1 | // Licensed to the .NET Foundation under one or more agreements. |
2 | // The .NET Foundation licenses this file to you under the MIT license. |
3 | // See the LICENSE file in the project root for more information. |
4 | // |
5 | |
6 | // |
7 | // =========================================================================== |
8 | // File: path.cpp |
9 | // |
10 | // Path APIs ported from shlwapi (especially for Fusion) |
11 | // =========================================================================== |
12 | |
13 | #include "common.h" |
14 | #include "strsafe.h" |
15 | |
16 | |
17 | #define CH_SLASH W('/') |
18 | #define CH_WHACK W('\\') |
19 | |
20 | // |
21 | // Inline function to check for a double-backslash at the |
22 | // beginning of a string |
23 | // |
24 | |
25 | static __inline BOOL DBL_BSLASH(LPCWSTR psz) |
26 | { |
27 | return (psz[0] == W('\\') && psz[1] == W('\\')); |
28 | } |
29 | |
30 | // |
31 | // Inline function to check for a path separator character. |
32 | // |
33 | |
34 | static __inline BOOL IsPathSeparator(WCHAR ch) |
35 | { |
36 | return (ch == CH_SLASH || ch == CH_WHACK); |
37 | } |
38 | |
39 | __inline BOOL ChrCmpW_inline(WCHAR w1, WCHAR wMatch) |
40 | { |
41 | return(!(w1 == wMatch)); |
42 | } |
43 | |
44 | STDAPI_(LPWSTR) StrRChrW(LPCWSTR lpStart, LPCWSTR lpEnd, WCHAR wMatch) |
45 | { |
46 | LPCWSTR lpFound = NULL; |
47 | |
48 | RIPMSG(lpStart && IS_VALID_STRING_PTRW(lpStart, -1), "StrRChrW: caller passed bad lpStart" ); |
49 | RIPMSG(!lpEnd || lpEnd <= lpStart + wcslen(lpStart), "StrRChrW: caller passed bad lpEnd" ); |
50 | // don't need to check for NULL lpStart |
51 | |
52 | if (!lpEnd) |
53 | lpEnd = lpStart + wcslen(lpStart); |
54 | |
55 | for ( ; lpStart < lpEnd; lpStart++) |
56 | { |
57 | if (!ChrCmpW_inline(*lpStart, wMatch)) |
58 | lpFound = lpStart; |
59 | } |
60 | return ((LPWSTR)lpFound); |
61 | } |
62 | |
63 | |
64 | // check if a path is a root |
65 | // |
66 | // returns: |
67 | // TRUE |
68 | // "\" "X:\" "\\" "\\foo" "\\foo\bar" |
69 | // |
70 | // FALSE for others including "\\foo\bar\" (!) |
71 | // |
72 | STDAPI_(BOOL) PathIsRootW(LPCWSTR pPath) |
73 | { |
74 | RIPMSG(pPath && IS_VALID_STRING_PTR(pPath, -1), "PathIsRoot: caller passed bad pPath" ); |
75 | |
76 | if (!pPath || !*pPath) |
77 | { |
78 | return FALSE; |
79 | } |
80 | |
81 | if (!lstrcmpiW(pPath + 1, W(":\\" ))) |
82 | { |
83 | return TRUE; // "X:\" case |
84 | } |
85 | |
86 | if (IsPathSeparator(*pPath) && (*(pPath + 1) == 0)) |
87 | { |
88 | return TRUE; // "/" or "\" case |
89 | } |
90 | |
91 | if (DBL_BSLASH(pPath)) // smells like UNC name |
92 | { |
93 | LPCWSTR p; |
94 | int cBackslashes = 0; |
95 | |
96 | for (p = pPath + 2; *p; p++) |
97 | { |
98 | if (*p == W('\\')) |
99 | { |
100 | // |
101 | // return FALSE for "\\server\share\dir" |
102 | // so just check if there is more than one slash |
103 | // |
104 | // "\\server\" without a share name causes |
105 | // problems for WNet APIs. we should return |
106 | // FALSE for this as well |
107 | // |
