1 | // Protocol Buffers - Google's data interchange format |
2 | // Copyright 2008 Google Inc. All rights reserved. |
3 | // https://developers.google.com/protocol-buffers/ |
4 | // |
5 | // Redistribution and use in source and binary forms, with or without |
6 | // modification, are permitted provided that the following conditions are |
7 | // met: |
8 | // |
9 | // * Redistributions of source code must retain the above copyright |
10 | // notice, this list of conditions and the following disclaimer. |
11 | // * Redistributions in binary form must reproduce the above |
12 | // copyright notice, this list of conditions and the following disclaimer |
13 | // in the documentation and/or other materials provided with the |
14 | // distribution. |
15 | // * Neither the name of Google Inc. nor the names of its |
16 | // contributors may be used to endorse or promote products derived from |
17 | // this software without specific prior written permission. |
18 | // |
19 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
20 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
21 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
22 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
23 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
24 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
25 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
26 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
27 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
28 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
29 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
30 | |
31 | // Author: laszlocsomor@google.com (Laszlo Csomor) |
32 | // Based on original Protocol Buffers design by |
33 | // Sanjay Ghemawat, Jeff Dean, and others. |
34 | |
35 | // Implementation for long-path-aware open/mkdir/access/etc. on Windows, as well |
36 | // as for the supporting utility functions. |
37 | // |
38 | // These functions convert the input path to an absolute Windows path |
39 | // with "\\?\" prefix, then pass that to _wopen/_wmkdir/_waccess/etc. |
40 | // (declared in <io.h>) respectively. This allows working with files/directories |
41 | // whose paths are longer than MAX_PATH (260 chars). |
42 | // |
43 | // This file is only used on Windows, it's empty on other platforms. |
44 | |
45 | #if defined(_WIN32) && !defined(_XBOX_ONE) |
46 | |
47 | // Comment this out to fall back to using the ANSI versions (open, mkdir, ...) |
48 | // instead of the Unicode ones (_wopen, _wmkdir, ...). Doing so can be useful to |
49 | // debug failing tests if that's caused by the long path support. |
50 | #define SUPPORT_LONGPATHS |
51 | |
52 | #include <google/protobuf/io/io_win32.h> |
53 | |
54 | #include <ctype.h> |
55 | #include <direct.h> |
56 | #include <errno.h> |
57 | #include <fcntl.h> |
58 | #include <io.h> |
59 | #include <sys/stat.h> |
60 | #include <sys/types.h> |
61 | #include <wctype.h> |
62 | |
63 | #ifndef WIN32_LEAN_AND_MEAN |
64 | #define WIN32_LEAN_AND_MEAN 1 |
65 | #endif |
66 | |
67 | #include <windows.h> |
68 | |
69 | #include <memory> |
70 | #include <sstream> |
71 | #include <string> |
72 | #include <vector> |
73 | |
74 | namespace google { |
75 | namespace protobuf { |
76 | namespace io { |
77 | namespace win32 { |
78 | namespace { |
79 | |
80 | using std::string; |
81 | using std::wstring; |
82 | |
83 | template <typename char_type> |
84 | struct CharTraits { |
85 | static bool is_alpha(char_type ch); |
86 | }; |
87 | |
88 | template <> |
89 | struct CharTraits<char> { |
90 | static bool is_alpha(char ch) { return isalpha(ch); } |
91 | }; |
92 | |
93 | template <> |
94 | struct CharTraits<wchar_t> { |
95 | static bool is_alpha(wchar_t ch) { return iswalpha(ch); } |
96 | }; |
97 | |
98 | template <typename char_type> |
99 | bool null_or_empty(const char_type* s) { |
100 | return s == nullptr || *s == 0; |
101 | } |
102 | |
103 | // Returns true if the path starts with a drive letter, e.g. "c:". |
