xref: /llvm-project/lldb/source/Utility/Args.cpp (revision 089227feaf0efb5e540783a5542655e25669e7d8)
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, size_t &shift) {
70   static const char *k_space_separators = " \t";
71   llvm::StringRef result = str.ltrim(k_space_separators);
72   shift = result.data() - str.data();
73   return result;
74 }
75 
76 // A helper function for SetCommandString. Parses a single argument from the
77 // command string, processing quotes and backslashes in a shell-like manner.
78 // The function returns a tuple consisting of the parsed argument, the quote
79 // char used, and the unparsed portion of the string starting at the first
80 // unqouted, unescaped whitespace character.
81 static std::tuple<std::string, char, llvm::StringRef>
82 ParseSingleArgument(llvm::StringRef command) {
83   // Argument can be split into multiple discontiguous pieces, for example:
84   //  "Hello ""World"
85   // this would result in a single argument "Hello World" (without the quotes)
86   // since the quotes would be removed and there is not space between the
87   // strings.
88   std::string arg;
89 
90   // Since we can have multiple quotes that form a single command in a command
91   // like: "Hello "world'!' (which will make a single argument "Hello world!")
92   // we remember the first quote character we encounter and use that for the
93   // quote character.
94   char first_quote_char = '\0';
95 
96   bool arg_complete = false;
97   do {
98     // Skip over regular characters and append them.
99     size_t regular = command.find_first_of(" \t\r\"'`\\");
100     arg += command.substr(0, regular);
101     command = command.substr(regular);
102 
103     if (command.empty())
104       break;
105 
106     char special = command.front();
107     command = command.drop_front();
108     switch (special) {
109     case '\\':
110       if (command.empty()) {
111         arg += '\\';
112         break;
113       }
114 
115       // If the character after the backslash is not an allowed escapable
116       // character, we leave the character sequence untouched.
117       if (strchr(" \t\\'\"`", command.front()) == nullptr)
118         arg += '\\';
119 
120       arg += command.front();
121       command = command.drop_front();
122 
123       break;
124 
125     case ' ':
126     case '\t':
127     case '\r':
128       // We are not inside any quotes, we just found a space after an argument.
129       // We are done.
130       arg_complete = true;
131       break;
132 
133     case '"':
134     case '\'':
135     case '`':
136       // We found the start of a quote scope.
137       if (first_quote_char == '\0')
138         first_quote_char = special;
139 
140       if (special == '"')
141         command = ParseDoubleQuotes(command, arg);
142       else {
143         // For single quotes, we simply skip ahead to the matching quote
144         // character (or the end of the string).
145         size_t quoted = command.find(special);
146         arg += command.substr(0, quoted);
147         command = command.substr(quoted);
148       }
149 
150       // If we found a closing quote, skip it.
151       if (!command.empty())
152         command = command.drop_front();
153 
154       break;
155     }
156   } while (!arg_complete);
157 
158   return std::make_tuple(arg, first_quote_char, command);
159 }
160 
161 Args::ArgEntry::ArgEntry(llvm::StringRef str, char quote,
162                          std::optional<uint16_t> column)
163     : quote(quote), column(column) {
164   size_t size = str.size();
165   ptr.reset(new char[size + 1]);
166 
167   ::memcpy(data(), str.data() ? str.data() : "", size);
168   ptr[size] = 0;
169 }
170 
171 // Args constructor
172 Args::Args(llvm::StringRef command) { SetCommandString(command); }
173 
174 Args::Args(const Args &rhs) { *this = rhs; }
