|
150
|
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.
|
|
|
29 FormatStringHandler::~FormatStringHandler() {}
|
|
|
30
|
|
|
31 //===----------------------------------------------------------------------===//
|
|
|
32 // Functions for parsing format strings components in both printf and
|
|
|
33 // scanf format strings.
|
|
|
34 //===----------------------------------------------------------------------===//
|
|
|
35
|
|
|
36 OptionalAmount
|
|
|
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
|
|
|
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
|
|
|
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
|
|
|
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
|
|
|
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
|
|
|
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
|
|
|
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
|
|
|
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
|
|
|
324 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:
|
|
152
|
426 #ifndef noCbC
|
|
|
427 case BuiltinType::__Code:
|
|
|
428 #endif
|
|
150
|
429 return Match;
|
|
|
430 default:
|
|
|
431 break;
|
|
|
432 }
|
|
|
433
|
|
|
434 return NoMatch;
|
|
|
435 }
|
|
|
436
|
|
|
437 case WCStrTy: {
|
|
|
438 const PointerType *PT = argTy->getAs<PointerType>();
|
|
|
439 if (!PT)
|
|
|
440 return NoMatch;
|
|
|
441 QualType pointeeTy =
|
|
|
442 C.getCanonicalType(PT->getPointeeType()).getUnqualifiedType();
|
|
|
443 return pointeeTy == C.getWideCharType() ? Match : NoMatch;
|
|
|
444 }
|
|
|
445
|
|
|
446 case WIntTy: {
|
|
|
447 QualType WInt = C.getCanonicalType(C.getWIntType()).getUnqualifiedType();
|
|
|
448
|
|
|
449 if (C.getCanonicalType(argTy).getUnqualifiedType() == WInt)
|
|
|
450 return Match;
|
|
|
451
|
|
|
452 QualType PromoArg = argTy->isPromotableIntegerType()
|
|
|
453 ? C.getPromotedIntegerType(argTy)
|
|
|
454 : argTy;
|
|
|
455 PromoArg = C.getCanonicalType(PromoArg).getUnqualifiedType();
|
|
|
456
|
|
|
457 // If the promoted argument is the corresponding signed type of the
|
|
|
458 // wint_t type, then it should match.
|
|
|
459 if (PromoArg->hasSignedIntegerRepresentation() &&
|
|
|
460 C.getCorrespondingUnsignedType(PromoArg) == WInt)
|
|
|
461 return Match;
|
|
|
462
|
|
|
463 return WInt == PromoArg ? Match : NoMatch;
|
|
|
464 }
|
|
|
465
|
|
|
466 case CPointerTy:
|
|
|
467 if (argTy->isVoidPointerType()) {
|
|
|
468 return Match;
|
|
|
469 } if (argTy->isPointerType() || argTy->isObjCObjectPointerType() ||
|
|
|
470 argTy->isBlockPointerType() || argTy->isNullPtrType()) {
|
|
|
471 return NoMatchPedantic;
|
|
|
472 } else {
|
|
|
473 return NoMatch;
|
|
|
474 }
|
|
|
475
|
|
|
476 case ObjCPointerTy: {
|
|
|
477 if (argTy->getAs<ObjCObjectPointerType>() ||
|
|
|
478 argTy->getAs<BlockPointerType>())
|
|
|
479 return Match;
|
|
|
480
|
|
|
481 // Handle implicit toll-free bridging.
|
|
|
482 if (const PointerType *PT = argTy->getAs<PointerType>()) {
|
|
|
483 // Things such as CFTypeRef are really just opaque pointers
|
|
|
484 // to C structs representing CF types that can often be bridged
|
|
|
485 // to Objective-C objects. Since the compiler doesn't know which
|
|
|
486 // structs can be toll-free bridged, we just accept them all.
|
|
|
487 QualType pointee = PT->getPointeeType();
|
|
|
488 if (pointee->getAsStructureType() || pointee->isVoidType())
|
|
|
489 return Match;
|
|
|
490 }
|
|
|
491 return NoMatch;
|
|
|
492 }
|
|
|
493 }
|
|
|
494
|
|
|
495 llvm_unreachable("Invalid ArgType Kind!");
|
|
|
496 }
|
|
|
497
|
|
|
498 ArgType ArgType::makeVectorType(ASTContext &C, unsigned NumElts) const {
|
|
|
499 // Check for valid vector element types.
