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::Void:
423 case BuiltinType::Char_U:
424 case BuiltinType::UChar:
425 case BuiltinType::Char_S:
426 case BuiltinType::SChar:
427 return Match;
428 default:
429 break;
430 }
431
432 return NoMatch;
433 }
434
435 case WCStrTy: {
436 const PointerType *PT = argTy->getAs<PointerType>();
437 if (!PT)
438 return NoMatch;
439 QualType pointeeTy =
440 C.getCanonicalType(PT->getPointeeType()).getUnqualifiedType();
441 return pointeeTy == C.getWideCharType() ? Match : NoMatch;
442 }
443
444 case WIntTy: {
445 QualType WInt = C.getCanonicalType(C.getWIntType()).getUnqualifiedType();
446
447 if (C.getCanonicalType(argTy).getUnqualifiedType() == WInt)
448 return Match;
449
450 QualType PromoArg = argTy->isPromotableIntegerType()
451 ? C.getPromotedIntegerType(argTy)
452 : argTy;
453 PromoArg = C.getCanonicalType(PromoArg).getUnqualifiedType();
454
455 // If the promoted argument is the corresponding signed type of the
456 // wint_t type, then it should match.
457 if (PromoArg->hasSignedIntegerRepresentation() &&
458 C.getCorrespondingUnsignedType(PromoArg) == WInt)
459 return Match;
460
461 return WInt == PromoArg ? Match : NoMatch;
462 }
463
464 case CPointerTy:
465 if (argTy->isVoidPointerType()) {
466 return Match;
467 } if (argTy->isPointerType() || argTy->isObjCObjectPointerType() ||
468 argTy->isBlockPointerType() || argTy->isNullPtrType()) {
469 return NoMatchPedantic;
470 } else {
471 return NoMatch;
472 }
473
474 case ObjCPointerTy: {
475 if (argTy->getAs<ObjCObjectPointerType>() ||
476 argTy->getAs<BlockPointerType>())
477 return Match;
478
479 // Handle implicit toll-free bridging.
480 if (const PointerType *PT = argTy->getAs<PointerType>()) {
481 // Things such as CFTypeRef are really just opaque pointers
482 // to C structs representing CF types that can often be bridged
483 // to Objective-C objects. Since the compiler doesn't know which
484 // structs can be toll-free bridged, we just accept them all.
485 QualType pointee = PT->getPointeeType();
486 if (pointee->getAsStructureType() || pointee->isVoidType())
487 return Match;
488 }
489 return NoMatch;
490 }
491 }
492
493 llvm_unreachable("Invalid ArgType Kind!");
494 }
495
makeVectorType(ASTContext & C,unsigned NumElts) const496 ArgType ArgType::makeVectorType(ASTContext &C, unsigned NumElts) const {
497 // Check for valid vector element types.
498 if (T.isNull())
499 return ArgType::Invalid();
500
501 QualType Vec = C.getExtVectorType(T, NumElts);
502 return ArgType(Vec, Name);
503 }
504
getRepresentativeType(ASTContext & C) const505 QualType ArgType::getRepresentativeType(ASTContext &C) const {
506 QualType Res;
507 switch (K) {
508 case InvalidTy:
509 llvm_unreachable("No representative type for Invalid ArgType");
510 case UnknownTy:
511 llvm_unreachable("No representative type for Unknown ArgType");
512 case AnyCharTy:
513 Res = C.CharTy;
514 break;
515 case SpecificTy:
516 Res = T;
517 break;
518 case CStrTy:
519 Res = C.getPointerType(C.CharTy);
520 break;
521 case WCStrTy:
522 Res = C.getPointerType(C.getWideCharType());
523 break;
524 case ObjCPointerTy:
525 Res = C.ObjCBuiltinIdTy;
526 break;
527 case CPointerTy:
528 Res = C.VoidPtrTy;
529 break;
530 case WIntTy: {
531 Res = C.getWIntType();
532 break;
533 }
534 }
535
536 if (Ptr)
537 Res = C.getPointerType(Res);
538 return Res;
539 }
540
getRepresentativeTypeName(ASTContext & C) const541 std::string ArgType::getRepresentativeTypeName(ASTContext &C) const {
542 std::string S = getRepresentativeType(C).getAsString();
543
544 std::string Alias;
545 if (Name) {
546 // Use a specific name for this type, e.g. "size_t".
