1 //===-- FileSpec.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/FileSpec.h"
10 #include "lldb/Utility/RegularExpression.h"
11 #include "lldb/Utility/Stream.h"
12
13 #include "llvm/ADT/SmallString.h"
14 #include "llvm/ADT/SmallVector.h"
15 #include "llvm/ADT/StringRef.h"
16 #include "llvm/ADT/Triple.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/Support/ErrorOr.h"
19 #include "llvm/Support/FileSystem.h"
20 #include "llvm/Support/Program.h"
21 #include "llvm/Support/raw_ostream.h"
22
23 #include <algorithm>
24 #include <system_error>
25 #include <vector>
26
27 #include <assert.h>
28 #include <limits.h>
29 #include <stdio.h>
30 #include <string.h>
31
32 using namespace lldb;
33 using namespace lldb_private;
34
35 namespace {
36
GetNativeStyle()37 static constexpr FileSpec::Style GetNativeStyle() {
38 #if defined(_WIN32)
39 return FileSpec::Style::windows;
40 #else
41 return FileSpec::Style::posix;
42 #endif
43 }
44
PathStyleIsPosix(FileSpec::Style style)45 bool PathStyleIsPosix(FileSpec::Style style) {
46 return (style == FileSpec::Style::posix ||
47 (style == FileSpec::Style::native &&
48 GetNativeStyle() == FileSpec::Style::posix));
49 }
50
GetPathSeparators(FileSpec::Style style)51 const char *GetPathSeparators(FileSpec::Style style) {
52 return llvm::sys::path::get_separator(style).data();
53 }
54
GetPreferredPathSeparator(FileSpec::Style style)55 char GetPreferredPathSeparator(FileSpec::Style style) {
56 return GetPathSeparators(style)[0];
57 }
58
Denormalize(llvm::SmallVectorImpl<char> & path,FileSpec::Style style)59 void Denormalize(llvm::SmallVectorImpl<char> &path, FileSpec::Style style) {
60 if (PathStyleIsPosix(style))
61 return;
62
63 std::replace(path.begin(), path.end(), '/', '\\');
64 }
65
66 } // end anonymous namespace
67
FileSpec()68 FileSpec::FileSpec() : m_style(GetNativeStyle()) {}
69
70 // Default constructor that can take an optional full path to a file on disk.
FileSpec(llvm::StringRef path,Style style)71 FileSpec::FileSpec(llvm::StringRef path, Style style) : m_style(style) {
72 SetFile(path, style);
73 }
74
FileSpec(llvm::StringRef path,const llvm::Triple & triple)75 FileSpec::FileSpec(llvm::StringRef path, const llvm::Triple &triple)
76 : FileSpec{path, triple.isOSWindows() ? Style::windows : Style::posix} {}
77
78 namespace {
79 /// Safely get a character at the specified index.
80 ///
81 /// \param[in] path
82 /// A full, partial, or relative path to a file.
83 ///
84 /// \param[in] i
85 /// An index into path which may or may not be valid.
86 ///
87 /// \return
88 /// The character at index \a i if the index is valid, or 0 if
89 /// the index is not valid.
safeCharAtIndex(const llvm::StringRef & path,size_t i)90 inline char safeCharAtIndex(const llvm::StringRef &path, size_t i) {
91 if (i < path.size())
92 return path[i];
93 return 0;
94 }
95
96 /// Check if a path needs to be normalized.
97 ///
98 /// Check if a path needs to be normalized. We currently consider a
99 /// path to need normalization if any of the following are true
100 /// - path contains "/./"
101 /// - path contains "/../"
102 /// - path contains "//"
103 /// - path ends with "/"
104 /// Paths that start with "./" or with "../" are not considered to
105 /// need normalization since we aren't trying to resolve the path,
106 /// we are just trying to remove redundant things from the path.
107 ///
108 /// \param[in] path
109 /// A full, partial, or relative path to a file.
110 ///
111 /// \return
112 /// Returns \b true if the path needs to be normalized.
needsNormalization(const llvm::StringRef & path)113 bool needsNormalization(const llvm::StringRef &path) {
114 if (path.empty())
115 return false;
116 // We strip off leading "." values so these paths need to be normalized
117 if (path[0] == '.')
