1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ****************************************************************************** |
5 | * |
6 | * Copyright (C) 1999-2016, International Business Machines |
7 | * Corporation and others. All Rights Reserved. |
8 | * |
9 | ****************************************************************************** |
10 | * file name: udata.cpp |
11 | * encoding: UTF-8 |
12 | * tab size: 8 (not used) |
13 | * indentation:4 |
14 | * |
15 | * created on: 1999oct25 |
16 | * created by: Markus W. Scherer |
17 | */ |
18 | |
19 | #include "unicode/utypes.h" /* U_PLATFORM etc. */ |
20 | |
21 | #ifdef __GNUC__ |
22 | /* if gcc |
23 | #define ATTRIBUTE_WEAK __attribute__ ((weak)) |
24 | might have to #include some other header |
25 | */ |
26 | #endif |
27 | |
28 | #include "unicode/putil.h" |
29 | #include "unicode/udata.h" |
30 | #include "unicode/uversion.h" |
31 | #include "charstr.h" |
32 | #include "cmemory.h" |
33 | #include "cstring.h" |
34 | #include "mutex.h" |
35 | #include "putilimp.h" |
36 | #include "restrace.h" |
37 | #include "uassert.h" |
38 | #include "ucln_cmn.h" |
39 | #include "ucmndata.h" |
40 | #include "udatamem.h" |
41 | #include "uhash.h" |
42 | #include "umapfile.h" |
43 | #include "umutex.h" |
44 | |
45 | /*********************************************************************** |
46 | * |
47 | * Notes on the organization of the ICU data implementation |
48 | * |
49 | * All of the public API is defined in udata.h |
50 | * |
51 | * The implementation is split into several files... |
52 | * |
53 | * - udata.c (this file) contains higher level code that knows about |
54 | * the search paths for locating data, caching opened data, etc. |
55 | * |
56 | * - umapfile.c contains the low level platform-specific code for actually loading |
57 | * (memory mapping, file reading, whatever) data into memory. |
58 | * |
59 | * - ucmndata.c deals with the tables of contents of ICU data items within |
60 | * an ICU common format data file. The implementation includes |
61 | * an abstract interface and support for multiple TOC formats. |
62 | * All knowledge of any specific TOC format is encapsulated here. |
63 | * |
64 | * - udatamem.c has code for managing UDataMemory structs. These are little |
65 | * descriptor objects for blocks of memory holding ICU data of |
66 | * various types. |
67 | */ |
68 | |
69 | /* configuration ---------------------------------------------------------- */ |
70 | |
71 | /* If you are excruciatingly bored turn this on .. */ |
72 | /* #define UDATA_DEBUG 1 */ |
73 | |
74 | #if defined(UDATA_DEBUG) |
75 | # include <stdio.h> |
76 | #endif |
77 | |
78 | U_NAMESPACE_USE |
79 | |
80 | /* |
81 | * Forward declarations |
82 | */ |
83 | static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err); |
84 | |
85 | /*********************************************************************** |
86 | * |
87 | * static (Global) data |
88 | * |
89 | ************************************************************************/ |
90 | |
91 | /* |
92 | * Pointers to the common ICU data. |
93 | * |
94 | * We store multiple pointers to ICU data packages and iterate through them |
95 | * when looking for a data item. |
96 | * |
97 | * It is possible to combine this with dependency inversion: |
98 | * One or more data package libraries may export |
99 | * functions that each return a pointer to their piece of the ICU data, |
100 | * and this file would import them as weak functions, without a |
101 | * strong linker dependency from the common library on the data library. |
102 | * |
103 | * Then we can have applications depend on only that part of ICU's data |
104 | * that they really need, reducing the size of binaries that take advantage |
105 | * of this. |
106 | */ |
107 | static UDataMemory *gCommonICUDataArray[10] = { nullptr }; // Access protected by icu global mutex. |
108 | |
109 | static u_atomic_int32_t gHaveTriedToLoadCommonData {0}; // See extendICUData(). |
110 | |
111 | static UHashtable *gCommonDataCache = nullptr; /* Global hash table of opened ICU data files. */ |
112 | static icu::UInitOnce gCommonDataCacheInitOnce {}; |
113 | |
114 | #if !defined(ICU_DATA_DIR_WINDOWS) |
115 | static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; // Access not synchronized. |
116 | // Modifying is documented as thread-unsafe. |
117 | #else |
118 | // If we are using the Windows data directory, then look in one spot only. |
119 | static UDataFileAccess gDataFileAccess = UDATA_NO_FILES; |
120 | #endif |
121 | |
122 | static UBool U_CALLCONV |
123 | udata_cleanup() |
124 | { |
125 | int32_t i; |
126 | |
127 | if (gCommonDataCache) { /* Delete the cache of user data mappings. */ |
128 | uhash_close(gCommonDataCache); /* Table owns the contents, and will delete them. */ |
129 | gCommonDataCache = nullptr; /* Cleanup is not thread safe. */ |
130 | } |
131 | gCommonDataCacheInitOnce.reset(); |
132 | |
133 | for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != nullptr; ++i) { |
134 | udata_close(gCommonICUDataArray[i]); |
135 | gCommonICUDataArray[i] = nullptr; |
136 | } |
137 | gHaveTriedToLoadCommonData = 0; |
138 | |
139 | return true; /* Everything was cleaned up */ |
140 | } |
141 | |
142 | static UBool U_CALLCONV |
143 | findCommonICUDataByName(const char *inBasename, UErrorCode &err) |
144 | { |
145 | UBool found = false; |
146 | int32_t i; |
147 | |
148 | UDataMemory *pData = udata_findCachedData(inBasename, err); |
149 | if (U_FAILURE(err) || pData == nullptr) |
150 | return false; |
151 | |
152 | { |
153 | Mutex lock; |
154 | for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) { |
155 | if ((gCommonICUDataArray[i] != nullptr) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) { |
156 | /* The data pointer is already in the array. */ |
157 | found = true; |
158 | break; |
159 | } |
160 | } |
161 | } |
162 | return found; |
163 | } |
164 | |
165 | |
166 | /* |
167 | * setCommonICUData. Set a UDataMemory to be the global ICU Data |
168 | */ |
169 | static UBool |
170 | setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to caller, we copy it. */ |
171 | UBool warn, /* If true, set USING_DEFAULT warning if ICUData was */ |
172 | /* changed by another thread before we got to it. */ |
173 | UErrorCode *pErr) |
174 | { |
175 | UDataMemory *newCommonData = UDataMemory_createNewInstance(pErr); |
176 | int32_t i; |
177 | UBool didUpdate = false; |
178 | if (U_FAILURE(*pErr)) { |
179 | return false; |
180 | } |
181 | |
182 | /* For the assignment, other threads must cleanly see either the old */ |
183 | /* or the new, not some partially initialized new. The old can not be */ |
184 | /* deleted - someone may still have a pointer to it lying around in */ |
