comparison lldb/source/Utility/Args.cpp @ 150:1d019706d866

LLVM10
author anatofuz
date Thu, 13 Feb 2020 15:10:13 +0900
parents
children 0572611fdcc8
comparison
equal deleted inserted replaced
147:c2174574ed3a 150:1d019706d866
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 a whitelisted 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 a whitelisted 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::operator=(const Args &rhs) {
179 Clear();
180
181 m_argv.clear();
182 m_entries.clear();
183 for (auto &entry : rhs.m_entries) {
184 m_entries.emplace_back(entry.ref(), entry.quote);
185 m_argv.push_back(m_entries.back().data());
186 }
187 m_argv.push_back(nullptr);
188 return *this;
189 }
190
191 // Destructor
192 Args::~Args() {}
193
194 void Args::Dump(Stream &s, const char *label_name) const {
195 if (!label_name)
196 return;
197
198 int i = 0;
199 for (auto &entry : m_entries) {
200 s.Indent();
201 s.Format("{0}[{1}]=\"{2}\"\n", label_name, i++, entry.ref());
202 }
203 s.Format("{0}[{1}]=NULL\n", label_name, i);
204 s.EOL();
205 }
206
207 bool Args::GetCommandString(std::string &command) const {
208 command.clear();
209
210 for (size_t i = 0; i < m_entries.size(); ++i) {
211 if (i > 0)
212 command += ' ';
213 command += m_entries[i].ref();
214 }
215
216 return !m_entries.empty();
217 }
218
219 bool Args::GetQuotedCommandString(std::string &command) const {
220 command.clear();
221
222 for (size_t i = 0; i < m_entries.size(); ++i) {
223 if (i > 0)
224 command += ' ';
225
226 if (m_entries[i].quote) {
227 command += m_entries[i].quote;
228 command += m_entries[i].ref();
229 command += m_entries[i].quote;
230 } else {
231 command += m_entries[i].ref();
232 }
233 }
234
235 return !m_entries.empty();
236 }
237
238 void Args::SetCommandString(llvm::StringRef command) {
239 Clear();
240 m_argv.clear();
241
242 command = ltrimForArgs(command);
243 std::string arg;
244 char quote;
245 while (!command.empty()) {
246 std::tie(arg, quote, command) = ParseSingleArgument(command);
247 m_entries.emplace_back(arg, quote);
248 m_argv.push_back(m_entries.back().data());
249 command = ltrimForArgs(command);
250 }
251 m_argv.push_back(nullptr);
252 }
253
254 size_t Args::GetArgumentCount() const { return m_entries.size(); }
255
256 const char *Args::GetArgumentAtIndex(size_t idx) const {
257 if (idx < m_argv.size())
258 return m_argv[idx];
259 return nullptr;
260 }
261
262 char **Args::GetArgumentVector() {
263 assert(!m_argv.empty());
264 // TODO: functions like execve and posix_spawnp exhibit undefined behavior
265 // when argv or envp is null. So the code below is actually wrong. However,
266 // other code in LLDB depends on it being null. The code has been acting
267 // this way for some time, so it makes sense to leave it this way until
268 // someone has the time to come along and fix it.