175 
176 Args::Args(const StringList &list) : Args() {
177   for (const std::string &arg : list)
178     AppendArgument(arg);
179 }
180 
181 Args::Args(llvm::ArrayRef<llvm::StringRef> args) : Args() {
182   for (llvm::StringRef arg : args)
183     AppendArgument(arg);
184 }
185 
186 Args &Args::operator=(const Args &rhs) {
187   Clear();
188 
189   m_argv.clear();
190   m_entries.clear();
191   for (auto &entry : rhs.m_entries) {
192     m_entries.emplace_back(entry.ref(), entry.quote, entry.column);
193     m_argv.push_back(m_entries.back().data());
194   }
195   m_argv.push_back(nullptr);
196   return *this;
197 }
198 
199 // Destructor
200 Args::~Args() = default;
201 
202 void Args::Dump(Stream &s, const char *label_name) const {
203   if (!label_name)
204     return;
205 
206   int i = 0;
207   for (auto &entry : m_entries) {
208     s.Indent();
209     s.Format("{0}[{1}]=\"{2}\"\n", label_name, i++, entry.ref());
210   }
211   s.Format("{0}[{1}]=NULL\n", label_name, i);
212   s.EOL();
213 }
214 
215 bool Args::GetCommandString(std::string &command) const {
216   command.clear();
217 
218   for (size_t i = 0; i < m_entries.size(); ++i) {
219     if (i > 0)
220       command += ' ';
221     char quote = m_entries[i].quote;
222     if (quote != '\0')
223      command += quote;
224     command += m_entries[i].ref();
225     if (quote != '\0')
226       command += quote;
227   }
228 
229   return !m_entries.empty();
230 }
231 
232 bool Args::GetQuotedCommandString(std::string &command) const {
233   command.clear();
234 
235   for (size_t i = 0; i < m_entries.size(); ++i) {
236     if (i > 0)
237       command += ' ';
238 
239     if (m_entries[i].quote) {
240       command += m_entries[i].quote;
241       command += m_entries[i].ref();
242       command += m_entries[i].quote;
243     } else {
244       command += m_entries[i].ref();
245     }
246   }
247 
248   return !m_entries.empty();
249 }
250 
251 void Args::SetCommandString(llvm::StringRef command) {
252   Clear();
253   m_argv.clear();
254 
255   uint16_t column = 1;
256   size_t shift = 0;
257   command = ltrimForArgs(command, shift);
258   column += shift;
259   std::string arg;
260   char quote;
261   while (!command.empty()) {
262     const char *prev = command.data();
263     std::tie(arg, quote, command) = ParseSingleArgument(command);
264     m_entries.emplace_back(arg, quote, column);
265     m_argv.push_back(m_entries.back().data());
266     command = ltrimForArgs(command, shift);
267     column += shift;
268     column += command.data() - prev;
269   }
270   m_argv.push_back(nullptr);
271 }
272 
273 const char *Args::GetArgumentAtIndex(size_t idx) const {
274   if (idx < m_argv.size())
275     return m_argv[idx];
276   return nullptr;
277 }
278 
279 char **Args::GetArgumentVector() {
280   assert(!m_argv.empty());
281   // TODO: functions like execve and posix_spawnp exhibit undefined behavior
282   // when argv or envp is null.  So the code below is actually wrong.  However,
283   // other code in LLDB depends on it being null.  The code has been acting
284   // this way for some time, so it makes sense to leave it this way until
285   // someone has the time to come along and fix it.
286   return (m_argv.size() > 1) ? m_argv.data() : nullptr;
287 }
288 
289 const char **Args::GetConstArgumentVector() const {
290   assert(!m_argv.empty());
291   return (m_argv.size() > 1) ? const_cast<const char **>(m_argv.data())
292                              : nullptr;
293 }
294 
295 void Args::Shift() {
296   // Don't pop the last NULL terminator from the argv array
297   if (m_entries.empty())
298     return;
299   m_argv.erase(m_argv.begin());
300   m_entries.erase(m_entries.begin());
301 }
302 
303 void Args::Unshift(llvm::StringRef arg_str, char quote_char) {
304   InsertArgumentAtIndex(0, arg_str, quote_char);
305 }
306 
