIndex, MissingArgCount, IncludedFlagsBitmask);
346 std::vector<std::string> Inputs = Args.getAllArgValues(OPT_INPUT);
347 return ModularizeUtilities::getCanonicalPath(Inputs.back());
348}
349
350// This arguments adjuster inserts "-include (file)" arguments for header
351// dependencies. It also inserts a "-w" option and a "-x c++",
352// if no other "-x" option is present.
353static ArgumentsAdjuster
354getModularizeArgumentsAdjuster(DependencyMap &Dependencies) {
355 return [&Dependencies](const CommandLineArguments &Args,
356 StringRef /*unused*/) {
357 std::string InputFile = findInputFile(Args);
358 DependentsVector &FileDependents = Dependencies[InputFile];
359 CommandLineArguments NewArgs(Args);
360 if (int Count = FileDependents.size()) {
361 for (int Index = 0; Index < Count; ++Index) {
362 NewArgs.push_back("-include");
363 std::string File(std::string("\"") + FileDependents[Index] +
364 std::string("\""));
365 NewArgs.push_back(FileDependents[Index]);
366 }
367 }
368 // Ignore warnings. (Insert after "clang_tool" at beginning.)
369 NewArgs.insert(NewArgs.begin() + 1, "-w");
370 // Since we are compiling .h files, assume C++ unless given a -x option.
371 if (!llvm::is_contained(NewArgs, "-x")) {
372 NewArgs.insert(NewArgs.begin() + 2, "-x");
373 NewArgs.insert(NewArgs.begin() + 3, "c++");
374 }
375 return NewArgs;
376 };
377}
378
379// FIXME: The Location class seems to be something that we might
380// want to design to be applicable to a wider range of tools, and stick it
381// somewhere into Tooling/ in mainline
382struct Location {
383 const FileEntry *File;
384 unsigned Line, Column;
385
386 Location() : File(), Line(), Column() {}
387
388 Location(SourceManager &SM, SourceLocation Loc) : File(), Line(), Column() {
389 Loc = SM.getExpansionLoc(Loc);
390 if (Loc.isInvalid())
391 return;
392
393 std::pair<FileID, unsigned> Decomposed = SM.getDecomposedLoc(Loc);
394 File = SM.getFileEntryForID(Decomposed.first);
395 if (!File)
396 return;
397
398 Line = SM.getLineNumber(Decomposed.first, Decomposed.second);
399 Column = SM.getColumnNumber(Decomposed.first, Decomposed.second);
400 }
401
402 operator bool() const { return File != nullptr; }
403
404 friend bool operator==(const Location &X, const Location &Y) {
405 return X.File == Y.File && X.Line == Y.Line && X.Column == Y.Column;
406 }
407
408 friend bool operator!=(const Location &X, const Location &Y) {
409 return !(X == Y);
410 }
411
412 friend bool operator<(const Location &X, const Location &Y) {
413 if (X.File != Y.File)
414 return X.File < Y.File;
415 if (X.Line != Y.Line)
416 return X.Line < Y.Line;
417 return X.Column < Y.Column;
418 }
419 friend bool operator>(const Location &X, const Location &Y) { return Y < X; }
420 friend bool operator<=(const Location &X, const Location &Y) {
421 return !(Y < X);
422 }
423 friend bool operator>=(const Location &X, const Location &Y) {
424 return !(X < Y);
425 }
426};
427
428struct Entry {
429 enum EntryKind {
430 EK_Tag,
431 EK_Value,
432 EK_Macro,
433
434 EK_NumberOfKinds
435 } Kind;
436
437 Location Loc;
438
439 StringRef getKindName() { return getKindName(Kind); }
440 static StringRef getKindName(EntryKind kind);
441};
442
443// Return a string representing the given kind.
