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 #include <optional>
19 
20 using clang::analyze_format_string::ArgType;
21 using clang::analyze_format_string::FormatStringHandler;
22 using clang::analyze_format_string::FormatSpecifier;
23 using clang::analyze_format_string::LengthModifier;
24 using clang::analyze_format_string::OptionalAmount;
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   // When using the format attribute in C++, you can receive a function or an
326   // array that will necessarily decay to a pointer when passed to the final
327   // format consumer. Apply decay before type comparison.
328   if (argTy->canDecayToPointerType())
329     argTy = C.getDecayedType(argTy);
330 
331   if (Ptr) {
332     // It has to be a pointer.
333     const PointerType *PT = argTy->getAs<PointerType>();
334     if (!PT)
335       return NoMatch;
336 
337     // We cannot write through a const qualified pointer.
338     if (PT->getPointeeType().isConstQualified())
339       return NoMatch;
340 
341     argTy = PT->getPointeeType();
342   }
343 
344   switch (K) {
345     case InvalidTy:
346       llvm_unreachable("ArgType must be valid");
347 
348     case UnknownTy:
349       return Match;
350 
351     case AnyCharTy: {
352       if (const auto *ETy = argTy->getAs<EnumType>()) {
353         // If the enum is incomplete we know nothing about the underlying type.
354         // Assume that it's 'int'. Do not use the underlying type for a scoped
355         // enumeration.
356         if (!ETy->getDecl()->isComplete())
357           return NoMatch;
358         if (ETy->isUnscopedEnumerationType())
359           argTy = ETy->getDecl()->getIntegerType();
360       }
361 
362       if (const auto *BT = argTy->getAs<BuiltinType>()) {
363         // The types are perfectly matched?
364         switch (BT->getKind()) {
365         default:
366           break;
367         case BuiltinType::Char_S:
368         case BuiltinType::SChar:
369         case BuiltinType::UChar:
370         case BuiltinType::Char_U:
371             return Match;
372         case BuiltinType::Bool:
373           if (!Ptr)
374             return Match;
375           break;
376         }
377         // "Partially matched" because of promotions?
378         if (!Ptr) {
379           switch (BT->getKind()) {
380           default:
381             break;
382           case BuiltinType::Int:
383           case BuiltinType::UInt:
384             return MatchPromotion;
385           case BuiltinType::Short:
386           case BuiltinType::UShort:
387           case BuiltinType::WChar_S:
388           case BuiltinType::WChar_U:
389             return NoMatchPromotionTypeConfusion;
390           }
391         }
392       }
393       return NoMatch;
394     }
395 
396     case SpecificTy: {
397       if (const EnumType *ETy = argTy->getAs<EnumType>()) {
398         // If the enum is incomplete we know nothing about the underlying type.
399         // Assume that it's 'int'. Do not use the underlying type for a scoped
400         // enumeration as that needs an exact match.
401         if (!ETy->getDecl()->isComplete())
402           argTy = C.IntTy;
403         else if (ETy->isUnscopedEnumerationType())
404           argTy = ETy->getDecl()->getIntegerType();
405       }
406       argTy = C.getCanonicalType(argTy).getUnqualifiedType();
407 
408       if (T == argTy)
409         return Match;
410       if (const auto *BT = argTy->getAs<BuiltinType>()) {
411         // Check if the only difference between them is signed vs unsigned
412         // if true, we consider they are compatible.