118 return true;
119 for (auto i = path.find_first_of("\\/"); i != llvm::StringRef::npos;
120 i = path.find_first_of("\\/", i + 1)) {
121 const auto next = safeCharAtIndex(path, i+1);
122 switch (next) {
123 case 0:
124 // path separator char at the end of the string which should be
125 // stripped unless it is the one and only character
126 return i > 0;
127 case '/':
128 case '\\':
129 // two path separator chars in the middle of a path needs to be
130 // normalized
131 if (i > 0)
132 return true;
133 ++i;
134 break;
135
136 case '.': {
137 const auto next_next = safeCharAtIndex(path, i+2);
138 switch (next_next) {
139 default: break;
140 case 0: return true; // ends with "/."
141 case '/':
142 case '\\':
143 return true; // contains "/./"
144 case '.': {
145 const auto next_next_next = safeCharAtIndex(path, i+3);
146 switch (next_next_next) {
147 default: break;
148 case 0: return true; // ends with "/.."
149 case '/':
150 case '\\':
151 return true; // contains "/../"
152 }
153 break;
154 }
155 }
156 }
157 break;
158
159 default:
160 break;
161 }
162 }
163 return false;
164 }
165
166
167 }
168
SetFile(llvm::StringRef pathname)169 void FileSpec::SetFile(llvm::StringRef pathname) { SetFile(pathname, m_style); }
170
171 // Update the contents of this object with a new path. The path will be split
172 // up into a directory and filename and stored as uniqued string values for
173 // quick comparison and efficient memory usage.
SetFile(llvm::StringRef pathname,Style style)174 void FileSpec::SetFile(llvm::StringRef pathname, Style style) {
175 m_filename.Clear();
176 m_directory.Clear();
177 m_is_resolved = false;
178 m_style = (style == Style::native) ? GetNativeStyle() : style;
179
180 if (pathname.empty())
181 return;
182
183 llvm::SmallString<128> resolved(pathname);
184
185 // Normalize the path by removing ".", ".." and other redundant components.
186 if (needsNormalization(resolved))
187 llvm::sys::path::remove_dots(resolved, true, m_style);
188
189 // Normalize back slashes to forward slashes
190 if (m_style == Style::windows)
191 std::replace(resolved.begin(), resolved.end(), '\\', '/');
192
193 if (resolved.empty()) {
194 // If we have no path after normalization set the path to the current
195 // directory. This matches what python does and also a few other path
196 // utilities.
197 m_filename.SetString(".");
198 return;
199 }
200
201 // Split path into filename and directory. We rely on the underlying char
202 // pointer to be nullptr when the components are empty.
203 llvm::StringRef filename = llvm::sys::path::filename(resolved, m_style);
204 if(!filename.empty())
205 m_filename.SetString(filename);
206
207 llvm::StringRef directory = llvm::sys::path::parent_path(resolved, m_style);
208 if(!directory.empty())
209 m_directory.SetString(directory);
210 }
211
SetFile(llvm::StringRef path,const llvm::Triple & triple)212 void FileSpec::SetFile(llvm::StringRef path, const llvm::Triple &triple) {
213 return SetFile(path, triple.isOSWindows() ? Style::windows : Style::posix);
214 }
215
216 // Convert to pointer operator. This allows code to check any FileSpec objects
217 // to see if they contain anything valid using code such as:
218 //
219 // if (file_spec)
220 // {}
operator bool() const221 FileSpec::operator bool() const { return m_filename || m_directory; }
222
223 // Logical NOT operator. This allows code to check any FileSpec objects to see
224 // if they are invalid using code such as:
225 //
226 // if (!file_spec)
227 // {}
operator !() const228 bool FileSpec::operator!() const { return !m_directory && !m_filename; }
229
DirectoryEquals(const FileSpec & rhs) const230 bool FileSpec::DirectoryEquals(const FileSpec &rhs) const {
231 const bool case_sensitive = IsCaseSensitive() || rhs.IsCaseSensitive();
232 return ConstString::Equals(m_directory, rhs.m_directory, case_sensitive);
233 }
234
FileEquals(const FileSpec & rhs) const235 bool FileSpec::FileEquals(const FileSpec &rhs) const {
236 const bool case_sensitive = IsCaseSensitive() || rhs.IsCaseSensitive();
237 return ConstString::Equals(m_filename, rhs.m_filename, case_sensitive);
238 }
239
240 // Equal to operator
operator ==(const FileSpec & rhs) const241 bool FileSpec::operator==(const FileSpec &rhs) const {
242 return FileEquals(rhs) && DirectoryEquals(rhs);
243 }
244
245 // Not equal to operator
operator !=(const FileSpec & rhs) const246 bool FileSpec::operator!=(const FileSpec &rhs) const { return !(*this == rhs); }
247
248 // Less than operator
operator <(const FileSpec & rhs) const249 bool FileSpec::operator<(const FileSpec &rhs) const {
250 return FileSpec::Compare(*this, rhs, true) < 0;
251 }
252
253 // Dump a FileSpec object to a stream
operator <<(Stream & s,const FileSpec & f)254 Stream &lldb_private::operator<<(Stream &s, const FileSpec &f) {
255 f.Dump(s.AsRawOstream());
256 return s;
257 }
258
259 // Clear this object by releasing both the directory and filename string values
260 // and making them both the empty string.
