1 //===--- PlistDiagnostics.cpp - Plist Diagnostics for Paths -----*- C++ -*-===//
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 defines the PlistDiagnostics object.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/Basic/FileManager.h"
15 #include "clang/Basic/PlistSupport.h"
16 #include "clang/Basic/SourceManager.h"
17 #include "clang/Basic/Version.h"
18 #include "clang/Lex/Preprocessor.h"
19 #include "clang/Rewrite/Core/HTMLRewrite.h"
20 #include "clang/StaticAnalyzer/Core/AnalyzerOptions.h"
21 #include "clang/StaticAnalyzer/Core/BugReporter/PathDiagnostic.h"
22 #include "clang/StaticAnalyzer/Core/IssueHash.h"
23 #include "clang/StaticAnalyzer/Core/PathDiagnosticConsumers.h"
24 #include "llvm/ADT/Statistic.h"
25 #include "llvm/ADT/SmallVector.h"
26 #include "llvm/Support/Casting.h"
27 using namespace clang;
28 using namespace ento;
29 using namespace markup;
30 
31 namespace {
32   class PlistDiagnostics : public PathDiagnosticConsumer {
33     const std::string OutputFile;
34     const LangOptions &LangOpts;
35     const bool SupportsCrossFileDiagnostics;
36     const bool SerializeStatistics;
37   public:
38     PlistDiagnostics(AnalyzerOptions &AnalyzerOpts,
39                      const std::string& prefix,
40                      const LangOptions &LangOpts,
41                      bool supportsMultipleFiles);
42 
43     ~PlistDiagnostics() override {}
44 
45     void FlushDiagnosticsImpl(std::vector<const PathDiagnostic *> &Diags,
46                               FilesMade *filesMade) override;
47 
48     StringRef getName() const override {
49       return "PlistDiagnostics";
50     }
51 
52     PathGenerationScheme getGenerationScheme() const override {
53       return Extensive;
54     }
55     bool supportsLogicalOpControlFlow() const override { return true; }
56     bool supportsCrossFileDiagnostics() const override {
57       return SupportsCrossFileDiagnostics;
58     }
59   };
60 } // end anonymous namespace
61 
62 PlistDiagnostics::PlistDiagnostics(AnalyzerOptions &AnalyzerOpts,
63                                    const std::string& output,
64                                    const LangOptions &LO,
65                                    bool supportsMultipleFiles)
66   : OutputFile(output),
67     LangOpts(LO),
68     SupportsCrossFileDiagnostics(supportsMultipleFiles),
69     SerializeStatistics(AnalyzerOpts.shouldSerializeStats()) {}
70 
71 void ento::createPlistDiagnosticConsumer(AnalyzerOptions &AnalyzerOpts,
72                                          PathDiagnosticConsumers &C,
73                                          const std::string& s,
74                                          const Preprocessor &PP) {
75   C.push_back(new PlistDiagnostics(AnalyzerOpts, s,
76                                    PP.getLangOpts(), false));
77 }
78 
79 void ento::createPlistMultiFileDiagnosticConsumer(AnalyzerOptions &AnalyzerOpts,
80                                                   PathDiagnosticConsumers &C,
81                                                   const std::string &s,
82                                                   const Preprocessor &PP) {
83   C.push_back(new PlistDiagnostics(AnalyzerOpts, s,
84                                    PP.getLangOpts(), true));
85 }
86 
87 static void ReportControlFlow(raw_ostream &o,
88                               const PathDiagnosticControlFlowPiece& P,
89                               const FIDMap& FM,
90                               const SourceManager &SM,
91                               const LangOptions &LangOpts,
92                               unsigned indent) {
93 
94   Indent(o, indent) << "<dict>\n";
95   ++indent;
96 
97   Indent(o, indent) << "<key>kind</key><string>control</string>\n";
98 
99   // Emit edges.