413         switch (BT->getKind()) {
414           default:
415             break;
416           case BuiltinType::Bool:
417             if (Ptr && (T == C.UnsignedCharTy || T == C.SignedCharTy))
418               return NoMatch;
419             [[fallthrough]];
420           case BuiltinType::Char_S:
421           case BuiltinType::SChar:
422           case BuiltinType::Char_U:
423           case BuiltinType::UChar:
424             if (T == C.UnsignedShortTy || T == C.ShortTy)
425               return NoMatchTypeConfusion;
426             if (T == C.UnsignedCharTy || T == C.SignedCharTy)
427               return Match;
428             break;
429           case BuiltinType::Short:
430             if (T == C.UnsignedShortTy)
431               return Match;
432             break;
433           case BuiltinType::UShort:
434             if (T == C.ShortTy)
435               return Match;
436             break;
437           case BuiltinType::Int:
438             if (T == C.UnsignedIntTy)
439               return Match;
440             break;
441           case BuiltinType::UInt:
442             if (T == C.IntTy)
443               return Match;
444             break;
445           case BuiltinType::Long:
446             if (T == C.UnsignedLongTy)
447               return Match;
448             break;
449           case BuiltinType::ULong:
450             if (T == C.LongTy)
451               return Match;
452             break;
453           case BuiltinType::LongLong:
454             if (T == C.UnsignedLongLongTy)
455               return Match;
456             break;
457           case BuiltinType::ULongLong:
458             if (T == C.LongLongTy)
459               return Match;
460             break;
461           }
462           // "Partially matched" because of promotions?
463           if (!Ptr) {
464             switch (BT->getKind()) {
465             default:
466               break;
467             case BuiltinType::Bool:
468               if (T == C.IntTy || T == C.UnsignedIntTy)
469                 return MatchPromotion;
470               break;
471             case BuiltinType::Int:
472             case BuiltinType::UInt:
473               if (T == C.SignedCharTy || T == C.UnsignedCharTy ||
474                   T == C.ShortTy || T == C.UnsignedShortTy || T == C.WCharTy ||
475                   T == C.WideCharTy)
476                 return MatchPromotion;
477               break;
478             case BuiltinType::Char_U:
479               if (T == C.UnsignedIntTy)
480                 return MatchPromotion;
481               if (T == C.UnsignedShortTy)
482                 return NoMatchPromotionTypeConfusion;
483               break;
484             case BuiltinType::Char_S:
485               if (T == C.IntTy)
486                 return MatchPromotion;
487               if (T == C.ShortTy)
488                 return NoMatchPromotionTypeConfusion;
489               break;
490             case BuiltinType::Half:
491             case BuiltinType::Float:
492               if (T == C.DoubleTy)
493                 return MatchPromotion;
494               break;
495             case BuiltinType::Short:
496             case BuiltinType::UShort:
497               if (T == C.SignedCharTy || T == C.UnsignedCharTy)
498                 return NoMatchPromotionTypeConfusion;
499               break;
500             case BuiltinType::WChar_U:
501             case BuiltinType::WChar_S:
502               if (T != C.WCharTy && T != C.WideCharTy)
503                 return NoMatchPromotionTypeConfusion;
504             }
505           }
506       }
507       return NoMatch;
508     }
509 
510     case CStrTy: {
511       const PointerType *PT = argTy->getAs<PointerType>();
512       if (!PT)
513         return NoMatch;
514       QualType pointeeTy = PT->getPointeeType();
515       if (const BuiltinType *BT = pointeeTy->getAs<BuiltinType>())
516         switch (BT->getKind()) {
517           case BuiltinType::Char_U:
518           case BuiltinType::UChar:
519           case BuiltinType::Char_S:
520           case BuiltinType::SChar:
521             return Match;
522           default:
523             break;
524         }
525 
526       return NoMatch;
527     }
528 
529     case WCStrTy: {
530       const PointerType *PT = argTy->getAs<PointerType>();
531       if (!PT)
532         return NoMatch;
533       QualType pointeeTy =
534         C.getCanonicalType(PT->getPointeeType()).getUnqualifiedType();
535       return pointeeTy == C.getWideCharType() ? Match : NoMatch;
536     }
537 
538     case WIntTy: {
539       QualType WInt = C.getCanonicalType(C.getWIntType()).getUnqualifiedType();
540 
541       if (C.getCanonicalType(argTy).getUnqualifiedType() == WInt)
542         return Match;
543 
544       QualType PromoArg = C.isPromotableIntegerType(argTy)
545                               ? C.getPromotedIntegerType(argTy)
546                               : argTy;
547       PromoArg = C.getCanonicalType(PromoArg).getUnqualifiedType();
548 
549       // If the promoted argument is the corresponding signed type of the
550       // wint_t type, then it should match.