108 | if ((++cBackslashes > 1) || !*(p+1)) |
109 | return FALSE; |
110 | } |
111 | } |
112 | // end of string with only 1 more backslash |
113 | // must be a bare UNC, which looks like a root dir |
114 | return TRUE; |
115 | } |
116 | return FALSE; |
117 | } |
118 | |
119 | /* |
120 | // rips the last part of the path off including the backslash |
121 | // C:\foo -> C:\ |
122 | // C:\foo\bar -> C:\foo |
123 | // C:\foo\ -> C:\foo |
124 | // \\x\y\x -> \\x\y |
125 | // \\x\y -> \\x |
126 | // \\x -> \\ (Just the double slash!) |
127 | // \foo -> \ (Just the slash!) |
128 | // |
129 | // in/out: |
130 | // pFile fully qualified path name |
131 | // returns: |
132 | // TRUE we stripped something |
133 | // FALSE didn't strip anything (root directory case) |
134 | // |
135 | */ |
136 | STDAPI_(BOOL) PathRemoveFileSpecW(LPWSTR pFile) |
137 | { |
138 | RIPMSG(pFile && IS_VALID_STRING_PTR(pFile, -1), "PathRemoveFileSpec: caller passed bad pFile" ); |
139 | |
140 | if (pFile) |
141 | { |
142 | LPWSTR pT; |
143 | LPWSTR pT2 = pFile; |
144 | |
145 | for (pT = pT2; *pT2; pT2++) |
146 | { |
147 | if (IsPathSeparator(*pT2)) |
148 | { |
149 | pT = pT2; // last "\" found, (we will strip here) |
150 | } |
151 | else if (*pT2 == W(':')) // skip ":\" so we don't |
152 | { |
153 | if (IsPathSeparator(pT2[1])) // strip the "\" from "C:\" |
154 | { |
155 | pT2++; |
156 | } |
157 | pT = pT2 + 1; |
158 | } |
159 | } |
160 | |
161 | if (*pT == 0) |
162 | { |
163 | // didn't strip anything |
164 | return FALSE; |
165 | } |
166 | else if (((pT == pFile) && IsPathSeparator(*pT)) || // is it the "\foo" case? |
167 | ((pT == pFile+1) && (*pT == CH_WHACK && *pFile == CH_WHACK))) // or the "\\bar" case? |
168 | { |
169 | // Is it just a '\'? |
170 | if (*(pT+1) != W('\0')) |
171 | { |
172 | // Nope. |
173 | *(pT+1) = W('\0'); |
174 | return TRUE; // stripped something |
175 | } |
176 | else |
177 | { |
178 | // Yep. |
179 | return FALSE; |
180 | } |
181 | } |
182 | else |
183 | { |
184 | *pT = 0; |
185 | return TRUE; // stripped something |
186 | } |
187 | } |
188 | return FALSE; |
189 | } |
190 | |
191 | // |
192 | // Return a pointer to the end of the next path component in the string. |
193 | // ie return a pointer to the next backslash or terminating NULL. |
194 | // |
195 | LPCWSTR GetPCEnd(LPCWSTR lpszStart) |
196 | { |
197 | LPCWSTR lpszEnd; |
198 | LPCWSTR lpszSlash; |
199 | |
200 | lpszEnd = StrChr(lpszStart, CH_WHACK); |
201 | lpszSlash = StrChr(lpszStart, CH_SLASH); |
202 | if ((lpszSlash && lpszSlash < lpszEnd) || |
203 | !lpszEnd) |
204 | { |
205 | lpszEnd = lpszSlash; |
206 | } |
207 | if (!lpszEnd) |
208 | { |
209 | lpszEnd = lpszStart + wcslen(lpszStart); |
210 | } |
211 | |
212 | return lpszEnd; |
213 | } |
214 | |
215 | // |
216 | // Given a pointer to the end of a path component, return a pointer to |
217 | // its begining. |
218 | // ie return a pointer to the previous backslash (or start of the string). |
219 | // |
220 | LPCWSTR PCStart(LPCWSTR lpszStart, LPCWSTR lpszEnd) |
221 | { |