100   Indent(o, indent) << "<key>edges</key>\n";
101   ++indent;
102   Indent(o, indent) << "<array>\n";
103   ++indent;
104   for (PathDiagnosticControlFlowPiece::const_iterator I=P.begin(), E=P.end();
105        I!=E; ++I) {
106     Indent(o, indent) << "<dict>\n";
107     ++indent;
108 
109     // Make the ranges of the start and end point self-consistent with adjacent edges
110     // by forcing to use only the beginning of the range.  This simplifies the layout
111     // logic for clients.
112     Indent(o, indent) << "<key>start</key>\n";
113     SourceRange StartEdge(
114         SM.getExpansionLoc(I->getStart().asRange().getBegin()));
115     EmitRange(o, SM, Lexer::getAsCharRange(StartEdge, SM, LangOpts), FM,
116               indent + 1);
117 
118     Indent(o, indent) << "<key>end</key>\n";
119     SourceRange EndEdge(SM.getExpansionLoc(I->getEnd().asRange().getBegin()));
120     EmitRange(o, SM, Lexer::getAsCharRange(EndEdge, SM, LangOpts), FM,
121               indent + 1);
122 
123     --indent;
124     Indent(o, indent) << "</dict>\n";
125   }
126   --indent;
127   Indent(o, indent) << "</array>\n";
128   --indent;
129 
130   // Output any helper text.
131   const auto &s = P.getString();
132   if (!s.empty()) {
133     Indent(o, indent) << "<key>alternate</key>";
134     EmitString(o, s) << '\n';
135   }
136 
137   --indent;
138   Indent(o, indent) << "</dict>\n";
139 }
140 
141 static void ReportEvent(raw_ostream &o, const PathDiagnosticPiece& P,
142                         const FIDMap& FM,
143                         const SourceManager &SM,
144                         const LangOptions &LangOpts,
145                         unsigned indent,
146                         unsigned depth,
147                         bool isKeyEvent = false) {
148 
149   Indent(o, indent) << "<dict>\n";
150   ++indent;
151 
152   Indent(o, indent) << "<key>kind</key><string>event</string>\n";
153 
154   if (isKeyEvent) {
155     Indent(o, indent) << "<key>key_event</key><true/>\n";
156   }
157 
158   // Output the location.
159   FullSourceLoc L = P.getLocation().asLocation();
160 
161   Indent(o, indent) << "<key>location</key>\n";
162   EmitLocation(o, SM, L, FM, indent);
163 
164   // Output the ranges (if any).
165   ArrayRef<SourceRange> Ranges = P.getRanges();
166 
167   if (!Ranges.empty()) {
168     Indent(o, indent) << "<key>ranges</key>\n";
169     Indent(o, indent) << "<array>\n";
170     ++indent;
171     for (auto &R : Ranges)
172       EmitRange(o, SM,
173                 Lexer::getAsCharRange(SM.getExpansionRange(R), SM, LangOpts),
174                 FM, indent + 1);
175     --indent;
176     Indent(o, indent) << "</array>\n";
177   }
178 
179   // Output the call depth.
180   Indent(o, indent) << "<key>depth</key>";
181   EmitInteger(o, depth) << '\n';
182 
183   // Output the text.
184   assert(!P.getString().empty());
185   Indent(o, indent) << "<key>extended_message</key>\n";
186   Indent(o, indent);
187   EmitString(o, P.getString()) << '\n';
188 
189   // Output the short text.
190   // FIXME: Really use a short string.
191   Indent(o, indent) << "<key>message</key>\n";
192   Indent(o, indent);
193   EmitString(o, P.getString()) << '\n';
194 
195   // Finish up.