|
|
|
500 if (T.isNull())
|
|
|
501 return ArgType::Invalid();
|
|
|
502
|
|
|
503 QualType Vec = C.getExtVectorType(T, NumElts);
|
|
|
504 return ArgType(Vec, Name);
|
|
|
505 }
|
|
|
506
|
|
|
507 QualType ArgType::getRepresentativeType(ASTContext &C) const {
|
|
|
508 QualType Res;
|
|
|
509 switch (K) {
|
|
|
510 case InvalidTy:
|
|
|
511 llvm_unreachable("No representative type for Invalid ArgType");
|
|
|
512 case UnknownTy:
|
|
|
513 llvm_unreachable("No representative type for Unknown ArgType");
|
|
|
514 case AnyCharTy:
|
|
|
515 Res = C.CharTy;
|
|
|
516 break;
|
|
|
517 case SpecificTy:
|
|
|
518 Res = T;
|
|
|
519 break;
|
|
|
520 case CStrTy:
|
|
|
521 Res = C.getPointerType(C.CharTy);
|
|
|
522 break;
|
|
|
523 case WCStrTy:
|
|
|
524 Res = C.getPointerType(C.getWideCharType());
|
|
|
525 break;
|
|
|
526 case ObjCPointerTy:
|
|
|
527 Res = C.ObjCBuiltinIdTy;
|
|
|
528 break;
|
|
|
529 case CPointerTy:
|
|
|
530 Res = C.VoidPtrTy;
|
|
|
531 break;
|
|
|
532 case WIntTy: {
|
|
|
533 Res = C.getWIntType();
|
|
|
534 break;
|
|
|
535 }
|
|
|
536 }
|
|
|
537
|
|
|
538 if (Ptr)
|
|
|
539 Res = C.getPointerType(Res);
|
|
|
540 return Res;
|
|
|
541 }
|
|
|
542
|
|
|
543 std::string ArgType::getRepresentativeTypeName(ASTContext &C) const {
|
|
173
|
544 std::string S = getRepresentativeType(C).getAsString(C.getPrintingPolicy());
|
|
150
|
545
|
|
|
546 std::string Alias;
|
|
|
547 if (Name) {
|
|
|
548 // Use a specific name for this type, e.g. "size_t".
|
|
|
549 Alias = Name;
|
|
|
550 if (Ptr) {
|
|
|
551 // If ArgType is actually a pointer to T, append an asterisk.
|
|
|
552 Alias += (Alias[Alias.size()-1] == '*') ? "*" : " *";
|
|
|
553 }
|
|
|
554 // If Alias is the same as the underlying type, e.g. wchar_t, then drop it.
|
|
|
555 if (S == Alias)
|
|
|
556 Alias.clear();
|
|
|
557 }
|
|
|
558
|
|
|
559 if (!Alias.empty())
|
|
|
560 return std::string("'") + Alias + "' (aka '" + S + "')";
|
|
|
561 return std::string("'") + S + "'";
|
|
|
562 }
|
|
|
563
|
|
|
564
|
|
|
565 //===----------------------------------------------------------------------===//
|
|
|
566 // Methods on OptionalAmount.
|
|
|
567 //===----------------------------------------------------------------------===//
|
|
|
568
|
|
|
569 ArgType
|
|
|
570 analyze_format_string::OptionalAmount::getArgType(ASTContext &Ctx) const {
|
|
|
571 return Ctx.IntTy;
|
|
|
572 }
|
|
|
573
|
|
|
574 //===----------------------------------------------------------------------===//
|
|
|
575 // Methods on LengthModifier.
