1 //== PrintfFormatString.cpp - Analysis of printf format strings --*- 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 // Handling of format string in printf and friends.  The structure of format
11 // strings for fprintf() are described in C99 7.19.6.1.
12 //
13 //===----------------------------------------------------------------------===//
14 
15 #include "clang/Analysis/Analyses/FormatString.h"
16 #include "FormatStringParsing.h"
17 #include "clang/Basic/TargetInfo.h"
18 
19 using clang::analyze_format_string::ArgType;
20 using clang::analyze_format_string::FormatStringHandler;
21 using clang::analyze_format_string::LengthModifier;
22 using clang::analyze_format_string::OptionalAmount;
23 using clang::analyze_format_string::ConversionSpecifier;
24 using clang::analyze_printf::PrintfSpecifier;
25 
26 using namespace clang;
27 
28 typedef clang::analyze_format_string::SpecifierResult<PrintfSpecifier>
29         PrintfSpecifierResult;
30 
31 //===----------------------------------------------------------------------===//
32 // Methods for parsing format strings.
33 //===----------------------------------------------------------------------===//
34 
35 using analyze_format_string::ParseNonPositionAmount;
36 
ParsePrecision(FormatStringHandler & H,PrintfSpecifier & FS,const char * Start,const char * & Beg,const char * E,unsigned * argIndex)37 static bool ParsePrecision(FormatStringHandler &H, PrintfSpecifier &FS,
38                            const char *Start, const char *&Beg, const char *E,
39                            unsigned *argIndex) {
40   if (argIndex) {
41     FS.setPrecision(ParseNonPositionAmount(Beg, E, *argIndex));
42   } else {
43     const OptionalAmount Amt = ParsePositionAmount(H, Start, Beg, E,
44                                            analyze_format_string::PrecisionPos);
45     if (Amt.isInvalid())
46       return true;
47     FS.setPrecision(Amt);
48   }
49   return false;
50 }
51 
ParseObjCFlags(FormatStringHandler & H,PrintfSpecifier & FS,const char * FlagBeg,const char * E,bool Warn)52 static bool ParseObjCFlags(FormatStringHandler &H, PrintfSpecifier &FS,
53                            const char *FlagBeg, const char *E, bool Warn) {
54    StringRef Flag(FlagBeg, E - FlagBeg);
55    // Currently there is only one flag.
56    if (Flag == "tt") {
57      FS.setHasObjCTechnicalTerm(FlagBeg);
58      return false;
59    }
60    // Handle either the case of no flag or an invalid flag.
61    if (Warn) {
62      if (Flag == "")
63        H.HandleEmptyObjCModifierFlag(FlagBeg, E  - FlagBeg);
64      else
65        H.HandleInvalidObjCModifierFlag(FlagBeg, E  - FlagBeg);
66    }
67    return true;
68 }
69 
ParsePrintfSpecifier(FormatStringHandler & H,const char * & Beg,const char * E,unsigned & argIndex,const LangOptions & LO,const TargetInfo & Target,bool Warn,bool isFreeBSDKPrintf)70 static PrintfSpecifierResult ParsePrintfSpecifier(FormatStringHandler &H,
71                                                   const char *&Beg,
72                                                   const char *E,
73                                                   unsigned &argIndex,
74                                                   const LangOptions &LO,
75                                                   const TargetInfo &Target,
76                                                   bool Warn,
77                                                   bool isFreeBSDKPrintf) {
78 
79   using namespace clang::analyze_format_string;
80   using namespace clang::analyze_printf;
81 
82   const char *I = Beg;
83   const char *Start = nullptr;
84   UpdateOnReturn <const char*> UpdateBeg(Beg, I);
85 
86   // Look for a '%' character that indicates the start of a format specifier.
87   for ( ; I != E ; ++I) {
88     char c = *I;
89     if (c == '\0') {
90       // Detect spurious null characters, which are likely errors.
91       H.HandleNullChar(I);
92       return true;
93     }
94     if (c == '%') {
95       Start = I++;  // Record the start of the format specifier.
96       break;
97     }
98   }
99 
100   // No format specifier found?
101   if (!Start)
102     return false;
103 
104   if (I == E) {
105     // No more characters left?
106     if (Warn)
107       H.HandleIncompleteSpecifier(Start, E - Start);
108     return true;
109   }
110 
111   PrintfSpecifier FS;
112   if (ParseArgPosition(H, FS, Start, I, E))
113     return true;
114 
115   if (I == E) {
116     // No more characters left?
117     if (Warn)
118       H.HandleIncompleteSpecifier(Start, E - Start);
119     return true;
120   }
121 
122   // Look for flags (if any).