551       if (PromoArg->hasSignedIntegerRepresentation() &&
552           C.getCorrespondingUnsignedType(PromoArg) == WInt)
553         return Match;
554 
555       return WInt == PromoArg ? Match : NoMatch;
556     }
557 
558     case CPointerTy:
559       if (argTy->isVoidPointerType()) {
560         return Match;
561       } if (argTy->isPointerType() || argTy->isObjCObjectPointerType() ||
562             argTy->isBlockPointerType() || argTy->isNullPtrType()) {
563         return NoMatchPedantic;
564       } else {
565         return NoMatch;
566       }
567 
568     case ObjCPointerTy: {
569       if (argTy->getAs<ObjCObjectPointerType>() ||
570           argTy->getAs<BlockPointerType>())
571         return Match;
572 
573       // Handle implicit toll-free bridging.
574       if (const PointerType *PT = argTy->getAs<PointerType>()) {
575         // Things such as CFTypeRef are really just opaque pointers
576         // to C structs representing CF types that can often be bridged
577         // to Objective-C objects.  Since the compiler doesn't know which
578         // structs can be toll-free bridged, we just accept them all.
579         QualType pointee = PT->getPointeeType();
580         if (pointee->getAsStructureType() || pointee->isVoidType())
581           return Match;
582       }
583       return NoMatch;
584     }
585   }
586 
587   llvm_unreachable("Invalid ArgType Kind!");
588 }
589 
makeVectorType(ASTContext & C,unsigned NumElts) const590 ArgType ArgType::makeVectorType(ASTContext &C, unsigned NumElts) const {
591   // Check for valid vector element types.
592   if (T.isNull())
593     return ArgType::Invalid();
594 
595   QualType Vec = C.getExtVectorType(T, NumElts);
596   return ArgType(Vec, Name);
597 }
598 
getRepresentativeType(ASTContext & C) const599 QualType ArgType::getRepresentativeType(ASTContext &C) const {
600   QualType Res;
601   switch (K) {
602     case InvalidTy:
603       llvm_unreachable("No representative type for Invalid ArgType");
604     case UnknownTy:
605       llvm_unreachable("No representative type for Unknown ArgType");
606     case AnyCharTy:
607       Res = C.CharTy;
608       break;
609     case SpecificTy:
610       Res = T;
611       break;
612     case CStrTy:
613       Res = C.getPointerType(C.CharTy);
614       break;
615     case WCStrTy:
616       Res = C.getPointerType(C.getWideCharType());
617       break;
618     case ObjCPointerTy:
619       Res = C.ObjCBuiltinIdTy;
620       break;
621     case CPointerTy:
622       Res = C.VoidPtrTy;
623       break;
624     case WIntTy: {
625       Res = C.getWIntType();
626       break;
627     }
628   }
629 
630   if (Ptr)
631     Res = C.getPointerType(Res);
632   return Res;
633 }
634 
getRepresentativeTypeName(ASTContext & C) const635 std::string ArgType::getRepresentativeTypeName(ASTContext &C) const {
636   std::string S = getRepresentativeType(C).getAsString(C.getPrintingPolicy());
637 
638   std::string Alias;
639   if (Name) {
640     // Use a specific name for this type, e.g. "size_t".
641     Alias = Name;
642     if (Ptr) {
643       // If ArgType is actually a pointer to T, append an asterisk.
644       Alias += (Alias[Alias.size()-1] == '*') ? "*" : " *";
645     }
646     // If Alias is the same as the underlying type, e.g. wchar_t, then drop it.
647     if (S == Alias)
648       Alias.clear();
649   }
650 
651   if (!Alias.empty())
652     return std::string("'") + Alias + "' (aka '" + S + "')";
653   return std::string("'") + S + "'";
654 }
655 
656 
657 //===----------------------------------------------------------------------===//
658 // Methods on OptionalAmount.