222 | LPCWSTR lpszBegin = StrRChrW(lpszStart, lpszEnd, CH_WHACK); |
223 | LPCWSTR lpszSlash = StrRChrW(lpszStart, lpszEnd, CH_SLASH); |
224 | if (lpszSlash > lpszBegin) |
225 | { |
226 | lpszBegin = lpszSlash; |
227 | } |
228 | if (!lpszBegin) |
229 | { |
230 | lpszBegin = lpszStart; |
231 | } |
232 | return lpszBegin; |
233 | } |
234 | |
235 | // |
236 | // Fix up a few special cases so that things roughly make sense. |
237 | // |
238 | void NearRootFixups(LPWSTR lpszPath, BOOL fUNC) |
239 | { |
240 | // Check for empty path. |
241 | if (lpszPath[0] == W('\0')) |
242 | { |
243 | // Fix up. |
244 | #ifndef PLATFORM_UNIX |
245 | lpszPath[0] = CH_WHACK; |
246 | #else |
247 | lpszPath[0] = CH_SLASH; |
248 | #endif |
249 | lpszPath[1] = W('\0'); |
250 | } |
251 | // Check for missing slash. |
252 | if (lpszPath[1] == W(':') && lpszPath[2] == W('\0')) |
253 | { |
254 | // Fix up. |
255 | lpszPath[2] = W('\\'); |
256 | lpszPath[3] = W('\0'); |
257 | } |
258 | // Check for UNC root. |
259 | if (fUNC && lpszPath[0] == W('\\') && lpszPath[1] == W('\0')) |
260 | { |
261 | // Fix up. |
262 | //lpszPath[0] = W('\\'); // already checked in if guard |
263 | lpszPath[1] = W('\\'); |
264 | lpszPath[2] = W('\0'); |
265 | } |
266 | } |
267 | |
268 | /*---------------------------------------------------------- |
269 | Purpose: Canonicalize a path. |
270 | |
271 | Returns: |
272 | Cond: -- |
273 | */ |
274 | STDAPI_(BOOL) PathCanonicalizeW(LPWSTR lpszDst, LPCWSTR lpszSrc) |
275 | { |
276 | LPCWSTR lpchSrc; |
277 | LPCWSTR lpchPCEnd; // Pointer to end of path component. |
278 | LPWSTR lpchDst; |
279 | BOOL fUNC; |
280 | int cchPC; |
281 | |
282 | RIPMSG(lpszDst && IS_VALID_WRITE_BUFFER(lpszDst, WCHAR, MAX_PATH), "PathCanonicalize: caller passed bad lpszDst" ); |
283 | RIPMSG(lpszSrc && IS_VALID_STRING_PTR(lpszSrc, -1), "PathCanonicalize: caller passed bad lpszSrc" ); |
284 | RIPMSG(lpszDst != lpszSrc, "PathCanonicalize: caller passed the same buffer for lpszDst and lpszSrc" ); |
285 | |
286 | if (!lpszDst || !lpszSrc) |
287 | { |
288 | SetLastError(ERROR_INVALID_PARAMETER); |
289 | return FALSE; |
290 | } |
291 | |
292 | *lpszDst = W('\0'); |
293 | |
294 | fUNC = PathIsUNCW(lpszSrc); // Check for UNCness. |
295 | |
296 | // Init. |
297 | lpchSrc = lpszSrc; |
298 | lpchDst = lpszDst; |
299 | |
300 | while (*lpchSrc) |
301 | { |
302 | lpchPCEnd = GetPCEnd(lpchSrc); |
303 | cchPC = (int) (lpchPCEnd - lpchSrc)+1; |
304 | |
305 | if (cchPC == 1 && IsPathSeparator(*lpchSrc)) // Check for slashes. |
306 | { |
307 | // Just copy them. |
308 | #ifndef PLATFORM_UNIX |
309 | *lpchDst = CH_WHACK; |
310 | #else |
311 | *lpchDst = CH_SLASH; |
312 | #endif |
313 | lpchDst++; |
314 | lpchSrc++; |
315 | } |
316 | else if (cchPC == 2 && *lpchSrc == W('.')) // Check for dots. |
317 | { |
318 | // Skip it... |
319 | // Are we at the end? |
320 | if (*(lpchSrc+1) == W('\0')) |
321 | { |
322 | lpchSrc++; |
323 | |
324 | // remove the last slash we copied (if we've copied one), but don't make a mal-formed root |