123   bool hasMore = true;
124   for ( ; I != E; ++I) {
125     switch (*I) {
126       default: hasMore = false; break;
127       case '\'':
128         // FIXME: POSIX specific.  Always accept?
129         FS.setHasThousandsGrouping(I);
130         break;
131       case '-': FS.setIsLeftJustified(I); break;
132       case '+': FS.setHasPlusPrefix(I); break;
133       case ' ': FS.setHasSpacePrefix(I); break;
134       case '#': FS.setHasAlternativeForm(I); break;
135       case '0': FS.setHasLeadingZeros(I); break;
136     }
137     if (!hasMore)
138       break;
139   }
140 
141   if (I == E) {
142     // No more characters left?
143     if (Warn)
144       H.HandleIncompleteSpecifier(Start, E - Start);
145     return true;
146   }
147 
148   // Look for the field width (if any).
149   if (ParseFieldWidth(H, FS, Start, I, E,
150                       FS.usesPositionalArg() ? nullptr : &argIndex))
151     return true;
152 
153   if (I == E) {
154     // No more characters left?
155     if (Warn)
156       H.HandleIncompleteSpecifier(Start, E - Start);
157     return true;
158   }
159 
160   // Look for the precision (if any).
161   if (*I == '.') {
162     ++I;
163     if (I == E) {
164       if (Warn)
165         H.HandleIncompleteSpecifier(Start, E - Start);
166       return true;
167     }
168 
169     if (ParsePrecision(H, FS, Start, I, E,
170                        FS.usesPositionalArg() ? nullptr : &argIndex))
171       return true;
172 
173     if (I == E) {
174       // No more characters left?
175       if (Warn)
176         H.HandleIncompleteSpecifier(Start, E - Start);
177       return true;
178     }
179   }
180 
181   // Look for the length modifier.
182   if (ParseLengthModifier(FS, I, E, LO) && I == E) {
183     // No more characters left?
184     if (Warn)
185       H.HandleIncompleteSpecifier(Start, E - Start);
186     return true;
187   }
188 
189   // Look for the Objective-C modifier flags, if any.
190   // We parse these here, even if they don't apply to
191   // the conversion specifier, and then emit an error
192   // later if the conversion specifier isn't '@'.  This
193   // enables better recovery, and we don't know if
194   // these flags are applicable until later.
195   const char *ObjCModifierFlagsStart = nullptr,
196              *ObjCModifierFlagsEnd = nullptr;
197   if (*I == '[') {
198     ObjCModifierFlagsStart = I;
199     ++I;
200     auto flagStart = I;
201     for (;; ++I) {
202       ObjCModifierFlagsEnd = I;
203       if (I == E) {
204         if (Warn)
205           H.HandleIncompleteSpecifier(Start, E - Start);
206         return true;
207       }
208       // Did we find the closing ']'?
209       if (*I == ']') {
210         if (ParseObjCFlags(H, FS, flagStart, I, Warn))
211           return true;
212         ++I;
213         break;
214       }
215       // There are no separators defined yet for multiple
216       // Objective-C modifier flags.  When those are
217       // defined, this is the place to check.
218     }
219   }
220 
221   if (*I == '\0') {
222     // Detect spurious null characters, which are likely errors.
223     H.HandleNullChar(I);
224     return true;
225   }
226 
227   // Finally, look for the conversion specifier.
228   const char *conversionPosition = I++;
229   ConversionSpecifier::Kind k = ConversionSpecifier::InvalidSpecifier;
230   switch (*conversionPosition) {
231     default:
232       break;
233     // C99: 7.19.6.1 (section 8).
234     case '%': k = ConversionSpecifier::PercentArg;   break;
235     case 'A': k = ConversionSpecifier::AArg; break;
236     case 'E': k = ConversionSpecifier::EArg; break;
237     case 'F': k = ConversionSpecifier::FArg; break;
238     case 'G': k = ConversionSpecifier::GArg; break;
239     case 'X': k = ConversionSpecifier::XArg; break;
240     case 'a': k = ConversionSpecifier::aArg; break;
241     case 'c': k = ConversionSpecifier::cArg; break;
242     case 'd': k = ConversionSpecifier::dArg; break;
243     case 'e': k = ConversionSpecifier::eArg; break;
244     case 'f': k = ConversionSpecifier::fArg; break;
245     case 'g': k = ConversionSpecifier::gArg; break;
246     case 'i': k = ConversionSpecifier::iArg; break;
247     case 'n': k = ConversionSpecifier::nArg; break;
248     case 'o': k = ConversionSpecifier::oArg; break;
249     case 'p': k = ConversionSpecifier::pArg; break;
250     case 's': k = ConversionSpecifier::sArg; break;
251     case 'u': k = ConversionSpecifier::uArg; break;
252     case 'x': k = ConversionSpecifier::xArg; break;
253     // POSIX specific.