547 Alias = Name;
548 if (Ptr) {
549 // If ArgType is actually a pointer to T, append an asterisk.
550 Alias += (Alias[Alias.size()-1] == '*') ? "*" : " *";
551 }
552 // If Alias is the same as the underlying type, e.g. wchar_t, then drop it.
553 if (S == Alias)
554 Alias.clear();
555 }
556
557 if (!Alias.empty())
558 return std::string("'") + Alias + "' (aka '" + S + "')";
559 return std::string("'") + S + "'";
560 }
561
562
563 //===----------------------------------------------------------------------===//
564 // Methods on OptionalAmount.
565 //===----------------------------------------------------------------------===//
566
567 ArgType
getArgType(ASTContext & Ctx) const568 analyze_format_string::OptionalAmount::getArgType(ASTContext &Ctx) const {
569 return Ctx.IntTy;
570 }
571
572 //===----------------------------------------------------------------------===//
573 // Methods on LengthModifier.
574 //===----------------------------------------------------------------------===//
575
576 const char *
toString() const577 analyze_format_string::LengthModifier::toString() const {
578 switch (kind) {
579 case AsChar:
580 return "hh";
581 case AsShort:
582 return "h";
583 case AsShortLong:
584 return "hl";
585 case AsLong: // or AsWideChar
586 return "l";
587 case AsLongLong:
588 return "ll";
589 case AsQuad:
590 return "q";
591 case AsIntMax:
592 return "j";
593 case AsSizeT:
594 return "z";
595 case AsPtrDiff:
596 return "t";
597 case AsInt32:
598 return "I32";
599 case AsInt3264:
600 return "I";
601 case AsInt64:
602 return "I64";
603 case AsLongDouble:
604 return "L";
605 case AsAllocate:
606 return "a";
607 case AsMAllocate:
608 return "m";
609 case AsWide:
610 return "w";
611 case None:
612 return "";
613 }
614 return nullptr;
615 }
616
617 //===----------------------------------------------------------------------===//
618 // Methods on ConversionSpecifier.
619 //===----------------------------------------------------------------------===//
620
toString() const621 const char *ConversionSpecifier::toString() const {
622 switch (kind) {
623 case dArg: return "d";
624 case DArg: return "D";
625 case iArg: return "i";
626 case oArg: return "o";
627 case OArg: return "O";
628 case uArg: return "u";
629 case UArg: return "U";
630 case xArg: return "x";
631 case XArg: return "X";
632 case fArg: return "f";
633 case FArg: return "F";
634 case eArg: return "e";
635 case EArg: return "E";
636 case gArg: return "g";
637 case GArg: return "G";
638 case aArg: return "a";
639 case AArg: return "A";
640 case cArg: return "c";
641 case sArg: return "s";
642 case pArg: return "p";
643 case PArg:
644 return "P";
645 case nArg: return "n";
646 case PercentArg: return "%";
647 case ScanListArg: return "[";
648 case InvalidSpecifier: return nullptr;
649
650 // POSIX unicode extensions.
651 case CArg: return "C";
652 case SArg: return "S";
653
654 // Objective-C specific specifiers.
655 case ObjCObjArg: return "@";
656
657 // FreeBSD kernel specific specifiers.
658 case FreeBSDbArg: return "b";
659 case FreeBSDDArg: return "D";
660 case FreeBSDrArg: return "r";
661 case FreeBSDyArg: return "y";
662
663 // GlibC specific specifiers.
664 case PrintErrno: return "m";
665
666 // MS specific specifiers.
667 case ZArg: return "Z";
668 }
669 return nullptr;
670 }
671
672 Optional<ConversionSpecifier>
getStandardSpecifier() const673 ConversionSpecifier::getStandardSpecifier() const {
674 ConversionSpecifier::Kind NewKind;
675
676 switch (getKind()) {
677 default:
678 return None;
679 case DArg:
680 NewKind = dArg;
681 break;
682 case UArg:
683 NewKind = uArg;
684 break;
685 case OArg:
686 NewKind = oArg;
687 break;
688 }
689
690 ConversionSpecifier FixedCS(*this);
691 FixedCS.setKind(NewKind);
692 return FixedCS;
693 }
694
695 //===----------------------------------------------------------------------===//
696 // Methods on OptionalAmount.