185 | /* their locals. */ |
186 | UDatamemory_assign(newCommonData, pData); |
187 | umtx_lock(nullptr); |
188 | for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) { |
189 | if (gCommonICUDataArray[i] == nullptr) { |
190 | gCommonICUDataArray[i] = newCommonData; |
191 | didUpdate = true; |
192 | break; |
193 | } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) { |
194 | /* The same data pointer is already in the array. */ |
195 | break; |
196 | } |
197 | } |
198 | umtx_unlock(nullptr); |
199 | |
200 | if (i == UPRV_LENGTHOF(gCommonICUDataArray) && warn) { |
201 | *pErr = U_USING_DEFAULT_WARNING; |
202 | } |
203 | if (didUpdate) { |
204 | ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
205 | } else { |
206 | uprv_free(newCommonData); |
207 | } |
208 | return didUpdate; |
209 | } |
210 | |
211 | #if !defined(ICU_DATA_DIR_WINDOWS) |
212 | |
213 | static UBool |
214 | setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) { |
215 | UDataMemory tData; |
216 | UDataMemory_init(&tData); |
217 | UDataMemory_setData(&tData, pData); |
218 | udata_checkCommonData(&tData, pErrorCode); |
219 | return setCommonICUData(&tData, false, pErrorCode); |
220 | } |
221 | |
222 | #endif |
223 | |
224 | static const char * |
225 | findBasename(const char *path) { |
226 | const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR); |
227 | if(basename==nullptr) { |
228 | return path; |
229 | } else { |
230 | return basename+1; |
231 | } |
232 | } |
233 | |
234 | #ifdef UDATA_DEBUG |
235 | static const char * |
236 | packageNameFromPath(const char *path) |
237 | { |
238 | if((path == nullptr) || (*path == 0)) { |
239 | return U_ICUDATA_NAME; |
240 | } |
241 | |
242 | path = findBasename(path); |
243 | |
244 | if((path == nullptr) || (*path == 0)) { |
245 | return U_ICUDATA_NAME; |
246 | } |
247 | |
248 | return path; |
249 | } |
250 | #endif |
251 | |
252 | /*----------------------------------------------------------------------* |
253 | * * |
254 | * Cache for common data * |
255 | * Functions for looking up or adding entries to a cache of * |
256 | * data that has been previously opened. Avoids a potentially * |
257 | * expensive operation of re-opening the data for subsequent * |
258 | * uses. * |
259 | * * |
260 | * Data remains cached for the duration of the process. * |
261 | * * |
262 | *----------------------------------------------------------------------*/ |
263 | |
264 | typedef struct DataCacheElement { |
265 | char *name; |
266 | UDataMemory *item; |
267 | } DataCacheElement; |
268 | |
269 | |
270 | |
271 | /* |
272 | * Deleter function for DataCacheElements. |
273 | * udata cleanup function closes the hash table; hash table in turn calls back to |
274 | * here for each entry. |
275 | */ |
276 | static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) { |
277 | DataCacheElement *p = (DataCacheElement *)pDCEl; |
278 | udata_close(p->item); /* unmaps storage */ |
279 | uprv_free(p->name); /* delete the hash key string. */ |
280 | uprv_free(pDCEl); /* delete 'this' */ |
281 | } |
282 | |
283 | static void U_CALLCONV udata_initHashTable(UErrorCode &err) { |
284 | U_ASSERT(gCommonDataCache == nullptr); |
285 | gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, nullptr, &err); |
286 | if (U_FAILURE(err)) { |
287 | return; |
288 | } |
289 | U_ASSERT(gCommonDataCache != nullptr); |
290 | uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter); |
291 | ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
292 | } |
293 | |
294 | /* udata_getCacheHashTable() |
295 | * Get the hash table used to store the data cache entries. |
296 | * Lazy create it if it doesn't yet exist. |
297 | */ |
298 | static UHashtable *udata_getHashTable(UErrorCode &err) { |
299 | umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable, err); |
300 | return gCommonDataCache; |
301 | } |
302 | |
303 | |
304 | |
305 | static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err) |
306 | { |
307 | UHashtable *htable; |
308 | UDataMemory *retVal = nullptr; |
309 | DataCacheElement *el; |
310 | const char *baseName; |
311 | |
312 | htable = udata_getHashTable(err); |
313 | if (U_FAILURE(err)) { |
314 | return nullptr; |
315 | } |
316 | |
317 | baseName = findBasename(path); /* Cache remembers only the base name, not the full path. */ |
318 | umtx_lock(nullptr); |
319 | el = (DataCacheElement *)uhash_get(htable, baseName); |
320 | umtx_unlock(nullptr); |
321 | if (el != nullptr) { |
322 | retVal = el->item; |
323 | } |
324 | #ifdef UDATA_DEBUG |
325 | fprintf(stderr, "Cache: [%s] -> %p\n" , baseName, (void*) retVal); |
326 | #endif |
327 | return retVal; |
328 | } |
329 | |
330 | |
331 | static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UErrorCode *pErr) { |
332 | DataCacheElement *newElement; |
333 | const char *baseName; |
334 | int32_t nameLen; |
335 | UHashtable *htable; |
336 | DataCacheElement *oldValue = nullptr; |
337 | UErrorCode subErr = U_ZERO_ERROR; |
338 | |
339 | htable = udata_getHashTable(*pErr); |
340 | if (U_FAILURE(*pErr)) { |
341 | return nullptr; |
342 | } |
343 | |
344 | /* Create a new DataCacheElement - the thingy we store in the hash table - |
345 | * and copy the supplied path and UDataMemoryItems into it. |
346 | */ |
347 | newElement = (DataCacheElement *)uprv_malloc(sizeof(DataCacheElement)); |
348 | if (newElement == nullptr) { |
349 | *pErr = U_MEMORY_ALLOCATION_ERROR; |
350 | return nullptr; |
351 | } |
352 | newElement->item = UDataMemory_createNewInstance(pErr); |
353 | if (U_FAILURE(*pErr)) { |
354 | uprv_free(newElement); |
355 | return nullptr; |
356 | } |
357 | UDatamemory_assign(newElement->item, item); |
358 | |
359 | baseName = findBasename(path); |
360 | nameLen = (int32_t)uprv_strlen(baseName); |
361 | newElement->name = (char *)uprv_malloc(nameLen+1); |
362 | if (newElement->name == nullptr) { |
363 | *pErr = U_MEMORY_ALLOCATION_ERROR; |
364 | uprv_free(newElement->item); |
365 | uprv_free(newElement); |
366 | return nullptr; |
367 | } |
368 | uprv_strcpy(newElement->name, baseName); |
369 | |
370 | /* Stick the new DataCacheElement into the hash table. |
371 | */ |
372 | umtx_lock(nullptr); |
373 | oldValue = (DataCacheElement *)uhash_get(htable, path); |
374 | if (oldValue != nullptr) { |
375 | subErr = U_USING_DEFAULT_WARNING; |
376 | } |
377 | else { |
378 | uhash_put( |
379 | htable, |
380 | newElement->name, /* Key */ |
381 | newElement, /* Value */ |
382 | &subErr); |
383 | } |
384 | umtx_unlock(nullptr); |
385 | |
386 | #ifdef UDATA_DEBUG |
387 | fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n" , newElement->name, |
388 | (void*) newElement->item, u_errorName(subErr), (void*) newElement->item->vFuncs); |
389 | #endif |
390 | |
391 | if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) { |