Clear()261 void FileSpec::Clear() {
262 m_directory.Clear();
263 m_filename.Clear();
264 }
265
266 // Compare two FileSpec objects. If "full" is true, then both the directory and
267 // the filename must match. If "full" is false, then the directory names for
268 // "a" and "b" are only compared if they are both non-empty. This allows a
269 // FileSpec object to only contain a filename and it can match FileSpec objects
270 // that have matching filenames with different paths.
271 //
272 // Return -1 if the "a" is less than "b", 0 if "a" is equal to "b" and "1" if
273 // "a" is greater than "b".
Compare(const FileSpec & a,const FileSpec & b,bool full)274 int FileSpec::Compare(const FileSpec &a, const FileSpec &b, bool full) {
275 int result = 0;
276
277 // case sensitivity of compare
278 const bool case_sensitive = a.IsCaseSensitive() || b.IsCaseSensitive();
279
280 // If full is true, then we must compare both the directory and filename.
281
282 // If full is false, then if either directory is empty, then we match on the
283 // basename only, and if both directories have valid values, we still do a
284 // full compare. This allows for matching when we just have a filename in one
285 // of the FileSpec objects.
286
287 if (full || (a.m_directory && b.m_directory)) {
288 result = ConstString::Compare(a.m_directory, b.m_directory, case_sensitive);
289 if (result)
290 return result;
291 }
292 return ConstString::Compare(a.m_filename, b.m_filename, case_sensitive);
293 }
294
Equal(const FileSpec & a,const FileSpec & b,bool full)295 bool FileSpec::Equal(const FileSpec &a, const FileSpec &b, bool full) {
296 if (full || (a.GetDirectory() && b.GetDirectory()))
297 return a == b;
298
299 return a.FileEquals(b);
300 }
301
Match(const FileSpec & pattern,const FileSpec & file)302 bool FileSpec::Match(const FileSpec &pattern, const FileSpec &file) {
303 if (pattern.GetDirectory())
304 return pattern == file;
305 if (pattern.GetFilename())
306 return pattern.FileEquals(file);
307 return true;
308 }
309
GuessPathStyle(llvm::StringRef absolute_path)310 llvm::Optional<FileSpec::Style> FileSpec::GuessPathStyle(llvm::StringRef absolute_path) {
311 if (absolute_path.startswith("/"))
312 return Style::posix;
313 if (absolute_path.startswith(R"(\\)"))
314 return Style::windows;
315 if (absolute_path.size() > 3 && llvm::isAlpha(absolute_path[0]) &&
316 absolute_path.substr(1, 2) == R"(:\)")
317 return Style::windows;
318 return llvm::None;
319 }
320
321 // Dump the object to the supplied stream. If the object contains a valid
322 // directory name, it will be displayed followed by a directory delimiter, and
323 // the filename.
Dump(llvm::raw_ostream & s) const324 void FileSpec::Dump(llvm::raw_ostream &s) const {
325 std::string path{GetPath(true)};
326 s << path;
327 char path_separator = GetPreferredPathSeparator(m_style);
328 if (!m_filename && !path.empty() && path.back() != path_separator)
329 s << path_separator;
330 }
331
GetPathStyle() const332 FileSpec::Style FileSpec::GetPathStyle() const { return m_style; }
333
334 // Directory string get accessor.
GetDirectory()335 ConstString &FileSpec::GetDirectory() { return m_directory; }
336
337 // Directory string const get accessor.
GetDirectory() const338 ConstString FileSpec::GetDirectory() const { return m_directory; }
339
340 // Filename string get accessor.
GetFilename()341 ConstString &FileSpec::GetFilename() { return m_filename; }
342
343 // Filename string const get accessor.
GetFilename() const344 ConstString FileSpec::GetFilename() const { return m_filename; }
345
346 // Extract the directory and path into a fixed buffer. This is needed as the
347 // directory and path are stored in separate string values.