269 return (m_argv.size() > 1) ? m_argv.data() : nullptr;
270 }
271
272 const char **Args::GetConstArgumentVector() const {
273 assert(!m_argv.empty());
274 return (m_argv.size() > 1) ? const_cast<const char **>(m_argv.data())
275 : nullptr;
276 }
277
278 void Args::Shift() {
279 // Don't pop the last NULL terminator from the argv array
280 if (m_entries.empty())
281 return;
282 m_argv.erase(m_argv.begin());
283 m_entries.erase(m_entries.begin());
284 }
285
286 void Args::Unshift(llvm::StringRef arg_str, char quote_char) {
287 InsertArgumentAtIndex(0, arg_str, quote_char);
288 }
289
290 void Args::AppendArguments(const Args &rhs) {
291 assert(m_argv.size() == m_entries.size() + 1);
292 assert(m_argv.back() == nullptr);
293 m_argv.pop_back();
294 for (auto &entry : rhs.m_entries) {
295 m_entries.emplace_back(entry.ref(), entry.quote);
296 m_argv.push_back(m_entries.back().data());
297 }
298 m_argv.push_back(nullptr);
299 }
300
301 void Args::AppendArguments(const char **argv) {
302 size_t argc = ArgvToArgc(argv);
303
304 assert(m_argv.size() == m_entries.size() + 1);
305 assert(m_argv.back() == nullptr);
306 m_argv.pop_back();
307 for (auto arg : llvm::makeArrayRef(argv, argc)) {
308 m_entries.emplace_back(arg, '\0');
309 m_argv.push_back(m_entries.back().data());
310 }
311
312 m_argv.push_back(nullptr);
313 }
314
315 void Args::AppendArgument(llvm::StringRef arg_str, char quote_char) {
316 InsertArgumentAtIndex(GetArgumentCount(), arg_str, quote_char);
317 }
318
319 void Args::InsertArgumentAtIndex(size_t idx, llvm::StringRef arg_str,
320 char quote_char) {
321 assert(m_argv.size() == m_entries.size() + 1);
322 assert(m_argv.back() == nullptr);
323
324 if (idx > m_entries.size())
325 return;
326 m_entries.emplace(m_entries.begin() + idx, arg_str, quote_char);
327 m_argv.insert(m_argv.begin() + idx, m_entries[idx].data());
328 }
329
330 void Args::ReplaceArgumentAtIndex(size_t idx, llvm::StringRef arg_str,
331 char quote_char) {
332 assert(m_argv.size() == m_entries.size() + 1);
333 assert(m_argv.back() == nullptr);
334
335 if (idx >= m_entries.size())
336 return;
337
338 m_entries[idx] = ArgEntry(arg_str, quote_char);
339 m_argv[idx] = m_entries[idx].data();
340 }
341
342 void Args::DeleteArgumentAtIndex(size_t idx) {
343 if (idx >= m_entries.size())
344 return;
345
346 m_argv.erase(m_argv.begin() + idx);
347 m_entries.erase(m_entries.begin() + idx);
348 }
349
350 void Args::SetArguments(size_t argc, const char **argv) {
351 Clear();
352
353 auto args = llvm::makeArrayRef(argv, argc);
354 m_entries.resize(argc);
355 m_argv.resize(argc + 1);
356 for (size_t i = 0; i < args.size(); ++i) {
357 char quote =
358 ((args[i][0] == '\'') || (args[i][0] == '"') || (args[i][0] == '`'))
359 ? args[i][0]
360 : '\0';
361
362 m_entries[i] = ArgEntry(args[i], quote);
363 m_argv[i] = m_entries[i].data();
364 }
365 }
366
367 void Args::SetArguments(const char **argv) {
368 SetArguments(ArgvToArgc(argv), argv);
369 }
370
371 void Args::Clear() {
372 m_entries.clear();
373 m_argv.clear();
374 m_argv.push_back(nullptr);
375 }
376
377 const char *Args::GetShellSafeArgument(const FileSpec &shell,
378 const char *unsafe_arg,
379 std::string &safe_arg) {
380 struct ShellDescriptor {
381 ConstString m_basename;
382 const char *m_escapables;
383 };
384
385 static ShellDescriptor g_Shells[] = {{ConstString("bash"), " '\"<>()&"},
386 {ConstString("tcsh"), " '\"<>()&$"},
387 {ConstString("sh"), " '\"<>()&"}};
388
389 // safe minimal set
390 const char *escapables = " '\"";
391
392 if (auto basename = shell.GetFilename()) {
393 for (const auto &Shell : g_Shells) {
394 if (Shell.m_basename == basename) {
395 escapables = Shell.m_escapables;
396 break;
397 }
398 }
399 }
400
401 safe_arg.assign(unsafe_arg);
402 size_t prev_pos = 0;
403 while (prev_pos < safe_arg.size()) {
404 // Escape spaces and quotes
405 size_t pos = safe_arg.find_first_of(escapables, prev_pos);
406 if (pos != std::string::npos) {
407 safe_arg.insert(pos, 1, '\\');
408 prev_pos = pos + 2;
409 } else
410 break;
411 }
412 return safe_arg.c_str();
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 (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
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 found_suffix = true;
684 m_suffix = std::string(original_args);
685 }
686 }