392 | *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. */ |
393 | uprv_free(newElement->name); |
394 | uprv_free(newElement->item); |
395 | uprv_free(newElement); |
396 | return oldValue ? oldValue->item : nullptr; |
397 | } |
398 | |
399 | return newElement->item; |
400 | } |
401 | |
402 | /*----------------------------------------------------------------------*============== |
403 | * * |
404 | * Path management. Could be shared with other tools/etc if need be * |
405 | * later on. * |
406 | * * |
407 | *----------------------------------------------------------------------*/ |
408 | |
409 | U_NAMESPACE_BEGIN |
410 | |
411 | class UDataPathIterator |
412 | { |
413 | public: |
414 | UDataPathIterator(const char *path, const char *pkg, |
415 | const char *item, const char *suffix, UBool doCheckLastFour, |
416 | UErrorCode *pErrorCode); |
417 | const char *next(UErrorCode *pErrorCode); |
418 | |
419 | private: |
420 | const char *path; /* working path (u_icudata_Dir) */ |
421 | const char *nextPath; /* path following this one */ |
422 | const char *basename; /* item's basename (icudt22e_mt.res)*/ |
423 | |
424 | StringPiece suffix; /* item suffix (can be null) */ |
425 | |
426 | uint32_t basenameLen; /* length of basename */ |
427 | |
428 | CharString itemPath; /* path passed in with item name */ |
429 | CharString pathBuffer; /* output path for this it'ion */ |
430 | CharString packageStub; /* example: "/icudt28b". Will ignore that leaf in set paths. */ |
431 | |
432 | UBool checkLastFour; /* if true then allow paths such as '/foo/myapp.dat' |
433 | * to match, checks last 4 chars of suffix with |
434 | * last 4 of path, then previous chars. */ |
435 | }; |
436 | |
437 | /** |
438 | * @param iter The iterator to be initialized. Its current state does not matter. |
439 | * @param inPath The full pathname to be iterated over. If nullptr, defaults to U_ICUDATA_NAME |
440 | * @param pkg Package which is being searched for, ex "icudt28l". Will ignore leaf directories such as /icudt28l |
441 | * @param item Item to be searched for. Can include full path, such as /a/b/foo.dat |
442 | * @param inSuffix Optional item suffix, if not-null (ex. ".dat") then 'path' can contain 'item' explicitly. |
443 | * Ex: 'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar/baz' as item #2. |
444 | * '/blarg/stuff.dat' would also be found. |
445 | * Note: inSuffix may also be the 'item' being searched for as well, (ex: "ibm-5348_P100-1997.cnv"), in which case |
446 | * the 'item' parameter is often the same as pkg. (Though sometimes might have a tree part as well, ex: "icudt62l-curr"). |
447 | */ |
448 | UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg, |
449 | const char *item, const char *inSuffix, UBool doCheckLastFour, |
450 | UErrorCode *pErrorCode) |
451 | { |
452 | #ifdef UDATA_DEBUG |
453 | fprintf(stderr, "SUFFIX1=%s PATH=%s\n" , inSuffix, inPath); |
454 | #endif |
455 | /** Path **/ |
456 | if(inPath == nullptr) { |
457 | path = u_getDataDirectory(); |
458 | } else { |
459 | path = inPath; |
460 | } |
461 | |
462 | /** Package **/ |
463 | if(pkg != nullptr) { |
464 | packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode); |
465 | #ifdef UDATA_DEBUG |
466 | fprintf(stderr, "STUB=%s [%d]\n" , packageStub.data(), packageStub.length()); |
467 | #endif |
468 | } |
469 | |
470 | /** Item **/ |
471 | basename = findBasename(item); |
472 | basenameLen = (int32_t)uprv_strlen(basename); |
473 | |
474 | /** Item path **/ |
475 | if(basename == item) { |
476 | nextPath = path; |
477 | } else { |
478 | itemPath.append(item, (int32_t)(basename-item), *pErrorCode); |
479 | nextPath = itemPath.data(); |
480 | } |
481 | #ifdef UDATA_DEBUG |
482 | fprintf(stderr, "SUFFIX=%s [%p]\n" , inSuffix, (void*) inSuffix); |
483 | #endif |
484 | |
485 | /** Suffix **/ |
486 | if(inSuffix != nullptr) { |
487 | suffix = inSuffix; |
488 | } else { |
489 | suffix = "" ; |
490 | } |
491 | |
492 | checkLastFour = doCheckLastFour; |
493 | |
494 | /* pathBuffer will hold the output path strings returned by this iterator */ |
495 | |
496 | #ifdef UDATA_DEBUG |
497 | fprintf(stderr, "0: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%s], [nextpath=%s], [checklast4=%s]\n" , |
498 | item, |
499 | path, |
500 | basename, |
501 | suffix.data(), |
502 | itemPath.data(), |
503 | nextPath, |
504 | checkLastFour?"true" :"false" ); |
505 | #endif |
506 | } |
507 | |
508 | /** |
509 | * Get the next path on the list. |
510 | * |
511 | * @param iter The Iter to be used |
512 | * @param len If set, pointer to the length of the returned path, for convenience. |
513 | * @return Pointer to the next path segment, or nullptr if there are no more. |
514 | */ |
515 | const char *UDataPathIterator::next(UErrorCode *pErrorCode) |
516 | { |
517 | if(U_FAILURE(*pErrorCode)) { |
518 | return nullptr; |
519 | } |
520 | |
521 | const char *currentPath = nullptr; |
522 | int32_t pathLen = 0; |
523 | const char *pathBasename; |
524 | |
525 | do |
526 | { |
527 | if( nextPath == nullptr ) { |
528 | break; |
529 | } |
530 | currentPath = nextPath; |
531 | |
532 | if(nextPath == itemPath.data()) { /* we were processing item's path. */ |
533 | nextPath = path; /* start with regular path next tm. */ |
534 | pathLen = (int32_t)uprv_strlen(currentPath); |
535 | } else { |
536 | /* fix up next for next time */ |
537 | nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR); |
538 | if(nextPath == nullptr) { |
539 | /* segment: entire path */ |
540 | pathLen = (int32_t)uprv_strlen(currentPath); |
541 | } else { |
542 | /* segment: until next segment */ |
543 | pathLen = (int32_t)(nextPath - currentPath); |
544 | /* skip divider */ |
545 | nextPath ++; |
546 | } |
547 | } |
548 | |
549 | if(pathLen == 0) { |
550 | continue; |
551 | } |
552 | |
553 | #ifdef UDATA_DEBUG |
554 | fprintf(stderr, "rest of path (IDD) = %s\n" , currentPath); |
555 | fprintf(stderr, " " ); |
556 | { |
557 | int32_t qqq; |
558 | for(qqq=0;qqq<pathLen;qqq++) |
559 | { |
560 | fprintf(stderr, " " ); |
561 | } |
562 | |
563 | fprintf(stderr, "^\n" ); |
564 | } |
565 | #endif |
566 | pathBuffer.clear().append(currentPath, pathLen, *pErrorCode); |
567 | |
568 | /* check for .dat files */ |
569 | pathBasename = findBasename(pathBuffer.data()); |
570 | |
571 | if(checkLastFour && |
572 | (pathLen>=4) && |
573 | uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix.data(), 4)==0 && /* suffix matches */ |
574 | uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)==0 && /* base matches */ |
575 | uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full len */ |
576 | |
577 | #ifdef UDATA_DEBUG |
578 | fprintf(stderr, "Have %s file on the path: %s\n" , suffix.data(), pathBuffer.data()); |
