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 char quote = m_entries[i].quote; 219 if (quote != '\0') 220 command += quote; 221 command += m_entries[i].ref(); 222 if (quote != '\0') 223 command += quote; 224 } 225 226 return !m_entries.empty(); 227 } 228 229 bool Args::GetQuotedCommandString(std::string &command) const { 230 command.clear(); 231 232 for (size_t i = 0; i < m_entries.size(); ++i) { 233 if (i > 0) 234 command += ' '; 235 236 if (m_entries[i].quote) { 237 command += m_entries[i].quote; 238 command += m_entries[i].ref(); 239 command += m_entries[i].quote; 240 } else { 241 command += m_entries[i].ref(); 242 } 243 } 244 245 return !m_entries.empty(); 246 } 247 248 void Args::SetCommandString(llvm::StringRef command) { 249 Clear(); 250 m_argv.clear(); 251 252 command = ltrimForArgs(command); 253 std::string arg; 254 char quote; 255 while (!command.empty()) { 256 std::tie(arg, quote, command) = ParseSingleArgument(command); 257 m_entries.emplace_back(arg, quote); 258 m_argv.push_back(m_entries.back().data()); 259 command = ltrimForArgs(command); 260 } 261 m_argv.push_back(nullptr); 262 } 263 264 const char *Args::GetArgumentAtIndex(size_t idx) const { 265 if (idx < m_argv.size()) 266 return m_argv[idx]; 267 return nullptr; 268 } 269 270 char **Args::GetArgumentVector() { 271 assert(!m_argv.empty()); 272 // TODO: functions like execve and posix_spawnp exhibit undefined behavior 273 // when argv or envp is null. So the code below is actually wrong. However, 274 // other code in LLDB depends on it being null. The code has been acting 275 // this way for some time, so it makes sense to leave it this way until 276 // someone has the time to come along and fix it. 277 return (m_argv.size() > 1) ? m_argv.data() : nullptr; 278 } 279 280 const char **Args::GetConstArgumentVector() const { 281 assert(!m_argv.empty()); 282 return (m_argv.size() > 1) ? const_cast<const char **>(m_argv.data()) 283 : nullptr; 284 } 285 286 void Args::Shift() { 287 // Don't pop the last NULL terminator from the argv array 288 if (m_entries.empty()) 289 return; 290 m_argv.erase(m_argv.begin()); 291 m_entries.erase(m_entries.begin()); 292 } 293 294 void Args::Unshift(llvm::StringRef arg_str, char quote_char) { 295 InsertArgumentAtIndex(0, arg_str, quote_char); 296 } 297 298 void Args::AppendArguments(const Args &rhs) { 299 assert(m_argv.size() == m_entries.size() + 1); 300 assert(m_argv.back() == nullptr); 301 m_argv.pop_back(); 302 for (auto &entry : rhs.m_entries) { 303 m_entries.emplace_back(entry.ref(), entry.quote); 304 m_argv.push_back(m_entries.back().data()); 305 } 306 m_argv.push_back(nullptr); 307 } 308 309 void Args::AppendArguments(const char **argv) { 310 size_t argc = ArgvToArgc(argv); 311 312 assert(m_argv.size() == m_entries.size() + 1); 313 assert(m_argv.back() == nullptr); 314 m_argv.pop_back(); 315 for (auto arg : llvm::ArrayRef(argv, argc)) { 316 m_entries.emplace_back(arg, '\0'); 317 m_argv.push_back(m_entries.back().data()); 318 } 319 320 m_argv.push_back(nullptr); 321 } 322 323 void Args::AppendArgument(llvm::StringRef arg_str, char quote_char) { 324 InsertArgumentAtIndex(GetArgumentCount(), arg_str, quote_char); 325 } 326 327 void