444StringRef Entry::getKindName(Entry::EntryKind kind) {
445 switch (kind) {
446 case EK_Tag:
447 return "tag";
448 case EK_Value:
449 return "value";
450 case EK_Macro:
451 return "macro";
452 case EK_NumberOfKinds:
453 break;
454 }
455 llvm_unreachable("invalid Entry kind");
456}
457
458struct HeaderEntry {
459 std::string Name;
460 Location Loc;
461
462 friend bool operator==(const HeaderEntry &X, const HeaderEntry &Y) {
463 return X.Loc == Y.Loc && X.Name == Y.Name;
464 }
465 friend bool operator!=(const HeaderEntry &X, const HeaderEntry &Y) {
466 return !(X == Y);
467 }
468 friend bool operator<(const HeaderEntry &X, const HeaderEntry &Y) {
469 return X.Loc < Y.Loc || (X.Loc == Y.Loc && X.Name < Y.Name);
470 }
471 friend bool operator>(const HeaderEntry &X, const HeaderEntry &Y) {
472 return Y < X;
473 }
474 friend bool operator<=(const HeaderEntry &X, const HeaderEntry &Y) {
475 return !(Y < X);
476 }
477 friend bool operator>=(const HeaderEntry &X, const HeaderEntry &Y) {
478 return !(X < Y);
479 }
480};
481
482typedef std::vector<HeaderEntry> HeaderContents;
483
484class EntityMap : public std::map<std::string, SmallVector<Entry, 2>> {
485public:
486 DenseMap<const FileEntry *, HeaderContents> HeaderContentMismatches;
487
488 void add(const std::string &Name, enum Entry::EntryKind Kind, Location Loc) {
489 // Record this entity in its header.
490 HeaderEntry HE = { Name, Loc };
491 CurHeaderContents[Loc.File].push_back(HE);
492
493 // Check whether we've seen this entry before.
494 SmallVector<Entry, 2> &Entries = (*this)[Name];
495 for (unsigned I = 0, N = Entries.size(); I != N; ++I) {
496 if (Entries[I].Kind == Kind && Entries[I].Loc == Loc)
497 return;
498 }
499
500 // We have not seen this entry before; record it.
501 Entry E = { Kind, Loc };
502 Entries.push_back(E);
503 }
504
505 void mergeCurHeaderContents() {
506 for (DenseMap<const FileEntry *, HeaderContents>::iterator
507 H = CurHeaderContents.begin(),
508 HEnd = CurHeaderContents.end();
509 H != HEnd; ++H) {
510 // Sort contents.
511 llvm::sort(H->second);
512
513 // Check whether we've seen this header before.
514 DenseMap<const FileEntry *, HeaderContents>::iterator KnownH =
515 AllHeaderContents.find(H->first);
516 if (KnownH == AllHeaderContents.end()) {
517 // We haven't seen this header before; record its contents.
518 AllHeaderContents.insert(*H);
519 continue;
520 }
521
522 // If the header contents are the same, we're done.
523 if (H->second == KnownH->second)
524 continue;
525
526 // Determine what changed.
527 std::set_symmetric_difference(
528 H->second.begin(), H->second.end(), KnownH->second.begin(),
529 KnownH->second.end(),
530 std::back_inserter(HeaderContentMismatches[H->first]));
531 }
532
533 CurHeaderContents.clear();
534 }
535
536private:
537 DenseMap<const FileEntry *, HeaderContents> CurHeaderContents;
538 DenseMap<const FileEntry *, HeaderContents> AllHeaderContents;
539};
540
541class CollectEntitiesVisitor
542 : public RecursiveASTVisitor<CollectEntitiesVisitor> {
543public:
544 CollectEntitiesVisitor(SourceManager &SM, EntityMap &Entities,
545 Preprocessor &PP, PreprocessorTracker &PPTracker,
546 int &HadErrors)
547 : SM(SM), Entities(Entities), PP(PP), PPTracker(PPTracker),
548 HadErrors(HadErrors) {}
549
550 bool TraverseStmt(Stmt *S) { return true; }
551 bool TraverseType(QualType T) { return true; }
552 bool TraverseTypeLoc(TypeLoc TL) { return true; }
553 bool TraverseNestedNameSpecifier(NestedNameSpecifier *NNS) { return true; }
554 bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS) {
555 return true;
556 }
557 bool TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo) {
558 return true;
559 }
560 bool TraverseTemplateName(TemplateName Template) { return true; }
561 bool TraverseTemplateArgument(const TemplateArgument &Arg) { return true; }
562 bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &ArgLoc) {
563 return true;
564 }
565 bool TraverseTemplateArguments(ArrayRef<TemplateArgument>) { return true; }
566 bool TraverseConstructorInitializer(CXXCtorInitializer *Init) { return true; }
567 bool TraverseLambdaCapture(LambdaExpr *LE, const LambdaCapture *C,
568 Expr *Init) {
569 return true;
570 }
571
572 // Check 'extern "*" {}' block for #include directives.