659 //===----------------------------------------------------------------------===//
660 
661 ArgType
getArgType(ASTContext & Ctx) const662 analyze_format_string::OptionalAmount::getArgType(ASTContext &Ctx) const {
663   return Ctx.IntTy;
664 }
665 
666 //===----------------------------------------------------------------------===//
667 // Methods on LengthModifier.
668 //===----------------------------------------------------------------------===//
669 
670 const char *
toString() const671 analyze_format_string::LengthModifier::toString() const {
672   switch (kind) {
673   case AsChar:
674     return "hh";
675   case AsShort:
676     return "h";
677   case AsShortLong:
678     return "hl";
679   case AsLong: // or AsWideChar
680     return "l";
681   case AsLongLong:
682     return "ll";
683   case AsQuad:
684     return "q";
685   case AsIntMax:
686     return "j";
687   case AsSizeT:
688     return "z";
689   case AsPtrDiff:
690     return "t";
691   case AsInt32:
692     return "I32";
693   case AsInt3264:
694     return "I";
695   case AsInt64:
696     return "I64";
697   case AsLongDouble:
698     return "L";
699   case AsAllocate:
700     return "a";
701   case AsMAllocate:
702     return "m";
703   case AsWide:
704     return "w";
705   case None:
706     return "";
707   }
708   return nullptr;
709 }
710 
711 //===----------------------------------------------------------------------===//
712 // Methods on ConversionSpecifier.
713 //===----------------------------------------------------------------------===//
714 
toString() const715 const char *ConversionSpecifier::toString() const {
716   switch (kind) {
717   case bArg: return "b";
718   case BArg: return "B";
719   case dArg: return "d";
720   case DArg: return "D";
721   case iArg: return "i";
722   case oArg: return "o";
723   case OArg: return "O";
724   case uArg: return "u";
725   case UArg: return "U";
726   case xArg: return "x";
727   case XArg: return "X";
728   case fArg: return "f";
729   case FArg: return "F";
730   case eArg: return "e";
731   case EArg: return "E";
732   case gArg: return "g";
733   case GArg: return "G";
734   case aArg: return "a";
735   case AArg: return "A";
736   case cArg: return "c";
737   case sArg: return "s";
738   case pArg: return "p";
739   case PArg:
740     return "P";
741   case nArg: return "n";
742   case PercentArg:  return "%";
743   case ScanListArg: return "[";
744   case InvalidSpecifier: return nullptr;
745 
746   // POSIX unicode extensions.
747   case CArg: return "C";
748   case SArg: return "S";
749 
750   // Objective-C specific specifiers.
751   case ObjCObjArg: return "@";
752 
753   // FreeBSD kernel specific specifiers.
754   case FreeBSDbArg: return "b";
755   case FreeBSDDArg: return "D";
756   case FreeBSDrArg: return "r";
757   case FreeBSDyArg: return "y";
758 
759   // GlibC specific specifiers.
760   case PrintErrno: return "m";
761 
762   // MS specific specifiers.
763   case ZArg: return "Z";
764   }
765   return nullptr;
766 }
767 
768 std::optional<ConversionSpecifier>
getStandardSpecifier() const769 ConversionSpecifier::getStandardSpecifier() const {
770   ConversionSpecifier::Kind NewKind;
771 
772   switch (getKind()) {
773   default:
774     return std::nullopt;
775   case DArg:
776     NewKind = dArg;
777     break;
778   case UArg:
779     NewKind = uArg;
780     break;
781   case OArg:
782     NewKind = oArg;
783     break;
784   }
785 
786   ConversionSpecifier FixedCS(*this);
787   FixedCS.setKind(NewKind);
788   return FixedCS;
789 }
790 
791 //===----------------------------------------------------------------------===//
792 // Methods on OptionalAmount.