104 | // Note that this won't check for the "\" after the drive letter, so this also |
105 | // returns true for "c:foo" (which is "c:\${PWD}\foo"). |
106 | // This check requires that a path not have a longpath prefix ("\\?\"). |
107 | template <typename char_type> |
108 | bool has_drive_letter(const char_type* ch) { |
109 | return CharTraits<char_type>::is_alpha(ch[0]) && ch[1] == ':'; |
110 | } |
111 | |
112 | // Returns true if the path starts with a longpath prefix ("\\?\"). |
113 | template <typename char_type> |
114 | bool has_longpath_prefix(const char_type* path) { |
115 | return path[0] == '\\' && path[1] == '\\' && path[2] == '?' && |
116 | path[3] == '\\'; |
117 | } |
118 | |
119 | template <typename char_type> |
120 | bool is_separator(char_type c) { |
121 | return c == '/' || c == '\\'; |
122 | } |
123 | |
124 | // Returns true if the path starts with a drive specifier (e.g. "c:\"). |
125 | template <typename char_type> |
126 | bool is_path_absolute(const char_type* path) { |
127 | return has_drive_letter(path) && is_separator(path[2]); |
128 | } |
129 | |
130 | template <typename char_type> |
131 | bool is_drive_relative(const char_type* path) { |
132 | return has_drive_letter(path) && (path[2] == 0 || !is_separator(path[2])); |
133 | } |
134 | |
135 | wstring join_paths(const wstring& path1, const wstring& path2) { |
136 | if (path1.empty() || is_path_absolute(path2.c_str()) || |
137 | has_longpath_prefix(path2.c_str())) { |
138 | return path2; |
139 | } |
140 | if (path2.empty()) { |
141 | return path1; |
142 | } |
143 | |
144 | if (is_separator(path1[path1.size() - 1])) { |
145 | return is_separator(path2[0]) ? (path1 + path2.substr(1)) |
146 | : (path1 + path2); |
147 | } else { |
148 | return is_separator(path2[0]) ? (path1 + path2) |
149 | : (path1 + L'\\' + path2); |
150 | } |
151 | } |
152 | |
153 | wstring normalize(wstring path) { |
154 | if (has_longpath_prefix(path.c_str())) { |
155 | path = path.substr(4); |
156 | } |
157 | |
158 | static const wstring dot(L"." ); |
159 | static const wstring dotdot(L".." ); |
160 | const WCHAR* p = path.c_str(); |
161 | |
162 | std::vector<wstring> segments; |
163 | int segment_start = -1; |
164 | // Find the path segments in `path` (separated by "/"). |
165 | for (int i = 0;; ++i) { |
166 | if (!is_separator(p[i]) && p[i] != L'\0') { |
167 | // The current character does not end a segment, so start one unless it's |
168 | // already started. |
169 | if (segment_start < 0) { |
170 | segment_start = i; |
171 | } |
172 | } else if (segment_start >= 0 && i > segment_start) { |
173 | // The current character is "/" or "\0", so this ends a segment. |
174 | // Add that to `segments` if there's anything to add; handle "." and "..". |
175 | wstring segment(p, segment_start, i - segment_start); |
176 | segment_start = -1; |
177 | if (segment == dotdot) { |
178 | if (!segments.empty() && |
179 | (!has_drive_letter(segments[0].c_str()) || segments.size() > 1)) { |
180 | segments.pop_back(); |
181 | } |
182 | } else if (segment != dot && !segment.empty()) { |
183 | segments.push_back(segment); |
184 | } |
185 | } |
186 | if (p[i] == L'\0') { |
187 | break; |
188 | } |
189 | } |
190 | |
191 | // Handle the case when `path` is just a drive specifier (or some degenerate |
192 | // form of it, e.g. "c:\.."). |
193 | if (segments.size() == 1 && segments[0].size() == 2 && |
194 | has_drive_letter(segments[0].c_str())) { |
195 | return segments[0] + L'\\'; |
196 | } |
197 | |
198 | // Join all segments. |
199 | bool first = true; |
200 | std::wstringstream result; |
201 | for (int i = 0; i < segments.size(); ++i) { |
202 | if (!first) { |
203 | result << L'\\'; |
204 | } |
205 | first = false; |
206 | result << segments[i]; |
207 | } |
208 | // Preserve trailing separator if the input contained it. |
209 | if (!path.empty() && is_separator(p[path.size() - 1])) { |
210 | result << L'\\'; |
211 | } |
212 | return result.str(); |
213 | } |
214 | |
215 | bool as_windows_path(const char* path, wstring* result) { |