573 bool VisitLinkageSpecDecl(LinkageSpecDecl *D) {
574 // Bail if not a block.
575 if (!D->hasBraces())
576 return true;
577 SourceRange BlockRange = D->getSourceRange();
578 const char *LinkageLabel;
579 switch (D->getLanguage()) {
580 case LinkageSpecDecl::lang_c:
581 LinkageLabel = "extern \"C\" {}";
582 break;
583 case LinkageSpecDecl::lang_cxx:
584 LinkageLabel = "extern \"C++\" {}";
585 break;
586 }
587 if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, LinkageLabel,
588 errs()))
589 HadErrors = 1;
590 return true;
591 }
592
593 // Check 'namespace (name) {}' block for #include directives.
594 bool VisitNamespaceDecl(const NamespaceDecl *D) {
595 SourceRange BlockRange = D->getSourceRange();
596 std::string Label("namespace ");
597 Label += D->getName();
598 Label += " {}";
599 if (!PPTracker.checkForIncludesInBlock(PP, BlockRange, Label.c_str(),
600 errs()))
601 HadErrors = 1;
602 return true;
603 }
604
605 // Collect definition entities.
606 bool VisitNamedDecl(NamedDecl *ND) {
607 // We only care about file-context variables.
608 if (!ND->getDeclContext()->isFileContext())
609 return true;
610
611 // Skip declarations that tend to be properly multiply-declared.
612 if (isa<NamespaceDecl>(ND) || isa<UsingDirectiveDecl>(ND) ||
613 isa<NamespaceAliasDecl>(ND) ||
614 isa<ClassTemplateSpecializationDecl>(ND) || isa<UsingDecl>(ND) ||
615 isa<ClassTemplateDecl>(ND) || isa<TemplateTypeParmDecl>(ND) ||
616 isa<TypeAliasTemplateDecl>(ND) || isa<UsingShadowDecl>(ND) ||
617 isa<FunctionDecl>(ND) || isa<FunctionTemplateDecl>(ND) ||
618 (isa<TagDecl>(ND) &&
619 !cast<TagDecl>(ND)->isThisDeclarationADefinition()))
620 return true;
621
622 // Skip anonymous declarations.
623 if (!ND->getDeclName())
624 return true;
625
626 // Get the qualified name.
627 std::string Name;
628 llvm::raw_string_ostream OS(Name);
629 ND->printQualifiedName(OS);
630 OS.flush();
631 if (Name.empty())
632 return true;
633
634 Location Loc(SM, ND->getLocation());
635 if (!Loc)
636 return true;
637
638 Entities.add(Name, isa<TagDecl>(ND) ? Entry::EK_Tag : Entry::EK_Value, Loc);
639 return true;
640 }
641
642private:
643 SourceManager &SM;
644 EntityMap &Entities;
645 Preprocessor &PP;
646 PreprocessorTracker &PPTracker;
647 int &HadErrors;
648};
649
650class CollectEntitiesConsumer : public ASTConsumer {
651public:
652 CollectEntitiesConsumer(EntityMap &Entities,
653 PreprocessorTracker &preprocessorTracker,
654 Preprocessor &PP, StringRef InFile, int &HadErrors)
655 : Entities(Entities), PPTracker(preprocessorTracker), PP(PP),
656 HadErrors(HadErrors) {
657 PPTracker.handlePreprocessorEntry(PP, InFile);
658 }
659
660 ~CollectEntitiesConsumer() override { PPTracker.handlePreprocessorExit(); }
661
662 void HandleTranslationUnit(ASTContext &Ctx) override {
663 SourceManager &SM = Ctx.getSourceManager();
664
665 // Collect declared entities.
666 CollectEntitiesVisitor(SM, Entities, PP, PPTracker, HadErrors)
667 .TraverseDecl(Ctx.getTranslationUnitDecl());
668
669 // Collect macro definitions.
