| 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. |
| 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 |
// 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'. |
| 355 |
if (!ETy->getDecl()->isComplete()) |
| 356 |
return NoMatch; |
| 357 |
argTy = ETy->getDecl()->getIntegerType(); |
| 358 |
} |
| 359 |
|
| 360 |
if (const auto *BT = argTy->getAs<BuiltinType>()) { |
| 361 |
// The types are perfectly matched? |
| 362 |
switch (BT->getKind()) { |
| 363 |
default: |
| 364 |
break; |
| 365 |
case BuiltinType::Char_S: |
| 366 |
case BuiltinType::SChar: |
| 367 |
case BuiltinType::UChar: |
| 368 |
case BuiltinType::Char_U: |
| 369 |
case BuiltinType::Bool: |
| 370 |
return Match; |
| 371 |
} |
| 372 |
// "Partially matched" because of promotions? |
| 373 |
if (!Ptr) { |
| 374 |
switch (BT->getKind()) { |
| 375 |
default: |
| 376 |
break; |
| 377 |
case BuiltinType::Int: |
| 378 |
case BuiltinType::UInt: |
| 379 |
return MatchPromotion; |
| 380 |
case BuiltinType::Short: |
| 381 |
case BuiltinType::UShort: |
| 382 |
case BuiltinType::WChar_S: |
| 383 |
case BuiltinType::WChar_U: |
| 384 |
return NoMatchPromotionTypeConfusion; |
| 385 |
} |
| 386 |
} |
| 387 |
} |
| 388 |
return NoMatch; |
| 389 |
} |
| 390 |
|
| 391 |
case SpecificTy: { |
| 392 |
if (const EnumType *ETy = argTy->getAs<EnumType>()) { |
| 393 |
// If the enum is incomplete we know nothing about the underlying type. |
| 394 |
// Assume that it's 'int'. |
| 395 |
if (!ETy->getDecl()->isComplete()) |
| 396 |
argTy = C.IntTy; |
| 397 |
else |
| 398 |
argTy = ETy->getDecl()->getIntegerType(); |
| 399 |
} |
| 400 |
argTy = C.getCanonicalType(argTy).getUnqualifiedType(); |
| 401 |
|
| 402 |
if (T == argTy) |
| 403 |
return Match; |
| 404 |
if (const auto *BT = argTy->getAs<BuiltinType>()) { |
| 405 |
// Check if the only difference between them is signed vs unsigned |
| 406 |
// if true, we consider they are compatible. |
| 407 |
switch (BT->getKind()) { |
| 408 |
default: |
| 409 |
break; |
| 410 |
case BuiltinType::Char_S: |
| 411 |
case BuiltinType::SChar: |
| 412 |
case BuiltinType::Char_U: |
| 413 |
case BuiltinType::UChar: |
| 414 |
case BuiltinType::Bool: |
| 415 |
if (T == C.UnsignedShortTy || T == C.ShortTy) |
| 416 |
return NoMatchTypeConfusion; |
| 417 |
if (T == C.UnsignedCharTy || T == C.SignedCharTy) |
| 418 |
return Match; |
| 419 |
break; |
| 420 |
case BuiltinType::Short: |
| 421 |
if (T == C.UnsignedShortTy) |
| 422 |
return Match; |
| 423 |
break; |
| 424 |
case BuiltinType::UShort: |
| 425 |
if (T == C.ShortTy) |
| 426 |
return Match; |
| 427 |
break; |
| 428 |
case BuiltinType::Int: |
| 429 |
if (T == C.UnsignedIntTy) |
| 430 |
return Match; |
| 431 |
break; |
| 432 |
case BuiltinType::UInt: |
| 433 |
if (T == C.IntTy) |
| 434 |
return Match; |
| 435 |
break; |
| 436 |
case BuiltinType::Long: |
| 437 |
if (T == C.UnsignedLongTy) |
| 438 |
return Match; |
| 439 |
break; |
| 440 |
case BuiltinType::ULong: |
| 441 |
if (T == C.LongTy) |
| 442 |
return Match; |
| 443 |
break; |
| 444 |
case BuiltinType::LongLong: |
| 445 |
if (T == C.UnsignedLongLongTy) |
| 446 |