GetPath(char * path,size_t path_max_len,bool denormalize) const348 size_t FileSpec::GetPath(char *path, size_t path_max_len,
349 bool denormalize) const {
350 if (!path)
351 return 0;
352
353 std::string result = GetPath(denormalize);
354 ::snprintf(path, path_max_len, "%s", result.c_str());
355 return std::min(path_max_len - 1, result.length());
356 }
357
GetPath(bool denormalize) const358 std::string FileSpec::GetPath(bool denormalize) const {
359 llvm::SmallString<64> result;
360 GetPath(result, denormalize);
361 return std::string(result.begin(), result.end());
362 }
363
GetCString(bool denormalize) const364 const char *FileSpec::GetCString(bool denormalize) const {
365 return ConstString{GetPath(denormalize)}.AsCString(nullptr);
366 }
367
GetPath(llvm::SmallVectorImpl<char> & path,bool denormalize) const368 void FileSpec::GetPath(llvm::SmallVectorImpl<char> &path,
369 bool denormalize) const {
370 path.append(m_directory.GetStringRef().begin(),
371 m_directory.GetStringRef().end());
372 // Since the path was normalized and all paths use '/' when stored in these
373 // objects, we don't need to look for the actual syntax specific path
374 // separator, we just look for and insert '/'.
375 if (m_directory && m_filename && m_directory.GetStringRef().back() != '/' &&
376 m_filename.GetStringRef().back() != '/')
377 path.insert(path.end(), '/');
378 path.append(m_filename.GetStringRef().begin(),
379 m_filename.GetStringRef().end());
380 if (denormalize && !path.empty())
381 Denormalize(path, m_style);
382 }
383
GetFileNameExtension() const384 ConstString FileSpec::GetFileNameExtension() const {
385 return ConstString(
386 llvm::sys::path::extension(m_filename.GetStringRef(), m_style));
387 }
388
GetFileNameStrippingExtension() const389 ConstString FileSpec::GetFileNameStrippingExtension() const {
390 return ConstString(llvm::sys::path::stem(m_filename.GetStringRef(), m_style));
391 }
392
393 // Return the size in bytes that this object takes in memory. This returns the
394 // size in bytes of this object, not any shared string values it may refer to.
MemorySize() const395 size_t FileSpec::MemorySize() const {
396 return m_filename.MemorySize() + m_directory.MemorySize();
397 }
398
399 FileSpec
CopyByAppendingPathComponent(llvm::StringRef component) const400 FileSpec::CopyByAppendingPathComponent(llvm::StringRef component) const {
401 FileSpec ret = *this;
402 ret.AppendPathComponent(component);
403 return ret;
404 }
405
CopyByRemovingLastPathComponent() const406 FileSpec FileSpec::CopyByRemovingLastPathComponent() const {
407 llvm::SmallString<64> current_path;
408 GetPath(current_path, false);
409 if (llvm::sys::path::has_parent_path(current_path, m_style))
410 return FileSpec(llvm::sys::path::parent_path(current_path, m_style),
411 m_style);
412 return *this;
413 }
414
GetLastPathComponent() const415 ConstString FileSpec::GetLastPathComponent() const {
416 llvm::SmallString<64> current_path;
417 GetPath(current_path, false);
418 return ConstString(llvm::sys::path::filename(current_path, m_style));
419 }
420
PrependPathComponent(llvm::StringRef component)421 void FileSpec::PrependPathComponent(llvm::StringRef component) {
422 llvm::SmallString<64> new_path(component);
423 llvm::SmallString<64> current_path;
424 GetPath(current_path, false);
425 llvm::sys::path::append(new_path,
426 llvm::sys::path::begin(current_path, m_style),
427 llvm::sys::path::end(current_path), m_style);
428 SetFile(new_path, m_style);
429 }
430
PrependPathComponent(const FileSpec & new_path)431 void FileSpec::PrependPathComponent(const FileSpec &new_path) {
432 return PrependPathComponent(new_path.GetPath(false));
433 }
434
AppendPathComponent(llvm::StringRef component)435 void FileSpec::AppendPathComponent(llvm::StringRef component) {
436 llvm::SmallString<64> current_path;
437 GetPath(current_path, false);
438 llvm::sys::path::append(current_path, m_style, component);
439 SetFile(current_path, m_style);
440 }
441
AppendPathComponent(const FileSpec & new_path)442 void FileSpec::AppendPathComponent(const FileSpec &new_path) {
443 return AppendPathComponent(new_path.GetPath(false));
444 }
445
RemoveLastPathComponent()446 bool FileSpec::RemoveLastPathComponent() {
447 llvm::SmallString<64> current_path;
448 GetPath(current_path, false);
449 if (llvm::sys::path::has_parent_path(current_path, m_style)) {
450 SetFile(llvm::sys::path::parent_path(current_path, m_style));
451 return true;
452 }
453 return false;
454 }
455 /// Returns true if the filespec represents an implementation source
456 /// file (files with a ".c", ".cpp", ".m", ".mm" (many more)
457 /// extension).