196   --indent;
197   Indent(o, indent); o << "</dict>\n";
198 }
199 
200 static void ReportPiece(raw_ostream &o,
201                         const PathDiagnosticPiece &P,
202                         const FIDMap& FM, const SourceManager &SM,
203                         const LangOptions &LangOpts,
204                         unsigned indent,
205                         unsigned depth,
206                         bool includeControlFlow,
207                         bool isKeyEvent = false);
208 
209 static void ReportCall(raw_ostream &o,
210                        const PathDiagnosticCallPiece &P,
211                        const FIDMap& FM, const SourceManager &SM,
212                        const LangOptions &LangOpts,
213                        unsigned indent,
214                        unsigned depth) {
215 
216   if (auto callEnter = P.getCallEnterEvent())
217     ReportPiece(o, *callEnter, FM, SM, LangOpts, indent, depth, true,
218                 P.isLastInMainSourceFile());
219 
220 
221   ++depth;
222 
223   if (auto callEnterWithinCaller = P.getCallEnterWithinCallerEvent())
224     ReportPiece(o, *callEnterWithinCaller, FM, SM, LangOpts,
225                 indent, depth, true);
226 
227   for (PathPieces::const_iterator I = P.path.begin(), E = P.path.end();I!=E;++I)
228     ReportPiece(o, **I, FM, SM, LangOpts, indent, depth, true);
229 
230   --depth;
231 
232   if (auto callExit = P.getCallExitEvent())
233     ReportPiece(o, *callExit, FM, SM, LangOpts, indent, depth, true);
234 }
235 
236 static void ReportMacro(raw_ostream &o,
237                         const PathDiagnosticMacroPiece& P,
238                         const FIDMap& FM, const SourceManager &SM,
239                         const LangOptions &LangOpts,
240                         unsigned indent,
241                         unsigned depth) {
242 
243   for (PathPieces::const_iterator I = P.subPieces.begin(), E=P.subPieces.end();
244        I!=E; ++I) {
245     ReportPiece(o, **I, FM, SM, LangOpts, indent, depth, false);
246   }
247 }
248 
249 static void ReportDiag(raw_ostream &o, const PathDiagnosticPiece& P,
250                        const FIDMap& FM, const SourceManager &SM,
251                        const LangOptions &LangOpts) {
252   ReportPiece(o, P, FM, SM, LangOpts, 4, 0, true);
253 }
254 
255 static void ReportPiece(raw_ostream &o,
256                         const PathDiagnosticPiece &P,
257                         const FIDMap& FM, const SourceManager &SM,
258                         const LangOptions &LangOpts,
259                         unsigned indent,
260                         unsigned depth,
261                         bool includeControlFlow,
262                         bool isKeyEvent) {
263   switch (P.getKind()) {
264     case PathDiagnosticPiece::ControlFlow:
265       if (includeControlFlow)
266         ReportControlFlow(o, cast<PathDiagnosticControlFlowPiece>(P), FM, SM,
267                           LangOpts, indent);
268       break;
269     case PathDiagnosticPiece::Call:
270       ReportCall(o, cast<PathDiagnosticCallPiece>(P), FM, SM, LangOpts,
271                  indent, depth);
272       break;
273     case PathDiagnosticPiece::Event:
274       ReportEvent(o, cast<PathDiagnosticSpotPiece>(P), FM, SM, LangOpts,
275                   indent, depth, isKeyEvent);
276       break;
277     case PathDiagnosticPiece::Macro:
278       ReportMacro(o, cast<PathDiagnosticMacroPiece>(P), FM, SM, LangOpts,
279                   indent, depth);
280       break;
281     case PathDiagnosticPiece::Note:
282       // FIXME: Extend the plist format to support those.
283       break;
284   }
285 }
286 
287 void PlistDiagnostics::FlushDiagnosticsImpl(
288                                     std::vector<const PathDiagnostic *> &Diags,
289                                     FilesMade *filesMade) {
290   // Build up a set of FIDs that we use by scanning the locations and
291   // ranges of the diagnostics.