return Match; |
| 447 |
break; |
| 448 |
case BuiltinType::ULongLong: |
| 449 |
if (T == C.LongLongTy) |
| 450 |
return Match; |
| 451 |
break; |
| 452 |
} |
| 453 |
// "Partially matched" because of promotions? |
| 454 |
if (!Ptr) { |
| 455 |
switch (BT->getKind()) { |
| 456 |
default: |
| 457 |
break; |
| 458 |
case BuiltinType::Int: |
| 459 |
case BuiltinType::UInt: |
| 460 |
if (T == C.SignedCharTy || T == C.UnsignedCharTy || |
| 461 |
T == C.ShortTy || T == C.UnsignedShortTy || T == C.WCharTy || |
| 462 |
T == C.WideCharTy) |
| 463 |
return MatchPromotion; |
| 464 |
break; |
| 465 |
case BuiltinType::Short: |
| 466 |
case BuiltinType::UShort: |
| 467 |
if (T == C.SignedCharTy || T == C.UnsignedCharTy) |
| 468 |
return NoMatchPromotionTypeConfusion; |
| 469 |
break; |
| 470 |
case BuiltinType::WChar_U: |
| 471 |
case BuiltinType::WChar_S: |
| 472 |
if (T != C.WCharTy && T != C.WideCharTy) |
| 473 |
return NoMatchPromotionTypeConfusion; |
| 474 |
} |
| 475 |
} |
| 476 |
} |
| 477 |
return NoMatch; |
| 478 |
} |
| 479 |
|
| 480 |
case CStrTy: { |
| 481 |
const PointerType *PT = argTy->getAs<PointerType>(); |
| 482 |
if (!PT) |
| 483 |
return NoMatch; |
| 484 |
QualType pointeeTy = PT->getPointeeType(); |
| 485 |
if (const BuiltinType *BT = pointeeTy->getAs<BuiltinType>()) |
| 486 |
switch (BT->getKind()) { |
| 487 |
case BuiltinType::Char_U: |
| 488 |
case BuiltinType::UChar: |
| 489 |
case BuiltinType::Char_S: |
| 490 |
case BuiltinType::SChar: |
| 491 |
return Match; |
| 492 |
default: |
| 493 |
break; |
| 494 |
} |
| 495 |
|
| 496 |
return NoMatch; |
| 497 |
} |
| 498 |
|
| 499 |
case WCStrTy: { |
| 500 |
const PointerType *PT = argTy->getAs<PointerType>(); |
| 501 |
if (!PT) |
| 502 |
return NoMatch; |
| 503 |
QualType pointeeTy = |
| 504 |
C.getCanonicalType(PT->getPointeeType()).getUnqualifiedType(); |
| 505 |
return pointeeTy == C.getWideCharType() ? Match : NoMatch; |
| 506 |
} |
| 507 |
|
| 508 |
case WIntTy: { |
| 509 |
QualType WInt = C.getCanonicalType(C.getWIntType()).getUnqualifiedType(); |
| 510 |
|
| 511 |
if (C.getCanonicalType(argTy).getUnqualifiedType() == WInt) |
| 512 |
return Match; |
| 513 |
|
| 514 |
QualType PromoArg = C.isPromotableIntegerType(argTy) |
| 515 |
? C.getPromotedIntegerType(argTy) |
| 516 |
: argTy; |
| 517 |
PromoArg = C.getCanonicalType(PromoArg).getUnqualifiedType(); |
| 518 |
|
| 519 |
// If the promoted argument is the corresponding signed type of the |
| 520 |
// wint_t type, then it should match. |
| 521 |
if (PromoArg->hasSignedIntegerRepresentation() && |
| 522 |
C.getCorrespondingUnsignedType(PromoArg) == WInt) |
| 523 |
return Match; |
| 524 |
|
| 525 |
return WInt == PromoArg ? Match : NoMatch; |
| 526 |
} |
| 527 |
|
| 528 |
case CPointerTy: |
| 529 |
if (argTy->isVoidPointerType()) { |
| 530 |
return Match; |
| 531 |
} if (argTy->isPointerType() || argTy->isObjCObjectPointerType() || |
| 532 |
argTy->isBlockPointerType() || argTy->isNullPtrType()) { |
| 533 |
return NoMatchPedantic; |
| 534 |
} else { |
| 535 |
return NoMatch; |
| 536 |
} |
| 537 |
|
| 538 |
case ObjCPointerTy: { |
| 539 |