579 | #endif |
580 | /* do nothing */ |
581 | } |
582 | else |
583 | { /* regular dir path */ |
584 | if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) { |
585 | if((pathLen>=4) && |
586 | uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat" , 4) == 0) |
587 | { |
588 | #ifdef UDATA_DEBUG |
589 | fprintf(stderr, "skipping non-directory .dat file %s\n" , pathBuffer.data()); |
590 | #endif |
591 | continue; |
592 | } |
593 | |
594 | /* Check if it is a directory with the same name as our package */ |
595 | if(!packageStub.isEmpty() && |
596 | (pathLen > packageStub.length()) && |
597 | !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length(), packageStub.data())) { |
598 | #ifdef UDATA_DEBUG |
599 | fprintf(stderr, "Found stub %s (will add package %s of len %d)\n" , packageStub.data(), basename, basenameLen); |
600 | #endif |
601 | pathBuffer.truncate(pathLen - packageStub.length()); |
602 | } |
603 | pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode); |
604 | } |
605 | |
606 | /* + basename */ |
607 | pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pErrorCode); |
608 | |
609 | if (!suffix.empty()) /* tack on suffix */ |
610 | { |
611 | if (suffix.length() > 4) { |
612 | // If the suffix is actually an item ("ibm-5348_P100-1997.cnv") and not an extension (".res") |
613 | // then we need to ensure that the path ends with a separator. |
614 | pathBuffer.ensureEndsWithFileSeparator(*pErrorCode); |
615 | } |
616 | pathBuffer.append(suffix, *pErrorCode); |
617 | } |
618 | } |
619 | |
620 | #ifdef UDATA_DEBUG |
621 | fprintf(stderr, " --> %s\n" , pathBuffer.data()); |
622 | #endif |
623 | |
624 | return pathBuffer.data(); |
625 | |
626 | } while(path); |
627 | |
628 | /* fell way off the end */ |
629 | return nullptr; |
630 | } |
631 | |
632 | U_NAMESPACE_END |
633 | |
634 | /* ==================================================================================*/ |
635 | |
636 | |
637 | /*----------------------------------------------------------------------* |
638 | * * |
639 | * Add a static reference to the common data library * |
640 | * Unless overridden by an explicit udata_setCommonData, this will be * |
641 | * our common data. * |
642 | * * |
643 | *----------------------------------------------------------------------*/ |
644 | #if !defined(ICU_DATA_DIR_WINDOWS) |
645 | // When using the Windows system data, we expect only a single data file. |
646 | extern "C" const DataHeader U_DATA_API U_ICUDATA_ENTRY_POINT; |
647 | #endif |
648 | |
649 | /* |
650 | * This would be a good place for weak-linkage declarations of |
651 | * partial-data-library access functions where each returns a pointer |
652 | * to its data package, if it is linked in. |
653 | */ |
654 | /* |
655 | extern const void *uprv_getICUData_collation() ATTRIBUTE_WEAK; |
656 | extern const void *uprv_getICUData_conversion() ATTRIBUTE_WEAK; |
657 | */ |
658 | |
659 | /*----------------------------------------------------------------------* |
660 | * * |
661 | * openCommonData Attempt to open a common format (.dat) file * |
662 | * Map it into memory (if it's not there already) * |
663 | * and return a UDataMemory object for it. * |
664 | * * |
665 | * If the requested data is already open and cached * |
666 | * just return the cached UDataMem object. * |
667 | * * |
668 | *----------------------------------------------------------------------*/ |
669 | static UDataMemory * |
670 | openCommonData(const char *path, /* Path from OpenChoice? */ |
671 | int32_t commonDataIndex, /* ICU Data (index >= 0) if path == nullptr */ |
672 | UErrorCode *pErrorCode) |
673 | { |
674 | UDataMemory tData; |
675 | const char *pathBuffer; |
676 | const char *inBasename; |
677 | |
678 | if (U_FAILURE(*pErrorCode)) { |
679 | return nullptr; |
680 | } |
681 | |
682 | UDataMemory_init(&tData); |
683 | |
684 | /* ??????? TODO revisit this */ |
685 | if (commonDataIndex >= 0) { |
686 | /* "mini-cache" for common ICU data */ |
687 | if(commonDataIndex >= UPRV_LENGTHOF(gCommonICUDataArray)) { |
688 | return nullptr; |
689 | } |
690 | { |
691 | Mutex lock; |
692 | if(gCommonICUDataArray[commonDataIndex] != nullptr) { |
693 | return gCommonICUDataArray[commonDataIndex]; |
694 | } |
695 | #if !defined(ICU_DATA_DIR_WINDOWS) |
696 | // When using the Windows system data, we expect only a single data file. |
697 | int32_t i; |
698 | for(i = 0; i < commonDataIndex; ++i) { |
699 | if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT) { |
700 | /* The linked-in data is already in the list. */ |
701 | return nullptr; |
702 | } |
703 | } |
704 | #endif |
705 | } |
706 | |
707 | /* Add the linked-in data to the list. */ |
708 | /* |
709 | * This is where we would check and call weakly linked partial-data-library |
710 | * access functions. |
711 | */ |
712 | /* |
713 | if (uprv_getICUData_collation) { |
714 | setCommonICUDataPointer(uprv_getICUData_collation(), false, pErrorCode); |
715 | } |
716 | if (uprv_getICUData_conversion) { |
717 | setCommonICUDataPointer(uprv_getICUData_conversion(), false, pErrorCode); |
718 | } |
719 | */ |
720 | #if !defined(ICU_DATA_DIR_WINDOWS) |
721 | // When using the Windows system data, we expect only a single data file. |
722 | setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, false, pErrorCode); |
723 | { |
724 | Mutex lock; |
725 | return gCommonICUDataArray[commonDataIndex]; |
726 | } |
727 | #endif |
728 | } |
729 | |
730 | |
731 | /* request is NOT for ICU Data. */ |
732 | |
733 | /* Find the base name portion of the supplied path. */ |
734 | /* inBasename will be left pointing somewhere within the original path string. */ |
735 | inBasename = findBasename(path); |
736 | #ifdef UDATA_DEBUG |
737 | fprintf(stderr, "inBasename = %s\n" , inBasename); |
738 | #endif |
739 | |
740 | if(*inBasename==0) { |
741 | /* no basename. This will happen if the original path was a directory name, */ |
742 | /* like "a/b/c/". (Fallback to separate files will still work.) */ |
743 | #ifdef UDATA_DEBUG |
744 | fprintf(stderr, "ocd: no basename in %s, bailing.\n" , path); |
745 | #endif |
746 | if (U_SUCCESS(*pErrorCode)) { |
747 | *pErrorCode=U_FILE_ACCESS_ERROR; |
748 | } |
749 | return nullptr; |
750 | } |
751 | |
752 | /* Is the requested common data file already open and cached? */ |
753 | /* Note that the cache is keyed by the base name only. The rest of the path, */ |
754 | /* if any, is not considered. */ |
755 | UDataMemory *dataToReturn = udata_findCachedData(inBasename, *pErrorCode); |
756 | if (dataToReturn != nullptr || U_FAILURE(*pErrorCode)) { |
757 | return dataToReturn; |
758 | } |
759 | |
760 | /* Requested item is not in the cache. |
761 | * Hunt it down, trying all the path locations |
762 | */ |
763 | |