292   FIDMap FM;
293   SmallVector<FileID, 10> Fids;
294   const SourceManager* SM = nullptr;
295 
296   if (!Diags.empty())
297     SM = &Diags.front()->path.front()->getLocation().getManager();
298 
299   auto AddPieceFID = [&FM, &Fids, SM](const PathDiagnosticPiece &Piece) {
300     AddFID(FM, Fids, *SM, Piece.getLocation().asLocation());
301     ArrayRef<SourceRange> Ranges = Piece.getRanges();
302     for (const SourceRange &Range : Ranges) {
303       AddFID(FM, Fids, *SM, Range.getBegin());
304       AddFID(FM, Fids, *SM, Range.getEnd());
305     }
306   };
307 
308   for (const PathDiagnostic *D : Diags) {
309 
310     SmallVector<const PathPieces *, 5> WorkList;
311     WorkList.push_back(&D->path);
312 
313     while (!WorkList.empty()) {
314       const PathPieces &Path = *WorkList.pop_back_val();
315 
316       for (const auto &Iter : Path) {
317         const PathDiagnosticPiece &Piece = *Iter;
318         AddPieceFID(Piece);
319 
320         if (const PathDiagnosticCallPiece *Call =
321                 dyn_cast<PathDiagnosticCallPiece>(&Piece)) {
322           if (auto CallEnterWithin = Call->getCallEnterWithinCallerEvent())
323             AddPieceFID(*CallEnterWithin);
324 
325           if (auto CallEnterEvent = Call->getCallEnterEvent())
326             AddPieceFID(*CallEnterEvent);
327 
328           WorkList.push_back(&Call->path);
329         } else if (const PathDiagnosticMacroPiece *Macro =
330                        dyn_cast<PathDiagnosticMacroPiece>(&Piece)) {
331           WorkList.push_back(&Macro->subPieces);
332         }
333       }
334     }
335   }
336 
337   // Open the file.
338   std::error_code EC;
339   llvm::raw_fd_ostream o(OutputFile, EC, llvm::sys::fs::F_Text);
340   if (EC) {
341     llvm::errs() << "warning: could not create file: " << EC.message() << '\n';
342     return;
343   }
344 
345   EmitPlistHeader(o);
346 
347   // Write the root object: a <dict> containing...
348   //  - "clang_version", the string representation of clang version
349   //  - "files", an <array> mapping from FIDs to file names
350   //  - "diagnostics", an <array> containing the path diagnostics
351   o << "<dict>\n" <<
352        " <key>clang_version</key>\n";
353   EmitString(o, getClangFullVersion()) << '\n';
354   o << " <key>files</key>\n"
355        " <array>\n";
356 
357   for (FileID FID : Fids)
358     EmitString(o << "  ", SM->getFileEntryForID(FID)->getName()) << '\n';
359 
360   o << " </array>\n"
361        " <key>diagnostics</key>\n"
362        " <array>\n";
363 
364   for (std::vector<const PathDiagnostic*>::iterator DI=Diags.begin(),
365        DE = Diags.end(); DI!=DE; ++DI) {
366 
367     o << "  <dict>\n"
368          "   <key>path</key>\n";
369 
370     const PathDiagnostic *D = *DI;
371 
372     o << "   <array>\n";
373 
374     for (PathPieces::const_iterator I = D->path.begin(), E = D->path.end();
375          I != E; ++I)
376       ReportDiag(o, **I, FM, *SM, LangOpts);
377 
378     o << "   </array>\n";
379 
380     // Output the bug type and bug category.
381     o << "   <key>description</key>";
382     EmitString(o, D->getShortDescription()) << '\n';
383     o << "   <key>category</key>";
384     EmitString(o, D->getCategory()) << '\n';
385     o << "   <key>type</key>";
386     EmitString(o, D->getBugType()) << '\n';
387     o << "   <key>check_name</key>";
388     EmitString(o, D->getCheckName()) << '\n';
389 
390     o << "   <!-- This hash is experimental and going to change! -->\n";
391     o << "   <key>issue_hash_content_of_line_in_context</key>";
392     PathDiagnosticLocation UPDLoc = D->getUniqueingLoc();
393     FullSourceLoc L(SM->getExpansionLoc(UPDLoc.isValid()
394                                             ? UPDLoc.asLocation()
395                                             : D->getLocation().asLocation()),
396                     *SM);
397     const Decl *DeclWithIssue = D->getDeclWithIssue();
398     EmitString(o, GetIssueHash(*SM, L, D->getCheckName(), D->getBugType(),
399                                DeclWithIssue, LangOpts))
400         << '\n';
401 
402     // Output information about the semantic context where
403     // the issue occurred.