if (argTy->getAs<ObjCObjectPointerType>() || |
| 540 |
argTy->getAs<BlockPointerType>()) |
| 541 |
return Match; |
| 542 |
|
| 543 |
// Handle implicit toll-free bridging. |
| 544 |
if (const PointerType *PT = argTy->getAs<PointerType>()) { |
| 545 |
// Things such as CFTypeRef are really just opaque pointers |
| 546 |
// to C structs representing CF types that can often be bridged |
| 547 |
// to Objective-C objects. Since the compiler doesn't know which |
| 548 |
// structs can be toll-free bridged, we just accept them all. |
| 549 |
QualType pointee = PT->getPointeeType(); |
| 550 |
if (pointee->getAsStructureType() || pointee->isVoidType()) |
| 551 |
return Match; |
| 552 |
} |
| 553 |
return NoMatch; |
| 554 |
} |
| 555 |
} |
| 556 |
|
| 557 |
llvm_unreachable("Invalid ArgType Kind!"); |
| 558 |
} |
| 559 |
|
| 560 |
ArgType ArgType::makeVectorType(ASTContext &C, unsigned NumElts) const { |
| 561 |
// Check for valid vector element types. |
| 562 |
if (T.isNull()) |
| 563 |
return ArgType::Invalid(); |
| 564 |
|
| 565 |
QualType Vec = C.getExtVectorType(T, NumElts); |
| 566 |
return ArgType(Vec, Name); |
| 567 |
} |
| 568 |
|
| 569 |
QualType ArgType::getRepresentativeType(ASTContext &C) const { |
| 570 |
QualType Res; |
| 571 |
switch (K) { |
| 572 |
case InvalidTy: |
| 573 |
llvm_unreachable("No representative type for Invalid ArgType"); |
| 574 |
case UnknownTy: |
| 575 |
llvm_unreachable("No representative type for Unknown ArgType"); |
| 576 |
case AnyCharTy: |
| 577 |
Res = C.CharTy; |
| 578 |
break; |
| 579 |
case SpecificTy: |
| 580 |
Res = T; |
| 581 |
break; |
| 582 |
case CStrTy: |
| 583 |
Res = C.getPointerType(C.CharTy); |
| 584 |
break; |
| 585 |
case WCStrTy: |
| 586 |
Res = C.getPointerType(C.getWideCharType()); |
| 587 |
break; |
| 588 |
case ObjCPointerTy: |
| 589 |
Res = C.ObjCBuiltinIdTy; |
| 590 |
break; |
| 591 |
case CPointerTy: |
| 592 |
Res = C.VoidPtrTy; |
| 593 |
break; |
| 594 |
case WIntTy: { |
| 595 |
Res = C.getWIntType(); |
| 596 |
break; |
| 597 |
} |
| 598 |
} |
| 599 |
|
| 600 |
if (Ptr) |
| 601 |
Res = C.getPointerType(Res); |
| 602 |
return Res; |
| 603 |
} |
| 604 |
|
| 605 |
std::string ArgType::getRepresentativeTypeName(ASTContext &C) const { |
| 606 |
std::string S = getRepresentativeType(C).getAsString(C.getPrintingPolicy()); |
| 607 |
|
| 608 |
std::string Alias; |
| 609 |
if (Name) { |
| 610 |
// Use a specific name for this type, e.g. "size_t". |
| 611 |
Alias = Name; |
| 612 |
if (Ptr) { |
| 613 |
// If ArgType is actually a pointer to T, append an asterisk. |
| 614 |
Alias += (Alias[Alias.size()-1] == '*') ? "*" : " *"; |
| 615 |
} |
| 616 |
// If Alias is the same as the underlying type, e.g. wchar_t, then drop it. |
| 617 |
if (S == Alias) |
| 618 |
Alias.clear(); |
| 619 |
} |
| 620 |
|
| 621 |
if (!Alias.empty()) |
| 622 |
return std::string("'") + Alias + "' (aka '" + S + "')"; |
| 623 |
return std::string("'") + S + "'"; |
| 624 |
} |
| 625 |
|
| 626 |
|
| 627 |
//===----------------------------------------------------------------------===// |
| 628 |
// Methods on OptionalAmount. |
| 629 |