793 //===----------------------------------------------------------------------===//
794 
toString(raw_ostream & os) const795 void OptionalAmount::toString(raw_ostream &os) const {
796   switch (hs) {
797   case Invalid:
798   case NotSpecified:
799     return;
800   case Arg:
801     if (UsesDotPrefix)
802         os << ".";
803     if (usesPositionalArg())
804       os << "*" << getPositionalArgIndex() << "$";
805     else
806       os << "*";
807     break;
808   case Constant:
809     if (UsesDotPrefix)
810         os << ".";
811     os << amt;
812     break;
813   }
814 }
815 
hasValidLengthModifier(const TargetInfo & Target,const LangOptions & LO) const816 bool FormatSpecifier::hasValidLengthModifier(const TargetInfo &Target,
817                                              const LangOptions &LO) const {
818   switch (LM.getKind()) {
819     case LengthModifier::None:
820       return true;
821 
822     // Handle most integer flags
823     case LengthModifier::AsShort:
824       // Length modifier only applies to FP vectors.
825       if (LO.OpenCL && CS.isDoubleArg())
826         return !VectorNumElts.isInvalid();
827 
828       if (Target.getTriple().isOSMSVCRT()) {
829         switch (CS.getKind()) {
830           case ConversionSpecifier::cArg:
831           case ConversionSpecifier::CArg:
832           case ConversionSpecifier::sArg:
833           case ConversionSpecifier::SArg:
834           case ConversionSpecifier::ZArg:
835             return true;
836           default:
837             break;
838         }
839       }
840       [[fallthrough]];
841     case LengthModifier::AsChar:
842     case LengthModifier::AsLongLong:
843     case LengthModifier::AsQuad:
844     case LengthModifier::AsIntMax:
845     case LengthModifier::AsSizeT:
846     case LengthModifier::AsPtrDiff:
847       switch (CS.getKind()) {
848         case ConversionSpecifier::bArg:
849         case ConversionSpecifier::BArg:
850         case ConversionSpecifier::dArg:
851         case ConversionSpecifier::DArg:
852         case ConversionSpecifier::iArg:
853         case ConversionSpecifier::oArg:
854         case ConversionSpecifier::OArg:
855         case ConversionSpecifier::uArg:
856         case ConversionSpecifier::UArg:
857         case ConversionSpecifier::xArg:
858         case ConversionSpecifier::XArg:
859         case ConversionSpecifier::nArg:
860           return true;
861         case ConversionSpecifier::FreeBSDrArg:
862         case ConversionSpecifier::FreeBSDyArg:
863           return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS();
864         default:
865           return false;
866       }
867 
868     case LengthModifier::AsShortLong:
869       return LO.OpenCL && !VectorNumElts.isInvalid();
870 
871     // Handle 'l' flag
872     case LengthModifier::AsLong: // or AsWideChar
873       if (CS.isDoubleArg()) {
874         // Invalid for OpenCL FP scalars.
875         if (LO.OpenCL && VectorNumElts.isInvalid())
876           return false;
877         return true;
878       }
879 
880       switch (CS.getKind()) {
881         case ConversionSpecifier::bArg:
882         case ConversionSpecifier::BArg:
883         case ConversionSpecifier::dArg:
884         case ConversionSpecifier::DArg:
885         case ConversionSpecifier::iArg:
886         case ConversionSpecifier::oArg:
887         case ConversionSpecifier::OArg:
888         case ConversionSpecifier::uArg:
889         case ConversionSpecifier::UArg:
890         case ConversionSpecifier::xArg:
891         case ConversionSpecifier::XArg:
892         case ConversionSpecifier::nArg:
893         case ConversionSpecifier::cArg:
894         case ConversionSpecifier::sArg:
895         case ConversionSpecifier::ScanListArg:
896         case ConversionSpecifier::ZArg:
897           return true;
898         case ConversionSpecifier::FreeBSDrArg:
899         case ConversionSpecifier::FreeBSDyArg:
900           return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS();
901         default:
902           return false;
903       }
904 
905     case LengthModifier::AsLongDouble:
906       switch (CS.getKind()) {
907         case ConversionSpecifier::aArg:
908         case ConversionSpecifier::AArg:
909         case ConversionSpecifier::fArg:
910         case ConversionSpecifier::FArg:
911         case ConversionSpecifier::eArg:
912         case ConversionSpecifier::EArg:
913         case ConversionSpecifier::gArg:
914         case ConversionSpecifier::GArg:
915           return true;
916         // GNU libc extension.