Args::InsertArgumentAtIndex(size_t idx, llvm::StringRef arg_str, 328 char quote_char) { 329 assert(m_argv.size() == m_entries.size() + 1); 330 assert(m_argv.back() == nullptr); 331 332 if (idx > m_entries.size()) 333 return; 334 m_entries.emplace(m_entries.begin() + idx, arg_str, quote_char); 335 m_argv.insert(m_argv.begin() + idx, m_entries[idx].data()); 336 } 337 338 void Args::ReplaceArgumentAtIndex(size_t idx, llvm::StringRef arg_str, 339 char quote_char) { 340 assert(m_argv.size() == m_entries.size() + 1); 341 assert(m_argv.back() == nullptr); 342 343 if (idx >= m_entries.size()) 344 return; 345 346 m_entries[idx] = ArgEntry(arg_str, quote_char); 347 m_argv[idx] = m_entries[idx].data(); 348 } 349 350 void Args::DeleteArgumentAtIndex(size_t idx) { 351 if (idx >= m_entries.size()) 352 return; 353 354 m_argv.erase(m_argv.begin() + idx); 355 m_entries.erase(m_entries.begin() + idx); 356 } 357 358 void Args::SetArguments(size_t argc, const char **argv) { 359 Clear(); 360 361 auto args = llvm::ArrayRef(argv, argc); 362 m_entries.resize(argc); 363 m_argv.resize(argc + 1); 364 for (size_t i = 0; i < args.size(); ++i) { 365 char quote = 366 ((args[i][0] == '\'') || (args[i][0] == '"') || (args[i][0] == '`')) 367 ? args[i][0] 368 : '\0'; 369 370 m_entries[i] = ArgEntry(args[i], quote); 371 m_argv[i] = m_entries[i].data(); 372 } 373 } 374 375 void Args::SetArguments(const char **argv) { 376 SetArguments(ArgvToArgc(argv), argv); 377 } 378 379 void Args::Clear() { 380 m_entries.clear(); 381 m_argv.clear(); 382 m_argv.push_back(nullptr); 383 } 384 385 std::string Args::GetShellSafeArgument(const FileSpec &shell, 386 llvm::StringRef unsafe_arg) { 387 struct ShellDescriptor { 388 ConstString m_basename; 389 llvm::StringRef m_escapables; 390 }; 391 392 static ShellDescriptor g_Shells[] = {{ConstString("bash"), " '\"<>()&;"}, 393 {ConstString("fish"), " '\"<>()&\\|;"}, 394 {ConstString("tcsh"), " '\"<>()&;"}, 395 {ConstString("zsh"), " '\"<>()&;\\|"}, 396 {ConstString("sh"), " '\"<>()&;"}}; 397 398 // safe minimal set 399 llvm::StringRef escapables = " '\""; 400 401 if (auto basename = shell.GetFilename()) { 402 for (const auto &Shell : g_Shells) { 403 if (Shell.m_basename == basename) { 404 escapables = Shell.m_escapables; 405 break; 406 } 407 } 408 } 409 410 std::string safe_arg; 411 safe_arg.reserve(unsafe_arg.size()); 412 // Add a \ before every character that needs to be escaped. 413 for (char c : unsafe_arg) { 414 if (escapables.contains(c)) 415 safe_arg.push_back('\\'); 416 safe_arg.push_back(c); 417 } 418 return safe_arg; 419 } 420 421 lldb::Encoding Args::StringToEncoding(llvm::StringRef s, 422 lldb::Encoding fail_value) { 423 return llvm::StringSwitch<lldb::Encoding>(s) 424 .Case("uint", eEncodingUint) 425 .Case("sint", eEncodingSint) 426 .Case("ieee754", eEncodingIEEE754) 427 .Case("vector", eEncodingVector) 428 .Default(fail_value); 429 } 430 431 uint32_t Args::StringToGenericRegister(llvm::StringRef s) { 432 if (s.empty()) 433 return LLDB_INVALID_REGNUM; 434 uint32_t result = llvm::StringSwitch<uint32_t>(s) 435 .Case("pc", LLDB_REGNUM_GENERIC_PC) 436 .Case("sp", LLDB_REGNUM_GENERIC_SP) 437 .Case("fp", LLDB_REGNUM_GENERIC_FP) 438 .Cases("ra", "lr", LLDB_REGNUM_GENERIC_RA) 