//===----------------------------------------------------------------------===// |
| 630 |
|
| 631 |
ArgType |
| 632 |
analyze_format_string::OptionalAmount::getArgType(ASTContext &Ctx) const { |
| 633 |
return Ctx.IntTy; |
| 634 |
} |
| 635 |
|
| 636 |
//===----------------------------------------------------------------------===// |
| 637 |
// Methods on LengthModifier. |
| 638 |
//===----------------------------------------------------------------------===// |
| 639 |
|
| 640 |
const char * |
| 641 |
analyze_format_string::LengthModifier::toString() const { |
| 642 |
switch (kind) { |
| 643 |
case AsChar: |
| 644 |
return "hh"; |
| 645 |
case AsShort: |
| 646 |
return "h"; |
| 647 |
case AsShortLong: |
| 648 |
return "hl"; |
| 649 |
case AsLong: // or AsWideChar |
| 650 |
return "l"; |
| 651 |
case AsLongLong: |
| 652 |
return "ll"; |
| 653 |
case AsQuad: |
| 654 |
return "q"; |
| 655 |
case AsIntMax: |
| 656 |
return "j"; |
| 657 |
case AsSizeT: |
| 658 |
return "z"; |
| 659 |
case AsPtrDiff: |
| 660 |
return "t"; |
| 661 |
case AsInt32: |
| 662 |
return "I32"; |
| 663 |
case AsInt3264: |
| 664 |
return "I"; |
| 665 |
case AsInt64: |
| 666 |
return "I64"; |
| 667 |
case AsLongDouble: |
| 668 |
return "L"; |
| 669 |
case AsAllocate: |
| 670 |
return "a"; |
| 671 |
case AsMAllocate: |
| 672 |
return "m"; |
| 673 |
case AsWide: |
| 674 |
return "w"; |
| 675 |
case None: |
| 676 |
return ""; |
| 677 |
} |
| 678 |
return nullptr; |
| 679 |
} |
| 680 |
|
| 681 |
//===----------------------------------------------------------------------===// |
| 682 |
// Methods on ConversionSpecifier. |
| 683 |
//===----------------------------------------------------------------------===// |
| 684 |
|
| 685 |
const char *ConversionSpecifier::toString() const { |
| 686 |
switch (kind) { |
| 687 |
case bArg: return "b"; |
| 688 |
case BArg: return "B"; |
| 689 |
case dArg: return "d"; |
| 690 |
case DArg: return "D"; |
| 691 |
case iArg: return "i"; |
| 692 |
case oArg: return "o"; |
| 693 |
case OArg: return "O"; |
| 694 |
case uArg: return "u"; |
| 695 |
case UArg: return "U"; |
| 696 |
case xArg: return "x"; |
| 697 |
case XArg: return "X"; |
| 698 |
case fArg: return "f"; |
| 699 |
case FArg: return "F"; |
| 700 |
case eArg: return "e"; |
| 701 |
case EArg: return "E"; |
| 702 |
case gArg: return "g"; |
| 703 |
case GArg: return "G"; |
| 704 |
case aArg: return "a"; |
| 705 |
case AArg: return "A"; |
| 706 |
case cArg: return "c"; |
| 707 |
case sArg: return "s"; |
| 708 |
case pArg: return "p"; |
| 709 |
case PArg: |
| 710 |
return "P"; |
| 711 |
case nArg: return "n"; |
| 712 |
case PercentArg: return "%"; |
| 713 |
case ScanListArg: return "["; |
| 714 |
case InvalidSpecifier: return nullptr; |
| 715 |
|
| 716 |
// POSIX unicode extensions. |
| 717 |
case CArg: return "C"; |
| 718 |
case SArg: return "S"; |
| 719 |
|
| 720 |
// Objective-C specific specifiers. |
| 721 |
case ObjCObjArg: return "@"; |
| 722 |
|
| 723 |
// FreeBSD kernel specific specifiers. |
| 724 |
case FreeBSDbArg: return "b"; |
| 725 |
case FreeBSDDArg: return "D"; |
| 726 |
case FreeBSDrArg: return "r"; |
| 727 |