|
|
|
576 //===----------------------------------------------------------------------===//
|
|
|
577
|
|
|
578 const char *
|
|
|
579 analyze_format_string::LengthModifier::toString() const {
|
|
|
580 switch (kind) {
|
|
|
581 case AsChar:
|
|
|
582 return "hh";
|
|
|
583 case AsShort:
|
|
|
584 return "h";
|
|
|
585 case AsShortLong:
|
|
|
586 return "hl";
|
|
|
587 case AsLong: // or AsWideChar
|
|
|
588 return "l";
|
|
|
589 case AsLongLong:
|
|
|
590 return "ll";
|
|
|
591 case AsQuad:
|
|
|
592 return "q";
|
|
|
593 case AsIntMax:
|
|
|
594 return "j";
|
|
|
595 case AsSizeT:
|
|
|
596 return "z";
|
|
|
597 case AsPtrDiff:
|
|
|
598 return "t";
|
|
|
599 case AsInt32:
|
|
|
600 return "I32";
|
|
|
601 case AsInt3264:
|
|
|
602 return "I";
|
|
|
603 case AsInt64:
|
|
|
604 return "I64";
|
|
|
605 case AsLongDouble:
|
|
|
606 return "L";
|
|
|
607 case AsAllocate:
|
|
|
608 return "a";
|
|
|
609 case AsMAllocate:
|
|
|
610 return "m";
|
|
|
611 case AsWide:
|
|
|
612 return "w";
|
|
|
613 case None:
|
|
|
614 return "";
|
|
|
615 }
|
|
|
616 return nullptr;
|
|
|
617 }
|
|
|
618
|
|
|
619 //===----------------------------------------------------------------------===//
|
|
|
620 // Methods on ConversionSpecifier.
|
|
|
621 //===----------------------------------------------------------------------===//
|
|
|
622
|
|
|
623 const char *ConversionSpecifier::toString() const {
|
|
|
624 switch (kind) {
|
|
|
625 case dArg: return "d";
|
|
|
626 case DArg: return "D";
|
|
|
627 case iArg: return "i";
|
|
|
628 case oArg: return "o";
|
|
|
629 case OArg: return "O";
|
|
|
630 case uArg: return "u";
|
|
|
631 case UArg: return "U";
|
|
|
632 case xArg: return "x";
|
|
|
633 case XArg: return "X";
|
|
|
634 case fArg: return "f";
|
|
|
635 case FArg: return "F";
|
|
|
636 case eArg: return "e";
|
|
|
637 case EArg: return "E";
|
|
|
638 case gArg: return "g";
|
|
|
639 case GArg: return "G";
|
|
|
640 case aArg: return "a";
|
|
|
641 case AArg: return "A";
|
|
|
642 case cArg: return "c";
|
|
|
643 case sArg: return "s";
|
|
|
644 case pArg: return "p";
|
|
|
645 case PArg:
|
|
|
646 return "P";
|
|
|
647 case nArg: return "n";
|
|
|
648 case PercentArg: return "%";
|
|
|
649 case ScanListArg: return "[";
|
|
|
650 case InvalidSpecifier: return nullptr;
|
|
|
651
|
|
|
652 // POSIX unicode extensions.
|
|
|
653 case CArg: return "C";
|
|
|
654 case SArg: return "S";
|
|
|
655
|
|
|
656 // Objective-C specific specifiers.
|
|
|
657 case ObjCObjArg: return "@";
|
|
|
658
|
|
|
659 // FreeBSD kernel specific specifiers.
|
|
|
660 case FreeBSDbArg: return "b";
|
|
|
661 case FreeBSDDArg: return "D";
|
|
|
662 case FreeBSDrArg: return "r";
|
|
|
663 case FreeBSDyArg: return "y";
|
|
|
664
|
|
|
665 // GlibC specific specifiers.
|
|
|
666 case PrintErrno: return "m";
|
|
|
667
|
|
|
668 // MS specific specifiers.
|
|
|
669 case ZArg: return "Z";
|
|
|
670 }
|
|
|
671 return nullptr;
|
|
|
672 }
|
|
|
673
|
|
|
674 Optional<ConversionSpecifier>
|
|
|
675 ConversionSpecifier::getStandardSpecifier() const {
|
|
|
676 ConversionSpecifier::Kind NewKind;
|
|
|
677
|
|
|
678 switch (getKind()) {
|
|
|
679 default:
|
|
|
680 return None;
|
|
|
681 case DArg:
|
|
|
682 NewKind = dArg;
|
|
|
683 break;
|
|
|
684 case UArg:
|
|
|
685 NewKind = uArg;
|
|
|
686 break;
|
|
|
687 case OArg:
|
|
|
688 NewKind = oArg;
|
|
|
689 break;
|
|
|
690 }
|
|
|
691
|
|
|
692 ConversionSpecifier FixedCS(*this);
|
|
|
693 FixedCS.setKind(NewKind);
|
|
|
694 return FixedCS;
|
|
|
695 }
|
|
|
696
|
|
|
697 //===----------------------------------------------------------------------===//
|
|
|
698 // Methods on OptionalAmount.