439 .Case("flags", LLDB_REGNUM_GENERIC_FLAGS) 440 .Case("arg1", LLDB_REGNUM_GENERIC_ARG1) 441 .Case("arg2", LLDB_REGNUM_GENERIC_ARG2) 442 .Case("arg3", LLDB_REGNUM_GENERIC_ARG3) 443 .Case("arg4", LLDB_REGNUM_GENERIC_ARG4) 444 .Case("arg5", LLDB_REGNUM_GENERIC_ARG5) 445 .Case("arg6", LLDB_REGNUM_GENERIC_ARG6) 446 .Case("arg7", LLDB_REGNUM_GENERIC_ARG7) 447 .Case("arg8", LLDB_REGNUM_GENERIC_ARG8) 448 .Default(LLDB_INVALID_REGNUM); 449 return result; 450 } 451 452 void Args::EncodeEscapeSequences(const char *src, std::string &dst) { 453 dst.clear(); 454 if (src) { 455 for (const char *p = src; *p != '\0'; ++p) { 456 size_t non_special_chars = ::strcspn(p, "\\"); 457 if (non_special_chars > 0) { 458 dst.append(p, non_special_chars); 459 p += non_special_chars; 460 if (*p == '\0') 461 break; 462 } 463 464 if (*p == '\\') { 465 ++p; // skip the slash 466 switch (*p) { 467 case 'a': 468 dst.append(1, '\a'); 469 break; 470 case 'b': 471 dst.append(1, '\b'); 472 break; 473 case 'f': 474 dst.append(1, '\f'); 475 break; 476 case 'n': 477 dst.append(1, '\n'); 478 break; 479 case 'r': 480 dst.append(1, '\r'); 481 break; 482 case 't': 483 dst.append(1, '\t'); 484 break; 485 case 'v': 486 dst.append(1, '\v'); 487 break; 488 case '\\': 489 dst.append(1, '\\'); 490 break; 491 case '\'': 492 dst.append(1, '\''); 493 break; 494 case '"': 495 dst.append(1, '"'); 496 break; 497 case '0': 498 // 1 to 3 octal chars 499 { 500 // Make a string that can hold onto the initial zero char, up to 3 501 // octal digits, and a terminating NULL. 502 char oct_str[5] = {'\0', '\0', '\0', '\0', '\0'}; 503 504 int i; 505 for (i = 0; (p[i] >= '0' && p[i] <= '7') && i < 4; ++i) 506 oct_str[i] = p[i]; 507 508 // We don't want to consume the last octal character since the main 509 // for loop will do this for us, so we advance p by one less than i 510 // (even if i is zero) 511 p += i - 1; 512 unsigned long octal_value = ::strtoul(oct_str, nullptr, 8); 513 if (octal_value <= UINT8_MAX) { 514 dst.append(1, static_cast<char>(octal_value)); 515 } 516 } 517 break; 518 519 case 'x': 520 // hex number in the format 521 if (isxdigit(p[1])) { 522 ++p; // Skip the 'x' 523 524 // Make a string that can hold onto two hex chars plus a 525 // NULL terminator 526 char hex_str[3] = {*p, '\0', '\0'}; 527 if (isxdigit(p[1])) { 528 ++p; // Skip the first of the two hex chars 529 hex_str[1] = *p; 530 } 531 532 unsigned long hex_value = strtoul(hex_str, nullptr, 16); 533 if (hex_value <= UINT8_MAX) 534 dst.append(1, static_cast<char>(hex_value)); 535 } else { 536 dst.append(1, 'x'); 537 } 538 break; 539 540 default: 541 // Just desensitize any other character by just printing what came 542 // after the '\' 543 dst.append(1, *p); 544 break; 545 } 546 } 547 } 548 } 549 } 550 551 void Args::ExpandEscapedCharacters(const char *src, std::string &dst) { 552 dst.clear(); 553 if (src) { 554 for (const char *p = src; *p != '\0'; ++p) { 555 if (llvm::isPrint(*p)) 556 dst.append(1, *p); 557 else { 558 switch (*p) { 559 case '\a': 560 dst.append("\\a"); 561 break; 562 case '\b': 563 dst.append("\\b"); 564 break; 565 case '\f': 566 dst.append("\\f"); 567 break; 568 case '\n': 569 dst.append("\\n"); 570 break; 571 