670 for (Preprocessor::macro_iterator M = PP.macro_begin(),
671 MEnd = PP.macro_end();
672 M != MEnd; ++M) {
673 Location Loc(SM, M->second.getLatest()->getLocation());
674 if (!Loc)
675 continue;
676
677 Entities.add(M->first->getName().str(), Entry::EK_Macro, Loc);
678 }
679
680 // Merge header contents.
681 Entities.mergeCurHeaderContents();
682 }
683
684private:
685 EntityMap &Entities;
686 PreprocessorTracker &PPTracker;
687 Preprocessor &PP;
688 int &HadErrors;
689};
690
691class CollectEntitiesAction : public SyntaxOnlyAction {
692public:
693 CollectEntitiesAction(EntityMap &Entities,
694 PreprocessorTracker &preprocessorTracker,
695 int &HadErrors)
696 : Entities(Entities), PPTracker(preprocessorTracker),
697 HadErrors(HadErrors) {}
698
699protected:
700 std::unique_ptr<clang::ASTConsumer>
701 CreateASTConsumer(CompilerInstance &CI, StringRef InFile) override {
702 return std::make_unique<CollectEntitiesConsumer>(
703 Entities, PPTracker, CI.getPreprocessor(), InFile, HadErrors);
704 }
705
706private:
707 EntityMap &Entities;
708 PreprocessorTracker &PPTracker;
709 int &HadErrors;
710};
711
712class ModularizeFrontendActionFactory : public FrontendActionFactory {
713public:
714 ModularizeFrontendActionFactory(EntityMap &Entities,
715 PreprocessorTracker &preprocessorTracker,
716 int &HadErrors)
717 : Entities(Entities), PPTracker(preprocessorTracker),
718 HadErrors(HadErrors) {}
719
720 std::unique_ptr<FrontendAction> create() override {
721 return std::make_unique<CollectEntitiesAction>(Entities, PPTracker,
722 HadErrors);
723 }
724
725private:
726 EntityMap &Entities;
727 PreprocessorTracker &PPTracker;
728 int &HadErrors;
729};
730
731class CompileCheckVisitor
732 : public RecursiveASTVisitor<CompileCheckVisitor> {
733public:
734 CompileCheckVisitor() {}
735
736 bool TraverseStmt(Stmt *S) { return true; }
737 bool TraverseType(QualType T) { return true; }
738 bool TraverseTypeLoc(TypeLoc TL) { return true; }
739 bool TraverseNestedNameSpecifier(NestedNameSpecifier *NNS) { return true; }
740 bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNS) {
741 return true;
742 }
743 bool TraverseDeclarationNameInfo(DeclarationNameInfo NameInfo) {
744 return true;
745 }
746 bool TraverseTemplateName(TemplateName Template) { return true; }
747 bool TraverseTemplateArgument(const TemplateArgument &Arg) { return true; }
748 bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &ArgLoc) {
749 return true;
750 }
751 bool TraverseTemplateArguments(ArrayRef<TemplateArgument>) { return true; }
752 bool TraverseConstructorInitializer(CXXCtorInitializer *Init) { return true; }
753 bool TraverseLambdaCapture(LambdaExpr *LE, const LambdaCapture *C,
754 Expr *Init) {
755 return true;
756 }
757
758 // Check 'extern "*" {}' block for #include directives.
759 bool VisitLinkageSpecDecl(LinkageSpecDecl *D) {
760 return true;
761 }
762
763 // Check 'namespace (name) {}' block for #include directives.
764 bool VisitNamespaceDecl(const NamespaceDecl *D) {
765 return true;
766 }
767
768 // Collect definition entities.
769 bool VisitNamedDecl(NamedDecl *ND) {
770 return true;
771 }
772};
773
774class CompileCheckConsumer : public ASTConsumer {
775public:
776 CompileCheckConsumer() {}
777
778 void HandleTranslationUnit(ASTContext &Ctx) override {
779 CompileCheckVisitor().TraverseDecl(Ctx.getTranslationUnitDecl());
780 }
781};
782
783class CompileCheckAction : public SyntaxOnlyAction {
784public:
785 CompileCheckAction() {}
786
787protected:
788 std::unique_ptr<clang::ASTConsumer>
789 CreateASTConsumer(CompilerInstance &CI, StringRef InFile) override {
790 return std::make_unique<CompileCheckConsumer>();
791 }
792};
793
794class CompileCheckFrontendActionFactory : public FrontendActionFactory {
795public:
796 CompileCheckFrontendActionFactory() {}
797
798 std::unique_ptr<FrontendAction> create() override {
799 return std::make_unique<CompileCheckAction>();
800 }
801};
802
803int main(int Argc, const char **Argv) {
804
805 // Save program name for error messages.