917         case ConversionSpecifier::dArg:
918         case ConversionSpecifier::iArg:
919         case ConversionSpecifier::oArg:
920         case ConversionSpecifier::uArg:
921         case ConversionSpecifier::xArg:
922         case ConversionSpecifier::XArg:
923           return !Target.getTriple().isOSDarwin() &&
924                  !Target.getTriple().isOSWindows();
925         default:
926           return false;
927       }
928 
929     case LengthModifier::AsAllocate:
930       switch (CS.getKind()) {
931         case ConversionSpecifier::sArg:
932         case ConversionSpecifier::SArg:
933         case ConversionSpecifier::ScanListArg:
934           return true;
935         default:
936           return false;
937       }
938 
939     case LengthModifier::AsMAllocate:
940       switch (CS.getKind()) {
941         case ConversionSpecifier::cArg:
942         case ConversionSpecifier::CArg:
943         case ConversionSpecifier::sArg:
944         case ConversionSpecifier::SArg:
945         case ConversionSpecifier::ScanListArg:
946           return true;
947         default:
948           return false;
949       }
950     case LengthModifier::AsInt32:
951     case LengthModifier::AsInt3264:
952     case LengthModifier::AsInt64:
953       switch (CS.getKind()) {
954         case ConversionSpecifier::dArg:
955         case ConversionSpecifier::iArg:
956         case ConversionSpecifier::oArg:
957         case ConversionSpecifier::uArg:
958         case ConversionSpecifier::xArg:
959         case ConversionSpecifier::XArg:
960           return Target.getTriple().isOSMSVCRT();
961         default:
962           return false;
963       }
964     case LengthModifier::AsWide:
965       switch (CS.getKind()) {
966         case ConversionSpecifier::cArg:
967         case ConversionSpecifier::CArg:
968         case ConversionSpecifier::sArg:
969         case ConversionSpecifier::SArg:
970         case ConversionSpecifier::ZArg:
971           return Target.getTriple().isOSMSVCRT();
972         default:
973           return false;
974       }
975   }
976   llvm_unreachable("Invalid LengthModifier Kind!");
977 }
978 
hasStandardLengthModifier() const979 bool FormatSpecifier::hasStandardLengthModifier() const {
980   switch (LM.getKind()) {
981     case LengthModifier::None:
982     case LengthModifier::AsChar:
983     case LengthModifier::AsShort:
984     case LengthModifier::AsLong:
985     case LengthModifier::AsLongLong:
986     case LengthModifier::AsIntMax:
987     case LengthModifier::AsSizeT:
988     case LengthModifier::AsPtrDiff:
989     case LengthModifier::AsLongDouble:
990       return true;
991     case LengthModifier::AsAllocate:
992     case LengthModifier::AsMAllocate:
993     case LengthModifier::AsQuad:
994     case LengthModifier::AsInt32:
995     case LengthModifier::AsInt3264:
996     case LengthModifier::AsInt64:
997     case LengthModifier::AsWide:
998     case LengthModifier::AsShortLong: // ???