697 //===----------------------------------------------------------------------===//
698
toString(raw_ostream & os) const699 void OptionalAmount::toString(raw_ostream &os) const {
700 switch (hs) {
701 case Invalid:
702 case NotSpecified:
703 return;
704 case Arg:
705 if (UsesDotPrefix)
706 os << ".";
707 if (usesPositionalArg())
708 os << "*" << getPositionalArgIndex() << "$";
709 else
710 os << "*";
711 break;
712 case Constant:
713 if (UsesDotPrefix)
714 os << ".";
715 os << amt;
716 break;
717 }
718 }
719
hasValidLengthModifier(const TargetInfo & Target,const LangOptions & LO) const720 bool FormatSpecifier::hasValidLengthModifier(const TargetInfo &Target,
721 const LangOptions &LO) const {
722 switch (LM.getKind()) {
723 case LengthModifier::None:
724 return true;
725
726 // Handle most integer flags
727 case LengthModifier::AsShort:
728 // Length modifier only applies to FP vectors.
729 if (LO.OpenCL && CS.isDoubleArg())
730 return !VectorNumElts.isInvalid();
731
732 if (Target.getTriple().isOSMSVCRT()) {
733 switch (CS.getKind()) {
734 case ConversionSpecifier::cArg:
735 case ConversionSpecifier::CArg:
736 case ConversionSpecifier::sArg:
737 case ConversionSpecifier::SArg:
738 case ConversionSpecifier::ZArg:
739 return true;
740 default:
741 break;
742 }
743 }
744 LLVM_FALLTHROUGH;
745 case LengthModifier::AsChar:
746 case LengthModifier::AsLongLong:
747 case LengthModifier::AsQuad:
748 case LengthModifier::AsIntMax:
749 case LengthModifier::AsSizeT:
750 case LengthModifier::AsPtrDiff:
751 switch (CS.getKind()) {
752 case ConversionSpecifier::dArg:
753 case ConversionSpecifier::DArg:
754 case ConversionSpecifier::iArg:
755 case ConversionSpecifier::oArg:
756 case ConversionSpecifier::OArg:
757 case ConversionSpecifier::uArg:
758 case ConversionSpecifier::UArg:
759 case ConversionSpecifier::xArg:
760 case ConversionSpecifier::XArg:
761 case ConversionSpecifier::nArg:
762 return true;
763 case ConversionSpecifier::FreeBSDrArg:
764 case ConversionSpecifier::FreeBSDyArg:
765 return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS4();
766 default:
767 return false;
768 }
769
770 case LengthModifier::AsShortLong:
771 return LO.OpenCL && !VectorNumElts.isInvalid();
772
773 // Handle 'l' flag
774 case LengthModifier::AsLong: // or AsWideChar
775 if (CS.isDoubleArg()) {
776 // Invalid for OpenCL FP scalars.
777 if (LO.OpenCL && VectorNumElts.isInvalid())
778 return false;
779 return true;
780 }
781
782 switch (CS.getKind()) {
783 case ConversionSpecifier::dArg:
784 case ConversionSpecifier::DArg:
785 case ConversionSpecifier::iArg:
786 case ConversionSpecifier::oArg:
787 case ConversionSpecifier::OArg:
788 case ConversionSpecifier::uArg:
789 case ConversionSpecifier::UArg:
790 case ConversionSpecifier::xArg:
791 case ConversionSpecifier::XArg:
792 case ConversionSpecifier::nArg:
793 case ConversionSpecifier::cArg:
794 case ConversionSpecifier::sArg:
795 case ConversionSpecifier::ScanListArg:
796 case ConversionSpecifier::ZArg:
797 return true;
798 case ConversionSpecifier::FreeBSDrArg:
799 case ConversionSpecifier::FreeBSDyArg:
800 return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS4();
801 default:
802 return false;
803 }
804
805 case LengthModifier::AsLongDouble:
806 switch (CS.getKind()) {
807 case ConversionSpecifier::aArg:
808 case ConversionSpecifier::AArg:
809 case ConversionSpecifier::fArg:
810 case ConversionSpecifier::FArg:
811 case ConversionSpecifier::eArg:
812 case ConversionSpecifier::EArg:
813 case ConversionSpecifier::gArg:
814 case ConversionSpecifier::GArg:
815 return true;
816 // GNU libc extension.