254     case 'C': k = ConversionSpecifier::CArg; break;
255     case 'S': k = ConversionSpecifier::SArg; break;
256     // Objective-C.
257     case '@': k = ConversionSpecifier::ObjCObjArg; break;
258     // Glibc specific.
259     case 'm': k = ConversionSpecifier::PrintErrno; break;
260     // FreeBSD kernel specific.
261     case 'b':
262       if (isFreeBSDKPrintf)
263         k = ConversionSpecifier::FreeBSDbArg; // int followed by char *
264       break;
265     case 'r':
266       if (isFreeBSDKPrintf)
267         k = ConversionSpecifier::FreeBSDrArg; // int
268       break;
269     case 'y':
270       if (isFreeBSDKPrintf)
271         k = ConversionSpecifier::FreeBSDyArg; // int
272       break;
273     // Apple-specific.
274     case 'D':
275       if (isFreeBSDKPrintf)
276         k = ConversionSpecifier::FreeBSDDArg; // void * followed by char *
277       else if (Target.getTriple().isOSDarwin())
278         k = ConversionSpecifier::DArg;
279       break;
280     case 'O':
281       if (Target.getTriple().isOSDarwin())
282         k = ConversionSpecifier::OArg;
283       break;
284     case 'U':
285       if (Target.getTriple().isOSDarwin())
286         k = ConversionSpecifier::UArg;
287       break;
288     // MS specific.
289     case 'Z':
290       if (Target.getTriple().isOSMSVCRT())
291         k = ConversionSpecifier::ZArg;
292   }
293 
294   // Check to see if we used the Objective-C modifier flags with
295   // a conversion specifier other than '@'.
296   if (k != ConversionSpecifier::ObjCObjArg &&
297       k != ConversionSpecifier::InvalidSpecifier &&
298       ObjCModifierFlagsStart) {
299     H.HandleObjCFlagsWithNonObjCConversion(ObjCModifierFlagsStart,
300                                            ObjCModifierFlagsEnd + 1,
301                                            conversionPosition);
302     return true;
303   }
304 
305   PrintfConversionSpecifier CS(conversionPosition, k);
306   FS.setConversionSpecifier(CS);
307   if (CS.consumesDataArgument() && !FS.usesPositionalArg())
308     FS.setArgIndex(argIndex++);
309   // FreeBSD kernel specific.
310   if (k == ConversionSpecifier::FreeBSDbArg ||
311       k == ConversionSpecifier::FreeBSDDArg)
312     argIndex++;
313 
314   if (k == ConversionSpecifier::InvalidSpecifier) {
315     // Assume the conversion takes one argument.
316     return !H.HandleInvalidPrintfConversionSpecifier(FS, Start, I - Start);
317   }
318   return PrintfSpecifierResult(Start, FS);
319 }
320 
ParsePrintfString(FormatStringHandler & H,const char * I,const char * E,const LangOptions & LO,const TargetInfo & Target,bool isFreeBSDKPrintf)321 bool clang::analyze_format_string::ParsePrintfString(FormatStringHandler &H,
322                                                      const char *I,
323                                                      const char *E,
324                                                      const LangOptions &LO,
325                                                      const TargetInfo &Target,
326                                                      bool isFreeBSDKPrintf) {
327 
328   unsigned argIndex = 0;
329 
330   // Keep looking for a format specifier until we have exhausted the string.
331   while (I != E) {
332     const PrintfSpecifierResult &FSR = ParsePrintfSpecifier(H, I, E, argIndex,
333                                                             LO, Target, true,
334                                                             isFreeBSDKPrintf);
335     // Did a fail-stop error of any kind occur when parsing the specifier?
336     // If so, don't do any more processing.
337     if (FSR.shouldStop())
338       return true;
339     // Did we exhaust the string or encounter an error that
340     // we can recover from?
341     if (!FSR.hasValue())
342       continue;
343     // We have a format specifier.  Pass it to the callback.
344     if (!H.HandlePrintfSpecifier(FSR.getValue(), FSR.getStart(),
345                                  I - FSR.getStart()))
346       return true;
347   }
348   assert(I == E && "Format string not exhausted");
349   return false;
350 }
351 
ParseFormatStringHasSArg(const char * I,const char * E,const LangOptions & LO,const TargetInfo & Target)352 bool clang::analyze_format_string::ParseFormatStringHasSArg(const char *I,
353                                                             const char *E,
354                                                             const LangOptions &LO,
355                                                             const TargetInfo &Target) {
356 
357   unsigned argIndex = 0;
358 
359   // Keep looking for a %s format specifier until we have exhausted the string.