307 void Args::AppendArguments(const Args &rhs) {
308   assert(m_argv.size() == m_entries.size() + 1);
309   assert(m_argv.back() == nullptr);
310   m_argv.pop_back();
311   for (auto &entry : rhs.m_entries) {
312     m_entries.emplace_back(entry.ref(), entry.quote, entry.column);
313     m_argv.push_back(m_entries.back().data());
314   }
315   m_argv.push_back(nullptr);
316 }
317 
318 void Args::AppendArguments(const char **argv) {
319   size_t argc = ArgvToArgc(argv);
320 
321   assert(m_argv.size() == m_entries.size() + 1);
322   assert(m_argv.back() == nullptr);
323   m_argv.pop_back();
324   for (auto arg : llvm::ArrayRef(argv, argc)) {
325     m_entries.emplace_back(arg, '\0', std::nullopt);
326     m_argv.push_back(m_entries.back().data());
327   }
328 
329   m_argv.push_back(nullptr);
330 }
331 
332 void Args::AppendArgument(llvm::StringRef arg_str, char quote_char) {
333   InsertArgumentAtIndex(GetArgumentCount(), arg_str, quote_char);
334 }
335 
336 void Args::InsertArgumentAtIndex(size_t idx, llvm::StringRef arg_str,
337                                  char quote_char) {
338   assert(m_argv.size() == m_entries.size() + 1);
339   assert(m_argv.back() == nullptr);
340 
341   if (idx > m_entries.size())
342     return;
343   m_entries.emplace(m_entries.begin() + idx, arg_str, quote_char, std::nullopt);
344   m_argv.insert(m_argv.begin() + idx, m_entries[idx].data());
345 }
346 
347 void Args::ReplaceArgumentAtIndex(size_t idx, llvm::StringRef arg_str,
348                                   char quote_char) {
349   assert(m_argv.size() == m_entries.size() + 1);
350   assert(m_argv.back() == nullptr);
351 
352   if (idx >= m_entries.size())
353     return;
354 
355   m_entries[idx] = ArgEntry(arg_str, quote_char, std::nullopt);
356   m_argv[idx] = m_entries[idx].data();
357 }
358 
359 void Args::DeleteArgumentAtIndex(size_t idx) {
360   if (idx >= m_entries.size())
361     return;
362 
363   m_argv.erase(m_argv.begin() + idx);
364   m_entries.erase(m_entries.begin() + idx);
365 }
366 
367 void Args::SetArguments(size_t argc, const char **argv) {
368   Clear();
369 
370   auto args = llvm::ArrayRef(argv, argc);
371   m_entries.resize(argc);
372   m_argv.resize(argc + 1);
373   for (size_t i = 0; i < args.size(); ++i) {
374     char quote =
375         ((args[i][0] == '\'') || (args[i][0] == '"') || (args[i][0] == '`'))
376             ? args[i][0]
377             : '\0';
378 
379     m_entries[i] = ArgEntry(args[i], quote, std::nullopt);
380     m_argv[i] = m_entries[i].data();
381   }
382 }
383 
384 void Args::SetArguments(const char **argv) {
385   SetArguments(ArgvToArgc(argv), argv);
386 }
387 
388 void Args::Clear() {
389   m_entries.clear();
390   m_argv.clear();
391   m_argv.push_back(nullptr);
392 }
393 
394 std::string Args::GetShellSafeArgument(const FileSpec &shell,
395                                        llvm::StringRef unsafe_arg) {
396   struct ShellDescriptor {
397     llvm::StringRef m_basename;
398     llvm::StringRef m_escapables;
399   };
400 
401   static ShellDescriptor g_Shells[] = {{"bash", " '\"<>()&;"},
402                                        {"fish", " '\"<>()&\\|;"},
403                                        {"tcsh", " '\"<>()&;"},
404                                        {"zsh", " '\"<>()&;\\|"},
405                                        {"sh", " '\"<>()&;"}};
406 
407   // safe minimal set
408   llvm::StringRef escapables = " '\"";
409 
410   auto basename = shell.GetFilename().GetStringRef();
411   if (!basename.empty()) {
412     for (const auto &Shell : g_Shells) {
413       if (Shell.m_basename == basename) {
414         escapables = Shell.m_escapables;
415         break;
416       }
417     }
418   }
419 
420   std::string safe_arg;
421   safe_arg.reserve(unsafe_arg.size());
422   // Add a \ before every character that needs to be escaped.