817 case ConversionSpecifier::dArg:
818 case ConversionSpecifier::iArg:
819 case ConversionSpecifier::oArg:
820 case ConversionSpecifier::uArg:
821 case ConversionSpecifier::xArg:
822 case ConversionSpecifier::XArg:
823 return !Target.getTriple().isOSDarwin() &&
824 !Target.getTriple().isOSWindows();
825 default:
826 return false;
827 }
828
829 case LengthModifier::AsAllocate:
830 switch (CS.getKind()) {
831 case ConversionSpecifier::sArg:
832 case ConversionSpecifier::SArg:
833 case ConversionSpecifier::ScanListArg:
834 return true;
835 default:
836 return false;
837 }
838
839 case LengthModifier::AsMAllocate:
840 switch (CS.getKind()) {
841 case ConversionSpecifier::cArg:
842 case ConversionSpecifier::CArg:
843 case ConversionSpecifier::sArg:
844 case ConversionSpecifier::SArg:
845 case ConversionSpecifier::ScanListArg:
846 return true;
847 default:
848 return false;
849 }
850 case LengthModifier::AsInt32:
851 case LengthModifier::AsInt3264:
852 case LengthModifier::AsInt64:
853 switch (CS.getKind()) {
854 case ConversionSpecifier::dArg:
855 case ConversionSpecifier::iArg:
856 case ConversionSpecifier::oArg:
857 case ConversionSpecifier::uArg:
858 case ConversionSpecifier::xArg:
859 case ConversionSpecifier::XArg:
860 return Target.getTriple().isOSMSVCRT();
861 default:
862 return false;
863 }
864 case LengthModifier::AsWide:
865 switch (CS.getKind()) {
866 case ConversionSpecifier::cArg:
867 case ConversionSpecifier::CArg:
868 case ConversionSpecifier::sArg:
869 case ConversionSpecifier::SArg:
870 case ConversionSpecifier::ZArg:
871 return Target.getTriple().isOSMSVCRT();
872 default:
873 return false;
874 }
875 }
876 llvm_unreachable("Invalid LengthModifier Kind!");
877 }
878
hasStandardLengthModifier() const879 bool FormatSpecifier::hasStandardLengthModifier() const {
880 switch (LM.getKind()) {
881 case LengthModifier::None:
882 case LengthModifier::AsChar:
883 case LengthModifier::AsShort:
884 case LengthModifier::AsLong:
885 case LengthModifier::AsLongLong:
886 case LengthModifier::AsIntMax:
887 case LengthModifier::AsSizeT:
888 case LengthModifier::AsPtrDiff:
889 case LengthModifier::AsLongDouble:
890 return true;
891 case LengthModifier::AsAllocate:
892 case LengthModifier::AsMAllocate:
893 case LengthModifier::AsQuad:
894 case LengthModifier::AsInt32:
895 case LengthModifier::AsInt3264:
896 case LengthModifier::AsInt64:
897 case LengthModifier::AsWide:
898 case LengthModifier::AsShortLong: // ???