case FreeBSDyArg: return "y"; |
| 728 |
|
| 729 |
// GlibC specific specifiers. |
| 730 |
case PrintErrno: return "m"; |
| 731 |
|
| 732 |
// MS specific specifiers. |
| 733 |
case ZArg: return "Z"; |
| 734 |
} |
| 735 |
return nullptr; |
| 736 |
} |
| 737 |
|
| 738 |
std::optional<ConversionSpecifier> |
| 739 |
ConversionSpecifier::getStandardSpecifier() const { |
| 740 |
ConversionSpecifier::Kind NewKind; |
| 741 |
|
| 742 |
switch (getKind()) { |
| 743 |
default: |
| 744 |
return std::nullopt; |
| 745 |
case DArg: |
| 746 |
NewKind = dArg; |
| 747 |
break; |
| 748 |
case UArg: |
| 749 |
NewKind = uArg; |
| 750 |
break; |
| 751 |
case OArg: |
| 752 |
NewKind = oArg; |
| 753 |
break; |
| 754 |
} |
| 755 |
|
| 756 |
ConversionSpecifier FixedCS(*this); |
| 757 |
FixedCS.setKind(NewKind); |
| 758 |
return FixedCS; |
| 759 |
} |
| 760 |
|
| 761 |
//===----------------------------------------------------------------------===// |
| 762 |
// Methods on OptionalAmount. |
| 763 |
//===----------------------------------------------------------------------===// |
| 764 |
|
| 765 |
void OptionalAmount::toString(raw_ostream &os) const { |
| 766 |
switch (hs) { |
| 767 |
case Invalid: |
| 768 |
case NotSpecified: |
| 769 |
return; |
| 770 |
case Arg: |
| 771 |
if (UsesDotPrefix) |
| 772 |
os << "."; |
| 773 |
if (usesPositionalArg()) |
| 774 |
os << "*" << getPositionalArgIndex() << "$"; |
| 775 |
else |
| 776 |
os << "*"; |
| 777 |
break; |
| 778 |
case Constant: |
| 779 |
if (UsesDotPrefix) |
| 780 |
os << "."; |
| 781 |
os << amt; |
| 782 |
break; |
| 783 |
} |
| 784 |
} |
| 785 |
|
| 786 |
bool FormatSpecifier::hasValidLengthModifier(const TargetInfo &Target, |
| 787 |
const LangOptions &LO) const { |
| 788 |
switch (LM.getKind()) { |
| 789 |
case LengthModifier::None: |
| 790 |
return true; |
| 791 |
|
| 792 |
// Handle most integer flags |
| 793 |
case LengthModifier::AsShort: |
| 794 |
// Length modifier only applies to FP vectors. |
| 795 |
if (LO.OpenCL && CS.isDoubleArg()) |
| 796 |
return !VectorNumElts.isInvalid(); |
| 797 |
|
| 798 |
if (Target.getTriple().isOSMSVCRT()) { |
| 799 |
switch (CS.getKind()) { |
| 800 |
case ConversionSpecifier::cArg: |
| 801 |
case ConversionSpecifier::CArg: |
| 802 |
case ConversionSpecifier::sArg: |
| 803 |
case ConversionSpecifier::SArg: |
| 804 |
case ConversionSpecifier::ZArg: |
| 805 |
return true; |
| 806 |
default: |
| 807 |
break; |
| 808 |
} |
| 809 |
} |
| 810 |
[[fallthrough]]; |
| 811 |
case LengthModifier::AsChar: |
| 812 |
case LengthModifier::AsLongLong: |
| 813 |
case LengthModifier::AsQuad: |
| 814 |
case LengthModifier::AsIntMax: |
| 815 |
case LengthModifier::AsSizeT: |
| 816 |
case LengthModifier::AsPtrDiff: |
| 817 |
switch (CS.getKind()) { |
| 818 |
case ConversionSpecifier::bArg: |
| 819 |
case ConversionSpecifier::BArg: |
| 820 |
case ConversionSpecifier::dArg: |
| 821 |
case ConversionSpecifier::DArg: |
| 822 |
case ConversionSpecifier::iArg: |
| 823 |
case ConversionSpecifier::oArg: |
| 824 |
case ConversionSpecifier::OArg: |
| 825 |
case ConversionSpecifier::uArg: |
| 826 |
case ConversionSpecifier::UArg: |