|
|
|
699 //===----------------------------------------------------------------------===//
|
|
|
700
|
|
|
701 void OptionalAmount::toString(raw_ostream &os) const {
|
|
|
702 switch (hs) {
|
|
|
703 case Invalid:
|
|
|
704 case NotSpecified:
|
|
|
705 return;
|
|
|
706 case Arg:
|
|
|
707 if (UsesDotPrefix)
|
|
|
708 os << ".";
|
|
|
709 if (usesPositionalArg())
|
|
|
710 os << "*" << getPositionalArgIndex() << "$";
|
|
|
711 else
|
|
|
712 os << "*";
|
|
|
713 break;
|
|
|
714 case Constant:
|
|
|
715 if (UsesDotPrefix)
|
|
|
716 os << ".";
|
|
|
717 os << amt;
|
|
|
718 break;
|
|
|
719 }
|
|
|
720 }
|
|
|
721
|
|
|
722 bool FormatSpecifier::hasValidLengthModifier(const TargetInfo &Target,
|
|
|
723 const LangOptions &LO) const {
|
|
|
724 switch (LM.getKind()) {
|
|
|
725 case LengthModifier::None:
|
|
|
726 return true;
|
|
|
727
|
|
|
728 // Handle most integer flags
|
|
|
729 case LengthModifier::AsShort:
|
|
|
730 // Length modifier only applies to FP vectors.
|
|
|
731 if (LO.OpenCL && CS.isDoubleArg())
|
|
|
732 return !VectorNumElts.isInvalid();
|
|
|
733
|
|
|
734 if (Target.getTriple().isOSMSVCRT()) {
|
|
|
735 switch (CS.getKind()) {
|
|
|
736 case ConversionSpecifier::cArg:
|
|
|
737 case ConversionSpecifier::CArg:
|
|
|
738 case ConversionSpecifier::sArg:
|
|
|
739 case ConversionSpecifier::SArg:
|
|
|
740 case ConversionSpecifier::ZArg:
|
|
|
741 return true;
|
|
|
742 default:
|
|
|
743 break;
|
|
|
744 }
|
|
|
745 }
|
|
|
746 LLVM_FALLTHROUGH;
|
|
|
747 case LengthModifier::AsChar:
|
|
|
748 case LengthModifier::AsLongLong:
|
|
|
749 case LengthModifier::AsQuad:
|
|
|
750 case LengthModifier::AsIntMax:
|
|
|
751 case LengthModifier::AsSizeT:
|
|
|
752 case LengthModifier::AsPtrDiff:
|
|
|
753 switch (CS.getKind()) {
|
|
|
754 case ConversionSpecifier::dArg:
|
|
|
755 case ConversionSpecifier::DArg:
|
|
|
756 case ConversionSpecifier::iArg:
|
|
|
757 case ConversionSpecifier::oArg:
|
|
|
758 case ConversionSpecifier::OArg:
|
|
|
759 case ConversionSpecifier::uArg:
|
|
|
760 case ConversionSpecifier::UArg:
|
|
|
761 case ConversionSpecifier::xArg:
|
|
|
762 case ConversionSpecifier::XArg:
|
|
|
763 case ConversionSpecifier::nArg:
|
|
|
764 return true;
|
|
|
765 case ConversionSpecifier::FreeBSDrArg:
|
|
|
766 case ConversionSpecifier::FreeBSDyArg:
|
|
|
767 return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS4();
|
|
|
768 default:
|
|
|
769 return false;
|
|
|
770 }
|
|
|
771
|
|
|
772 case LengthModifier::AsShortLong:
|
|
|
773 return LO.OpenCL && !VectorNumElts.isInvalid();
|
|
|
774
|
|
|
775 // Handle 'l' flag
|
|
|
776 case LengthModifier::AsLong: // or AsWideChar
|
|
|
777 if (CS.isDoubleArg()) {
|
|
|
778 // Invalid for OpenCL FP scalars.