423   for (char c : unsafe_arg) {
424     if (escapables.contains(c))
425       safe_arg.push_back('\\');
426     safe_arg.push_back(c);
427   }
428   return safe_arg;
429 }
430 
431 lldb::Encoding Args::StringToEncoding(llvm::StringRef s,
432                                       lldb::Encoding fail_value) {
433   return llvm::StringSwitch<lldb::Encoding>(s)
434       .Case("uint", eEncodingUint)
435       .Case("sint", eEncodingSint)
436       .Case("ieee754", eEncodingIEEE754)
437       .Case("vector", eEncodingVector)
438       .Default(fail_value);
439 }
440 
441 uint32_t Args::StringToGenericRegister(llvm::StringRef s) {
442   if (s.empty())
443     return LLDB_INVALID_REGNUM;
444   uint32_t result = llvm::StringSwitch<uint32_t>(s)
445                         .Case("pc", LLDB_REGNUM_GENERIC_PC)
446                         .Case("sp", LLDB_REGNUM_GENERIC_SP)
447                         .Case("fp", LLDB_REGNUM_GENERIC_FP)
448                         .Cases("ra", "lr", LLDB_REGNUM_GENERIC_RA)
449                         .Case("flags", LLDB_REGNUM_GENERIC_FLAGS)
450                         .Case("arg1", LLDB_REGNUM_GENERIC_ARG1)
451                         .Case("arg2", LLDB_REGNUM_GENERIC_ARG2)
452                         .Case("arg3", LLDB_REGNUM_GENERIC_ARG3)
453                         .Case("arg4", LLDB_REGNUM_GENERIC_ARG4)
454                         .Case("arg5", LLDB_REGNUM_GENERIC_ARG5)
455                         .Case("arg6", LLDB_REGNUM_GENERIC_ARG6)
456                         .Case("arg7", LLDB_REGNUM_GENERIC_ARG7)
457                         .Case("arg8", LLDB_REGNUM_GENERIC_ARG8)
458                         .Case("tp", LLDB_REGNUM_GENERIC_TP)
459                         .Default(LLDB_INVALID_REGNUM);
460   return result;
461 }
462 
463 void Args::EncodeEscapeSequences(const char *src, std::string &dst) {
464   dst.clear();
465   if (src) {
466     for (const char *p = src; *p != '\0'; ++p) {
467       size_t non_special_chars = ::strcspn(p, "\\");
468       if (non_special_chars > 0) {
469         dst.append(p, non_special_chars);
470         p += non_special_chars;
471         if (*p == '\0')
472           break;
473       }
474 
475       if (*p == '\\') {
476         ++p; // skip the slash
477         switch (*p) {
478         case 'a':
479           dst.append(1, '\a');
480           break;
481         case 'b':
482           dst.append(1, '\b');
483           break;
484         case 'f':
485           dst.append(1, '\f');
486           break;
487         case 'n':
488           dst.append(1, '\n');
489           break;
490         case 'r':
491           dst.append(1, '\r');
492           break;
493         case 't':
494           dst.append(1, '\t');
495           break;
496         case 'v':
497           dst.append(1, '\v');
498           break;
499         case '\\':
500           dst.append(1, '\\');
501           break;
502         case '\'':
503           dst.append(1, '\'');
504           break;
505         case '"':
506           dst.append(1, '"');
507           break;
508         case '0':
509           // 1 to 3 octal chars
510           {
511             // Make a string that can hold onto the initial zero char, up to 3
512             // octal digits, and a terminating NULL.
513             char oct_str[5] = {'\0', '\0', '\0', '\0', '\0'};
514 
515             int i;
516             for (i = 0; (p[i] >= '0' && p[i] <= '7') && i < 4; ++i)
517               oct_str[i] = p[i];
518 
519             // We don't want to consume the last octal character since the main
520             // for loop will do this for us, so we advance p by one less than i
521             // (even if i is zero)
522             p += i - 1;
523             unsigned long octal_value = ::strtoul(oct_str, nullptr, 8);
524             if (octal_value <= UINT8_MAX) {
525               dst.append(1, static_cast<char>(octal_value));
526             }
527           }
528           break;
529 
530         case 'x':
531           // hex number in the format
532           if (isxdigit(p[1])) {
533             ++p; // Skip the 'x'
534 
535             // Make a string that can hold onto two hex chars plus a
536             // NULL terminator
537             char hex_str[3] = {*p, '\0', '\0'};
538             if (isxdigit(p[1])) {
539               ++p; // Skip the first of the two hex chars
540               hex_str[1] = *p;
541             }
542 
543             unsigned long hex_value = strtoul(hex_str, nullptr, 16);
544             if (hex_value <= UINT8_MAX)
545               dst.append(1, static_cast<char>(hex_value));
546           } else {
547             dst.append(1, 'x');
548           }