| 827 |
case ConversionSpecifier::xArg: |
| 828 |
case ConversionSpecifier::XArg: |
| 829 |
case ConversionSpecifier::nArg: |
| 830 |
return true; |
| 831 |
case ConversionSpecifier::FreeBSDrArg: |
| 832 |
case ConversionSpecifier::FreeBSDyArg: |
| 833 |
return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS(); |
| 834 |
default: |
| 835 |
return false; |
| 836 |
} |
| 837 |
|
| 838 |
case LengthModifier::AsShortLong: |
| 839 |
return LO.OpenCL && !VectorNumElts.isInvalid(); |
| 840 |
|
| 841 |
// Handle 'l' flag |
| 842 |
case LengthModifier::AsLong: // or AsWideChar |
| 843 |
if (CS.isDoubleArg()) { |
| 844 |
// Invalid for OpenCL FP scalars. |
| 845 |
if (LO.OpenCL && VectorNumElts.isInvalid()) |
| 846 |
return false; |
| 847 |
return true; |
| 848 |
} |
| 849 |
|
| 850 |
switch (CS.getKind()) { |
| 851 |
case ConversionSpecifier::dArg: |
| 852 |
case ConversionSpecifier::DArg: |
| 853 |
case ConversionSpecifier::iArg: |
| 854 |
case ConversionSpecifier::oArg: |
| 855 |
case ConversionSpecifier::OArg: |
| 856 |
case ConversionSpecifier::uArg: |
| 857 |
case ConversionSpecifier::UArg: |
| 858 |
case ConversionSpecifier::xArg: |
| 859 |
case ConversionSpecifier::XArg: |
| 860 |
case ConversionSpecifier::nArg: |
| 861 |
case ConversionSpecifier::cArg: |
| 862 |
case ConversionSpecifier::sArg: |
| 863 |
case ConversionSpecifier::ScanListArg: |
| 864 |
case ConversionSpecifier::ZArg: |
| 865 |
return true; |
| 866 |
case ConversionSpecifier::FreeBSDrArg: |
| 867 |
case ConversionSpecifier::FreeBSDyArg: |
| 868 |
return Target.getTriple().isOSFreeBSD() || Target.getTriple().isPS(); |
| 869 |
default: |
| 870 |
return false; |
| 871 |
} |
| 872 |
|
| 873 |
case LengthModifier::AsLongDouble: |
| 874 |
switch (CS.getKind()) { |
| 875 |
case ConversionSpecifier::aArg: |
| 876 |
case ConversionSpecifier::AArg: |
| 877 |
case ConversionSpecifier::fArg: |
| 878 |
case ConversionSpecifier::FArg: |
| 879 |
case ConversionSpecifier::eArg: |
| 880 |
case ConversionSpecifier::EArg: |
| 881 |
case ConversionSpecifier::gArg: |
| 882 |
case ConversionSpecifier::GArg: |
| 883 |
return true; |
| 884 |
// GNU libc extension. |
| 885 |
case ConversionSpecifier::dArg: |
| 886 |
case ConversionSpecifier::iArg: |
| 887 |
case ConversionSpecifier::oArg: |
| 888 |
case ConversionSpecifier::uArg: |
| 889 |
case ConversionSpecifier::xArg: |
| 890 |
case ConversionSpecifier::XArg: |
| 891 |
return !Target.getTriple().isOSDarwin() && |
| 892 |
!Target.getTriple().isOSWindows(); |
| 893 |
default: |
| 894 |
return false; |
| 895 |
} |
| 896 |
|
| 897 |
case LengthModifier::AsAllocate: |
| 898 |
switch (CS.getKind()) { |
| 899 |
case ConversionSpecifier::sArg: |
| 900 |
case ConversionSpecifier::SArg: |
| 901 |
case ConversionSpecifier::ScanListArg: |
| 902 |
return true; |
| 903 |
default: |
| 904 |
return false; |
| 905 |
} |
| 906 |
|
| 907 |
case LengthModifier::AsMAllocate: |
| 908 |
switch (CS.getKind()) { |
| 909 |
case ConversionSpecifier::cArg: |
| 910 |
case ConversionSpecifier::CArg: |
| 911 |
case ConversionSpecifier::sArg: |
| 912 |
case ConversionSpecifier::SArg: |
