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() {} 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 size_t Args::GetArgumentCount() const { return m_entries.size(); } 260 261 const char *Args::GetArgumentAtIndex(size_t idx) const { 262 if (idx < m_argv.size()) 263 return m_argv[idx]; 264 return nullptr; 265 } 266 267 char **Args::GetArgumentVector() { 268 assert(!m_argv.empty()); 269 // TODO: functions like execve and posix_spawnp exhibit undefined behavior 270 // when argv or envp is null. So the code below is actually wrong. However, 271 // other code in LLDB depends on it being null. The code has been acting 272 // this way for some time, so it makes sense to leave it this way until 273 // someone has the time to come along and fix it. 274 return (m_argv.size() > 1) ? m_argv.data() : nullptr; 275 } 276 277 const char **Args::GetConstArgumentVector() const { 278 assert(!m_argv.empty()); 279 return (m_argv.size() > 1) ? const_cast<const char **>(m_argv.data()) 280 : nullptr; 281 } 282 283 void Args::Shift() { 284 // Don't pop the last NULL terminator from the argv array 285 if (m_entries.empty()) 286 return; 287 m_argv.erase(m_argv.begin()); 288 m_entries.erase(m_entries.begin()); 289 } 290 291 void Args::Unshift(llvm::StringRef arg_str, char quote_char) { 292 InsertArgumentAtIndex(0, arg_str, quote_char); 293 } 294 295 void Args::AppendArguments(const Args &rhs) { 296 assert(m_argv.size() == m_entries.size() + 1); 297 assert(m_argv.back() == nullptr); 298 m_argv.pop_back(); 299 for (auto &entry : rhs.m_entries) { 300 m_entries.emplace_back(entry.ref(), entry.quote); 301 m_argv.push_back(m_entries.back().data()); 302 } 303 m_argv.push_back(nullptr); 304 } 305 306 void Args::AppendArguments(const char **argv) { 307 size_t argc = ArgvToArgc(argv); 308 309 assert(m_argv.size() == m_entries.size() + 1); 310 assert(m_argv.back() == nullptr); 311 m_argv.pop_back(); 312 for (auto arg : llvm::makeArrayRef(argv, argc)) { 313 m_entries.emplace_back(arg, '\0'); 314 m_argv.push_back(m_entries.back().data()); 315 } 316 317 m_argv.push_back(nullptr); 318 } 319 320 void Args::AppendArgument(llvm::StringRef arg_str, char quote_char) { 321 InsertArgumentAtIndex(GetArgumentCount(), arg_str, quote_char); 322 } 323 324 void Args::InsertArgumentAtIndex(size_t idx, llvm::StringRef arg_str, 325 char quote_char) { 326 assert(m_argv.size() == m_entries.size() + 1); 327 assert(m_argv.back() == nullptr); 328 329 if (idx > m_entries.size()) 330 return; 331 m_entries.emplace(m_entries.begin() + idx, arg_str, quote_char); 332 m_argv.insert(m_argv.begin() + idx, m_entries[idx].data()); 333 } 334 335 void Args::ReplaceArgumentAtIndex(size_t idx, llvm::StringRef arg_str, 336 char quote_char) { 337 assert(m_argv.size() == m_entries.size() + 1); 338 assert(m_argv.back() == nullptr); 339 340 if (idx >= m_entries.size()) 341 return; 342 343 m_entries[idx] = ArgEntry(arg_str, quote_char); 344 m_argv[idx] = m_entries[idx].data(); 345 } 346 347 void Args::DeleteArgumentAtIndex(size_t idx) { 348 if (idx >= m_entries.size()) 349 return; 350 351 m_argv.erase(m_argv.begin() + idx); 352 m_entries.erase(m_entries.begin() + idx); 353 } 354 355 void Args::SetArguments(size_t argc, const char **argv) { 356 Clear(); 357 358 auto args = llvm::makeArrayRef(argv, argc); 359 m_entries.resize(argc); 360 m_argv.resize(argc + 1); 361 for (size_t i = 0; i < args.size(); ++i) { 362 char quote = 363 ((args[i][0] == '\'') || (args[i][0] == '"') || (args[i][0] == '`')) 364 ? args[i][0] 365 : '\0'; 366 367 m_entries[i] = ArgEntry(args[i], quote); 368 m_argv[i] = m_entries[i].data(); 369 } 370 } 371 372 void Args::SetArguments(const char **argv) { 373 SetArguments(ArgvToArgc(argv), argv); 374 } 375 376 void Args::Clear() { 377 m_entries.clear(); 378 m_argv.clear(); 379 m_argv.push_back(nullptr); 380 } 381 382 std::string Args::GetShellSafeArgument(const FileSpec &shell, 383 llvm::StringRef unsafe_arg) { 384 struct ShellDescriptor { 385 ConstString m_basename; 386 llvm::StringRef m_escapables; 387 }; 388 389 static ShellDescriptor g_Shells[] = {{ConstString("bash"), " '\"<>()&"}, 390 {ConstString("tcsh"), " '\"<>()&$"}, 391 {ConstString("sh"), " '\"<>()&"}}; 392 393 // safe minimal set 394 llvm::StringRef escapables = " '\""; 395 396 if (auto basename = shell.GetFilename()) { 397 for (const auto &Shell : g_Shells) { 398 if (Shell.m_basename == basename) { 399 escapables = Shell.m_escapables; 400 break; 401 } 402 } 403 } 404 405 std::string safe_arg; 406 safe_arg.reserve(unsafe_arg.size()); 407 // Add a \ before every character that needs to be escaped. 