360   FormatStringHandler H;
361   while (I != E) {
362     const PrintfSpecifierResult &FSR = ParsePrintfSpecifier(H, I, E, argIndex,
363                                                             LO, Target, false,
364                                                             false);
365     // Did a fail-stop error of any kind occur when parsing the specifier?
366     // If so, don't do any more processing.
367     if (FSR.shouldStop())
368       return false;
369     // Did we exhaust the string or encounter an error that
370     // we can recover from?
371     if (!FSR.hasValue())
372       continue;
373     const analyze_printf::PrintfSpecifier &FS = FSR.getValue();
374     // Return true if this a %s format specifier.
375     if (FS.getConversionSpecifier().getKind() == ConversionSpecifier::Kind::sArg)
376       return true;
377   }
378   return false;
379 }
380 
381 //===----------------------------------------------------------------------===//
382 // Methods on PrintfSpecifier.
383 //===----------------------------------------------------------------------===//
384 
getArgType(ASTContext & Ctx,bool IsObjCLiteral) const385 ArgType PrintfSpecifier::getArgType(ASTContext &Ctx,
386                                     bool IsObjCLiteral) const {
387   const PrintfConversionSpecifier &CS = getConversionSpecifier();
388 
389   if (!CS.consumesDataArgument())
390     return ArgType::Invalid();
391 
392   if (CS.getKind() == ConversionSpecifier::cArg)
393     switch (LM.getKind()) {
394       case LengthModifier::None:
395         return Ctx.IntTy;
396       case LengthModifier::AsLong:
397       case LengthModifier::AsWide:
398         return ArgType(ArgType::WIntTy, "wint_t");
399       case LengthModifier::AsShort:
400         if (Ctx.getTargetInfo().getTriple().isOSMSVCRT())
401           return Ctx.IntTy;
402       default:
403         return ArgType::Invalid();
404     }
405 
406   if (CS.isIntArg())
407     switch (LM.getKind()) {
408       case LengthModifier::AsLongDouble:
409         // GNU extension.
410         return Ctx.LongLongTy;
411       case LengthModifier::None:
412         return Ctx.IntTy;
413       case LengthModifier::AsInt32:
414         return ArgType(Ctx.IntTy, "__int32");
415       case LengthModifier::AsChar: return ArgType::AnyCharTy;
416       case LengthModifier::AsShort: return Ctx.ShortTy;
417       case LengthModifier::AsLong: return Ctx.LongTy;
418       case LengthModifier::AsLongLong:
419       case LengthModifier::AsQuad:
420         return Ctx.LongLongTy;
421       case LengthModifier::AsInt64:
422         return ArgType(Ctx.LongLongTy, "__int64");
423       case LengthModifier::AsIntMax:
424         return ArgType(Ctx.getIntMaxType(), "intmax_t");
425       case LengthModifier::AsSizeT:
426         // FIXME: How to get the corresponding signed version of size_t?
427         return ArgType();
428       case LengthModifier::AsInt3264:
429         return Ctx.getTargetInfo().getTriple().isArch64Bit()
430                    ? ArgType(Ctx.LongLongTy, "__int64")
431                    : ArgType(Ctx.IntTy, "__int32");
432       case LengthModifier::AsPtrDiff:
433         return ArgType(Ctx.getPointerDiffType(), "ptrdiff_t");
434       case LengthModifier::AsAllocate:
435       case LengthModifier::AsMAllocate:
436       case LengthModifier::AsWide:
437         return ArgType::Invalid();
438     }
439 
440   if (CS.isUIntArg())
441     switch (LM.getKind()) {
442       case LengthModifier::AsLongDouble:
443         // GNU extension.
444         return Ctx.UnsignedLongLongTy;
445       case LengthModifier::None:
446         return Ctx.UnsignedIntTy;
447       case LengthModifier::AsInt32:
448         return ArgType(Ctx.UnsignedIntTy, "unsigned __int32");
449       case LengthModifier::AsChar: return Ctx.UnsignedCharTy;
450       case LengthModifier::AsShort: return Ctx.UnsignedShortTy;
451       case LengthModifier::AsLong: return Ctx.UnsignedLongTy;
452       case LengthModifier::AsLongLong:
453       case LengthModifier::AsQuad:
454         return Ctx.UnsignedLongLongTy;
455       case LengthModifier::AsInt64:
456         return ArgType(Ctx.UnsignedLongLongTy, "unsigned __int64");
457       case LengthModifier::AsIntMax:
458         return ArgType(Ctx.getUIntMaxType(), "uintmax_t");
459       case LengthModifier::AsSizeT:
460         return ArgType(Ctx.getSizeType(), "size_t");
461       case LengthModifier::AsInt3264:
462         return Ctx.getTargetInfo().getTriple().isArch64Bit()
463                    ? ArgType(Ctx.UnsignedLongLongTy, "unsigned __int64")
464                    : ArgType(Ctx.UnsignedIntTy, "unsigned __int32");
465       case LengthModifier::AsPtrDiff:
466         // FIXME: How to get the corresponding unsigned
467         // version of ptrdiff_t?
