1 // FormatString.cpp - Common stuff for handling printf/scanf formats -*- C++ -*-
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 // Shared details for processing format strings of printf and scanf
10 // (and friends).
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "FormatStringParsing.h"
15 #include "clang/Basic/LangOptions.h"
16 #include "clang/Basic/TargetInfo.h"
17 #include "llvm/Support/ConvertUTF.h"
18 
19 using clang::analyze_format_string::ArgType;
20 using clang::analyze_format_string::FormatStringHandler;
21 using clang::analyze_format_string::FormatSpecifier;
22 using clang::analyze_format_string::LengthModifier;
23 using clang::analyze_format_string::OptionalAmount;
24 using clang::analyze_format_string::PositionContext;
25 using clang::analyze_format_string::ConversionSpecifier;
26 using namespace clang;
27 
28 // Key function to FormatStringHandler.
~FormatStringHandler()29 FormatStringHandler::~FormatStringHandler() {}
30 
31 //===----------------------------------------------------------------------===//
32 // Functions for parsing format strings components in both printf and
33 // scanf format strings.
34 //===----------------------------------------------------------------------===//
35 
36 OptionalAmount
ParseAmount(const char * & Beg,const char * E)37 clang::analyze_format_string::ParseAmount(const char *&Beg, const char *E) {
38   const char *I = Beg;
39   UpdateOnReturn <const char*> UpdateBeg(Beg, I);
40 
41   unsigned accumulator = 0;
42   bool hasDigits = false;
43 
44   for ( ; I != E; ++I) {
45     char c = *I;
46     if (c >= '0' && c <= '9') {
47       hasDigits = true;
48       accumulator = (accumulator * 10) + (c - '0');
49       continue;
50     }
51 
52     if (hasDigits)
53       return OptionalAmount(OptionalAmount::Constant, accumulator, Beg, I - Beg,
54           false);
55 
56     break;
57   }
58 
59   return OptionalAmount();
60 }
61 
62 OptionalAmount
ParseNonPositionAmount(const char * & Beg,const char * E,unsigned & argIndex)63 clang::analyze_format_string::ParseNonPositionAmount(const char *&Beg,
64                                                      const char *E,
65                                                      unsigned &argIndex) {
66   if (*Beg == '*') {
67     ++Beg;
68     return OptionalAmount(OptionalAmount::Arg, argIndex++, Beg, 0, false);
69   }
70 
71   return ParseAmount(Beg, E);
72 }
73 
74 OptionalAmount
ParsePositionAmount(FormatStringHandler & H,const char * Start,const char * & Beg,const char * E,PositionContext p)75 clang::analyze_format_string::ParsePositionAmount(FormatStringHandler &H,
76                                                   const char *Start,
77                                                   const char *&Beg,
78                                                   const char *E,
79                                                   PositionContext p) {
80   if (*Beg == '*') {
81     const char *I = Beg + 1;
82     const OptionalAmount &Amt = ParseAmount(I, E);
83 
84     if (Amt.getHowSpecified() == OptionalAmount::NotSpecified) {
85       H.HandleInvalidPosition(Beg, I - Beg, p);
86       return OptionalAmount(false);
87     }
88 
89     if (I == E) {
90       // No more characters left?
91       H.HandleIncompleteSpecifier(Start, E - Start);
92       return OptionalAmount(false);
93     }
94 
95     assert(Amt.getHowSpecified() == OptionalAmount::Constant);
96 
97     if (*I == '$') {
98       // Handle positional arguments
99 
100       // Special case: '*0$', since this is an easy mistake.
101       if (Amt.getConstantAmount() == 0) {
102         H.HandleZeroPosition(Beg, I - Beg + 1);
103         return OptionalAmount(false);
104       }
105 
106       const char *Tmp = Beg;
107       Beg = ++I;
108 
109       return OptionalAmount(OptionalAmount::Arg, Amt.getConstantAmount() - 1,
110                             Tmp, 0, true);
111     }
112 
113     H.HandleInvalidPosition(Beg, I - Beg, p);
114     return OptionalAmount(false);
115   }
116 
117   return ParseAmount(Beg, E);
118 }
119 
120 
121 bool
ParseFieldWidth(FormatStringHandler & H,FormatSpecifier & CS,const char * Start,const char * & Beg,const char * E,unsigned * argIndex)122 clang::analyze_format_string::ParseFieldWidth(FormatStringHandler &H,
123                                               FormatSpecifier &CS,
124                                               const char *Start,
125                                               const char *&Beg, const char *E,
126                                               unsigned *argIndex) {
127   // FIXME: Support negative field widths.