806 Argv0 = Argv[0];
807
808 // Save program arguments for use in module.modulemap comment.
809 CommandLine = std::string(sys::path::stem(sys::path::filename(Argv0)));
810 for (int ArgIndex = 1; ArgIndex < Argc; ArgIndex++) {
811 CommandLine.append(" ");
812 CommandLine.append(Argv[ArgIndex]);
813 }
814
815 // This causes options to be parsed.
816 cl::ParseCommandLineOptions(Argc, Argv, "modularize.\n");
817
818 // No go if we have no header list file.
819 if (ListFileNames.size() == 0) {
820 cl::PrintHelpMessage();
821 return 1;
822 }
823
824 std::unique_ptr<ModularizeUtilities> ModUtil;
825 int HadErrors = 0;
826
827 ModUtil.reset(
828 ModularizeUtilities::createModularizeUtilities(
829 ListFileNames, HeaderPrefix, ProblemFilesList));
830
831 // Get header file names and dependencies.
832 if (ModUtil->loadAllHeaderListsAndDependencies())
833 HadErrors = 1;
834
835 // If we are in assistant mode, output the module map and quit.
836 if (ModuleMapPath.length() != 0) {
837 if (!createModuleMap(ModuleMapPath, ModUtil->HeaderFileNames,
838 ModUtil->ProblemFileNames,
839 ModUtil->Dependencies, HeaderPrefix, RootModule))
840 return 1; // Failed.
841 return 0; // Success - Skip checks in assistant mode.
842 }
843
844 // If we're doing module maps.
845 if (!NoCoverageCheck && ModUtil->HasModuleMap) {
846 // Do coverage check.
847 if (ModUtil->doCoverageCheck(IncludePaths, CommandLine))
848 HadErrors = 1;
849 }
850
851 // Bail early if only doing the coverage check.
852 if (CoverageCheckOnly)
853 return HadErrors;
854
855 // Create the compilation database.
856 SmallString<256> PathBuf;
857 sys::fs::current_path(PathBuf);
858 std::unique_ptr<CompilationDatabase> Compilations;
859 Compilations.reset(
860 new FixedCompilationDatabase(Twine(PathBuf), CC1Arguments));
861
862 // Create preprocessor tracker, to watch for macro and conditional problems.
863 std::unique_ptr<PreprocessorTracker> PPTracker(
864 PreprocessorTracker::create(ModUtil->HeaderFileNames,
865 BlockCheckHeaderListOnly));
866
867 // Coolect entities here.
868 EntityMap Entities;
869
870 // Because we can't easily determine which files failed
871 // during the tool run, if we're collecting the file lists
872 // for display, we do a first compile pass on individual
873 // files to find which ones don't compile stand-alone.
874 if (DisplayFileLists) {
875 // First, make a pass to just get compile errors.
876 for (auto &CompileCheckFile : ModUtil->HeaderFileNames) {
877 llvm::SmallVector<std::string, 32> CompileCheckFileArray;
878 CompileCheckFileArray.push_back(CompileCheckFile);
879 ClangTool CompileCheckTool(*Compilations, CompileCheckFileArray);
880 CompileCheckTool.appendArgumentsAdjuster(
881 getModularizeArgumentsAdjuster(ModUtil->Dependencies));
882 int CompileCheckFileErrors = 0;
883 // FIXME: use newFrontendActionFactory.
884 CompileCheckFrontendActionFactory CompileCheckFactory;
885 CompileCheckFileErrors |= CompileCheckTool.run(&CompileCheckFactory);
886 if (CompileCheckFileErrors != 0) {
887 ModUtil->addUniqueProblemFile(CompileCheckFile); // Save problem file.
888 HadErrors |= 1;
889 }
890 else
891 ModUtil->addNoCompileErrorsFile(CompileCheckFile); // Save good file.
892 }
893 }
894
895 // Then we make another pass on the good files to do the rest of the work.