case '\r': 572 dst.append("\\r"); 573 break; 574 case '\t': 575 dst.append("\\t"); 576 break; 577 case '\v': 578 dst.append("\\v"); 579 break; 580 case '\'': 581 dst.append("\\'"); 582 break; 583 case '"': 584 dst.append("\\\""); 585 break; 586 case '\\': 587 dst.append("\\\\"); 588 break; 589 default: { 590 // Just encode as octal 591 dst.append("\\0"); 592 char octal_str[32]; 593 snprintf(octal_str, sizeof(octal_str), "%o", *p); 594 dst.append(octal_str); 595 } break; 596 } 597 } 598 } 599 } 600 } 601 602 std::string Args::EscapeLLDBCommandArgument(const std::string &arg, 603 char quote_char) { 604 const char *chars_to_escape = nullptr; 605 switch (quote_char) { 606 case '\0': 607 chars_to_escape = " \t\\'\"`"; 608 break; 609 case '"': 610 chars_to_escape = "$\"`\\"; 611 break; 612 case '`': 613 case '\'': 614 return arg; 615 default: 616 assert(false && "Unhandled quote character"); 617 return arg; 618 } 619 620 std::string res; 621 res.reserve(arg.size()); 622 for (char c : arg) { 623 if (::strchr(chars_to_escape, c)) 624 res.push_back('\\'); 625 res.push_back(c); 626 } 627 return res; 628 } 629 630 OptionsWithRaw::OptionsWithRaw(llvm::StringRef arg_string) { 631 SetFromString(arg_string); 632 } 633 634 void OptionsWithRaw::SetFromString(llvm::StringRef arg_string) { 635 const llvm::StringRef original_args = arg_string; 636 637 arg_string = ltrimForArgs(arg_string); 638 std::string arg; 639 char quote; 640 641 // If the string doesn't start with a dash, we just have no options and just 642 // a raw part. 643 if (!arg_string.startswith("-")) { 644 m_suffix = std::string(original_args); 645 return; 646 } 647 648 bool found_suffix = false; 649 while (!arg_string.empty()) { 650 // The length of the prefix before parsing. 651 std::size_t prev_prefix_length = original_args.size() - arg_string.size(); 652 653 // Parse the next argument from the remaining string. 654 std::tie(arg, quote, arg_string) = ParseSingleArgument(arg_string); 655 656 // If we get an unquoted '--' argument, then we reached the suffix part 657 // of the command. 658 Args::ArgEntry entry(arg, quote); 659 if (!entry.IsQuoted() && arg == "--") { 660 // The remaining line is the raw suffix, and the line we parsed so far 661 // needs to be interpreted as arguments. 662 m_has_args = true; 663 m_suffix = std::string(arg_string); 664 found_suffix = true; 665 666 // The length of the prefix after parsing. 667 std::size_t prefix_length = original_args.size() - arg_string.size(); 668 669 // Take the string we know contains all the arguments and actually parse 670 // it as proper arguments. 671 llvm::StringRef prefix = original_args.take_front(prev_prefix_length); 672 m_args = Args(prefix); 673 m_arg_string = prefix; 674 675 // We also record the part of the string that contains the arguments plus 676 // the delimiter. 677 m_arg_string_with_delimiter = original_args.take_front(prefix_length); 678 679 // As the rest of the string became the raw suffix, we are done here. 680 break; 681 } 682 683 arg_string = ltrimForArgs(arg_string); 684 } 685 686 // If we didn't find a suffix delimiter, the whole string is the raw suffix. 687 if (!found_suffix) 688 m_suffix = std::string(original_args); 689 } 690