404     if (const Decl *DeclWithIssue = D->getDeclWithIssue()) {
405       // FIXME: handle blocks, which have no name.
406       if (const NamedDecl *ND = dyn_cast<NamedDecl>(DeclWithIssue)) {
407         StringRef declKind;
408         switch (ND->getKind()) {
409           case Decl::CXXRecord:
410             declKind = "C++ class";
411             break;
412           case Decl::CXXMethod:
413             declKind = "C++ method";
414             break;
415           case Decl::ObjCMethod:
416             declKind = "Objective-C method";
417             break;
418           case Decl::Function:
419             declKind = "function";
420             break;
421           default:
422             break;
423         }
424         if (!declKind.empty()) {
425           const std::string &declName = ND->getDeclName().getAsString();
426           o << "  <key>issue_context_kind</key>";
427           EmitString(o, declKind) << '\n';
428           o << "  <key>issue_context</key>";
429           EmitString(o, declName) << '\n';
430         }
431 
432         // Output the bug hash for issue unique-ing. Currently, it's just an
433         // offset from the beginning of the function.
434         if (const Stmt *Body = DeclWithIssue->getBody()) {
435 
436           // If the bug uniqueing location exists, use it for the hash.
437           // For example, this ensures that two leaks reported on the same line
438           // will have different issue_hashes and that the hash will identify
439           // the leak location even after code is added between the allocation
440           // site and the end of scope (leak report location).
441           if (UPDLoc.isValid()) {
442             FullSourceLoc UFunL(SM->getExpansionLoc(
443               D->getUniqueingDecl()->getBody()->getLocStart()), *SM);
444             o << "  <key>issue_hash_function_offset</key><string>"
445               << L.getExpansionLineNumber() - UFunL.getExpansionLineNumber()
446               << "</string>\n";
447 
448           // Otherwise, use the location on which the bug is reported.
449           } else {
450             FullSourceLoc FunL(SM->getExpansionLoc(Body->getLocStart()), *SM);
451             o << "  <key>issue_hash_function_offset</key><string>"
452               << L.getExpansionLineNumber() - FunL.getExpansionLineNumber()
453               << "</string>\n";
454           }
455 
456         }
457       }
458     }
459 
460     // Output the location of the bug.
461     o << "  <key>location</key>\n";
462     EmitLocation(o, *SM, D->getLocation().asLocation(), FM, 2);
463 
464     // Output the diagnostic to the sub-diagnostic client, if any.
465     if (!filesMade->empty()) {
466       StringRef lastName;
467       PDFileEntry::ConsumerFiles *files = filesMade->getFiles(*D);
468       if (files) {
469         for (PDFileEntry::ConsumerFiles::const_iterator CI = files->begin(),
470                 CE = files->end(); CI != CE; ++CI) {
471           StringRef newName = CI->first;
472           if (newName != lastName) {
473             if (!lastName.empty()) {
474               o << "  </array>\n";
475             }
476             lastName = newName;
477             o <<  "  <key>" << lastName << "_files</key>\n";
478             o << "  <array>\n";
479           }
480           o << "   <string>" << CI->second << "</string>\n";
481         }
482         o << "  </array>\n";
483       }
484     }
485 
486     // Close up the entry.
487     o << "  </dict>\n";
488   }
489 
490   o << " </array>\n";
491 
492   if (llvm::AreStatisticsEnabled() && SerializeStatistics) {
493     o << " <key>statistics</key>\n";
494     std::string stats;
495     llvm::raw_string_ostream os(stats);
496     llvm::PrintStatisticsJSON(os);
497     os.flush();
498     EmitString(o, html::EscapeText(stats)) << '\n';
499   }
500 
501   // Finish.
502   o << "</dict>\n</plist>";
503 }
504