896 ClangTool Tool(*Compilations,
897 (DisplayFileLists ? ModUtil->GoodFileNames : ModUtil->HeaderFileNames));
898 Tool.appendArgumentsAdjuster(
899 getModularizeArgumentsAdjuster(ModUtil->Dependencies));
900 ModularizeFrontendActionFactory Factory(Entities, *PPTracker, HadErrors);
901 HadErrors |= Tool.run(&Factory);
902
903 // Create a place to save duplicate entity locations, separate bins per kind.
904 typedef SmallVector<Location, 8> LocationArray;
905 typedef SmallVector<LocationArray, Entry::EK_NumberOfKinds> EntryBinArray;
906 EntryBinArray EntryBins;
907 int KindIndex;
908 for (KindIndex = 0; KindIndex < Entry::EK_NumberOfKinds; ++KindIndex) {
909 LocationArray Array;
910 EntryBins.push_back(Array);
911 }
912
913 // Check for the same entity being defined in multiple places.
914 for (EntityMap::iterator E = Entities.begin(), EEnd = Entities.end();
915 E != EEnd; ++E) {
916 // If only one occurrence, exit early.
917 if (E->second.size() == 1)
918 continue;
919 // Clear entity locations.
920 for (EntryBinArray::iterator CI = EntryBins.begin(), CE = EntryBins.end();
921 CI != CE; ++CI) {
922 CI->clear();
923 }
924 // Walk the entities of a single name, collecting the locations,
925 // separated into separate bins.
926 for (unsigned I = 0, N = E->second.size(); I != N; ++I) {
927 EntryBins[E->second[I].Kind].push_back(E->second[I].Loc);
928 }
929 // Report any duplicate entity definition errors.
930 int KindIndex = 0;
931 for (EntryBinArray::iterator DI = EntryBins.begin(), DE = EntryBins.end();
932 DI != DE; ++DI, ++KindIndex) {
933 int ECount = DI->size();
934 // If only 1 occurrence of this entity, skip it, we only report duplicates.
935 if (ECount <= 1)
936 continue;
937 LocationArray::iterator FI = DI->begin();
938 StringRef kindName = Entry::getKindName((Entry::EntryKind)KindIndex);
939 errs() << "error: " << kindName << " '" << E->first
940 << "' defined at multiple locations:\n";
941 for (LocationArray::iterator FE = DI->end(); FI != FE; ++FI) {
942 errs() << " " << FI->File->getName() << ":" << FI->Line << ":"
943 << FI->Column << "\n";
944 ModUtil->addUniqueProblemFile(std::string(FI->File->getName()));
945 }
946 HadErrors = 1;
947 }
948 }
949
950 // Complain about macro instance in header files that differ based on how
951 // they are included.
952 if (PPTracker->reportInconsistentMacros(errs()))
953 HadErrors = 1;
954
955 // Complain about preprocessor conditional directives in header files that
956 // differ based on how they are included.
957 if (PPTracker->reportInconsistentConditionals(errs()))
958 HadErrors = 1;
959
960 // Complain about any headers that have contents that differ based on how
961 // they are included.
962 // FIXME: Could we provide information about which preprocessor conditionals
963 // are involved?
964 for (DenseMap<const FileEntry *, HeaderContents>::iterator
965 H = Entities.HeaderContentMismatches.begin(),
966 HEnd = Entities.HeaderContentMismatches.end();
967 H != HEnd; ++H) {
968 if (H->second.empty()) {
969 errs() << "internal error: phantom header content mismatch\n";
970 continue;
971 }
972
973 HadErrors = 1;
974 ModUtil->addUniqueProblemFile(std::string(H->first->getName()));
975 errs() << "error: header '" << H->first->getName()
976 << "' has different contents depending on how it was included.\n";
977 for (unsigned I = 0, N = H->second.size(); I != N; ++I) {
978 errs() << "note: '" << H->second[I].Name << "' in "
979 << H->second[I].Loc.File->getName() << " at "
980 << H->second[I].Loc.Line << ":" << H->second[I].Loc.Column
981 << " not always provided\n";
982 }
983 }
984
985 if (DisplayFileLists) {
986 ModUtil->displayProblemFiles();
987 ModUtil->displayGoodFiles();
988 ModUtil->displayCombinedFiles();
989 }
990
991 return HadErrors;
992}
993