764 | UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat" , true, pErrorCode); |
765 | |
766 | while ((UDataMemory_isLoaded(&tData)==false) && (pathBuffer = iter.next(pErrorCode)) != nullptr) |
767 | { |
768 | #ifdef UDATA_DEBUG |
769 | fprintf(stderr, "ocd: trying path %s - " , pathBuffer); |
770 | #endif |
771 | uprv_mapFile(&tData, pathBuffer, pErrorCode); |
772 | #ifdef UDATA_DEBUG |
773 | fprintf(stderr, "%s\n" , UDataMemory_isLoaded(&tData)?"LOADED" :"not loaded" ); |
774 | #endif |
775 | } |
776 | if (U_FAILURE(*pErrorCode)) { |
777 | return nullptr; |
778 | } |
779 | |
780 | #if defined(OS390_STUBDATA) && defined(OS390BATCH) |
781 | if (!UDataMemory_isLoaded(&tData)) { |
782 | char ourPathBuffer[1024]; |
783 | /* One more chance, for extendCommonData() */ |
784 | uprv_strncpy(ourPathBuffer, path, 1019); |
785 | ourPathBuffer[1019]=0; |
786 | uprv_strcat(ourPathBuffer, ".dat" ); |
787 | uprv_mapFile(&tData, ourPathBuffer, pErrorCode); |
788 | } |
789 | #endif |
790 | |
791 | if (U_FAILURE(*pErrorCode)) { |
792 | return nullptr; |
793 | } |
794 | if (!UDataMemory_isLoaded(&tData)) { |
795 | /* no common data */ |
796 | *pErrorCode=U_FILE_ACCESS_ERROR; |
797 | return nullptr; |
798 | } |
799 | |
800 | /* we have mapped a file, check its header */ |
801 | udata_checkCommonData(&tData, pErrorCode); |
802 | |
803 | |
804 | /* Cache the UDataMemory struct for this .dat file, |
805 | * so we won't need to hunt it down and map it again next time |
806 | * something is needed from it. */ |
807 | return udata_cacheDataItem(inBasename, &tData, pErrorCode); |
808 | } |
809 | |
810 | |
811 | /*----------------------------------------------------------------------* |
812 | * * |
813 | * extendICUData If the full set of ICU data was not loaded at * |
814 | * program startup, load it now. This function will * |
815 | * be called when the lookup of an ICU data item in * |
816 | * the common ICU data fails. * |
817 | * * |
818 | * return true if new data is loaded, false otherwise.* |
819 | * * |
820 | *----------------------------------------------------------------------*/ |
821 | static UBool extendICUData(UErrorCode *pErr) |
822 | { |
823 | UDataMemory *pData; |
824 | UDataMemory copyPData; |
825 | UBool didUpdate = false; |
826 | |
827 | /* |
828 | * There is a chance for a race condition here. |
829 | * Normally, ICU data is loaded from a DLL or via mmap() and |
830 | * setCommonICUData() will detect if the same address is set twice. |
831 | * If ICU is built with data loading via fread() then the address will |
832 | * be different each time the common data is loaded and we may add |
833 | * multiple copies of the data. |
834 | * In this case, use a mutex to prevent the race. |
835 | * Use a specific mutex to avoid nested locks of the global mutex. |
836 | */ |
837 | #if MAP_IMPLEMENTATION==MAP_STDIO |
838 | static UMutex extendICUDataMutex; |
839 | umtx_lock(&extendICUDataMutex); |
840 | #endif |
841 | if(!umtx_loadAcquire(gHaveTriedToLoadCommonData)) { |
842 | /* See if we can explicitly open a .dat file for the ICUData. */ |
843 | pData = openCommonData( |
844 | U_ICUDATA_NAME, /* "icudt20l" , for example. */ |
845 | -1, /* Pretend we're not opening ICUData */ |
846 | pErr); |
847 | |
848 | /* How about if there is no pData, eh... */ |
849 | |
850 | UDataMemory_init(©PData); |
851 | if(pData != nullptr) { |
852 | UDatamemory_assign(©PData, pData); |
853 | copyPData.map = 0; /* The mapping for this data is owned by the hash table */ |
854 | copyPData.mapAddr = 0; /* which will unmap it when ICU is shut down. */ |
855 | /* CommonICUData is also unmapped when ICU is shut down.*/ |
856 | /* To avoid unmapping the data twice, zero out the map */ |
857 | /* fields in the UDataMemory that we're assigning */ |
858 | /* to CommonICUData. */ |
859 | |
860 | didUpdate = /* no longer using this result */ |
861 | setCommonICUData(©PData,/* The new common data. */ |
862 | false, /* No warnings if write didn't happen */ |
863 | pErr); /* setCommonICUData honors errors; NOP if error set */ |
864 | } |
865 | |
866 | umtx_storeRelease(gHaveTriedToLoadCommonData, 1); |
867 | } |
868 | |
869 | didUpdate = findCommonICUDataByName(U_ICUDATA_NAME, *pErr); /* Return 'true' when a racing writes out the extended */ |
870 | /* data after another thread has failed to see it (in openCommonData), so */ |
871 | /* extended data can be examined. */ |
872 | /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */ |
873 | |
874 | #if MAP_IMPLEMENTATION==MAP_STDIO |
875 | umtx_unlock(&extendICUDataMutex); |
876 | #endif |
877 | return didUpdate; /* Return true if ICUData pointer was updated. */ |
878 | /* (Could potentially have been done by another thread racing */ |
879 | /* us through here, but that's fine, we still return true */ |
880 | /* so that current thread will also examine extended data. */ |
881 | } |
882 | |
883 | /*----------------------------------------------------------------------* |
884 | * * |
885 | * udata_setCommonData * |
886 | * * |
887 | *----------------------------------------------------------------------*/ |
888 | U_CAPI void U_EXPORT2 |
889 | udata_setCommonData(const void *data, UErrorCode *pErrorCode) { |
890 | UDataMemory dataMemory; |
891 | |
892 | if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) { |
893 | return; |
894 | } |
895 | |
896 | if(data==nullptr) { |
897 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
898 | return; |
899 | } |
900 | |
901 | /* set the data pointer and test for validity */ |
902 | UDataMemory_init(&dataMemory); |
903 | UDataMemory_setData(&dataMemory, data); |
904 | udata_checkCommonData(&dataMemory, pErrorCode); |
905 | if (U_FAILURE(*pErrorCode)) {return;} |
906 | |
907 | /* we have good data */ |
908 | /* Set it up as the ICU Common Data. */ |
909 | setCommonICUData(&dataMemory, true, pErrorCode); |
910 | } |
911 | |
912 | /*--------------------------------------------------------------------------- |
913 | * |
914 | * udata_setAppData |
915 | * |
916 | *---------------------------------------------------------------------------- */ |
917 | U_CAPI void U_EXPORT2 |
918 | udata_setAppData(const char *path, const void *data, UErrorCode *err) |
919 | { |
920 | UDataMemory udm; |
921 | |
922 | if(err==nullptr || U_FAILURE(*err)) { |
923 | return; |
924 | } |
925 | if(data==nullptr) { |
926 | *err=U_ILLEGAL_ARGUMENT_ERROR; |
927 | return; |
928 | } |
929 | |
930 | UDataMemory_init(&udm); |
931 | UDataMemory_setData(&udm, data); |
932 | udata_checkCommonData(&udm, err); |
933 | udata_cacheDataItem(path, &udm, err); |
934 | } |
935 | |
936 | /*----------------------------------------------------------------------------* |
937 | * * |
938 | * checkDataItem Given a freshly located/loaded data item, either * |