| 913 |
case ConversionSpecifier::ScanListArg: |
| 914 |
return true; |
| 915 |
default: |
| 916 |
return false; |
| 917 |
} |
| 918 |
case LengthModifier::AsInt32: |
| 919 |
case LengthModifier::AsInt3264: |
| 920 |
case LengthModifier::AsInt64: |
| 921 |
switch (CS.getKind()) { |
| 922 |
case ConversionSpecifier::dArg: |
| 923 |
case ConversionSpecifier::iArg: |
| 924 |
case ConversionSpecifier::oArg: |
| 925 |
case ConversionSpecifier::uArg: |
| 926 |
case ConversionSpecifier::xArg: |
| 927 |
case ConversionSpecifier::XArg: |
| 928 |
return Target.getTriple().isOSMSVCRT(); |
| 929 |
default: |
| 930 |
return false; |
| 931 |
} |
| 932 |
case LengthModifier::AsWide: |
| 933 |
switch (CS.getKind()) { |
| 934 |
case ConversionSpecifier::cArg: |
| 935 |
case ConversionSpecifier::CArg: |
| 936 |
case ConversionSpecifier::sArg: |
| 937 |
case ConversionSpecifier::SArg: |
| 938 |
case ConversionSpecifier::ZArg: |
| 939 |
return Target.getTriple().isOSMSVCRT(); |
| 940 |
default: |
| 941 |
return false; |
| 942 |
} |
| 943 |
} |
| 944 |
llvm_unreachable("Invalid LengthModifier Kind!"); |
| 945 |
} |
| 946 |
|
| 947 |
bool FormatSpecifier::hasStandardLengthModifier() const { |
| 948 |
switch (LM.getKind()) { |
| 949 |
case LengthModifier::None: |
| 950 |
case LengthModifier::AsChar: |
| 951 |
case LengthModifier::AsShort: |
| 952 |
case LengthModifier::AsLong: |
| 953 |
case LengthModifier::AsLongLong: |
| 954 |
case LengthModifier::AsIntMax: |
| 955 |
case LengthModifier::AsSizeT: |
| 956 |
case LengthModifier::AsPtrDiff: |
| 957 |
case LengthModifier::AsLongDouble: |
| 958 |
return true; |
| 959 |
case LengthModifier::AsAllocate: |
| 960 |
case LengthModifier::AsMAllocate: |
| 961 |
case LengthModifier::AsQuad: |
| 962 |
case LengthModifier::AsInt32: |
| 963 |
case LengthModifier::AsInt3264: |
| 964 |
case LengthModifier::AsInt64: |
| 965 |
case LengthModifier::AsWide: |
| 966 |
case LengthModifier::AsShortLong: // ??? |
| 967 |
return false; |
| 968 |
} |
| 969 |
llvm_unreachable("Invalid LengthModifier Kind!"); |
| 970 |
} |
| 971 |
|
| 972 |
bool FormatSpecifier::hasStandardConversionSpecifier( |
| 973 |
const LangOptions &LangOpt) const { |
| 974 |
switch (CS.getKind()) { |
| 975 |
case ConversionSpecifier::bArg: |
| 976 |
case ConversionSpecifier::BArg: |
| 977 |
case ConversionSpecifier::cArg: |
| 978 |
case ConversionSpecifier::dArg: |
| 979 |
case ConversionSpecifier::iArg: |
| 980 |
case ConversionSpecifier::oArg: |
| 981 |
case ConversionSpecifier::uArg: |
| 982 |
case ConversionSpecifier::xArg: |
| 983 |
case ConversionSpecifier::XArg: |
| 984 |
case ConversionSpecifier::fArg: |
| 985 |
case ConversionSpecifier::FArg: |
| 986 |
case ConversionSpecifier::eArg: |
| 987 |
case ConversionSpecifier::EArg: |
| 988 |
case ConversionSpecifier::gArg: |
| 989 |
case ConversionSpecifier::GArg: |
| 990 |
case ConversionSpecifier::aArg: |
| 991 |
case ConversionSpecifier::AArg: |
| 992 |
case ConversionSpecifier::sArg: |
| 993 |
case ConversionSpecifier::pArg: |
| 994 |
case ConversionSpecifier::nArg: |
| 995 |
case ConversionSpecifier::ObjCObjArg: |