216 | if (null_or_empty(path)) { |
217 | result->clear(); |
218 | return true; |
219 | } |
220 | wstring wpath; |
221 | if (!strings::utf8_to_wcs(path, &wpath)) { |
222 | return false; |
223 | } |
224 | if (has_longpath_prefix(wpath.c_str())) { |
225 | *result = wpath; |
226 | return true; |
227 | } |
228 | if (is_separator(path[0]) || is_drive_relative(path)) { |
229 | return false; |
230 | } |
231 | |
232 | |
233 | if (!is_path_absolute(wpath.c_str())) { |
234 | int size = ::GetCurrentDirectoryW(0, nullptr); |
235 | if (size == 0 && GetLastError() != ERROR_INSUFFICIENT_BUFFER) { |
236 | return false; |
237 | } |
238 | std::unique_ptr<WCHAR[]> wcwd(new WCHAR[size]); |
239 | ::GetCurrentDirectoryW(size, wcwd.get()); |
240 | wpath = join_paths(wcwd.get(), wpath); |
241 | } |
242 | wpath = normalize(wpath); |
243 | if (!has_longpath_prefix(wpath.c_str())) { |
244 | // Add the "\\?\" prefix unconditionally. This way we prevent the Win32 API |
245 | // from processing the path and "helpfully" removing trailing dots from the |
246 | // path, for example. |
247 | // See https://github.com/bazelbuild/bazel/issues/2935 |
248 | wpath = wstring(L"\\\\?\\" ) + wpath; |
249 | } |
250 | *result = wpath; |
251 | return true; |
252 | } |
253 | |
254 | } // namespace |
255 | |
256 | int open(const char* path, int flags, int mode) { |
257 | #ifdef SUPPORT_LONGPATHS |
258 | wstring wpath; |
259 | if (!as_windows_path(path, &wpath)) { |
260 | errno = ENOENT; |
261 | return -1; |
262 | } |
263 | return ::_wopen(wpath.c_str(), flags, mode); |
264 | #else |
265 | return ::_open(path, flags, mode); |
266 | #endif |
267 | } |
268 | |
269 | int mkdir(const char* path, int /*_mode*/) { |
270 | #ifdef SUPPORT_LONGPATHS |
271 | wstring wpath; |
272 | if (!as_windows_path(path, &wpath)) { |
273 | errno = ENOENT; |
274 | return -1; |
275 | } |
276 | return ::_wmkdir(wpath.c_str()); |
277 | #else // not SUPPORT_LONGPATHS |
278 | return ::_mkdir(path); |
279 | #endif // not SUPPORT_LONGPATHS |
280 | } |
281 | |
282 | int access(const char* path, int mode) { |
283 | #ifdef SUPPORT_LONGPATHS |
284 | wstring wpath; |
285 | if (!as_windows_path(path, &wpath)) { |
286 | errno = ENOENT; |
287 | return -1; |
288 | } |
289 | return ::_waccess(wpath.c_str(), mode); |
290 | #else |
291 | return ::_access(path, mode); |
292 | #endif |
293 | } |
294 | |
295 | int chdir(const char* path) { |
296 | #ifdef SUPPORT_LONGPATHS |
297 | wstring wpath; |
298 | if (!as_windows_path(path, &wpath)) { |
299 | errno = ENOENT; |
300 | return -1; |
301 | } |
302 | return ::_wchdir(wpath.c_str()); |
303 | #else |
304 | return ::_chdir(path); |
305 | #endif |
306 | } |
307 | |
308 | int stat(const char* path, struct _stat* buffer) { |
309 | #ifdef SUPPORT_LONGPATHS |
310 | wstring wpath; |
311 | if (!as_windows_path(path, &wpath)) { |
312 | errno = ENOENT; |
313 | return -1; |
314 | } |
315 | return ::_wstat(wpath.c_str(), buffer); |
316 | #else // not SUPPORT_LONGPATHS |
317 | return ::_stat(path, buffer); |
318 | #endif // not SUPPORT_LONGPATHS |
319 | } |
320 | |
321 | FILE* fopen(const char* path, const char* mode) { |
322 | #ifdef SUPPORT_LONGPATHS |
323 | if (null_or_empty(path)) { |
324 | errno = EINVAL; |
325 | return nullptr; |
326 | } |
327 | wstring wpath; |
328 | if (!as_windows_path(path, &wpath)) { |
329 | errno = ENOENT; |
330 | return nullptr; |
331 | } |
332 | wstring wmode; |
333 | if (!strings::utf8_to_wcs(mode, &wmode)) { |
334 | errno = EINVAL; |
335 | return nullptr; |
336 | } |
337 | return ::_wfopen(wpath.c_str(), wmode.c_str()); |
338 | #else |
339 | return ::fopen(path, mode); |
340 | #endif |
341 | } |
342 | |
343 | int close(int fd) { return ::_close(fd); } |
344 | |
345 | int dup(int fd) { return ::_dup(fd); } |
346 | |
347 | int dup2(int fd1, int fd2) { return ::_dup2(fd1, fd2); } |
348 | |
349 | int read(int fd, void* buffer, size_t size) { |
350 | return ::_read(fd, buffer, size); |
351 | } |
352 | |
353 | int setmode(int fd, int mode) { return ::_setmode(fd, mode); } |