899 return false;
900 }
901 llvm_unreachable("Invalid LengthModifier Kind!");
902 }
903
hasStandardConversionSpecifier(const LangOptions & LangOpt) const904 bool FormatSpecifier::hasStandardConversionSpecifier(
905 const LangOptions &LangOpt) const {
906 switch (CS.getKind()) {
907 case ConversionSpecifier::cArg:
908 case ConversionSpecifier::dArg:
909 case ConversionSpecifier::iArg:
910 case ConversionSpecifier::oArg:
911 case ConversionSpecifier::uArg:
912 case ConversionSpecifier::xArg:
913 case ConversionSpecifier::XArg:
914 case ConversionSpecifier::fArg:
915 case ConversionSpecifier::FArg:
916 case ConversionSpecifier::eArg:
917 case ConversionSpecifier::EArg:
918 case ConversionSpecifier::gArg:
919 case ConversionSpecifier::GArg:
920 case ConversionSpecifier::aArg:
921 case ConversionSpecifier::AArg:
922 case ConversionSpecifier::sArg:
923 case ConversionSpecifier::pArg:
924 case ConversionSpecifier::nArg:
925 case ConversionSpecifier::ObjCObjArg:
926 case ConversionSpecifier::ScanListArg:
927 case ConversionSpecifier::PercentArg:
928 case ConversionSpecifier::PArg:
929 return true;
930 case ConversionSpecifier::CArg:
931 case ConversionSpecifier::SArg:
932 return LangOpt.ObjC;
933 case ConversionSpecifier::InvalidSpecifier:
934 case ConversionSpecifier::FreeBSDbArg:
935 case ConversionSpecifier::FreeBSDDArg:
936 case ConversionSpecifier::FreeBSDrArg:
937 case ConversionSpecifier::FreeBSDyArg:
938 case ConversionSpecifier::PrintErrno:
939 case ConversionSpecifier::DArg:
940 case ConversionSpecifier::OArg:
941 case ConversionSpecifier::UArg:
942 case ConversionSpecifier::ZArg:
943 return false;
944 }
945 llvm_unreachable("Invalid ConversionSpecifier Kind!");
946 }
947
hasStandardLengthConversionCombination() const948 bool FormatSpecifier::hasStandardLengthConversionCombination() const {
949 if (LM.getKind() == LengthModifier::AsLongDouble) {
950 switch(CS.getKind()) {
951 case ConversionSpecifier::dArg:
952 case ConversionSpecifier::iArg:
953 case ConversionSpecifier::oArg:
954 case ConversionSpecifier::uArg:
955 case ConversionSpecifier::xArg:
956 case ConversionSpecifier::XArg:
957 return false;
958 default:
959 return true;
960 }
961 }
962 return true;
963 }
964
getCorrectedLengthModifier() const965 Optional<LengthModifier> FormatSpecifier::getCorrectedLengthModifier() const {
966 if (CS.isAnyIntArg() || CS.getKind() == ConversionSpecifier::nArg) {
967 if (LM.getKind() == LengthModifier::AsLongDouble ||
968 LM.getKind() == LengthModifier::AsQuad) {
969 LengthModifier FixedLM(LM);
970 FixedLM.setKind(LengthModifier::AsLongLong);
971 return FixedLM;
972 }
973 }
974
975 return None;
976 }
977
namedTypeToLengthModifier(QualType QT,LengthModifier & LM)978 bool FormatSpecifier::namedTypeToLengthModifier(QualType QT,
979 LengthModifier &LM) {
980 assert(isa<TypedefType>(QT) && "Expected a TypedefType");
981 const TypedefNameDecl *Typedef = cast<TypedefType>(QT)->getDecl();
982
983 for (;;) {
984 const IdentifierInfo *Identifier = Typedef->getIdentifier();
985 if (Identifier->getName() == "size_t") {
986 LM.setKind(LengthModifier::AsSizeT);
987 return true;
988 } else if (Identifier->getName() == "ssize_t") {
989 // Not C99, but common in Unix.
990 LM.setKind(LengthModifier::AsSizeT);
991 return true;
992 } else if (Identifier->getName() == "intmax_t") {
993 LM.setKind(LengthModifier::AsIntMax);
994 return true;
995 } else if (Identifier->getName() == "uintmax_t") {
996 LM.setKind(LengthModifier::AsIntMax);
997 return true;
998 } else if (Identifier->getName() == "ptrdiff_t") {
999 LM.setKind(LengthModifier::AsPtrDiff);
1000 return true;
1001 }
1002
1003 QualType T = Typedef->getUnderlyingType();
1004 if (!isa<TypedefType>(T))
1005 break;
1006
1007 Typedef = cast<TypedefType>(T)->getDecl();
1008 }
1009 return false;
1010 }
1011