1 //===-- Args.cpp ----------------------------------------------------------===// 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 #include "lldb/Utility/Args.h" 10 #include "lldb/Utility/ConstString.h" 11 #include "lldb/Utility/FileSpec.h" 12 #include "lldb/Utility/Stream.h" 13 #include "lldb/Utility/StringList.h" 14 #include "llvm/ADT/StringSwitch.h" 15 16 using namespace lldb; 17 using namespace lldb_private; 18 19 // A helper function for argument parsing. 20 // Parses the initial part of the first argument using normal double quote 21 // rules: backslash escapes the double quote and itself. The parsed string is 22 // appended to the second argument. The function returns the unparsed portion 23 // of the string, starting at the closing quote. 24 static llvm::StringRef ParseDoubleQuotes(llvm::StringRef quoted, 25 std::string &result) { 26 // Inside double quotes, '\' and '"' are special. 27 static const char *k_escapable_characters = "\"\\"; 28 while (true) { 29 // Skip over over regular characters and append them. 30 size_t regular = quoted.find_first_of(k_escapable_characters); 31 result += quoted.substr(0, regular); 32 quoted = quoted.substr(regular); 33 34 // If we have reached the end of string or the closing quote, we're done. 35 if (quoted.empty() || quoted.front() == '"') 36 break; 37 38 // We have found a backslash. 39 quoted = quoted.drop_front(); 40 41 if (quoted.empty()) { 42 // A lone backslash at the end of string, let's just append it. 43 result += '\\'; 44 break; 45 } 46 47 // If the character after the backslash is not an allowed escapable 48 // character, we leave the character sequence untouched. 49 if (strchr(k_escapable_characters, quoted.front()) == nullptr) 50 result += '\\'; 51 52 result += quoted.front(); 53 quoted = quoted.drop_front(); 54 } 55 56 return quoted; 57 } 58 59 static size_t ArgvToArgc(const char **argv) { 60 if (!argv) 61 return 0; 62 size_t count = 0; 63 while (*argv++) 64 ++count; 65 return count; 66 } 67 68 // Trims all whitespace that can separate command line arguments from the left 69 // side of the string. 70 static llvm::StringRef ltrimForArgs(llvm::StringRef str) { 71 static const char *k_space_separators = " \t"; 72 return str.ltrim(k_space_separators); 73 } 74 75 // A helper function for SetCommandString. Parses a single argument from the 76 // command string, processing quotes and backslashes in a shell-like manner. 77 // The function returns a tuple consisting of the parsed argument, the quote 78 // char used, and the unparsed portion of the string starting at the first 79 // unqouted, unescaped whitespace character. 80 static std::tuple<std::string, char, llvm::StringRef> 81 ParseSingleArgument(llvm::StringRef command) { 82 // Argument can be split into multiple discontiguous pieces, for example: 83 // "Hello ""World" 84 // this would result in a single argument "Hello World" (without the quotes) 85 // since the quotes would be removed and there is not space between the 86 // strings. 87 std::string arg; 88 89 // Since we can have multiple quotes that form a single command in a command 90 // like: "Hello "world'!' (which will make a single argument "Hello world!") 91 // we remember the first quote character we encounter and use that for the 92 // quote character. 