468         return ArgType();
469       case LengthModifier::AsAllocate:
470       case LengthModifier::AsMAllocate:
471       case LengthModifier::AsWide:
472         return ArgType::Invalid();
473     }
474 
475   if (CS.isDoubleArg()) {
476     if (LM.getKind() == LengthModifier::AsLongDouble)
477       return Ctx.LongDoubleTy;
478     return Ctx.DoubleTy;
479   }
480 
481   if (CS.getKind() == ConversionSpecifier::nArg) {
482     switch (LM.getKind()) {
483       case LengthModifier::None:
484         return ArgType::PtrTo(Ctx.IntTy);
485       case LengthModifier::AsChar:
486         return ArgType::PtrTo(Ctx.SignedCharTy);
487       case LengthModifier::AsShort:
488         return ArgType::PtrTo(Ctx.ShortTy);
489       case LengthModifier::AsLong:
490         return ArgType::PtrTo(Ctx.LongTy);
491       case LengthModifier::AsLongLong:
492       case LengthModifier::AsQuad:
493         return ArgType::PtrTo(Ctx.LongLongTy);
494       case LengthModifier::AsIntMax:
495         return ArgType::PtrTo(ArgType(Ctx.getIntMaxType(), "intmax_t"));
496       case LengthModifier::AsSizeT:
497         return ArgType(); // FIXME: ssize_t
498       case LengthModifier::AsPtrDiff:
499         return ArgType::PtrTo(ArgType(Ctx.getPointerDiffType(), "ptrdiff_t"));
500       case LengthModifier::AsLongDouble:
501         return ArgType(); // FIXME: Is this a known extension?
502       case LengthModifier::AsAllocate:
503       case LengthModifier::AsMAllocate:
504       case LengthModifier::AsInt32:
505       case LengthModifier::AsInt3264:
506       case LengthModifier::AsInt64:
507       case LengthModifier::AsWide:
508         return ArgType::Invalid();
509     }
510   }
511 
512   switch (CS.getKind()) {
513     case ConversionSpecifier::sArg:
514       if (LM.getKind() == LengthModifier::AsWideChar) {
515         if (IsObjCLiteral)
516           return ArgType(Ctx.getPointerType(Ctx.UnsignedShortTy.withConst()),
517                          "const unichar *");
518         return ArgType(ArgType::WCStrTy, "wchar_t *");
519       }
520       if (LM.getKind() == LengthModifier::AsWide)
521         return ArgType(ArgType::WCStrTy, "wchar_t *");
522       return ArgType::CStrTy;
523     case ConversionSpecifier::SArg:
524       if (IsObjCLiteral)
525         return ArgType(Ctx.getPointerType(Ctx.UnsignedShortTy.withConst()),
526                        "const unichar *");
527       if (Ctx.getTargetInfo().getTriple().isOSMSVCRT() &&
528           LM.getKind() == LengthModifier::AsShort)
529         return ArgType::CStrTy;
530       return ArgType(ArgType::WCStrTy, "wchar_t *");
531     case ConversionSpecifier::CArg:
532       if (IsObjCLiteral)
533         return ArgType(Ctx.UnsignedShortTy, "unichar");
534       if (Ctx.getTargetInfo().getTriple().isOSMSVCRT() &&
535           LM.getKind() == LengthModifier::AsShort)
536         return Ctx.IntTy;
537       return ArgType(Ctx.WideCharTy, "wchar_t");
538     case ConversionSpecifier::pArg:
539       return ArgType::CPointerTy;
540     case ConversionSpecifier::ObjCObjArg:
541       return ArgType::ObjCPointerTy;
542     default:
543       break;
544   }
545 
546   // FIXME: Handle other cases.
547   return ArgType();
548 }
549 
fixType(QualType QT,const LangOptions & LangOpt,ASTContext & Ctx,bool IsObjCLiteral)550 bool PrintfSpecifier::fixType(QualType QT, const LangOptions &LangOpt,
551                               ASTContext &Ctx, bool IsObjCLiteral) {
552   // %n is different from other conversion specifiers; don't try to fix it.
553   if (CS.getKind() == ConversionSpecifier::nArg)
554     return false;
555 
556   // Handle Objective-C objects first. Note that while the '%@' specifier will
557   // not warn for structure pointer or void pointer arguments (because that's
558   // how CoreFoundation objects are implemented), we only show a fixit for '%@'
559   // if we know it's an object (block, id, class, or __attribute__((NSObject))).