128   if (argIndex) {
129     CS.setFieldWidth(ParseNonPositionAmount(Beg, E, *argIndex));
130   }
131   else {
132     const OptionalAmount Amt =
133       ParsePositionAmount(H, Start, Beg, E,
134                           analyze_format_string::FieldWidthPos);
135 
136     if (Amt.isInvalid())
137       return true;
138     CS.setFieldWidth(Amt);
139   }
140   return false;
141 }
142 
143 bool
ParseArgPosition(FormatStringHandler & H,FormatSpecifier & FS,const char * Start,const char * & Beg,const char * E)144 clang::analyze_format_string::ParseArgPosition(FormatStringHandler &H,
145                                                FormatSpecifier &FS,
146                                                const char *Start,
147                                                const char *&Beg,
148                                                const char *E) {
149   const char *I = Beg;
150 
151   const OptionalAmount &Amt = ParseAmount(I, E);
152 
153   if (I == E) {
154     // No more characters left?
155     H.HandleIncompleteSpecifier(Start, E - Start);
156     return true;
157   }
158 
159   if (Amt.getHowSpecified() == OptionalAmount::Constant && *(I++) == '$') {
160     // Warn that positional arguments are non-standard.
161     H.HandlePosition(Start, I - Start);
162 
163     // Special case: '%0$', since this is an easy mistake.
164     if (Amt.getConstantAmount() == 0) {
165       H.HandleZeroPosition(Start, I - Start);
166       return true;
167     }
168 
169     FS.setArgIndex(Amt.getConstantAmount() - 1);
170     FS.setUsesPositionalArg();
171     // Update the caller's pointer if we decided to consume
172     // these characters.
173     Beg = I;
174     return false;
175   }
176 
177   return false;
178 }
179 
180 bool
ParseVectorModifier(FormatStringHandler & H,FormatSpecifier & FS,const char * & I,const char * E,const LangOptions & LO)181 clang::analyze_format_string::ParseVectorModifier(FormatStringHandler &H,
182                                                   FormatSpecifier &FS,
183                                                   const char *&I,
184                                                   const char *E,
185                                                   const LangOptions &LO) {
186   if (!LO.OpenCL)
187     return false;
188 
189   const char *Start = I;
190   if (*I == 'v') {
191     ++I;
192 
193     if (I == E) {
194       H.HandleIncompleteSpecifier(Start, E - Start);
195       return true;
196     }
197 
198     OptionalAmount NumElts = ParseAmount(I, E);
199     if (NumElts.getHowSpecified() != OptionalAmount::Constant) {
200       H.HandleIncompleteSpecifier(Start, E - Start);
201       return true;
202     }
203 
204     FS.setVectorNumElts(NumElts);
205   }
206 
207   return false;
208 }
209 
210 bool
ParseLengthModifier(FormatSpecifier & FS,const char * & I,const char * E,const LangOptions & LO,bool IsScanf)211 clang::analyze_format_string::ParseLengthModifier(FormatSpecifier &FS,
212                                                   const char *&I,
213                                                   const char *E,
214                                                   const LangOptions &LO,
215                                                   bool IsScanf) {
216   LengthModifier::Kind lmKind = LengthModifier::None;
217   const char *lmPosition = I;
218   switch (*I) {
219     default:
220       return false;
221     case 'h':
222       ++I;
223       if (I != E && *I == 'h') {
224         ++I;
225         lmKind = LengthModifier::AsChar;
226       } else if (I != E && *I == 'l' && LO.OpenCL) {
227         ++I;
228         lmKind = LengthModifier::AsShortLong;
229       } else {
230         lmKind = LengthModifier::AsShort;
231       }
232       break;
233     case 'l':
234       ++I;
235       if (I != E && *I == 'l') {
236         ++I;
237         lmKind = LengthModifier::AsLongLong;
238       } else {
239         lmKind = LengthModifier::AsLong;
240       }
241       break;
242     case 'j': lmKind = LengthModifier::AsIntMax;     ++I; break;
243     case 'z': lmKind = LengthModifier::AsSizeT;      ++I; break;
244     case 't': lmKind = LengthModifier::AsPtrDiff;    ++I; break;
245     case 'L': lmKind = LengthModifier::AsLongDouble; ++I; break;
246     case 'q': lmKind = LengthModifier::AsQuad;       ++I; break;
247     case 'a':
248       if (IsScanf && !LO.C99 && !LO.CPlusPlus11) {
249         // For scanf in C90, look at the next character to see if this should
250         // be parsed as the GNU extension 'a' length modifier. If not, this
251         // will be parsed as a conversion specifier.