|
|
|
779 if (LO.OpenCL && VectorNumElts.isInvalid())
|
|
|
780 return false;
|
|
|
781 return true;
|
|
|
782 }
|
|
|
783
|
|
|
784 switch (CS.getKind()) {
|
|
|
785 case ConversionSpecifier::dArg:
|
|
|
786 case ConversionSpecifier::DArg:
|
|
|
787 case ConversionSpecifier::iArg:
|
|
|
788 case ConversionSpecifier::oArg:
|
|
|
789 case ConversionSpecifier::OArg:
|
|
|
790 case ConversionSpecifier::uArg:
|
|
|
791 case ConversionSpecifier::UArg:
|
|
|
792 case ConversionSpecifier::xArg:
|
|
|
793 case ConversionSpecifier::XArg:
|
|
|
794 case ConversionSpecifier::nArg:
|
|
|
795 case ConversionSpecifier::cArg:
|
|
|
796 case ConversionSpecifier::sArg:
|
|
|
797 case ConversionSpecifier::ScanListArg:
|
|
|
798 case ConversionSpecifier::ZArg:
|
|
|
799 return true;
|
|
|
800 case ConversionSpecifier::FreeBSDrArg:
|
|
|
801 case ConversionSpecifier::FreeBSDyArg:
|
|
|
802 return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS4();
|
|
|
803 default:
|
|
|
804 return false;
|
|
|
805 }
|
|
|
806
|
|
|
807 case LengthModifier::AsLongDouble:
|
|
|
808 switch (CS.getKind()) {
|
|
|
809 case ConversionSpecifier::aArg:
|
|
|
810 case ConversionSpecifier::AArg:
|
|
|
811 case ConversionSpecifier::fArg:
|
|
|
812 case ConversionSpecifier::FArg:
|
|
|
813 case ConversionSpecifier::eArg:
|
|
|
814 case ConversionSpecifier::EArg:
|
|
|
815 case ConversionSpecifier::gArg:
|
|
|
816 case ConversionSpecifier::GArg:
|
|
|
817 return true;
|
|
|
818 // GNU libc extension.
|
|
|
819 case ConversionSpecifier::dArg:
|
|
|
820 case ConversionSpecifier::iArg:
|
|
|
821 case ConversionSpecifier::oArg:
|
|
|
822 case ConversionSpecifier::uArg:
|
|
|
823 case ConversionSpecifier::xArg:
|
|
|
824 case ConversionSpecifier::XArg:
|
|
|
825 return !Target.getTriple().isOSDarwin() &&
|
|
|
826 !Target.getTriple().isOSWindows();
|
|
|
827 default:
|
|
|
828 return false;
|
|
|
829 }
|
|
|
830
|
|
|
831 case LengthModifier::AsAllocate:
|
|
|
832 switch (CS.getKind()) {
|
|
|
833 case ConversionSpecifier::sArg:
|
|
|
834 case ConversionSpecifier::SArg:
|
|
|
835 case ConversionSpecifier::ScanListArg:
|
|
|
836 return true;
|
|
|
837 default:
|
|
|
838 return false;
|
|
|
839 }
|
|
|
840
|
|
|
841 case LengthModifier::AsMAllocate:
|
|
|
842 switch (CS.getKind()) {
|
|
|
843 case ConversionSpecifier::cArg:
|
|
|
844 case ConversionSpecifier::CArg:
|
|
|
845 case ConversionSpecifier::sArg:
|
|
|
846 case ConversionSpecifier::SArg:
|
|
|
847 case ConversionSpecifier::ScanListArg:
|
|
|
848 return true;
|
|
|
849 default:
|
|
|
850 return false;
|
|
|
851 }
|
|
|
852 case LengthModifier::AsInt32:
|
|
|
853 case LengthModifier::AsInt3264:
|
|
|
854 case LengthModifier::AsInt64:
|
|
|
855 switch (CS.getKind()) {
|
|
|
856 case ConversionSpecifier::dArg:
|
|
|
857 case ConversionSpecifier::iArg:
|
|
|
858 case ConversionSpecifier::oArg:
|
|
|
859 case ConversionSpecifier::uArg:
|
|
|
860 case ConversionSpecifier::xArg:
|
|
|
861 case ConversionSpecifier::XArg:
|
|
|
862 return Target.getTriple().isOSMSVCRT();
|
|
|
863 default:
|
|
|
864 return false;
|
|
|
865 }
|
|
|
866 case LengthModifier::AsWide:
|
|
|
867 switch (CS.getKind()) {
|
|
|
868 case ConversionSpecifier::cArg:
|
|
|
869 case ConversionSpecifier::CArg:
|
|
|
870 case ConversionSpecifier::sArg:
|
|
|
871 case ConversionSpecifier::SArg:
|
|
|
872 case ConversionSpecifier::ZArg:
|
|
|
873 return Target.getTriple().isOSMSVCRT();
|
|
|
874 default:
|
|
|
875 return false;
|
|
|
876 }
|
|
|
877 }
|
|
|
878 llvm_unreachable("Invalid LengthModifier Kind!");
|
|
|
879 }
|
|
|
880
|
|
|
881 bool FormatSpecifier::hasStandardLengthModifier() const {
|
|
|
882 switch (LM.getKind()) {
|
|
|
883 case LengthModifier::None:
|
|
|
884 case LengthModifier::AsChar:
|
|
|
885 case LengthModifier::AsShort:
|
|
|
886 case LengthModifier::AsLong:
|
|
|
887 case LengthModifier::AsLongLong:
|
|
|
888 case LengthModifier::AsIntMax:
|
|
|
889 case LengthModifier::AsSizeT:
|
|
|
890 case LengthModifier::AsPtrDiff:
|
|
|
891 case LengthModifier::AsLongDouble:
|
|
|
892 return true;
|
|
|
893 case LengthModifier::AsAllocate:
|
|
|
894 case LengthModifier::AsMAllocate:
|
|
|
895 case LengthModifier::AsQuad:
|
|
|
896 case LengthModifier::AsInt32:
|
|
|
897 case LengthModifier::AsInt3264:
|
|
|
898 case LengthModifier::AsInt64:
|
|
|
899 case LengthModifier::AsWide:
|
|
|
900 case LengthModifier::AsShortLong: // ???
|
|
|
901 return false;
|
|
|
902 }
|
|
|
903 llvm_unreachable("Invalid LengthModifier Kind!");
|
|
|
904 }
|
|
|
905
|
|
|
906 bool FormatSpecifier::hasStandardConversionSpecifier(
|
|
|
907 const LangOptions &LangOpt) const {
|
|
|
908 switch (CS.getKind()) {
|
|
|
909 case ConversionSpecifier::cArg:
|
|
|
910 case ConversionSpecifier::dArg:
|
|
|
911 case ConversionSpecifier::iArg:
|
|
|
912 case ConversionSpecifier::oArg:
|
|
|
913 case ConversionSpecifier::uArg:
|
|
|
914 case ConversionSpecifier::xArg:
|
|
|
915 case ConversionSpecifier::XArg:
|
|
|
916 case ConversionSpecifier::fArg:
|
|
|
917 case ConversionSpecifier::FArg:
|
|
|
918 case ConversionSpecifier::eArg:
|
|
|
919 case ConversionSpecifier::EArg:
|
|
|
920 case ConversionSpecifier::gArg:
|
|
|
921 case ConversionSpecifier::GArg:
|
|
|
922 case ConversionSpecifier::aArg:
|
|
|
923 case ConversionSpecifier::AArg:
|
|
|
924 case ConversionSpecifier::sArg:
|
|
|
925 case ConversionSpecifier::pArg:
|
|
|
926 case ConversionSpecifier::nArg:
|
|
|
927 case ConversionSpecifier::ObjCObjArg:
|
|
|
928 case ConversionSpecifier::ScanListArg:
|
|
|
929 case ConversionSpecifier::PercentArg:
|
|
|
930 case ConversionSpecifier::PArg:
|
|
|
931 return true;
|
|
|
932 case ConversionSpecifier::CArg:
|
|
|
933 case ConversionSpecifier::SArg:
|
|
|
934 return LangOpt.ObjC;
|
|
|
935 case ConversionSpecifier::InvalidSpecifier:
|
|
|
936 case ConversionSpecifier::FreeBSDbArg:
|
|
|
937 case ConversionSpecifier::FreeBSDDArg:
|
|
|
938 case ConversionSpecifier::FreeBSDrArg:
|
|
|
939 case ConversionSpecifier::FreeBSDyArg:
|
|
|
940 case ConversionSpecifier::PrintErrno:
|
|
|
941 case ConversionSpecifier::DArg:
|
|
|
942 case ConversionSpecifier::OArg:
|
|
|
943 case ConversionSpecifier::UArg:
|
|
|
944 case ConversionSpecifier::ZArg:
|
|
|
945 return false;
|
|
|
946 }
|
|
|
947 llvm_unreachable("Invalid ConversionSpecifier Kind!");
|
|
|
948 }
|
|
|
949
|
|
|
950 bool FormatSpecifier::hasStandardLengthConversionCombination() const {
|
|
|
951 if (LM.getKind() == LengthModifier::AsLongDouble) {
|
|
|
952 switch(CS.getKind()) {
|
|
|
953 case ConversionSpecifier::dArg:
|
|
|
954 case ConversionSpecifier::iArg:
|
|
|
955 case ConversionSpecifier::oArg:
|
|
|
956 case ConversionSpecifier::uArg:
|
|
|
957 case ConversionSpecifier::xArg:
|
|
|
958 case ConversionSpecifier::XArg:
|
|
|
959 return false;
|
|
|
960 default:
|
|
|
961 return true;
|
|
|
962 }
|
|
|
963 }
|
|
|
964 return true;
|
|
|
965 }
|
|
|
966
|
|
|
967 Optional<LengthModifier> FormatSpecifier::getCorrectedLengthModifier() const {
|
|
|
968 if (CS.isAnyIntArg() || CS.getKind() == ConversionSpecifier::nArg) {
|
|
|
969 if (LM.getKind() == LengthModifier::AsLongDouble ||
|
|
|
970 LM.getKind() == LengthModifier::AsQuad) {
|
|
|
971 LengthModifier FixedLM(LM);
|
|
|
972 FixedLM.setKind(LengthModifier::AsLongLong);
|
|
|
973 return FixedLM;
|
|
|
974 }
|
|
|
975 }
|
|
|
976
|
|
|
977 return None;
|
|
|
978 }
|
|
|
979
|
|
|
980 bool FormatSpecifier::namedTypeToLengthModifier(QualType QT,
|
|
|
981 LengthModifier &LM) {
|
|
|
982 assert(isa<TypedefType>(QT) && "Expected a TypedefType");
|
|
|
983 const TypedefNameDecl *Typedef = cast<TypedefType>(QT)->getDecl();
|
|
|
984
|
|
|
985 for (;;) {
|
|
|
986 const IdentifierInfo *Identifier = Typedef->getIdentifier();
|
|
|
987 if (Identifier->getName() == "size_t") {
|
|
|
988 LM.setKind(LengthModifier::AsSizeT);
|
|
|
989 return true;
|
|
|
990 } else if (Identifier->getName() == "ssize_t") {
|
|
|
991 // Not C99, but common in Unix.
|
|
|
992 LM.setKind(LengthModifier::AsSizeT);
|
|
|
993 return true;
|
|
|
994 } else if (Identifier->getName() == "intmax_t") {
|
|
|
995 LM.setKind(LengthModifier::AsIntMax);
|
|
|
996 return true;
|
|
|
997 } else if (Identifier->getName() == "uintmax_t") {
|
|
|
998 LM.setKind(LengthModifier::AsIntMax);
|
|
|
999 return true;
|
|
|
1000 } else if (Identifier->getName() == "ptrdiff_t") {
|
|
|
1001 LM.setKind(LengthModifier::AsPtrDiff);
|
|
|
1002 return true;
|
|
|
1003 }
|
|
|
1004
|
|
|
1005 QualType T = Typedef->getUnderlyingType();
|
|
|
1006 if (!isa<TypedefType>(T))
|
|
|
1007 break;
|
|
|
1008
|
|
|
1009 Typedef = cast<TypedefType>(T)->getDecl();
|
|
|
1010 }
|
|
|
1011 return false;
|
|
|
1012 }
|