93 char first_quote_char = '\0'; 94 95 bool arg_complete = false; 96 do { 97 // Skip over over regular characters and append them. 98 size_t regular = command.find_first_of(" \t\r\"'`\\"); 99 arg += command.substr(0, regular); 100 command = command.substr(regular); 101 102 if (command.empty()) 103 break; 104 105 char special = command.front(); 106 command = command.drop_front(); 107 switch (special) { 108 case '\\': 109 if (command.empty()) { 110 arg += '\\'; 111 break; 112 } 113 114 // If the character after the backslash is not an allowed escapable 115 // character, we leave the character sequence untouched. 116 if (strchr(" \t\\'\"`", command.front()) == nullptr) 117 arg += '\\'; 118 119 arg += command.front(); 120 command = command.drop_front(); 121 122 break; 123 124 case ' ': 125 case '\t': 126 case '\r': 127 // We are not inside any quotes, we just found a space after an argument. 128 // We are done. 129 arg_complete = true; 130 break; 131 132 case '"': 133 case '\'': 134 case '`': 135 // We found the start of a quote scope. 136 if (first_quote_char == '\0') 137 first_quote_char = special; 138 139 if (special == '"') 140 command = ParseDoubleQuotes(command, arg); 141 else { 142 // For single quotes, we simply skip ahead to the matching quote 143 // character (or the end of the string). 144 size_t quoted = command.find(special); 145 arg += command.substr(0, quoted); 146 command = command.substr(quoted); 147 } 148 149 // If we found a closing quote, skip it. 150 if (!command.empty()) 151 command = command.drop_front(); 152 153 break; 154 } 155 } while (!arg_complete); 156 157 return std::make_tuple(arg, first_quote_char, command); 158 } 159 160 Args::ArgEntry::ArgEntry(llvm::StringRef str, char quote) : quote(quote) { 161 size_t size = str.size(); 162 ptr.reset(new char[size + 1]); 163 164 ::memcpy(data(), str.data() ? str.data() : "", size); 165 ptr[size] = 0; 166 } 167 168 // Args constructor 169 Args::Args(llvm::StringRef command) { SetCommandString(command); } 170 171 Args::Args(const Args &rhs) { *this = rhs; } 172 173 Args::Args(const StringList &list) : Args() { 174 for (const std::string &arg : list) 175 AppendArgument(arg); 176 } 177 178 Args::Args(llvm::ArrayRef<llvm::StringRef> args) : Args() { 179 for (llvm::StringRef arg : args) 180 AppendArgument(arg); 181 } 182 183 Args &Args::operator=(const Args &rhs) { 184 Clear(); 185 186 m_argv.clear(); 187 m_entries.clear(); 188 for (auto &entry : rhs.m_entries) { 189 m_entries.emplace_back(entry.ref(), entry.quote); 190 m_argv.push_back(m_entries.back().data()); 191 } 192 m_argv.push_back(nullptr); 193 return *this; 194 } 195 196 // Destructor 197 Args::~Args() = default; 198 199 void Args::Dump(Stream &s, const char *label_name) const { 200 if (!label_name) 201 return; 202 203 int i = 0; 204 for (auto &entry : m_entries) { 205 s.Indent(); 206 s.Format("{0}[{1}]=\"{2}\"\n", label_name, i++, entry.ref()); 207 } 208 s.Format("{0}[{1}]=NULL\n", label_name, i); 209 s.EOL(); 210 } 211 212 bool Args::GetCommandString(std::string &command) const { 213 command.clear(); 214 215 for (size_t i = 0; i < m_entries.size(); ++i) { 216 if (i > 0) 217 command += ' '; 218 command += m_entries[i].ref(); 219 } 220 221 return !m_entries.empty(); 222 } 223 224 