252         ++I;
253         if (I != E && (*I == 's' || *I == 'S' || *I == '[')) {
254           lmKind = LengthModifier::AsAllocate;
255           break;
256         }
257         --I;
258       }
259       return false;
260     case 'm':
261       if (IsScanf) {
262         lmKind = LengthModifier::AsMAllocate;
263         ++I;
264         break;
265       }
266       return false;
267     // printf: AsInt64, AsInt32, AsInt3264
268     // scanf:  AsInt64
269     case 'I':
270       if (I + 1 != E && I + 2 != E) {
271         if (I[1] == '6' && I[2] == '4') {
272           I += 3;
273           lmKind = LengthModifier::AsInt64;
274           break;
275         }
276         if (IsScanf)
277           return false;
278 
279         if (I[1] == '3' && I[2] == '2') {
280           I += 3;
281           lmKind = LengthModifier::AsInt32;
282           break;
283         }
284       }
285       ++I;
286       lmKind = LengthModifier::AsInt3264;
287       break;
288     case 'w':
289       lmKind = LengthModifier::AsWide; ++I; break;
290   }
291   LengthModifier lm(lmPosition, lmKind);
292   FS.setLengthModifier(lm);
293   return true;
294 }
295 
ParseUTF8InvalidSpecifier(const char * SpecifierBegin,const char * FmtStrEnd,unsigned & Len)296 bool clang::analyze_format_string::ParseUTF8InvalidSpecifier(
297     const char *SpecifierBegin, const char *FmtStrEnd, unsigned &Len) {
298   if (SpecifierBegin + 1 >= FmtStrEnd)
299     return false;
300 
301   const llvm::UTF8 *SB =
302       reinterpret_cast<const llvm::UTF8 *>(SpecifierBegin + 1);
303   const llvm::UTF8 *SE = reinterpret_cast<const llvm::UTF8 *>(FmtStrEnd);
304   const char FirstByte = *SB;
305 
306   // If the invalid specifier is a multibyte UTF-8 string, return the
307   // total length accordingly so that the conversion specifier can be
308   // properly updated to reflect a complete UTF-8 specifier.
309   unsigned NumBytes = llvm::getNumBytesForUTF8(FirstByte);
310   if (NumBytes == 1)
311     return false;
312   if (SB + NumBytes > SE)
313     return false;
314 
315   Len = NumBytes + 1;
316   return true;
317 }
318 
319 //===----------------------------------------------------------------------===//
320 // Methods on ArgType.
321 //===----------------------------------------------------------------------===//
322 
323 clang::analyze_format_string::ArgType::MatchKind
matchesType(ASTContext & C,QualType argTy) const324 ArgType::matchesType(ASTContext &C, QualType argTy) const {
325   if (Ptr) {
326     // It has to be a pointer.
327     const PointerType *PT = argTy->getAs<PointerType>();
328     if (!PT)
329       return NoMatch;
330 
331     // We cannot write through a const qualified pointer.
332     if (PT->getPointeeType().isConstQualified())
333       return NoMatch;
334 
335     argTy = PT->getPointeeType();
336   }
337 
338   switch (K) {
339     case InvalidTy:
340       llvm_unreachable("ArgType must be valid");
341 
342     case UnknownTy:
343       return Match;
344 
345     case AnyCharTy: {
346       if (const EnumType *ETy = argTy->getAs<EnumType>()) {
347         // If the enum is incomplete we know nothing about the underlying type.
348         // Assume that it's 'int'.
349         if (!ETy->getDecl()->isComplete())
350           return NoMatch;
351         argTy = ETy->getDecl()->getIntegerType();
352       }
353 
354       if (const BuiltinType *BT = argTy->getAs<BuiltinType>())
355         switch (BT->getKind()) {
356           default:
357             break;
358           case BuiltinType::Char_S:
359           case BuiltinType::SChar:
360           case BuiltinType::UChar:
361           case BuiltinType::Char_U:
362           case BuiltinType::Bool:
363             return Match;
364         }
365       return NoMatch;
366     }
367 
368     case SpecificTy: {
369       if (const EnumType *ETy = argTy->getAs<EnumType>()) {
370         // If the enum is incomplete we know nothing about the underlying type.
371         // Assume that it's 'int'.
372         if (!ETy->getDecl()->isComplete())
373           argTy = C.IntTy;
374         else
375           argTy = ETy->getDecl()->getIntegerType();
376       }
377       argTy = C.getCanonicalType(argTy).getUnqualifiedType();
378 
379       if (T == argTy)
380         return Match;
381       // Check for "compatible types".