325 | if ((lpchDst > lpszDst) && !PathIsRootW(lpszDst)) |
326 | lpchDst--; |
327 | } |
328 | else |
329 | { |
330 | lpchSrc += 2; |
331 | } |
332 | } |
333 | else if (cchPC == 3 && *lpchSrc == W('.') && *(lpchSrc + 1) == W('.')) // Check for dot dot. |
334 | { |
335 | // make sure we aren't already at the root |
336 | if (!PathIsRootW(lpszDst)) |
337 | { |
338 | // Go up... Remove the previous path component. |
339 | lpchDst = (LPWSTR)PCStart(lpszDst, lpchDst - 1); |
340 | } |
341 | else |
342 | { |
343 | // When we can't back up, skip the trailing backslash |
344 | // so we don't copy one again. (C:\..\FOO would otherwise |
345 | // turn into C:\\FOO). |
346 | if (IsPathSeparator(*(lpchSrc + 2))) |
347 | { |
348 | lpchSrc++; |
349 | } |
350 | } |
351 | |
352 | // skip ".." |
353 | lpchSrc += 2; |
354 | } |
355 | else // Everything else |
356 | { |
357 | // Just copy it. |
358 | int cchRemainingBuffer = MAX_PATH - (lpszDst - lpchDst); |
359 | StringCchCopyNW(lpchDst, cchRemainingBuffer, lpchSrc, cchPC); |
360 | lpchDst += cchPC - 1; |
361 | lpchSrc += cchPC - 1; |
362 | } |
363 | |
364 | // Keep everything nice and tidy. |
365 | *lpchDst = W('\0'); |
366 | } |
367 | |
368 | // Check for weirdo root directory stuff. |
369 | NearRootFixups(lpszDst, fUNC); |
370 | |
371 | return TRUE; |
372 | } |
373 | |
374 | // Modifies: |
375 | // pszRoot |
376 | // |
377 | // Returns: |
378 | // TRUE if a drive root was found |
379 | // FALSE otherwise |
380 | // |
381 | STDAPI_(BOOL) PathStripToRootW(LPWSTR pszRoot) |
382 | { |
383 | RIPMSG(pszRoot && IS_VALID_STRING_PTR(pszRoot, -1), "PathStripToRoot: caller passed bad pszRoot" ); |
384 | |
385 | if (pszRoot) |
386 | { |
387 | while (!PathIsRootW(pszRoot)) |
388 | { |
389 | if (!PathRemoveFileSpecW(pszRoot)) |
390 | { |
391 | // If we didn't strip anything off, |
392 | // must be current drive |
393 | return FALSE; |
394 | } |
395 | } |
396 | return TRUE; |
397 | } |
398 | return FALSE; |
399 | } |
400 | |
401 | |
402 | |
403 | /*---------------------------------------------------------- |
404 | Purpose: Concatenate lpszDir and lpszFile into a properly formed |
405 | path and canonicalize any relative path pieces. |
406 | |
407 | lpszDest and lpszFile can be the same buffer |
408 | lpszDest and lpszDir can be the same buffer |
409 | |
410 | Returns: pointer to lpszDest |
411 | */ |
412 | STDAPI_(LPWSTR) PathCombineW(LPWSTR lpszDest, LPCWSTR lpszDir, LPCWSTR lpszFile) |
413 | { |
414 | #ifdef DEBUG |
415 | RIPMSG(lpszDest && IS_VALID_WRITE_BUFFER(lpszDest, TCHAR, MAX_LONGPATH), "PathCombine: caller passed bad lpszDest" ); |
416 | RIPMSG(!lpszDir || IS_VALID_STRING_PTR(lpszDir, -1), "PathCombine: caller passed bad lpszDir" ); |
417 | RIPMSG(!lpszFile || IS_VALID_STRING_PTR(lpszFile, -1), "PathCombine: caller passed bad lpszFile" ); |
418 | RIPMSG(lpszDir || lpszFile, "PathCombine: caller neglected to pass lpszDir or lpszFile" ); |
419 | #endif // DEBUG |
420 | |
421 | |
422 | if (lpszDest) |
423 | { |
424 | TCHAR szTemp[MAX_LONGPATH]; |
425 | LPWSTR pszT; |