bool Args::GetQuotedCommandString(std::string &command) const { 225 command.clear(); 226 227 for (size_t i = 0; i < m_entries.size(); ++i) { 228 if (i > 0) 229 command += ' '; 230 231 if (m_entries[i].quote) { 232 command += m_entries[i].quote; 233 command += m_entries[i].ref(); 234 command += m_entries[i].quote; 235 } else { 236 command += m_entries[i].ref(); 237 } 238 } 239 240 return !m_entries.empty(); 241 } 242 243 void Args::SetCommandString(llvm::StringRef command) { 244 Clear(); 245 m_argv.clear(); 246 247 command = ltrimForArgs(command); 248 std::string arg; 249 char quote; 250 while (!command.empty()) { 251 std::tie(arg, quote, command) = ParseSingleArgument(command); 252 m_entries.emplace_back(arg, quote); 253 m_argv.push_back(m_entries.back().data()); 254 command = ltrimForArgs(command); 255 } 256 m_argv.push_back(nullptr); 257 } 258 259 const char *Args::GetArgumentAtIndex(size_t idx) const { 260 if (idx < m_argv.size()) 261 return m_argv[idx]; 262 return nullptr; 263 } 264 265 char **Args::GetArgumentVector() { 266 assert(!m_argv.empty()); 267 // TODO: functions like execve and posix_spawnp exhibit undefined behavior 268 // when argv or envp is null. So the code below is actually wrong. However, 269 // other code in LLDB depends on it being null. The code has been acting 270 // this way for some time, so it makes sense to leave it this way until 271 // someone has the time to come along and fix it. 272 return (m_argv.size() > 1) ? m_argv.data() : nullptr; 273 } 274 275 const char **Args::GetConstArgumentVector() const { 276 assert(!m_argv.empty()); 277 return (m_argv.size() > 1) ? const_cast<const char **>(m_argv.data()) 278 : nullptr; 279 } 280 281 void Args::Shift() { 282 // Don't pop the last NULL terminator from the argv array 283 if (m_entries.empty()) 284 return; 285 m_argv.erase(m_argv.begin()); 286 m_entries.erase(m_entries.begin()); 287 } 288 289 void Args::Unshift(llvm::StringRef arg_str, char quote_char) { 290 InsertArgumentAtIndex(0, arg_str, quote_char); 291 } 292 293 void Args::AppendArguments(const Args &rhs) { 294 assert(m_argv.size() == m_entries.size() + 1); 295 assert(m_argv.back() == nullptr); 296 m_argv.pop_back(); 297 for (auto &entry : rhs.m_entries) { 298 m_entries.emplace_back(entry.ref(), entry.quote); 299 m_argv.push_back(m_entries.back().data()); 300 } 301 m_argv.push_back(nullptr); 302 } 303 304 void Args::AppendArguments(const char **argv) { 305 size_t argc = ArgvToArgc(argv); 306 307 assert(m_argv.size() == m_entries.size() + 1); 308 assert(m_argv.back() == nullptr); 309 m_argv.pop_back(); 310 for (auto arg : llvm::makeArrayRef(argv, argc)) { 311 m_entries.emplace_back(arg, '\0'); 312 m_argv.push_back(m_entries.back().data()); 313 } 314 315 m_argv.push_back(nullptr); 316 } 317 318 void Args::AppendArgument(llvm::StringRef arg_str, char quote_char) { 319 InsertArgumentAtIndex(GetArgumentCount(), arg_str, quote_char); 320 } 321 322 void Args::InsertArgumentAtIndex(size_t idx, llvm::StringRef arg_str, 323 char quote_char) { 324 assert(m_argv.size() == m_entries.size() + 1); 325 assert(m_argv.back() == nullptr); 326 327 if (idx > m_entries.size()) 328 return; 329 m_entries.emplace(m_entries.begin() + idx, arg_str, quote_char); 330 m_argv.insert(m_argv.begin() + idx, m_entries[idx].data()); 