382       if (const BuiltinType *BT = argTy->getAs<BuiltinType>())
383         switch (BT->getKind()) {
384           default:
385             break;
386           case BuiltinType::Char_S:
387           case BuiltinType::SChar:
388           case BuiltinType::Char_U:
389           case BuiltinType::UChar:
390           case BuiltinType::Bool:
391             if (T == C.UnsignedShortTy || T == C.ShortTy)
392               return NoMatchTypeConfusion;
393             return T == C.UnsignedCharTy || T == C.SignedCharTy ? Match
394                                                                 : NoMatch;
395           case BuiltinType::Short:
396             return T == C.UnsignedShortTy ? Match : NoMatch;
397           case BuiltinType::UShort:
398             return T == C.ShortTy ? Match : NoMatch;
399           case BuiltinType::Int:
400             return T == C.UnsignedIntTy ? Match : NoMatch;
401           case BuiltinType::UInt:
402             return T == C.IntTy ? Match : NoMatch;
403           case BuiltinType::Long:
404             return T == C.UnsignedLongTy ? Match : NoMatch;
405           case BuiltinType::ULong:
406             return T == C.LongTy ? Match : NoMatch;
407           case BuiltinType::LongLong:
408             return T == C.UnsignedLongLongTy ? Match : NoMatch;
409           case BuiltinType::ULongLong:
410             return T == C.LongLongTy ? Match : NoMatch;
411         }
412       return NoMatch;
413     }
414 
415     case CStrTy: {
416       const PointerType *PT = argTy->getAs<PointerType>();
417       if (!PT)
418         return NoMatch;
419       QualType pointeeTy = PT->getPointeeType();
420       if (const BuiltinType *BT = pointeeTy->getAs<BuiltinType>())
421         switch (BT->getKind()) {
422           case BuiltinType::Char_U:
423           case BuiltinType::UChar:
424           case BuiltinType::Char_S:
425           case BuiltinType::SChar:
426             return Match;
427           default:
428             break;
429         }
430 
431       return NoMatch;
432     }
433 
434     case WCStrTy: {
435       const PointerType *PT = argTy->getAs<PointerType>();
436       if (!PT)
437         return NoMatch;
438       QualType pointeeTy =
439         C.getCanonicalType(PT->getPointeeType()).getUnqualifiedType();
440       return pointeeTy == C.getWideCharType() ? Match : NoMatch;
441     }
442 
443     case WIntTy: {
444       QualType WInt = C.getCanonicalType(C.getWIntType()).getUnqualifiedType();
445 
446       if (C.getCanonicalType(argTy).getUnqualifiedType() == WInt)
447         return Match;
448 
449       QualType PromoArg = argTy->isPromotableIntegerType()
450                               ? C.getPromotedIntegerType(argTy)
451                               : argTy;
452       PromoArg = C.getCanonicalType(PromoArg).getUnqualifiedType();
453 
454       // If the promoted argument is the corresponding signed type of the
455       // wint_t type, then it should match.
456       if (PromoArg->hasSignedIntegerRepresentation() &&
457           C.getCorrespondingUnsignedType(PromoArg) == WInt)
458         return Match;
459 
460       return WInt == PromoArg ? Match : NoMatch;
461     }
462 
463     case CPointerTy:
464       if (argTy->isVoidPointerType()) {
465         return Match;
466       } if (argTy->isPointerType() || argTy->isObjCObjectPointerType() ||
467             argTy->isBlockPointerType() || argTy->isNullPtrType()) {
468         return NoMatchPedantic;
469       } else {
470         return NoMatch;
471       }
472 
473     case ObjCPointerTy: {
474       if (argTy->getAs<ObjCObjectPointerType>() ||
475           argTy->getAs<BlockPointerType>())
476         return Match;
477 
478       // Handle implicit toll-free bridging.
479       if (const PointerType *PT = argTy->getAs<PointerType>()) {
480         // Things such as CFTypeRef are really just opaque pointers
481         // to C structs representing CF types that can often be bridged
482         // to Objective-C objects.  Since the compiler doesn't know which
483         // structs can be toll-free bridged, we just accept them all.
484         QualType pointee = PT->getPointeeType();
485         if (pointee->getAsStructureType() || pointee->isVoidType())
486           return Match;
487       }
488       return NoMatch;
489     }
490   }
491 
492   llvm_unreachable("Invalid ArgType Kind!");
493 }
494 
makeVectorType(ASTContext & C,unsigned NumElts) const495 ArgType ArgType::makeVectorType(ASTContext &C, unsigned NumElts) const {
496   // Check for valid vector element types.