354 | |
355 | int write(int fd, const void* buffer, size_t size) { |
356 | return ::_write(fd, buffer, size); |
357 | } |
358 | |
359 | wstring testonly_utf8_to_winpath(const char* path) { |
360 | wstring wpath; |
361 | return as_windows_path(path, &wpath) ? wpath : wstring(); |
362 | } |
363 | |
364 | ExpandWildcardsResult ExpandWildcards( |
365 | const string& path, std::function<void(const string&)> consume) { |
366 | if (path.find_first_of("*?" ) == string::npos) { |
367 | // There are no wildcards in the path, we don't need to expand it. |
368 | consume(path); |
369 | return ExpandWildcardsResult::kSuccess; |
370 | } |
371 | |
372 | wstring wpath; |
373 | if (!as_windows_path(path.c_str(), &wpath)) { |
374 | return ExpandWildcardsResult::kErrorInputPathConversion; |
375 | } |
376 | |
377 | static const wstring kDot = L"." ; |
378 | static const wstring kDotDot = L".." ; |
379 | WIN32_FIND_DATAW metadata; |
380 | HANDLE handle = ::FindFirstFileW(wpath.c_str(), &metadata); |
381 | if (handle == INVALID_HANDLE_VALUE) { |
382 | // The pattern does not match any files (or directories). |
383 | return ExpandWildcardsResult::kErrorNoMatchingFile; |
384 | } |
385 | |
386 | string::size_type pos = path.find_last_of("\\/" ); |
387 | string dirname; |
388 | if (pos != string::npos) { |
389 | dirname = path.substr(0, pos + 1); |
390 | } |
391 | |
392 | ExpandWildcardsResult matched = ExpandWildcardsResult::kErrorNoMatchingFile; |
393 | do { |
394 | // Ignore ".", "..", and directories. |
395 | if ((metadata.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) == 0 && |
396 | kDot != metadata.cFileName && kDotDot != metadata.cFileName) { |
397 | matched = ExpandWildcardsResult::kSuccess; |
398 | string filename; |
399 | if (!strings::wcs_to_utf8(metadata.cFileName, &filename)) { |
400 | matched = ExpandWildcardsResult::kErrorOutputPathConversion; |
401 | break; |
402 | } |
403 | |
404 | if (dirname.empty()) { |
405 | consume(filename); |
406 | } else { |
407 | consume(dirname + filename); |
408 | } |
409 | } |
410 | } while (::FindNextFileW(handle, &metadata)); |
411 | FindClose(handle); |
412 | return matched; |
413 | } |
414 | |
415 | namespace strings { |
416 | |
417 | bool wcs_to_mbs(const WCHAR* s, string* out, bool outUtf8) { |
418 | if (null_or_empty(s)) { |
419 | out->clear(); |
420 | return true; |
421 | } |
422 | BOOL usedDefaultChar = FALSE; |
423 | SetLastError(0); |
424 | int size = WideCharToMultiByte( |
425 | outUtf8 ? CP_UTF8 : CP_ACP, 0, s, -1, nullptr, 0, nullptr, |
426 | outUtf8 ? nullptr : &usedDefaultChar); |
427 | if ((size == 0 && GetLastError() != ERROR_INSUFFICIENT_BUFFER) |
428 | || usedDefaultChar) { |
429 | return false; |
430 | } |
431 | std::unique_ptr<CHAR[]> astr(new CHAR[size]); |
432 | WideCharToMultiByte( |
433 | outUtf8 ? CP_UTF8 : CP_ACP, 0, s, -1, astr.get(), size, nullptr, nullptr); |
434 | out->assign(astr.get()); |
435 | return true; |
436 | } |
437 | |
438 | bool mbs_to_wcs(const char* s, wstring* out, bool inUtf8) { |
439 | if (null_or_empty(s)) { |
440 | out->clear(); |
441 | return true; |
442 | } |
443 | |
444 | SetLastError(0); |
445 | int size = |
446 | MultiByteToWideChar(inUtf8 ? CP_UTF8 : CP_ACP, 0, s, -1, nullptr, 0); |
447 | if (size == 0 && GetLastError() != ERROR_INSUFFICIENT_BUFFER) { |
448 | return false; |
449 | } |
450 | std::unique_ptr<WCHAR[]> wstr(new WCHAR[size]); |
451 | MultiByteToWideChar( |
452 | inUtf8 ? CP_UTF8 : CP_ACP, 0, s, -1, wstr.get(), size + 1); |
453 | out->assign(wstr.get()); |
454 | return true; |
455 | } |
456 | |
457 | bool utf8_to_wcs(const char* input, wstring* out) { |
458 | return mbs_to_wcs(input, out, true); |
459 | } |
460 | |
461 | bool wcs_to_utf8(const wchar_t* input, string* out) { |
462 | return wcs_to_mbs(input, out, true); |
463 | } |
464 | |
465 | } // namespace strings |
466 | } // namespace win32 |
467 | } // namespace io |
468 | } // namespace protobuf |
469 | } // namespace google |
470 | |
471 | #endif // defined(_WIN32) |
472 | |