939 | * an entry in a common file or a separately loaded file, * |
940 | * sanity check its header, and see if the data is * |
941 | * acceptable to the app. * |
942 | * If the data is good, create and return a UDataMemory * |
943 | * object that can be returned to the application. * |
944 | * Return nullptr on any sort of failure. * |
945 | * * |
946 | *----------------------------------------------------------------------------*/ |
947 | static UDataMemory * |
948 | |
949 | ( |
950 | const DataHeader *, /* The data item to be checked. */ |
951 | UDataMemoryIsAcceptable *isAcceptable, /* App's call-back function */ |
952 | void *context, /* pass-thru param for above. */ |
953 | const char *type, /* pass-thru param for above. */ |
954 | const char *name, /* pass-thru param for above. */ |
955 | UErrorCode *nonFatalErr, /* Error code if this data was not acceptable */ |
956 | /* but openChoice should continue with */ |
957 | /* trying to get data from fallback path. */ |
958 | UErrorCode *fatalErr /* Bad error, caller should return immediately */ |
959 | ) |
960 | { |
961 | UDataMemory *rDataMem = nullptr; /* the new UDataMemory, to be returned. */ |
962 | |
963 | if (U_FAILURE(*fatalErr)) { |
964 | return nullptr; |
965 | } |
966 | |
967 | if(pHeader->dataHeader.magic1==0xda && |
968 | pHeader->dataHeader.magic2==0x27 && |
969 | (isAcceptable==nullptr || isAcceptable(context, type, name, &pHeader->info)) |
970 | ) { |
971 | rDataMem=UDataMemory_createNewInstance(fatalErr); |
972 | if (U_FAILURE(*fatalErr)) { |
973 | return nullptr; |
974 | } |
975 | rDataMem->pHeader = pHeader; |
976 | } else { |
977 | /* the data is not acceptable, look further */ |
978 | /* If we eventually find something good, this errorcode will be */ |
979 | /* cleared out. */ |
980 | *nonFatalErr=U_INVALID_FORMAT_ERROR; |
981 | } |
982 | return rDataMem; |
983 | } |
984 | |
985 | /** |
986 | * @return 0 if not loaded, 1 if loaded or err |
987 | */ |
988 | static UDataMemory *doLoadFromIndividualFiles(const char *pkgName, |
989 | const char *dataPath, const char *tocEntryPathSuffix, |
990 | /* following arguments are the same as doOpenChoice itself */ |
991 | const char *path, const char *type, const char *name, |
992 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
993 | UErrorCode *subErrorCode, |
994 | UErrorCode *pErrorCode) |
995 | { |
996 | const char *pathBuffer; |
997 | UDataMemory dataMemory; |
998 | UDataMemory *pEntryData; |
999 | |
1000 | /* look in ind. files: package\nam.typ ========================= */ |
1001 | /* init path iterator for individual files */ |
1002 | UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, false, pErrorCode); |
1003 | |
1004 | while ((pathBuffer = iter.next(pErrorCode)) != nullptr) |
1005 | { |
1006 | #ifdef UDATA_DEBUG |
1007 | fprintf(stderr, "UDATA: trying individual file %s\n" , pathBuffer); |
1008 | #endif |
1009 | if (uprv_mapFile(&dataMemory, pathBuffer, pErrorCode)) |
1010 | { |
1011 | pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode); |
1012 | if (pEntryData != nullptr) { |
1013 | /* Data is good. |
1014 | * Hand off ownership of the backing memory to the user's UDataMemory. |
1015 | * and return it. */ |
1016 | pEntryData->mapAddr = dataMemory.mapAddr; |
1017 | pEntryData->map = dataMemory.map; |
1018 | |
1019 | #ifdef UDATA_DEBUG |
1020 | fprintf(stderr, "** Mapped file: %s\n" , pathBuffer); |
1021 | #endif |
1022 | return pEntryData; |
1023 | } |
1024 | |
1025 | /* the data is not acceptable, or some error occurred. Either way, unmap the memory */ |
1026 | udata_close(&dataMemory); |
1027 | |
1028 | /* If we had a nasty error, bail out completely. */ |
1029 | if (U_FAILURE(*pErrorCode)) { |
1030 | return nullptr; |
1031 | } |
1032 | |
1033 | /* Otherwise remember that we found data but didn't like it for some reason */ |
1034 | *subErrorCode=U_INVALID_FORMAT_ERROR; |
1035 | } |
1036 | #ifdef UDATA_DEBUG |
1037 | fprintf(stderr, "%s\n" , UDataMemory_isLoaded(&dataMemory)?"LOADED" :"not loaded" ); |
1038 | #endif |
1039 | } |
1040 | return nullptr; |
1041 | } |
1042 | |
1043 | /** |
1044 | * @return 0 if not loaded, 1 if loaded or err |
1045 | */ |
1046 | static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName*/, |
1047 | const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const char *tocEntryName, |
1048 | /* following arguments are the same as doOpenChoice itself */ |
1049 | const char *path, const char *type, const char *name, |
1050 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
1051 | UErrorCode *subErrorCode, |
1052 | UErrorCode *pErrorCode) |
1053 | { |
1054 | UDataMemory *pEntryData; |
1055 | const DataHeader *; |
1056 | UDataMemory *pCommonData; |
1057 | int32_t commonDataIndex; |
1058 | UBool checkedExtendedICUData = false; |
1059 | /* try to get common data. The loop is for platforms such as the 390 that do |
1060 | * not initially load the full set of ICU data. If the lookup of an ICU data item |
1061 | * fails, the full (but slower to load) set is loaded, the and the loop repeats, |
1062 | * trying the lookup again. Once the full set of ICU data is loaded, the loop wont |
1063 | * repeat because the full set will be checked the first time through. |
1064 | * |
1065 | * The loop also handles the fallback to a .dat file if the application linked |
1066 | * to the stub data library rather than a real library. |
1067 | */ |
1068 | for (commonDataIndex = isICUData ? 0 : -1;;) { |
1069 | pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** search for pkg **/ |
1070 | |
1071 | if(U_SUCCESS(*subErrorCode) && pCommonData!=nullptr) { |
1072 | int32_t length; |
1073 | |
1074 | /* look up the data piece in the common data */ |
1075 | pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &length, subErrorCode); |
1076 | #ifdef UDATA_DEBUG |
1077 | fprintf(stderr, "%s: pHeader=%p - %s\n" , tocEntryName, (void*) pHeader, u_errorName(*subErrorCode)); |
1078 | #endif |
1079 | |
1080 | if(pHeader!=nullptr) { |
1081 | pEntryData = checkDataItem(pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode); |
1082 | #ifdef UDATA_DEBUG |
1083 | fprintf(stderr, "pEntryData=%p\n" , (void*) pEntryData); |
1084 | #endif |
1085 | if (U_FAILURE(*pErrorCode)) { |
1086 | return nullptr; |
1087 | } |
1088 | if (pEntryData != nullptr) { |
1089 | pEntryData->length = length; |
1090 | return pEntryData; |
1091 | } |
1092 | } |
1093 | } |
1094 | // If we failed due to being out-of-memory, then stop early and report the error. |
1095 | if (*subErrorCode == U_MEMORY_ALLOCATION_ERROR) { |
1096 | *pErrorCode = *subErrorCode; |
1097 | return nullptr; |
1098 | } |
1099 | /* Data wasn't found. If we were looking for an ICUData item and there is |