331 } 332 333 void Args::ReplaceArgumentAtIndex(size_t idx, llvm::StringRef arg_str, 334 char quote_char) { 335 assert(m_argv.size() == m_entries.size() + 1); 336 assert(m_argv.back() == nullptr); 337 338 if (idx >= m_entries.size()) 339 return; 340 341 m_entries[idx] = ArgEntry(arg_str, quote_char); 342 m_argv[idx] = m_entries[idx].data(); 343 } 344 345 void Args::DeleteArgumentAtIndex(size_t idx) { 346 if (idx >= m_entries.size()) 347 return; 348 349 m_argv.erase(m_argv.begin() + idx); 350 m_entries.erase(m_entries.begin() + idx); 351 } 352 353 void Args::SetArguments(size_t argc, const char **argv) { 354 Clear(); 355 356 auto args = llvm::makeArrayRef(argv, argc); 357 m_entries.resize(argc); 358 m_argv.resize(argc + 1); 359 for (size_t i = 0; i < args.size(); ++i) { 360 char quote = 361 ((args[i][0] == '\'') || (args[i][0] == '"') || (args[i][0] == '`')) 362 ? args[i][0] 363 : '\0'; 364 365 m_entries[i] = ArgEntry(args[i], quote); 366 m_argv[i] = m_entries[i].data(); 367 } 368 } 369 370 void Args::SetArguments(const char **argv) { 371 SetArguments(ArgvToArgc(argv), argv); 372 } 373 374 void Args::Clear() { 375 m_entries.clear(); 376 m_argv.clear(); 377 m_argv.push_back(nullptr); 378 } 379 380 std::string Args::GetShellSafeArgument(const FileSpec &shell, 381 llvm::StringRef unsafe_arg) { 382 struct ShellDescriptor { 383 ConstString m_basename; 384 llvm::StringRef m_escapables; 385 }; 386 387 static ShellDescriptor g_Shells[] = {{ConstString("bash"), " '\"<>()&;"}, 388 {ConstString("tcsh"), " '\"<>()&;"}, 389 {ConstString("zsh"), " '\"<>()&;\\|"}, 390 {ConstString("sh"), " '\"<>()&;"}}; 391 392 // safe minimal set 393 llvm::StringRef escapables = " '\""; 394 395 if (auto basename = shell.GetFilename()) { 396 for (const auto &Shell : g_Shells) { 397 if (Shell.m_basename == basename) { 398 escapables = Shell.m_escapables; 399 break; 400 } 401 } 402 } 403 404 std::string safe_arg; 405 safe_arg.reserve(unsafe_arg.size()); 406 // Add a \ before every character that needs to be escaped. 407 for (char c : unsafe_arg) { 408 if (escapables.contains(c)) 409 safe_arg.push_back('\\'); 410 safe_arg.push_back(c); 411 } 412 return safe_arg; 413 } 414 415 lldb::Encoding Args::StringToEncoding(llvm::StringRef s, 416 lldb::Encoding fail_value) { 417 return llvm::StringSwitch<lldb::Encoding>(s) 418 .Case("uint", eEncodingUint) 419 .Case("sint", eEncodingSint) 420 .Case("ieee754", eEncodingIEEE754) 421 .Case("vector", eEncodingVector) 422 .Default(fail_value); 423 } 424 425 uint32_t Args::StringToGenericRegister(llvm::StringRef s) { 426 if (s.empty()) 427 return LLDB_INVALID_REGNUM; 428 uint32_t result = llvm::StringSwitch<uint32_t>(s) 429 .Case("pc", LLDB_REGNUM_GENERIC_PC) 430 .Case("sp", LLDB_REGNUM_GENERIC_SP) 431 .Case("fp", LLDB_REGNUM_GENERIC_FP) 432 .Cases("ra", "lr", LLDB_REGNUM_GENERIC_RA) 433 .Case("flags", LLDB_REGNUM_GENERIC_FLAGS) 434 .Case("arg1", LLDB_REGNUM_GENERIC_ARG1) 435 .Case("arg2", LLDB_REGNUM_GENERIC_ARG2) 436 .Case("arg3", LLDB_REGNUM_GENERIC_ARG3) 437 .Case("arg4", LLDB_REGNUM_GENERIC_ARG4) 438 .Case("arg5", LLDB_REGNUM_GENERIC_ARG5) 439 .Case("arg6", LLDB_REGNUM_GENERIC_ARG6) 440 .Case("arg7", LLDB_REGNUM_GENERIC_ARG7) 441 .Case("arg8", LLDB_REGNUM_GENERIC_ARG8) 442 .Default(LLDB_INVALID_REGNUM); 443 return result; 444 } 445 446 void Args::EncodeEscapeSequences(const char *src, std::string &dst) { 447 dst.clear(); 448 if (src) { 449 for (const char *p = src; *p != '\0'; ++p) { 450 size_t non_special_chars = ::strcspn(p, "\\"); 451 if (non_special_chars > 0) { 452 dst.append(p, non_special_chars); 453 p += non_special_chars; 454 if (*p == '\0') 455 break; 456 } 457 458 if (*p == '\\') { 459 ++p; // skip the slash 460 switch (*p) { 461 case 'a': 462 dst.append(1, '\a'); 463 break; 464 case 'b': 465 dst.append(1, '\b'); 466 break; 467 case 'f': 468 dst.append(1, '\f'); 469 break; 470 case 'n': 471 dst.append(1, '\n'); 472 break; 473 case 'r': 474 dst.append(1, '\r'); 475 break; 476 case 't': 477 dst.append(1, '\t'); 478 break; 479 case 'v': 480 dst.append(1, '\v'); 481 break; 482 case '\\': 483 dst.append(1, '\\'); 484 break; 485 case '\'': 486 dst.append(1, '\''); 487 break; 488 case '"': 489 dst.append(1, '"'); 490 break; 491 case '0': 492 // 1 to 3 octal chars 493 { 494 // Make a string that can hold onto the initial zero char, up to 3 495 // octal digits, and a terminating NULL. 496 char oct_str[5] = {'\0', '\0', '\0', '\0', '\0'}; 497 498 int i; 499 for (i = 0; (p[i] >= '0' && p[i] <= '7') && i < 4; ++i) 500 oct_str[i] = p[i]; 501 502 // We don't want to consume the last octal character since the main 503 // for loop will do this for us, so we advance p by one less than i 504 // (even if i is zero) 505 p += i - 1; 506 unsigned long octal_value = ::strtoul(oct_str, nullptr, 8); 507 if (octal_value <= UINT8_MAX) { 508 dst.append(1, static_cast<char>(octal_value)); 509 } 510 } 511 break; 512 513 case 'x': 514 // hex number in the format 515 if (isxdigit(p[1])) { 516 ++p; // Skip the 'x' 517 518 // Make a string that can hold onto two hex chars plus a 519 // NULL terminator 520 char hex_str[3] = {*p, '\0', '\0'}; 521 if (isxdigit(p[1])) { 522 ++p; // Skip the first of the two hex chars 523 hex_str[1] = *p; 524 } 525 526 unsigned long hex_value = strtoul(hex_str, nullptr, 16); 527 if (hex_value <= UINT8_MAX) 528 dst.append(1, static_cast<char>(hex_value)); 529 } else { 530 dst.append(1, 'x'); 531 } 532 break; 533 534 default: 535 // Just desensitize any other character by just printing what came 536 // after the '\' 537 dst.append(1, *p); 538 break; 539 } 540 } 541 } 542 } 543 } 544 545 void Args::ExpandEscapedCharacters(const char *src, std::string &dst) { 546 dst.clear(); 547 if (src) { 548 for (const char *p = src; *p != '\0'; ++p) { 549 if (llvm::isPrint(*p)) 550 dst.append(1, *p); 551 else { 552 switch (*p) { 553 case '\a': 554 dst.append("\\a"); 555 break; 556 case '\b': 557 dst.append("\\b"); 558 break; 559 case '\f': 560 dst.append("\\f"); 561 break; 562 case '\n': 563 dst.append("\\n"); 564 break; 565 case '\r': 566 dst.append("\\r"); 567 break; 568 case '\t': 569 dst.append("\\t"); 570 break; 571 case '\v': 572 dst.append("\\v"); 573 break; 574 case '\'': 575 dst.append("\\'"); 576 break; 577 case '"': 578 dst.append("\\\""); 579 break; 580 case '\\': 581 dst.append("\\\\"); 582 break; 583 default: { 584 // Just encode as octal 585 dst.append("\\0"); 586 char octal_str[32]; 587 snprintf(octal_str, sizeof(octal_str), "%o", *p); 588 dst.append(octal_str); 589 } break; 590 } 591 } 