| 996 |
case ConversionSpecifier::ScanListArg: |
| 997 |
case ConversionSpecifier::PercentArg: |
| 998 |
case ConversionSpecifier::PArg: |
| 999 |
return true; |
| 1000 |
case ConversionSpecifier::CArg: |
| 1001 |
case ConversionSpecifier::SArg: |
| 1002 |
return LangOpt.ObjC; |
| 1003 |
case ConversionSpecifier::InvalidSpecifier: |
| 1004 |
case ConversionSpecifier::FreeBSDbArg: |
| 1005 |
case ConversionSpecifier::FreeBSDDArg: |
| 1006 |
case ConversionSpecifier::FreeBSDrArg: |
| 1007 |
case ConversionSpecifier::FreeBSDyArg: |
| 1008 |
case ConversionSpecifier::PrintErrno: |
| 1009 |
case ConversionSpecifier::DArg: |
| 1010 |
case ConversionSpecifier::OArg: |
| 1011 |
case ConversionSpecifier::UArg: |
| 1012 |
case ConversionSpecifier::ZArg: |
| 1013 |
return false; |
| 1014 |
} |
| 1015 |
llvm_unreachable("Invalid ConversionSpecifier Kind!"); |
| 1016 |
} |
| 1017 |
|
| 1018 |
bool FormatSpecifier::hasStandardLengthConversionCombination() const { |
| 1019 |
if (LM.getKind() == LengthModifier::AsLongDouble) { |
| 1020 |
switch(CS.getKind()) { |
| 1021 |
case ConversionSpecifier::dArg: |
| 1022 |
case ConversionSpecifier::iArg: |
| 1023 |
case ConversionSpecifier::oArg: |
| 1024 |
case ConversionSpecifier::uArg: |
| 1025 |
case ConversionSpecifier::xArg: |
| 1026 |
case ConversionSpecifier::XArg: |
| 1027 |
return false; |
| 1028 |
default: |
| 1029 |
return true; |
| 1030 |
} |
| 1031 |
} |
| 1032 |
return true; |
| 1033 |
} |
| 1034 |
|
| 1035 |
std::optional<LengthModifier> |
| 1036 |
FormatSpecifier::getCorrectedLengthModifier() const { |
| 1037 |
if (CS.isAnyIntArg() || CS.getKind() == ConversionSpecifier::nArg) { |
| 1038 |
if (LM.getKind() == LengthModifier::AsLongDouble || |
| 1039 |
LM.getKind() == LengthModifier::AsQuad) { |
| 1040 |
LengthModifier FixedLM(LM); |
| 1041 |
FixedLM.setKind(LengthModifier::AsLongLong); |
| 1042 |
return FixedLM; |
| 1043 |
} |
| 1044 |
} |
| 1045 |
|
| 1046 |
return std::nullopt; |
| 1047 |
} |
| 1048 |
|
| 1049 |
bool FormatSpecifier::namedTypeToLengthModifier(QualType QT, |
| 1050 |
LengthModifier &LM) { |
| 1051 |
for (/**/; const auto *TT = QT->getAs<TypedefType>(); |
| 1052 |
QT = TT->getDecl()->getUnderlyingType()) { |
| 1053 |
const TypedefNameDecl *Typedef = TT->getDecl(); |
| 1054 |
const IdentifierInfo *Identifier = Typedef->getIdentifier(); |
| 1055 |
if (Identifier->getName() == "size_t") { |
| 1056 |
LM.setKind(LengthModifier::AsSizeT); |
| 1057 |
return true; |
| 1058 |
} else if (Identifier->getName() == "ssize_t") { |
| 1059 |
// Not C99, but common in Unix. |
| 1060 |
LM.setKind(LengthModifier::AsSizeT); |
| 1061 |
return true; |
| 1062 |
} else if (Identifier->getName() == "intmax_t") { |
| 1063 |
LM.setKind(LengthModifier::AsIntMax); |
| 1064 |
return true; |
| 1065 |
} else if (Identifier->getName() == "uintmax_t") { |
| 1066 |
LM.setKind(LengthModifier::AsIntMax); |
| 1067 |
return true; |
| 1068 |
} else if (Identifier->getName() == "ptrdiff_t") { |
| 1069 |
LM.setKind(LengthModifier::AsPtrDiff); |
| 1070 |
return true; |
| 1071 |
} |
| 1072 |
} |
| 1073 |
return false; |
| 1074 |
} |