1100 | * more data available, load it and try again, |
1101 | * otherwise break out of this loop. */ |
1102 | if (!isICUData) { |
1103 | return nullptr; |
1104 | } else if (pCommonData != nullptr) { |
1105 | ++commonDataIndex; /* try the next data package */ |
1106 | } else if ((!checkedExtendedICUData) && extendICUData(subErrorCode)) { |
1107 | checkedExtendedICUData = true; |
1108 | /* try this data package slot again: it changed from nullptr to non-nullptr */ |
1109 | } else { |
1110 | return nullptr; |
1111 | } |
1112 | } |
1113 | } |
1114 | |
1115 | /* |
1116 | * Identify the Time Zone resources that are subject to special override data loading. |
1117 | */ |
1118 | static UBool isTimeZoneFile(const char *name, const char *type) { |
1119 | return ((uprv_strcmp(type, "res" ) == 0) && |
1120 | (uprv_strcmp(name, "zoneinfo64" ) == 0 || |
1121 | uprv_strcmp(name, "timezoneTypes" ) == 0 || |
1122 | uprv_strcmp(name, "windowsZones" ) == 0 || |
1123 | uprv_strcmp(name, "metaZones" ) == 0)); |
1124 | } |
1125 | |
1126 | /* |
1127 | * A note on the ownership of Mapped Memory |
1128 | * |
1129 | * For common format files, ownership resides with the UDataMemory object |
1130 | * that lives in the cache of opened common data. These UDataMemorys are private |
1131 | * to the udata implementation, and are never seen directly by users. |
1132 | * |
1133 | * The UDataMemory objects returned to users will have the address of some desired |
1134 | * data within the mapped region, but they wont have the mapping info itself, and thus |
1135 | * won't cause anything to be removed from memory when they are closed. |
1136 | * |
1137 | * For individual data files, the UDataMemory returned to the user holds the |
1138 | * information necessary to unmap the data on close. If the user independently |
1139 | * opens the same data file twice, two completely independent mappings will be made. |
1140 | * (There is no cache of opened data items from individual files, only a cache of |
1141 | * opened Common Data files, that is, files containing a collection of data items.) |
1142 | * |
1143 | * For common data passed in from the user via udata_setAppData() or |
1144 | * udata_setCommonData(), ownership remains with the user. |
1145 | * |
1146 | * UDataMemory objects themselves, as opposed to the memory they describe, |
1147 | * can be anywhere - heap, stack/local or global. |
1148 | * They have a flag to indicate when they're heap allocated and thus |
1149 | * must be deleted when closed. |
1150 | */ |
1151 | |
1152 | |
1153 | /*----------------------------------------------------------------------------* |
1154 | * * |
1155 | * main data loading functions * |
1156 | * * |
1157 | *----------------------------------------------------------------------------*/ |
1158 | static UDataMemory * |
1159 | doOpenChoice(const char *path, const char *type, const char *name, |
1160 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
1161 | UErrorCode *pErrorCode) |
1162 | { |
1163 | UDataMemory *retVal = nullptr; |
1164 | |
1165 | const char *dataPath; |
1166 | |
1167 | int32_t tocEntrySuffixIndex; |
1168 | const char *tocEntryPathSuffix; |
1169 | UErrorCode subErrorCode=U_ZERO_ERROR; |
1170 | const char *treeChar; |
1171 | |
1172 | UBool isICUData = false; |
1173 | |
1174 | |
1175 | FileTracer::traceOpen(path, type, name); |
1176 | |
1177 | |
1178 | /* Is this path ICU data? */ |
1179 | if(path == nullptr || |
1180 | !strcmp(path, U_ICUDATA_ALIAS) || /* "ICUDATA" */ |
1181 | !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-" */ |
1182 | uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) || |
1183 | !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-" */ |
1184 | uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) { |
1185 | isICUData = true; |
1186 | } |
1187 | |
1188 | #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR) /* Windows: try "foo\bar" and "foo/bar" */ |
1189 | /* remap from alternate path char to the main one */ |
1190 | CharString altSepPath; |
1191 | if(path) { |
1192 | if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != nullptr) { |
1193 | altSepPath.append(path, *pErrorCode); |
1194 | char *p; |
1195 | while ((p = uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR)) != nullptr) { |
1196 | *p = U_FILE_SEP_CHAR; |
1197 | } |
1198 | #if defined (UDATA_DEBUG) |
1199 | fprintf(stderr, "Changed path from [%s] to [%s]\n" , path, altSepPath.s); |
1200 | #endif |
1201 | path = altSepPath.data(); |
1202 | } |
1203 | } |
1204 | #endif |
1205 | |
1206 | CharString tocEntryName; /* entry name in tree format. ex: 'icudt28b/coll/ar.res' */ |
1207 | CharString tocEntryPath; /* entry name in path format. ex: 'icudt28b\\coll\\ar.res' */ |
1208 | |
1209 | CharString pkgName; |
1210 | CharString treeName; |
1211 | |
1212 | /* ======= Set up strings */ |
1213 | if(path==nullptr) { |
1214 | pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
1215 | } else { |
1216 | const char *pkg; |
1217 | const char *first; |
1218 | pkg = uprv_strrchr(path, U_FILE_SEP_CHAR); |
1219 | first = uprv_strchr(path, U_FILE_SEP_CHAR); |
1220 | if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash in the path- not a tree name */ |
1221 | /* see if this is an /absolute/path/to/package path */ |
1222 | if(pkg) { |
1223 | pkgName.append(pkg+1, *pErrorCode); |
1224 | } else { |
1225 | pkgName.append(path, *pErrorCode); |
1226 | } |
1227 | } else { |
1228 | treeChar = uprv_strchr(path, U_TREE_SEPARATOR); |
1229 | if(treeChar) { |
1230 | treeName.append(treeChar+1, *pErrorCode); /* following '-' */ |
1231 | if(isICUData) { |
1232 | pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
1233 | } else { |
1234 | pkgName.append(path, (int32_t)(treeChar-path), *pErrorCode); |
1235 | if (first == nullptr) { |
1236 | /* |
1237 | This user data has no path, but there is a tree name. |
1238 | Look up the correct path from the data cache later. |
1239 | */ |
1240 | path = pkgName.data(); |
1241 | } |
1242 | } |
1243 | } else { |
1244 | if(isICUData) { |
1245 | pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
1246 | } else { |
1247 | pkgName.append(path, *pErrorCode); |
1248 | } |
1249 | } |
1250 | } |
1251 | } |
1252 | |
1253 | #ifdef UDATA_DEBUG |
1254 | fprintf(stderr, " P=%s T=%s\n" , pkgName.data(), treeName.data()); |
1255 | #endif |
1256 | |
1257 | /* setting up the entry name and file name |
1258 | * Make up a full name by appending the type to the supplied |
1259 | * name, assuming that a type was supplied. |
1260 | */ |
1261 | |
1262 | /* prepend the package */ |
1263 | tocEntryName.append(pkgName, *pErrorCode); |
1264 | tocEntryPath.append(pkgName, *pErrorCode); |
1265 | tocEntrySuffixIndex = tocEntryName.length(); |
1266 | |
1267 | if(!treeName.isEmpty()) { |