408 for (char c : unsafe_arg) { 409 if (escapables.contains(c)) 410 safe_arg.push_back('\\'); 411 safe_arg.push_back(c); 412 } 413 return safe_arg; 414 } 415 416 lldb::Encoding Args::StringToEncoding(llvm::StringRef s, 417 lldb::Encoding fail_value) { 418 return llvm::StringSwitch<lldb::Encoding>(s) 419 .Case("uint", eEncodingUint) 420 .Case("sint", eEncodingSint) 421 .Case("ieee754", eEncodingIEEE754) 422 .Case("vector", eEncodingVector) 423 .Default(fail_value); 424 } 425 426 uint32_t Args::StringToGenericRegister(llvm::StringRef s) { 427 if (s.empty()) 428 return LLDB_INVALID_REGNUM; 429 uint32_t result = llvm::StringSwitch<uint32_t>(s) 430 .Case("pc", LLDB_REGNUM_GENERIC_PC) 431 .Case("sp", LLDB_REGNUM_GENERIC_SP) 432 .Case("fp", LLDB_REGNUM_GENERIC_FP) 433 .Cases("ra", "lr", LLDB_REGNUM_GENERIC_RA) 434 .Case("flags", LLDB_REGNUM_GENERIC_FLAGS) 435 .Case("arg1", LLDB_REGNUM_GENERIC_ARG1) 436 .Case("arg2", LLDB_REGNUM_GENERIC_ARG2) 437 .Case("arg3", LLDB_REGNUM_GENERIC_ARG3) 438 .Case("arg4", LLDB_REGNUM_GENERIC_ARG4) 439 .Case("arg5", LLDB_REGNUM_GENERIC_ARG5) 440 .Case("arg6", LLDB_REGNUM_GENERIC_ARG6) 441 .Case("arg7", LLDB_REGNUM_GENERIC_ARG7) 442 .Case("arg8", LLDB_REGNUM_GENERIC_ARG8) 443 .Default(LLDB_INVALID_REGNUM); 444 return result; 445 } 446 447 void Args::EncodeEscapeSequences(const char *src, std::string &dst) { 448 dst.clear(); 449 if (src) { 450 for (const char *p = src; *p != '\0'; ++p) { 451 size_t non_special_chars = ::strcspn(p, "\\"); 452 if (non_special_chars > 0) { 453 dst.append(p, non_special_chars); 454 p += non_special_chars; 455 if (*p == '\0') 456 break; 457 } 458 459 if (*p == '\\') { 460 ++p; // skip the slash 461 switch (*p) { 462 case 'a': 463 dst.append(1, '\a'); 464 break; 465 case 'b': 466 dst.append(1, '\b'); 467 break; 468 case 'f': 469 dst.append(1, '\f'); 470 break; 471 case 'n': 472 dst.append(1, '\n'); 473 break; 474 case 'r': 475 dst.append(1, '\r'); 476 break; 477 case 't': 478 dst.append(1, '\t'); 479 break; 480 case 'v': 481 dst.append(1, '\v'); 482 break; 483 case '\\': 484 dst.append(1, '\\'); 485 break; 486 case '\'': 487 dst.append(1, '\''); 488 break; 489 case '"': 490 dst.append(1, '"'); 491 break; 492 case '0': 493 // 1 to 3 octal chars 494 { 495 // Make a string that can hold onto the initial zero char, up to 3 496 // octal digits, and a terminating NULL. 497 char oct_str[5] = {'\0', '\0', '\0', '\0', '\0'}; 498 499 int i; 500 for (i = 0; (p[i] >= '0' && p[i] <= '7') && i < 4; ++i) 501 oct_str[i] = p[i]; 502 503 // We don't want to consume the last octal character since the main 504 // for loop will do this for us, so we advance p by one less than i 505 // (even if i is zero) 506 p += i - 1; 507 unsigned long octal_value = ::strtoul(oct_str, nullptr, 8); 508 if (octal_value <= UINT8_MAX) { 509 dst.append(1, static_cast<char>(octal_value)); 510 } 511 } 512 break; 513 514 case 'x': 515 // hex number in the format 516 if (isxdigit(p[1])) { 517 ++p; // Skip the 'x' 518 519 // Make a string that can hold onto two hex chars plus a 520 // NULL terminator 521 char hex_str[3] = {*p, '\0', '\0'}; 522 if (isxdigit(p[1])) { 523 ++p; // Skip the first of the two hex chars 524 hex_str[1] = *p; 525 } 526 527 unsigned long hex_value = strtoul(hex_str, nullptr, 16); 528 if (hex_value <= UINT8_MAX) 529 dst.append(1, static_cast<char>(hex_value)); 530 } else { 531 dst.append(1, 'x'); 532 } 533 break; 534 535 default: 536 // Just desensitize any other character by just printing what came 537 // after the '\' 538 dst.append(1, *p); 539 break; 540 } 541 } 542 } 543 } 544 } 545 546 void Args::ExpandEscapedCharacters(const char *src, std::string &dst) { 