560   if (QT->isObjCRetainableType()) {
561     if (!IsObjCLiteral)
562       return false;
563 
564     CS.setKind(ConversionSpecifier::ObjCObjArg);
565 
566     // Disable irrelevant flags
567     HasThousandsGrouping = false;
568     HasPlusPrefix = false;
569     HasSpacePrefix = false;
570     HasAlternativeForm = false;
571     HasLeadingZeroes = false;
572     Precision.setHowSpecified(OptionalAmount::NotSpecified);
573     LM.setKind(LengthModifier::None);
574 
575     return true;
576   }
577 
578   // Handle strings next (char *, wchar_t *)
579   if (QT->isPointerType() && (QT->getPointeeType()->isAnyCharacterType())) {
580     CS.setKind(ConversionSpecifier::sArg);
581 
582     // Disable irrelevant flags
583     HasAlternativeForm = 0;
584     HasLeadingZeroes = 0;
585 
586     // Set the long length modifier for wide characters
587     if (QT->getPointeeType()->isWideCharType())
588       LM.setKind(LengthModifier::AsWideChar);
589     else
590       LM.setKind(LengthModifier::None);
591 
592     return true;
593   }
594 
595   // If it's an enum, get its underlying type.
596   if (const EnumType *ETy = QT->getAs<EnumType>())
597     QT = ETy->getDecl()->getIntegerType();
598 
599   // We can only work with builtin types.
600   const BuiltinType *BT = QT->getAs<BuiltinType>();
601   if (!BT)
602     return false;
603 
604   // Set length modifier
605   switch (BT->getKind()) {
606   case BuiltinType::Bool:
607   case BuiltinType::WChar_U:
608   case BuiltinType::WChar_S:
609   case BuiltinType::Char16:
610   case BuiltinType::Char32:
611   case BuiltinType::UInt128:
612   case BuiltinType::Int128:
613   case BuiltinType::Half:
614     // Various types which are non-trivial to correct.
615     return false;
616 
617 #define SIGNED_TYPE(Id, SingletonId)
618 #define UNSIGNED_TYPE(Id, SingletonId)
619 #define FLOATING_TYPE(Id, SingletonId)
620 #define BUILTIN_TYPE(Id, SingletonId) \
621   case BuiltinType::Id:
622 #include "clang/AST/BuiltinTypes.def"
623     // Misc other stuff which doesn't make sense here.
624     return false;
625 
626   case BuiltinType::UInt:
627   case BuiltinType::Int:
628   case BuiltinType::Float:
629   case BuiltinType::Double:
630     LM.setKind(LengthModifier::None);
631     break;
632 
633   case BuiltinType::Char_U:
634   case BuiltinType::UChar:
635   case BuiltinType::Char_S:
636   case BuiltinType::SChar:
637     LM.setKind(LengthModifier::AsChar);
638     break;
639 
640   case BuiltinType::Short:
641   case BuiltinType::UShort:
642     LM.setKind(LengthModifier::AsShort);
643     break;
644 
645   case BuiltinType::Long:
646   case BuiltinType::ULong:
647     LM.setKind(LengthModifier::AsLong);
648     break;
649 
650   case BuiltinType::LongLong:
651   case BuiltinType::ULongLong:
652     LM.setKind(LengthModifier::AsLongLong);
653     break;
654 
655   case BuiltinType::LongDouble:
656     LM.setKind(LengthModifier::AsLongDouble);
657     break;
658   }
659 
660   // Handle size_t, ptrdiff_t, etc. that have dedicated length modifiers in C99.
661   if (isa<TypedefType>(QT) && (LangOpt.C99 || LangOpt.CPlusPlus11))
662     namedTypeToLengthModifier(QT, LM);
663 
664   // If fixing the length modifier was enough, we might be done.
665   if (hasValidLengthModifier(Ctx.getTargetInfo())) {
666     // If we're going to offer a fix anyway, make sure the sign matches.
667     switch (CS.getKind()) {
668     case ConversionSpecifier::uArg:
669     case ConversionSpecifier::UArg:
670       if (QT->isSignedIntegerType())
671         CS.setKind(clang::analyze_format_string::ConversionSpecifier::dArg);
672       break;
673     case ConversionSpecifier::dArg:
674     case ConversionSpecifier::DArg:
675     case ConversionSpecifier::iArg:
676       if (QT->isUnsignedIntegerType() && !HasPlusPrefix)
677         CS.setKind(clang::analyze_format_string::ConversionSpecifier::uArg);
678       break;
679     default:
680       // Other specifiers do not have signed/unsigned variants.