497   if (T.isNull())
498     return ArgType::Invalid();
499 
500   QualType Vec = C.getExtVectorType(T, NumElts);
501   return ArgType(Vec, Name);
502 }
503 
getRepresentativeType(ASTContext & C) const504 QualType ArgType::getRepresentativeType(ASTContext &C) const {
505   QualType Res;
506   switch (K) {
507     case InvalidTy:
508       llvm_unreachable("No representative type for Invalid ArgType");
509     case UnknownTy:
510       llvm_unreachable("No representative type for Unknown ArgType");
511     case AnyCharTy:
512       Res = C.CharTy;
513       break;
514     case SpecificTy:
515       Res = T;
516       break;
517     case CStrTy:
518       Res = C.getPointerType(C.CharTy);
519       break;
520     case WCStrTy:
521       Res = C.getPointerType(C.getWideCharType());
522       break;
523     case ObjCPointerTy:
524       Res = C.ObjCBuiltinIdTy;
525       break;
526     case CPointerTy:
527       Res = C.VoidPtrTy;
528       break;
529     case WIntTy: {
530       Res = C.getWIntType();
531       break;
532     }
533   }
534 
535   if (Ptr)
536     Res = C.getPointerType(Res);
537   return Res;
538 }
539 
getRepresentativeTypeName(ASTContext & C) const540 std::string ArgType::getRepresentativeTypeName(ASTContext &C) const {
541   std::string S = getRepresentativeType(C).getAsString(C.getPrintingPolicy());
542 
543   std::string Alias;
544   if (Name) {
545     // Use a specific name for this type, e.g. "size_t".
546     Alias = Name;
547     if (Ptr) {
548       // If ArgType is actually a pointer to T, append an asterisk.
549       Alias += (Alias[Alias.size()-1] == '*') ? "*" : " *";
550     }
551     // If Alias is the same as the underlying type, e.g. wchar_t, then drop it.
552     if (S == Alias)
553       Alias.clear();
554   }
555 
556   if (!Alias.empty())
557     return std::string("'") + Alias + "' (aka '" + S + "')";
558   return std::string("'") + S + "'";
559 }
560 
561 
562 //===----------------------------------------------------------------------===//
563 // Methods on OptionalAmount.
564 //===----------------------------------------------------------------------===//
565 
566 ArgType
getArgType(ASTContext & Ctx) const567 analyze_format_string::OptionalAmount::getArgType(ASTContext &Ctx) const {
568   return Ctx.IntTy;
569 }
570 
571 //===----------------------------------------------------------------------===//
572 // Methods on LengthModifier.
573 //===----------------------------------------------------------------------===//
574 
575 const char *
toString() const576 analyze_format_string::LengthModifier::toString() const {
577   switch (kind) {
578   case AsChar:
579     return "hh";
580   case AsShort:
581     return "h";
582   case AsShortLong:
583     return "hl";
584   case AsLong: // or AsWideChar
585     return "l";
586   case AsLongLong:
587     return "ll";
588   case AsQuad:
589     return "q";
590   case AsIntMax:
591     return "j";
592   case AsSizeT:
593     return "z";
594   case AsPtrDiff:
595     return "t";
596   case AsInt32:
597     return "I32";
598   case AsInt3264:
599     return "I";
600   case AsInt64:
601     return "I64";
602   case AsLongDouble:
603     return "L";
604   case AsAllocate:
605     return "a";
606   case AsMAllocate:
607     return "m";
608   case AsWide:
609     return "w";
610   case None:
611     return "";
612   }
613   return nullptr;
614 }
615 
616 //===----------------------------------------------------------------------===//
617 // Methods on ConversionSpecifier.
618 //===----------------------------------------------------------------------===//
619 
toString() const620 const char *ConversionSpecifier::toString() const {
621   switch (kind) {
622   case dArg: return "d";
623   case DArg: return "D";
624   case iArg: return "i";
625   case oArg: return "o";
626   case OArg: return "O";
627   case uArg: return "u";
628   case UArg: return "U";
629   case xArg: return "x";
630   case XArg: return "X";
631   case fArg: return "f";
632   case FArg: return "F";
633   case eArg: return "e";
634   case EArg: return "E";
635   case gArg: return "g";
636   case GArg: return "G";
637   case aArg: return "a";
638   case AArg: return "A";
639   case cArg: return "c";
640   case sArg: return "s";
641   case pArg: return "p";
642   case PArg:
643     return "P";
644   case nArg: return "n";
645   case PercentArg:  return "%";
646   case ScanListArg: return "[";
647   case InvalidSpecifier: return nullptr;
648 
649   // POSIX unicode extensions.