547 dst.clear(); 548 if (src) { 549 for (const char *p = src; *p != '\0'; ++p) { 550 if (llvm::isPrint(*p)) 551 dst.append(1, *p); 552 else { 553 switch (*p) { 554 case '\a': 555 dst.append("\\a"); 556 break; 557 case '\b': 558 dst.append("\\b"); 559 break; 560 case '\f': 561 dst.append("\\f"); 562 break; 563 case '\n': 564 dst.append("\\n"); 565 break; 566 case '\r': 567 dst.append("\\r"); 568 break; 569 case '\t': 570 dst.append("\\t"); 571 break; 572 case '\v': 573 dst.append("\\v"); 574 break; 575 case '\'': 576 dst.append("\\'"); 577 break; 578 case '"': 579 dst.append("\\\""); 580 break; 581 case '\\': 582 dst.append("\\\\"); 583 break; 584 default: { 585 // Just encode as octal 586 dst.append("\\0"); 587 char octal_str[32]; 588 snprintf(octal_str, sizeof(octal_str), "%o", *p); 589 dst.append(octal_str); 590 } break; 591 } 592 } 593 } 594 } 595 } 596 597 std::string Args::EscapeLLDBCommandArgument(const std::string &arg, 598 char quote_char) { 599 const char *chars_to_escape = nullptr; 600 switch (quote_char) { 601 case '\0': 602 chars_to_escape = " \t\\'\"`"; 603 break; 604 case '"': 605 chars_to_escape = "$\"`\\"; 606 break; 607 case '`': 608 case '\'': 609 return arg; 610 default: 611 assert(false && "Unhandled quote character"); 612 return arg; 613 } 614 615 std::string res; 616 res.reserve(arg.size()); 617 for (char c : arg) { 618 if (::strchr(chars_to_escape, c)) 619 res.push_back('\\'); 620 res.push_back(c); 621 } 622 return res; 623 } 624 625 OptionsWithRaw::OptionsWithRaw(llvm::StringRef arg_string) { 626 SetFromString(arg_string); 627 } 628 629 void OptionsWithRaw::SetFromString(llvm::StringRef arg_string) { 630 const llvm::StringRef original_args = arg_string; 631 632 arg_string = ltrimForArgs(arg_string); 633 std::string arg; 634 char quote; 635 636 // If the string doesn't start with a dash, we just have no options and just 637 // a raw part. 638 if (!arg_string.startswith("-")) { 639 m_suffix = std::string(original_args); 640 return; 641 } 642 643 bool found_suffix = false; 644 while (!arg_string.empty()) { 645 // The length of the prefix before parsing. 646 std::size_t prev_prefix_length = original_args.size() - arg_string.size(); 647 648 // Parse the next argument from the remaining string. 649 std::tie(arg, quote, arg_string) = ParseSingleArgument(arg_string); 650 651 // If we get an unquoted '--' argument, then we reached the suffix part 652 // of the command. 653 Args::ArgEntry entry(arg, quote); 654 if (!entry.IsQuoted() && arg == "--") { 655 // The remaining line is the raw suffix, and the line we parsed so far 656 // needs to be interpreted as arguments. 657 m_has_args = true; 658 m_suffix = std::string(arg_string); 659 found_suffix = true; 660 661 // The length of the prefix after parsing. 662 std::size_t prefix_length = original_args.size() - arg_string.size(); 663 664 // Take the string we know contains all the arguments and actually parse 665 // it as proper arguments. 666 llvm::StringRef prefix = original_args.take_front(prev_prefix_length); 667 m_args = Args(prefix); 668 m_arg_string = prefix; 669 670 // We also record the part of the string that contains the arguments plus 671 // the delimiter. 672 m_arg_string_with_delimiter = original_args.take_front(prefix_length); 673 674 // As the rest of the string became the raw suffix, we are done here. 675 break; 676 } 677 678 arg_string = ltrimForArgs(arg_string); 679 } 680 681 // If we didn't find a suffix delimiter, the whole string is the raw suffix. 682 if (!found_suffix) 683 m_suffix = std::string(original_args); 684 } 685 686 void llvm::yaml::MappingTraits<Args::ArgEntry>::mapping(IO &io, 687 Args::ArgEntry &v) { 688 MappingNormalization<NormalizedArgEntry, Args::ArgEntry> keys(io, v); 689 io.mapRequired("value", keys->value); 690 io.mapRequired("quote", keys->quote); 691 } 692 693 void llvm::yaml::MappingTraits<Args>::mapping(IO &io, Args &v) { 694 io.mapRequired("entries", v.m_entries); 695 696 // Recompute m_argv vector. 697 v.m_argv.clear(); 698 for (auto &entry : v.m_entries) 699 v.m_argv.push_back(entry.data()); 700 v.m_argv.push_back(nullptr); 701 } 702