681       break;
682     }
683 
684     const analyze_printf::ArgType &ATR = getArgType(Ctx, IsObjCLiteral);
685     if (ATR.isValid() && ATR.matchesType(Ctx, QT))
686       return true;
687   }
688 
689   // Set conversion specifier and disable any flags which do not apply to it.
690   // Let typedefs to char fall through to int, as %c is silly for uint8_t.
691   if (!isa<TypedefType>(QT) && QT->isCharType()) {
692     CS.setKind(ConversionSpecifier::cArg);
693     LM.setKind(LengthModifier::None);
694     Precision.setHowSpecified(OptionalAmount::NotSpecified);
695     HasAlternativeForm = 0;
696     HasLeadingZeroes = 0;
697     HasPlusPrefix = 0;
698   }
699   // Test for Floating type first as LongDouble can pass isUnsignedIntegerType
700   else if (QT->isRealFloatingType()) {
701     CS.setKind(ConversionSpecifier::fArg);
702   }
703   else if (QT->isSignedIntegerType()) {
704     CS.setKind(ConversionSpecifier::dArg);
705     HasAlternativeForm = 0;
706   }
707   else if (QT->isUnsignedIntegerType()) {
708     CS.setKind(ConversionSpecifier::uArg);
709     HasAlternativeForm = 0;
710     HasPlusPrefix = 0;
711   } else {
712     llvm_unreachable("Unexpected type");
713   }
714 
715   return true;
716 }
717 
toString(raw_ostream & os) const718 void PrintfSpecifier::toString(raw_ostream &os) const {
719   // Whilst some features have no defined order, we are using the order
720   // appearing in the C99 standard (ISO/IEC 9899:1999 (E) 7.19.6.1)
721   os << "%";
722 
723   // Positional args
724   if (usesPositionalArg()) {
725     os << getPositionalArgIndex() << "$";
726   }
727 
728   // Conversion flags
729   if (IsLeftJustified)    os << "-";
730   if (HasPlusPrefix)      os << "+";
731   if (HasSpacePrefix)     os << " ";
732   if (HasAlternativeForm) os << "#";
733   if (HasLeadingZeroes)   os << "0";
734 
735   // Minimum field width
736   FieldWidth.toString(os);
737   // Precision
738   Precision.toString(os);
739   // Length modifier
740   os << LM.toString();
741   // Conversion specifier
742   os << CS.toString();
743 }
744 
hasValidPlusPrefix() const745 bool PrintfSpecifier::hasValidPlusPrefix() const {
746   if (!HasPlusPrefix)
747     return true;
748 
749   // The plus prefix only makes sense for signed conversions
750   switch (CS.getKind()) {
751   case ConversionSpecifier::dArg:
752   case ConversionSpecifier::DArg:
753   case ConversionSpecifier::iArg:
754   case ConversionSpecifier::fArg:
755   case ConversionSpecifier::FArg:
756   case ConversionSpecifier::eArg:
757   case ConversionSpecifier::EArg:
758   case ConversionSpecifier::gArg:
759   case ConversionSpecifier::GArg:
760   case ConversionSpecifier::aArg:
761   case ConversionSpecifier::AArg:
762   case ConversionSpecifier::FreeBSDrArg:
763   case ConversionSpecifier::FreeBSDyArg:
764     return true;
765 
766   default:
767     return false;
768   }
769 }
770 
hasValidAlternativeForm() const771 bool PrintfSpecifier::hasValidAlternativeForm() const {
772   if (!HasAlternativeForm)
773     return true;
774 
775   // Alternate form flag only valid with the oxXaAeEfFgG conversions
776   switch (CS.getKind()) {
777   case ConversionSpecifier::oArg:
778   case ConversionSpecifier::OArg:
779   case ConversionSpecifier::xArg:
780   case ConversionSpecifier::XArg:
781   case ConversionSpecifier::aArg:
782   case ConversionSpecifier::AArg:
783   case ConversionSpecifier::eArg:
784   case ConversionSpecifier::EArg:
785   case ConversionSpecifier::fArg:
786   case ConversionSpecifier::FArg:
787   case ConversionSpecifier::gArg:
788   case ConversionSpecifier::GArg:
789   case ConversionSpecifier::FreeBSDrArg:
790   case ConversionSpecifier::FreeBSDyArg:
791     return true;
792 
793   default:
794     return false;
795   }
796 }
797 
hasValidLeadingZeros() const798 bool PrintfSpecifier::hasValidLeadingZeros() const {
799   if (!HasLeadingZeroes)
800     return true;
801 
802   // Leading zeroes flag only valid with the diouxXaAeEfFgG conversions
803   switch (CS.getKind()) {
804   case ConversionSpecifier::dArg:
805   case ConversionSpecifier::DArg:
806   case ConversionSpecifier::iArg:
807   case ConversionSpecifier::oArg:
808   case ConversionSpecifier::OArg:
809   case ConversionSpecifier::uArg:
810   case ConversionSpecifier::UArg:
811   case ConversionSpecifier::xArg:
812   case ConversionSpecifier::XArg:
813   case ConversionSpecifier::aArg:
814   case ConversionSpecifier::AArg:
815   case ConversionSpecifier::eArg:
816   case ConversionSpecifier::EArg:
817   case ConversionSpecifier::fArg:
818   case ConversionSpecifier::FArg:
819   case ConversionSpecifier::gArg:
820   case ConversionSpecifier::GArg:
821   case ConversionSpecifier::FreeBSDrArg:
822   case ConversionSpecifier::FreeBSDyArg:
823     return true;
824 
825   default:
826     return false;
827   }
828 }
829 
hasValidSpacePrefix() const830 bool PrintfSpecifier::hasValidSpacePrefix() const {
831   if (!HasSpacePrefix)
832     return true;
833 
834   // The space prefix only makes sense for signed conversions
835   switch (CS.getKind()) {
836   case ConversionSpecifier::dArg:
837   case ConversionSpecifier::DArg:
838   case ConversionSpecifier::iArg:
839   case ConversionSpecifier::fArg:
840   case ConversionSpecifier::FArg:
841   case ConversionSpecifier::eArg:
842   case ConversionSpecifier::EArg:
843   case ConversionSpecifier::gArg:
844   case ConversionSpecifier::GArg:
845   case ConversionSpecifier::aArg:
846   case ConversionSpecifier::AArg:
847   case ConversionSpecifier::FreeBSDrArg:
848   case ConversionSpecifier::FreeBSDyArg:
849     return true;
850 
851   default:
852     return false;
853   }
854 }
855 
hasValidLeftJustified() const856 bool PrintfSpecifier::hasValidLeftJustified() const {
857   if (!IsLeftJustified)
858     return true;
859 
860   // The left justified flag is valid for all conversions except n
861   switch (CS.getKind()) {
862   case ConversionSpecifier::nArg:
863     return false;
864 
865   default:
866     return true;
867   }
868 }
869 
hasValidThousandsGroupingPrefix() const870 bool PrintfSpecifier::hasValidThousandsGroupingPrefix() const {
871   if (!HasThousandsGrouping)
872     return true;
873 
874   switch (CS.getKind()) {
875     case ConversionSpecifier::dArg:
876     case ConversionSpecifier::DArg:
877     case ConversionSpecifier::iArg:
878     case ConversionSpecifier::uArg:
879     case ConversionSpecifier::UArg:
880     case ConversionSpecifier::fArg:
881     case ConversionSpecifier::FArg:
882     case ConversionSpecifier::gArg:
883     case ConversionSpecifier::GArg:
884       return true;
885     default:
886       return false;
887   }
888 }
889 
hasValidPrecision() const890 bool PrintfSpecifier::hasValidPrecision() const {
891   if (Precision.getHowSpecified() == OptionalAmount::NotSpecified)
892     return true;
893 
894   // Precision is only valid with the diouxXaAeEfFgGs conversions
895   switch (CS.getKind()) {
896   case ConversionSpecifier::dArg:
897   case ConversionSpecifier::DArg:
898   case ConversionSpecifier::iArg:
899   case ConversionSpecifier::oArg:
900   case ConversionSpecifier::OArg:
901   case ConversionSpecifier::uArg:
902   case ConversionSpecifier::UArg:
903   case ConversionSpecifier::xArg:
904   case ConversionSpecifier::XArg:
905   case ConversionSpecifier::aArg:
906   case ConversionSpecifier::AArg:
907   case ConversionSpecifier::eArg:
908   case ConversionSpecifier::EArg:
909   case ConversionSpecifier::fArg:
910   case ConversionSpecifier::FArg:
911   case ConversionSpecifier::gArg:
912   case ConversionSpecifier::GArg:
913   case ConversionSpecifier::sArg:
914   case ConversionSpecifier::FreeBSDrArg:
915   case ConversionSpecifier::FreeBSDyArg:
916     return true;
917 
918   default:
919     return false;
920   }
921 }
hasValidFieldWidth() const922 bool PrintfSpecifier::hasValidFieldWidth() const {
923   if (FieldWidth.getHowSpecified() == OptionalAmount::NotSpecified)
924       return true;
925 
926   // The field width is valid for all conversions except n
927   switch (CS.getKind()) {
928   case ConversionSpecifier::nArg:
929     return false;
930 
931   default:
932     return true;
933   }
934 }
935