650   case CArg: return "C";
651   case SArg: return "S";
652 
653   // Objective-C specific specifiers.
654   case ObjCObjArg: return "@";
655 
656   // FreeBSD kernel specific specifiers.
657   case FreeBSDbArg: return "b";
658   case FreeBSDDArg: return "D";
659   case FreeBSDrArg: return "r";
660   case FreeBSDyArg: return "y";
661 
662   // GlibC specific specifiers.
663   case PrintErrno: return "m";
664 
665   // MS specific specifiers.
666   case ZArg: return "Z";
667   }
668   return nullptr;
669 }
670 
671 Optional<ConversionSpecifier>
getStandardSpecifier() const672 ConversionSpecifier::getStandardSpecifier() const {
673   ConversionSpecifier::Kind NewKind;
674 
675   switch (getKind()) {
676   default:
677     return None;
678   case DArg:
679     NewKind = dArg;
680     break;
681   case UArg:
682     NewKind = uArg;
683     break;
684   case OArg:
685     NewKind = oArg;
686     break;
687   }
688 
689   ConversionSpecifier FixedCS(*this);
690   FixedCS.setKind(NewKind);
691   return FixedCS;
692 }
693 
694 //===----------------------------------------------------------------------===//
695 // Methods on OptionalAmount.
696 //===----------------------------------------------------------------------===//
697 
toString(raw_ostream & os) const698 void OptionalAmount::toString(raw_ostream &os) const {
699   switch (hs) {
700   case Invalid:
701   case NotSpecified:
702     return;
703   case Arg:
704     if (UsesDotPrefix)
705         os << ".";
706     if (usesPositionalArg())
707       os << "*" << getPositionalArgIndex() << "$";
708     else
709       os << "*";
710     break;
711   case Constant:
712     if (UsesDotPrefix)
713         os << ".";
714     os << amt;
715     break;
716   }
717 }
718 
hasValidLengthModifier(const TargetInfo & Target,const LangOptions & LO) const719 bool FormatSpecifier::hasValidLengthModifier(const TargetInfo &Target,
720                                              const LangOptions &LO) const {
721   switch (LM.getKind()) {
722     case LengthModifier::None:
723       return true;
724 
725     // Handle most integer flags
726     case LengthModifier::AsShort:
727       // Length modifier only applies to FP vectors.
728       if (LO.OpenCL && CS.isDoubleArg())
729         return !VectorNumElts.isInvalid();
730 
731       if (Target.getTriple().isOSMSVCRT()) {
732         switch (CS.getKind()) {
733           case ConversionSpecifier::cArg:
734           case ConversionSpecifier::CArg:
735           case ConversionSpecifier::sArg:
736           case ConversionSpecifier::SArg:
737           case ConversionSpecifier::ZArg:
738             return true;
739           default:
740             break;
741         }
742       }
743       LLVM_FALLTHROUGH;
744     case LengthModifier::AsChar:
745     case LengthModifier::AsLongLong:
746     case LengthModifier::AsQuad:
747     case LengthModifier::AsIntMax:
748     case LengthModifier::AsSizeT:
749     case LengthModifier::AsPtrDiff:
750       switch (CS.getKind()) {
751         case ConversionSpecifier::dArg:
752         case ConversionSpecifier::DArg:
753         case ConversionSpecifier::iArg:
754         case ConversionSpecifier::oArg:
755         case ConversionSpecifier::OArg:
756         case ConversionSpecifier::uArg:
757         case ConversionSpecifier::UArg:
758         case ConversionSpecifier::xArg:
759         case ConversionSpecifier::XArg:
760         case ConversionSpecifier::nArg:
761           return true;
762         case ConversionSpecifier::FreeBSDrArg:
763         case ConversionSpecifier::FreeBSDyArg:
764           return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS4();
765         default:
766           return false;
767       }
768 
769     case LengthModifier::AsShortLong:
770       return LO.OpenCL && !VectorNumElts.isInvalid();
771 
772     // Handle 'l' flag
773     case LengthModifier::AsLong: // or AsWideChar
774       if (CS.isDoubleArg()) {
775         // Invalid for OpenCL FP scalars.