458 ///
459 /// \return
460 /// \b true if the filespec represents an implementation source
461 /// file, \b false otherwise.
IsSourceImplementationFile() const462 bool FileSpec::IsSourceImplementationFile() const {
463 ConstString extension(GetFileNameExtension());
464 if (!extension)
465 return false;
466
467 static RegularExpression g_source_file_regex(llvm::StringRef(
468 "^.([cC]|[mM]|[mM][mM]|[cC][pP][pP]|[cC]\\+\\+|[cC][xX][xX]|[cC][cC]|["
469 "cC][pP]|[sS]|[aA][sS][mM]|[fF]|[fF]77|[fF]90|[fF]95|[fF]03|[fF][oO]["
470 "rR]|[fF][tT][nN]|[fF][pP][pP]|[aA][dD][aA]|[aA][dD][bB]|[aA][dD][sS])"
471 "$"));
472 return g_source_file_regex.Execute(extension.GetStringRef());
473 }
474
IsRelative() const475 bool FileSpec::IsRelative() const {
476 return !IsAbsolute();
477 }
478
IsAbsolute() const479 bool FileSpec::IsAbsolute() const {
480 llvm::SmallString<64> current_path;
481 GetPath(current_path, false);
482
483 // Early return if the path is empty.
484 if (current_path.empty())
485 return false;
486
487 // We consider paths starting with ~ to be absolute.
488 if (current_path[0] == '~')
489 return true;
490
491 return llvm::sys::path::is_absolute(current_path, m_style);
492 }
493
MakeAbsolute(const FileSpec & dir)494 void FileSpec::MakeAbsolute(const FileSpec &dir) {
495 if (IsRelative())
496 PrependPathComponent(dir);
497 }
498
format(const FileSpec & F,raw_ostream & Stream,StringRef Style)499 void llvm::format_provider<FileSpec>::format(const FileSpec &F,
500 raw_ostream &Stream,
501 StringRef Style) {
502 assert(
503 (Style.empty() || Style.equals_lower("F") || Style.equals_lower("D")) &&
504 "Invalid FileSpec style!");
505
506 StringRef dir = F.GetDirectory().GetStringRef();
507 StringRef file = F.GetFilename().GetStringRef();
508
509 if (dir.empty() && file.empty()) {
510 Stream << "(empty)";
511 return;
512 }
513
514 if (Style.equals_lower("F")) {
515 Stream << (file.empty() ? "(empty)" : file);
516 return;
517 }
518
519 // Style is either D or empty, either way we need to print the directory.
520 if (!dir.empty()) {
521 // Directory is stored in normalized form, which might be different than
522 // preferred form. In order to handle this, we need to cut off the
523 // filename, then denormalize, then write the entire denorm'ed directory.
524 llvm::SmallString<64> denormalized_dir = dir;
525 Denormalize(denormalized_dir, F.GetPathStyle());
526 Stream << denormalized_dir;
527 Stream << GetPreferredPathSeparator(F.GetPathStyle());
528 }
529
530 if (Style.equals_lower("D")) {
531 // We only want to print the directory, so now just exit.
532 if (dir.empty())
533 Stream << "(empty)";
534 return;
535 }
536
537 if (!file.empty())
538 Stream << file;
539 }
540
enumeration(IO & io,FileSpecStyle & value)541 void llvm::yaml::ScalarEnumerationTraits<FileSpecStyle>::enumeration(
542 IO &io, FileSpecStyle &value) {
543 io.enumCase(value, "windows", FileSpecStyle(FileSpec::Style::windows));
544 io.enumCase(value, "posix", FileSpecStyle(FileSpec::Style::posix));
545 io.enumCase(value, "native", FileSpecStyle(FileSpec::Style::native));
546 }
547
mapping(IO & io,FileSpec & f)548 void llvm::yaml::MappingTraits<FileSpec>::mapping(IO &io, FileSpec &f) {
549 io.mapRequired("directory", f.m_directory);
550 io.mapRequired("file", f.m_filename);
551 io.mapRequired("resolved", f.m_is_resolved);
552 FileSpecStyle style = f.m_style;
553 io.mapRequired("style", style);
554 f.m_style = style;
555 }
556