426 | |
427 | *szTemp = W('\0'); |
428 | |
429 | if (lpszDir && *lpszDir) |
430 | { |
431 | if (!lpszFile || *lpszFile==W('\0')) |
432 | { |
433 | // lpszFile is empty |
434 | StringCchCopyNW(szTemp, ARRAYSIZE(szTemp), lpszDir, ARRAYSIZE(szTemp)); |
435 | } |
436 | else if (PathIsRelativeW(lpszFile)) |
437 | { |
438 | StringCchCopyNW(szTemp, ARRAYSIZE(szTemp), lpszDir, ARRAYSIZE(szTemp)); |
439 | pszT = PathAddBackslashW(szTemp); |
440 | if (pszT) |
441 | { |
442 | size_t iRemaining = ARRAYSIZE(szTemp) - (pszT - szTemp); |
443 | |
444 | if (wcslen(lpszFile) < iRemaining) |
445 | { |
446 | StringCchCopyNW(pszT, iRemaining, lpszFile, iRemaining); |
447 | } |
448 | else |
449 | { |
450 | *szTemp = W('\0'); |
451 | } |
452 | } |
453 | else |
454 | { |
455 | *szTemp = W('\0'); |
456 | } |
457 | } |
458 | else if (IsPathSeparator(*lpszFile) && !PathIsUNCW(lpszFile)) |
459 | { |
460 | StringCchCopyNW(szTemp, ARRAYSIZE(szTemp), lpszDir, ARRAYSIZE(szTemp)); |
461 | // FEATURE: Note that we do not check that an actual root is returned; |
462 | // it is assumed that we are given valid parameters |
463 | PathStripToRootW(szTemp); |
464 | |
465 | pszT = PathAddBackslashW(szTemp); |
466 | if (pszT) |
467 | { |
468 | // Skip the backslash when copying |
469 | // Note: We don't support strings longer than 4GB, but that's |
470 | // okay because we already barf at MAX_PATH |
471 | int iRemaining = (int)(ARRAYSIZE(szTemp) - (pszT - szTemp)); |
472 | StringCchCopyNW(pszT, iRemaining, lpszFile+1, iRemaining); |
473 | } |
474 | else |
475 | { |
476 | *szTemp = W('\0'); |
477 | } |
478 | } |
479 | else |
480 | { |
481 | // already fully qualified file part |
482 | StringCchCopyNW(szTemp, ARRAYSIZE(szTemp), lpszFile, ARRAYSIZE(szTemp)); |
483 | } |
484 | } |
485 | else if (lpszFile && *lpszFile) |
486 | { |
487 | // no dir just use file. |
488 | StringCchCopyNW(szTemp, ARRAYSIZE(szTemp), lpszFile, ARRAYSIZE(szTemp)); |
489 | } |
490 | |
491 | // |
492 | // if szTemp has something in it we succeeded. Also if szTemp is empty and |
493 | // the input strings are empty we succeed and PathCanonicalize() will |
494 | // return "\" |
495 | // |
496 | if (*szTemp || ((lpszDir || lpszFile) && !((lpszDir && *lpszDir) || (lpszFile && *lpszFile)))) |
497 | { |
498 | PathCanonicalizeW(lpszDest, szTemp); // this deals with .. and . stuff |
499 | // returns "\" on empty szTemp |
500 | } |
501 | else |
502 | { |
503 | *lpszDest = W('\0'); // set output buffer to empty string. |
504 | lpszDest = NULL; // return failure. |
505 | } |
506 | } |
507 | |
508 | return lpszDest; |
509 | } |
510 | |
511 | // add a backslash to a qualified path |
512 | // |
513 | // in: |
514 | // lpszPath path (A:, C:\foo, etc) |
515 | // |
516 | // out: |
517 | // lpszPath A:\, C:\foo\ ; |
518 | // |
519 | // returns: |
520 | // pointer to the NULL that terminates the path |
521 | // |
522 | STDAPI_(LPWSTR) PathAddBackslashW(LPWSTR lpszPath) |
523 | { |
524 | LPWSTR lpszRet = NULL; |
525 | |