549           break;
550 
551         default:
552           // Just desensitize any other character by just printing what came
553           // after the '\'
554           dst.append(1, *p);
555           break;
556         }
557       }
558     }
559   }
560 }
561 
562 void Args::ExpandEscapedCharacters(const char *src, std::string &dst) {
563   dst.clear();
564   if (src) {
565     for (const char *p = src; *p != '\0'; ++p) {
566       if (llvm::isPrint(*p))
567         dst.append(1, *p);
568       else {
569         switch (*p) {
570         case '\a':
571           dst.append("\\a");
572           break;
573         case '\b':
574           dst.append("\\b");
575           break;
576         case '\f':
577           dst.append("\\f");
578           break;
579         case '\n':
580           dst.append("\\n");
581           break;
582         case '\r':
583           dst.append("\\r");
584           break;
585         case '\t':
586           dst.append("\\t");
587           break;
588         case '\v':
589           dst.append("\\v");
590           break;
591         case '\'':
592           dst.append("\\'");
593           break;
594         case '"':
595           dst.append("\\\"");
596           break;
597         case '\\':
598           dst.append("\\\\");
599           break;
600         default: {
601           // Just encode as octal
602           dst.append("\\0");
603           char octal_str[32];
604           snprintf(octal_str, sizeof(octal_str), "%o", *p);
605           dst.append(octal_str);
606         } break;
607         }
608       }
609     }
610   }
611 }
612 
613 std::string Args::EscapeLLDBCommandArgument(const std::string &arg,
614                                             char quote_char) {
615   const char *chars_to_escape = nullptr;
616   switch (quote_char) {
617   case '\0':
618     chars_to_escape = " \t\\'\"`";
619     break;
620   case '"':
621     chars_to_escape = "$\"`\\";
622     break;
623   case '`':
624   case '\'':
625     return arg;
626   default:
627     assert(false && "Unhandled quote character");
628     return arg;
629   }
630 
631   std::string res;
632   res.reserve(arg.size());
633   for (char c : arg) {
634     if (::strchr(chars_to_escape, c))
635       res.push_back('\\');
636     res.push_back(c);
637   }
638   return res;
639 }
640 
641 OptionsWithRaw::OptionsWithRaw(llvm::StringRef arg_string) {
642   SetFromString(arg_string);
643 }
644 
645 void OptionsWithRaw::SetFromString(llvm::StringRef arg_string) {
646   const llvm::StringRef original_args = arg_string;
647 
648   size_t shift;
649   arg_string = ltrimForArgs(arg_string, shift);
650   std::string arg;
651   char quote;
652 
653   // If the string doesn't start with a dash, we just have no options and just
654   // a raw part.
655   if (!arg_string.starts_with("-")) {
656     m_suffix = std::string(original_args);
657     return;
658   }
659 
660   bool found_suffix = false;
661   while (!arg_string.empty()) {
662     // The length of the prefix before parsing.
663     std::size_t prev_prefix_length = original_args.size() - arg_string.size();
664 
665     // Parse the next argument from the remaining string.
666     std::tie(arg, quote, arg_string) = ParseSingleArgument(arg_string);
667 
668     // If we get an unquoted '--' argument, then we reached the suffix part
669     // of the command.
670     Args::ArgEntry entry(arg, quote, std::nullopt);
671     if (!entry.IsQuoted() && arg == "--") {
672       // The remaining line is the raw suffix, and the line we parsed so far
673       // needs to be interpreted as arguments.
674       m_has_args = true;
675       m_suffix = std::string(arg_string);
676       found_suffix = true;
677 
678       // The length of the prefix after parsing.
679       std::size_t prefix_length = original_args.size() - arg_string.size();
680 
681       // Take the string we know contains all the arguments and actually parse
682       // it as proper arguments.
683       llvm::StringRef prefix = original_args.take_front(prev_prefix_length);
684       m_args = Args(prefix);
685       m_arg_string = prefix;
686 
687       // We also record the part of the string that contains the arguments plus
688       // the delimiter.
689       m_arg_string_with_delimiter = original_args.take_front(prefix_length);
690 
691       // As the rest of the string became the raw suffix, we are done here.
692       break;
693     }
694 
695     arg_string = ltrimForArgs(arg_string, shift);
696   }
697 
698   // If we didn't find a suffix delimiter, the whole string is the raw suffix.
699   if (!found_suffix)
700     m_suffix = std::string(original_args);
701 }
702