592 } 593 } 594 } 595 596 std::string Args::EscapeLLDBCommandArgument(const std::string &arg, 597 char quote_char) { 598 const char *chars_to_escape = nullptr; 599 switch (quote_char) { 600 case '\0': 601 chars_to_escape = " \t\\'\"`"; 602 break; 603 case '"': 604 chars_to_escape = "$\"`\\"; 605 break; 606 case '`': 607 case '\'': 608 return arg; 609 default: 610 assert(false && "Unhandled quote character"); 611 return arg; 612 } 613 614 std::string res; 615 res.reserve(arg.size()); 616 for (char c : arg) { 617 if (::strchr(chars_to_escape, c)) 618 res.push_back('\\'); 619 res.push_back(c); 620 } 621 return res; 622 } 623 624 OptionsWithRaw::OptionsWithRaw(llvm::StringRef arg_string) { 625 SetFromString(arg_string); 626 } 627 628 void OptionsWithRaw::SetFromString(llvm::StringRef arg_string) { 629 const llvm::StringRef original_args = arg_string; 630 631 arg_string = ltrimForArgs(arg_string); 632 std::string arg; 633 char quote; 634 635 // If the string doesn't start with a dash, we just have no options and just 636 // a raw part. 637 if (!arg_string.startswith("-")) { 638 m_suffix = std::string(original_args); 639 return; 640 } 641 642 bool found_suffix = false; 643 while (!arg_string.empty()) { 644 // The length of the prefix before parsing. 645 std::size_t prev_prefix_length = original_args.size() - arg_string.size(); 646 647 // Parse the next argument from the remaining string. 648 std::tie(arg, quote, arg_string) = ParseSingleArgument(arg_string); 649 650 // If we get an unquoted '--' argument, then we reached the suffix part 651 // of the command. 652 Args::ArgEntry entry(arg, quote); 653 if (!entry.IsQuoted() && arg == "--") { 654 // The remaining line is the raw suffix, and the line we parsed so far 655 // needs to be interpreted as arguments. 656 m_has_args = true; 657 m_suffix = std::string(arg_string); 658 found_suffix = true; 659 660 // The length of the prefix after parsing. 661 std::size_t prefix_length = original_args.size() - arg_string.size(); 662 663 // Take the string we know contains all the arguments and actually parse 664 // it as proper arguments. 665 llvm::StringRef prefix = original_args.take_front(prev_prefix_length); 666 m_args = Args(prefix); 667 m_arg_string = prefix; 668 669 // We also record the part of the string that contains the arguments plus 670 // the delimiter. 671 m_arg_string_with_delimiter = original_args.take_front(prefix_length); 672 673 // As the rest of the string became the raw suffix, we are done here. 674 break; 675 } 676 677 arg_string = ltrimForArgs(arg_string); 678 } 679 680 // If we didn't find a suffix delimiter, the whole string is the raw suffix. 681 if (!found_suffix) 682 m_suffix = std::string(original_args); 683 } 684 685 void llvm::yaml::MappingTraits<Args::ArgEntry>::mapping(IO &io, 686 Args::ArgEntry &v) { 687 MappingNormalization<NormalizedArgEntry, Args::ArgEntry> keys(io, v); 688 io.mapRequired("value", keys->value); 689 io.mapRequired("quote", keys->quote); 690 } 691 692 void llvm::yaml::MappingTraits<Args>::mapping(IO &io, Args &v) { 693 io.mapRequired("entries", v.m_entries); 694 695 // Recompute m_argv vector. 696 v.m_argv.clear(); 697 for (auto &entry : v.m_entries) 698 v.m_argv.push_back(entry.data()); 699 v.m_argv.push_back(nullptr); 700 } 701