999       return false;
1000   }
1001   llvm_unreachable("Invalid LengthModifier Kind!");
1002 }
1003 
hasStandardConversionSpecifier(const LangOptions & LangOpt) const1004 bool FormatSpecifier::hasStandardConversionSpecifier(
1005     const LangOptions &LangOpt) const {
1006   switch (CS.getKind()) {
1007     case ConversionSpecifier::bArg:
1008     case ConversionSpecifier::BArg:
1009     case ConversionSpecifier::cArg:
1010     case ConversionSpecifier::dArg:
1011     case ConversionSpecifier::iArg:
1012     case ConversionSpecifier::oArg:
1013     case ConversionSpecifier::uArg:
1014     case ConversionSpecifier::xArg:
1015     case ConversionSpecifier::XArg:
1016     case ConversionSpecifier::fArg:
1017     case ConversionSpecifier::FArg:
1018     case ConversionSpecifier::eArg:
1019     case ConversionSpecifier::EArg:
1020     case ConversionSpecifier::gArg:
1021     case ConversionSpecifier::GArg:
1022     case ConversionSpecifier::aArg:
1023     case ConversionSpecifier::AArg:
1024     case ConversionSpecifier::sArg:
1025     case ConversionSpecifier::pArg:
1026     case ConversionSpecifier::nArg:
1027     case ConversionSpecifier::ObjCObjArg:
1028     case ConversionSpecifier::ScanListArg:
1029     case ConversionSpecifier::PercentArg:
1030     case ConversionSpecifier::PArg:
1031       return true;
1032     case ConversionSpecifier::CArg:
1033     case ConversionSpecifier::SArg:
1034       return LangOpt.ObjC;
1035     case ConversionSpecifier::InvalidSpecifier:
1036     case ConversionSpecifier::FreeBSDbArg:
1037     case ConversionSpecifier::FreeBSDDArg:
1038     case ConversionSpecifier::FreeBSDrArg:
1039     case ConversionSpecifier::FreeBSDyArg:
1040     case ConversionSpecifier::PrintErrno:
1041     case ConversionSpecifier::DArg:
1042     case ConversionSpecifier::OArg:
1043     case ConversionSpecifier::UArg:
1044     case ConversionSpecifier::ZArg:
1045       return false;
1046   }
1047   llvm_unreachable("Invalid ConversionSpecifier Kind!");
1048 }
1049 
hasStandardLengthConversionCombination() const1050 bool FormatSpecifier::hasStandardLengthConversionCombination() const {
1051   if (LM.getKind() == LengthModifier::AsLongDouble) {
1052     switch(CS.getKind()) {
1053         case ConversionSpecifier::dArg:
1054         case ConversionSpecifier::iArg:
1055         case ConversionSpecifier::oArg:
1056         case ConversionSpecifier::uArg:
1057         case ConversionSpecifier::xArg:
1058         case ConversionSpecifier::XArg:
1059           return false;
1060         default:
1061           return true;
1062     }
1063   }
1064   return true;
1065 }
1066 
1067 std::optional<LengthModifier>
getCorrectedLengthModifier() const1068 FormatSpecifier::getCorrectedLengthModifier() const {
1069   if (CS.isAnyIntArg() || CS.getKind() == ConversionSpecifier::nArg) {
1070     if (LM.getKind() == LengthModifier::AsLongDouble ||
1071         LM.getKind() == LengthModifier::AsQuad) {
1072       LengthModifier FixedLM(LM);
1073       FixedLM.setKind(LengthModifier::AsLongLong);
1074       return FixedLM;
1075     }
1076   }
1077 
1078   return std::nullopt;
1079 }
1080 
namedTypeToLengthModifier(QualType QT,LengthModifier & LM)1081 bool FormatSpecifier::namedTypeToLengthModifier(QualType QT,
1082                                                 LengthModifier &LM) {
1083   for (/**/; const auto *TT = QT->getAs<TypedefType>();
1084        QT = TT->getDecl()->getUnderlyingType()) {
1085     const TypedefNameDecl *Typedef = TT->getDecl();
1086     const IdentifierInfo *Identifier = Typedef->getIdentifier();
1087     if (Identifier->getName() == "size_t") {
1088       LM.setKind(LengthModifier::AsSizeT);
1089       return true;
1090     } else if (Identifier->getName() == "ssize_t") {
1091       // Not C99, but common in Unix.
1092       LM.setKind(LengthModifier::AsSizeT);
1093       return true;
1094     } else if (Identifier->getName() == "intmax_t") {
1095       LM.setKind(LengthModifier::AsIntMax);
1096       return true;
1097     } else if (Identifier->getName() == "uintmax_t") {
1098       LM.setKind(LengthModifier::AsIntMax);
1099       return true;
1100     } else if (Identifier->getName() == "ptrdiff_t") {
1101       LM.setKind(LengthModifier::AsPtrDiff);
1102       return true;
1103     }
1104   }
1105   return false;
1106 }
1107