776         if (LO.OpenCL && VectorNumElts.isInvalid())
777           return false;
778         return true;
779       }
780 
781       switch (CS.getKind()) {
782         case ConversionSpecifier::dArg:
783         case ConversionSpecifier::DArg:
784         case ConversionSpecifier::iArg:
785         case ConversionSpecifier::oArg:
786         case ConversionSpecifier::OArg:
787         case ConversionSpecifier::uArg:
788         case ConversionSpecifier::UArg:
789         case ConversionSpecifier::xArg:
790         case ConversionSpecifier::XArg:
791         case ConversionSpecifier::nArg:
792         case ConversionSpecifier::cArg:
793         case ConversionSpecifier::sArg:
794         case ConversionSpecifier::ScanListArg:
795         case ConversionSpecifier::ZArg:
796           return true;
797         case ConversionSpecifier::FreeBSDrArg:
798         case ConversionSpecifier::FreeBSDyArg:
799           return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS4();
800         default:
801           return false;
802       }
803 
804     case LengthModifier::AsLongDouble:
805       switch (CS.getKind()) {
806         case ConversionSpecifier::aArg:
807         case ConversionSpecifier::AArg:
808         case ConversionSpecifier::fArg:
809         case ConversionSpecifier::FArg:
810         case ConversionSpecifier::eArg:
811         case ConversionSpecifier::EArg:
812         case ConversionSpecifier::gArg:
813         case ConversionSpecifier::GArg:
814           return true;
815         // GNU libc extension.
816         case ConversionSpecifier::dArg:
817         case ConversionSpecifier::iArg:
818         case ConversionSpecifier::oArg:
819         case ConversionSpecifier::uArg:
820         case ConversionSpecifier::xArg:
821         case ConversionSpecifier::XArg:
822           return !Target.getTriple().isOSDarwin() &&
823                  !Target.getTriple().isOSWindows();
824         default:
825           return false;
826       }
827 
828     case LengthModifier::AsAllocate:
829       switch (CS.getKind()) {
830         case ConversionSpecifier::sArg:
831         case ConversionSpecifier::SArg:
832         case ConversionSpecifier::ScanListArg:
833           return true;
834         default:
835           return false;
836       }
837 
838     case LengthModifier::AsMAllocate:
839       switch (CS.getKind()) {
840         case ConversionSpecifier::cArg:
841         case ConversionSpecifier::CArg:
842         case ConversionSpecifier::sArg:
843         case ConversionSpecifier::SArg:
844         case ConversionSpecifier::ScanListArg:
845           return true;
846         default:
847           return false;
848       }
849     case LengthModifier::AsInt32:
850     case LengthModifier::AsInt3264:
851     case LengthModifier::AsInt64:
852       switch (CS.getKind()) {
853         case ConversionSpecifier::dArg:
854         case ConversionSpecifier::iArg:
855         case ConversionSpecifier::oArg:
856         case ConversionSpecifier::uArg:
857         case ConversionSpecifier::xArg:
858         case ConversionSpecifier::XArg:
859           return Target.getTriple().isOSMSVCRT();
860         default:
861           return false;
862       }
863     case LengthModifier::AsWide:
864       switch (CS.getKind()) {
865         case ConversionSpecifier::cArg:
866         case ConversionSpecifier::CArg:
867         case ConversionSpecifier::sArg:
868         case ConversionSpecifier::SArg:
869         case ConversionSpecifier::ZArg:
870           return Target.getTriple().isOSMSVCRT();
871         default:
872           return false;
873       }
874   }
875   llvm_unreachable("Invalid LengthModifier Kind!");
876 }
877 
hasStandardLengthModifier() const878 bool FormatSpecifier::hasStandardLengthModifier() const {
879   switch (LM.getKind()) {
880     case LengthModifier::None:
881     case LengthModifier::AsChar:
882     case LengthModifier::AsShort:
883     case LengthModifier::AsLong:
884     case LengthModifier::AsLongLong:
885     case LengthModifier::AsIntMax:
886     case LengthModifier::AsSizeT:
887     case LengthModifier::AsPtrDiff:
888     case LengthModifier::AsLongDouble:
889       return true;
890     case LengthModifier::AsAllocate:
891     case LengthModifier::AsMAllocate:
892     case LengthModifier::AsQuad:
893     case LengthModifier::AsInt32:
894     case LengthModifier::AsInt3264:
895     case LengthModifier::AsInt64:
896     case LengthModifier::AsWide:
897     case LengthModifier::AsShortLong: // ???