526 | RIPMSG(lpszPath && IS_VALID_STRING_PTR(lpszPath, -1), "PathAddBackslash: caller passed bad lpszPath" ); |
527 | |
528 | if (lpszPath) |
529 | { |
530 | size_t ichPath = wcslen(lpszPath); |
531 | LPWSTR lpszEnd = lpszPath + ichPath; |
532 | |
533 | if (ichPath) |
534 | { |
535 | |
536 | // Get the end of the source directory |
537 | switch(*(lpszEnd-1)) |
538 | { |
539 | case CH_SLASH: |
540 | case CH_WHACK: |
541 | break; |
542 | |
543 | default: |
544 | // try to keep us from tromping over MAX_PATH in size. |
545 | // if we find these cases, return NULL. Note: We need to |
546 | // check those places that call us to handle their GP fault |
547 | // if they try to use the NULL! |
548 | if (ichPath >= (MAX_PATH - 2)) // -2 because ichPath doesn't include NULL, and we're adding a CH_WHACK. |
549 | { |
550 | return(NULL); |
551 | } |
552 | |
553 | *lpszEnd++ = CH_WHACK; |
554 | *lpszEnd = W('\0'); |
555 | } |
556 | } |
557 | |
558 | lpszRet = lpszEnd; |
559 | } |
560 | |
561 | return lpszRet; |
562 | } |
563 | |
564 | |
565 | |
566 | |
567 | //--------------------------------------------------------------------------- |
568 | // Returns TRUE if the given string is a UNC path. |
569 | // |
570 | // TRUE |
571 | // "\\foo\bar" |
572 | // "\\foo" <- careful |
573 | // "\\" |
574 | // FALSE |
575 | // "\foo" |
576 | // "foo" |
577 | // "c:\foo" |
578 | // |
579 | // |
580 | STDAPI_(BOOL) PathIsUNCW(LPCWSTR pszPath) |
581 | { |
582 | RIPMSG(pszPath && IS_VALID_STRING_PTR(pszPath, -1), "PathIsUNC: caller passed bad pszPath" ); |
583 | |
584 | if (pszPath) |
585 | { |
586 | return DBL_BSLASH(pszPath); |
587 | } |
588 | return FALSE; |
589 | } |
590 | |
591 | |
592 | |
593 | |
594 | |
595 | //--------------------------------------------------------------------------- |
596 | // Return TRUE if the path isn't absoulte. |
597 | // |
598 | // TRUE |
599 | // "foo.exe" |
600 | // ".\foo.exe" |
601 | // "..\boo\foo.exe" |
602 | // |
603 | // FALSE |
604 | // "\foo" |
605 | // "c:bar" <- be careful |
606 | // "c:\bar" |
607 | // "\\foo\bar" |
608 | // |
609 | STDAPI_(BOOL) PathIsRelativeW(LPCWSTR lpszPath) |
610 | { |
611 | RIPMSG(lpszPath && IS_VALID_STRING_PTR(lpszPath, -1), "PathIsRelative: caller passed bad lpszPath" ); |
612 | |
613 | if (!lpszPath || *lpszPath == 0) |
614 | { |
615 | // The NULL path is assumed relative |
616 | return TRUE; |
617 | } |
618 | |
619 | if (IsPathSeparator(lpszPath[0])) |
620 | { |
621 | // Does it begin with a slash ? |
622 | return FALSE; |
623 | } |
624 | else if (lpszPath[1] == W(':')) |
625 | { |
626 | // Does it begin with a drive and a colon ? |
627 | return FALSE; |
628 | } |
629 | else |
630 | { |
631 | // Probably relative. |
632 | return TRUE; |
633 | } |
634 | } |
635 | |
636 | // find the next slash or null terminator |
637 | LPWSTR StrSlash(LPCWSTR psz) |
638 | { |
639 | for (; *psz && !IsPathSeparator(*psz); psz++); |
640 | |
641 | // Cast to a non-const string to mimic the behavior |
642 | // of wcschr/StrChr and strchr. |
643 | return (LPWSTR) psz; |
644 | } |
645 | |
646 | |
647 | |
648 | |