lldb/source/Utility/Args.cpp

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