898       return false;
899   }
900   llvm_unreachable("Invalid LengthModifier Kind!");
901 }
902 
hasStandardConversionSpecifier(const LangOptions & LangOpt) const903 bool FormatSpecifier::hasStandardConversionSpecifier(
904     const LangOptions &LangOpt) const {
905   switch (CS.getKind()) {
906     case ConversionSpecifier::cArg:
907     case ConversionSpecifier::dArg:
908     case ConversionSpecifier::iArg:
909     case ConversionSpecifier::oArg:
910     case ConversionSpecifier::uArg:
911     case ConversionSpecifier::xArg:
912     case ConversionSpecifier::XArg:
913     case ConversionSpecifier::fArg:
914     case ConversionSpecifier::FArg:
915     case ConversionSpecifier::eArg:
916     case ConversionSpecifier::EArg:
917     case ConversionSpecifier::gArg:
918     case ConversionSpecifier::GArg:
919     case ConversionSpecifier::aArg:
920     case ConversionSpecifier::AArg:
921     case ConversionSpecifier::sArg:
922     case ConversionSpecifier::pArg:
923     case ConversionSpecifier::nArg:
924     case ConversionSpecifier::ObjCObjArg:
925     case ConversionSpecifier::ScanListArg:
926     case ConversionSpecifier::PercentArg:
927     case ConversionSpecifier::PArg:
928       return true;
929     case ConversionSpecifier::CArg:
930     case ConversionSpecifier::SArg:
931       return LangOpt.ObjC;
932     case ConversionSpecifier::InvalidSpecifier:
933     case ConversionSpecifier::FreeBSDbArg:
934     case ConversionSpecifier::FreeBSDDArg:
935     case ConversionSpecifier::FreeBSDrArg:
936     case ConversionSpecifier::FreeBSDyArg:
937     case ConversionSpecifier::PrintErrno:
938     case ConversionSpecifier::DArg:
939     case ConversionSpecifier::OArg:
940     case ConversionSpecifier::UArg:
941     case ConversionSpecifier::ZArg:
942       return false;
943   }
944   llvm_unreachable("Invalid ConversionSpecifier Kind!");
945 }
946 
hasStandardLengthConversionCombination() const947 bool FormatSpecifier::hasStandardLengthConversionCombination() const {
948   if (LM.getKind() == LengthModifier::AsLongDouble) {
949     switch(CS.getKind()) {
950         case ConversionSpecifier::dArg:
951         case ConversionSpecifier::iArg:
952         case ConversionSpecifier::oArg:
953         case ConversionSpecifier::uArg:
954         case ConversionSpecifier::xArg:
955         case ConversionSpecifier::XArg:
956           return false;
957         default:
958           return true;
959     }
960   }
961   return true;
962 }
963 
getCorrectedLengthModifier() const964 Optional<LengthModifier> FormatSpecifier::getCorrectedLengthModifier() const {
965   if (CS.isAnyIntArg() || CS.getKind() == ConversionSpecifier::nArg) {
966     if (LM.getKind() == LengthModifier::AsLongDouble ||
967         LM.getKind() == LengthModifier::AsQuad) {
968       LengthModifier FixedLM(LM);
969       FixedLM.setKind(LengthModifier::AsLongLong);
970       return FixedLM;
971     }
972   }
973 
974   return None;
975 }
976 
namedTypeToLengthModifier(QualType QT,LengthModifier & LM)977 bool FormatSpecifier::namedTypeToLengthModifier(QualType QT,
978                                                 LengthModifier &LM) {
979   assert(isa<TypedefType>(QT) && "Expected a TypedefType");
980   const TypedefNameDecl *Typedef = cast<TypedefType>(QT)->getDecl();
981 
982   for (;;) {
983     const IdentifierInfo *Identifier = Typedef->getIdentifier();
984     if (Identifier->getName() == "size_t") {
985       LM.setKind(LengthModifier::AsSizeT);
986       return true;
987     } else if (Identifier->getName() == "ssize_t") {
988       // Not C99, but common in Unix.
989       LM.setKind(LengthModifier::AsSizeT);
990       return true;
991     } else if (Identifier->getName() == "intmax_t") {
992       LM.setKind(LengthModifier::AsIntMax);
993       return true;
994     } else if (Identifier->getName() == "uintmax_t") {
995       LM.setKind(LengthModifier::AsIntMax);
996       return true;
997     } else if (Identifier->getName() == "ptrdiff_t") {
998       LM.setKind(LengthModifier::AsPtrDiff);
999       return true;
1000     }
1001 
1002     QualType T = Typedef->getUnderlyingType();
1003     if (!isa<TypedefType>(T))
1004       break;
1005 
1006     Typedef = cast<TypedefType>(T)->getDecl();
1007   }
1008   return false;
1009 }
1010