1268 | tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode); |
1269 | tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode); |
1270 | } |
1271 | |
1272 | tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
1273 | tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
1274 | if(type!=nullptr && *type!=0) { |
1275 | tocEntryName.append("." , *pErrorCode).append(type, *pErrorCode); |
1276 | tocEntryPath.append("." , *pErrorCode).append(type, *pErrorCode); |
1277 | } |
1278 | // The +1 is for the U_FILE_SEP_CHAR that is always appended above. |
1279 | tocEntryPathSuffix = tocEntryPath.data() + tocEntrySuffixIndex + 1; /* suffix starts here */ |
1280 | |
1281 | #ifdef UDATA_DEBUG |
1282 | fprintf(stderr, " tocEntryName = %s\n" , tocEntryName.data()); |
1283 | fprintf(stderr, " tocEntryPath = %s\n" , tocEntryName.data()); |
1284 | #endif |
1285 | |
1286 | #if !defined(ICU_DATA_DIR_WINDOWS) |
1287 | if(path == nullptr) { |
1288 | path = COMMON_DATA_NAME; /* "icudt26e" */ |
1289 | } |
1290 | #else |
1291 | // When using the Windows system data, we expects only a single data file. |
1292 | path = COMMON_DATA_NAME; /* "icudt26e" */ |
1293 | #endif |
1294 | |
1295 | /************************ Begin loop looking for ind. files ***************/ |
1296 | #ifdef UDATA_DEBUG |
1297 | fprintf(stderr, "IND: inBasename = %s, pkg=%s\n" , "(n/a)" , packageNameFromPath(path)); |
1298 | #endif |
1299 | |
1300 | /* End of dealing with a null basename */ |
1301 | dataPath = u_getDataDirectory(); |
1302 | |
1303 | /**** Time zone individual files override */ |
1304 | if (isICUData && isTimeZoneFile(name, type)) { |
1305 | const char *tzFilesDir = u_getTimeZoneFilesDirectory(pErrorCode); |
1306 | if (tzFilesDir[0] != 0) { |
1307 | #ifdef UDATA_DEBUG |
1308 | fprintf(stderr, "Trying Time Zone Files directory = %s\n" , tzFilesDir); |
1309 | #endif |
1310 | retVal = doLoadFromIndividualFiles(/* pkgName.data() */ "" , tzFilesDir, tocEntryPathSuffix, |
1311 | /* path */ "" , type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
1312 | if((retVal != nullptr) || U_FAILURE(*pErrorCode)) { |
1313 | return retVal; |
1314 | } |
1315 | } |
1316 | } |
1317 | |
1318 | /**** COMMON PACKAGE - only if packages are first. */ |
1319 | if(gDataFileAccess == UDATA_PACKAGES_FIRST) { |
1320 | #ifdef UDATA_DEBUG |
1321 | fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n" ); |
1322 | #endif |
1323 | /* #2 */ |
1324 | retVal = doLoadFromCommonData(isICUData, |
1325 | pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), |
1326 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
1327 | if((retVal != nullptr) || U_FAILURE(*pErrorCode)) { |
1328 | return retVal; |
1329 | } |
1330 | } |
1331 | |
1332 | /**** INDIVIDUAL FILES */ |
1333 | if((gDataFileAccess==UDATA_PACKAGES_FIRST) || |
1334 | (gDataFileAccess==UDATA_FILES_FIRST)) { |
1335 | #ifdef UDATA_DEBUG |
1336 | fprintf(stderr, "Trying individual files\n" ); |
1337 | #endif |
1338 | /* Check to make sure that there is a dataPath to iterate over */ |
1339 | if ((dataPath && *dataPath) || !isICUData) { |
1340 | retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix, |
1341 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
1342 | if((retVal != nullptr) || U_FAILURE(*pErrorCode)) { |
1343 | return retVal; |
1344 | } |
1345 | } |
1346 | } |
1347 | |
1348 | /**** COMMON PACKAGE */ |
1349 | if((gDataFileAccess==UDATA_ONLY_PACKAGES) || |
1350 | (gDataFileAccess==UDATA_FILES_FIRST)) { |
1351 | #ifdef UDATA_DEBUG |
1352 | fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n" ); |
1353 | #endif |
1354 | retVal = doLoadFromCommonData(isICUData, |
1355 | pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), |
1356 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
1357 | if((retVal != nullptr) || U_FAILURE(*pErrorCode)) { |
1358 | return retVal; |
1359 | } |
1360 | } |
1361 | |
1362 | /* Load from DLL. If we haven't attempted package load, we also haven't had any chance to |
1363 | try a DLL (static or setCommonData/etc) load. |
1364 | If we ever have a "UDATA_ONLY_FILES", add it to the or list here. */ |
1365 | if(gDataFileAccess==UDATA_NO_FILES) { |
1366 | #ifdef UDATA_DEBUG |
1367 | fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n" ); |
1368 | #endif |
1369 | retVal = doLoadFromCommonData(isICUData, |
1370 | pkgName.data(), "" , tocEntryPathSuffix, tocEntryName.data(), |
1371 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
1372 | if((retVal != nullptr) || U_FAILURE(*pErrorCode)) { |
1373 | return retVal; |
1374 | } |
1375 | } |
1376 | |
1377 | /* data not found */ |
1378 | if(U_SUCCESS(*pErrorCode)) { |
1379 | if(U_SUCCESS(subErrorCode)) { |
1380 | /* file not found */ |
1381 | *pErrorCode=U_FILE_ACCESS_ERROR; |
1382 | } else { |
1383 | /* entry point not found or rejected */ |
1384 | *pErrorCode=subErrorCode; |
1385 | } |
1386 | } |
1387 | return retVal; |
1388 | } |
1389 | |
1390 | |
1391 | |
1392 | /* API ---------------------------------------------------------------------- */ |
1393 | |
1394 | U_CAPI UDataMemory * U_EXPORT2 |
1395 | udata_open(const char *path, const char *type, const char *name, |
1396 | UErrorCode *pErrorCode) { |
1397 | #ifdef UDATA_DEBUG |
1398 | fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n" , (path?path:"nullptr" ), name, type); |
1399 | fflush(stderr); |
1400 | #endif |
1401 | |
1402 | if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) { |
1403 | return nullptr; |
1404 | } else if(name==nullptr || *name==0) { |
1405 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
1406 | return nullptr; |
1407 | } else { |
1408 | return doOpenChoice(path, type, name, nullptr, nullptr, pErrorCode); |
1409 | } |
1410 | } |
1411 | |
1412 | |
1413 | |
1414 | U_CAPI UDataMemory * U_EXPORT2 |
1415 | udata_openChoice(const char *path, const char *type, const char *name, |
1416 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
1417 | UErrorCode *pErrorCode) { |
1418 | #ifdef UDATA_DEBUG |
1419 | fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n" , (path?path:"nullptr" ), name, type); |
1420 | #endif |
1421 | |
1422 | if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) { |
1423 | return nullptr; |
1424 | } else if(name==nullptr || *name==0 || isAcceptable==nullptr) { |
1425 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
1426 | return nullptr; |
1427 | } else { |
1428 | return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode); |
1429 | } |
1430 | } |
1431 | |
1432 | |
1433 | |
1434 | U_CAPI void U_EXPORT2 |
1435 | udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) { |
1436 | if(pInfo!=nullptr) { |
1437 | if(pData!=nullptr && pData->pHeader!=nullptr) { |
1438 | const UDataInfo *info=&pData->pHeader->info; |
1439 | uint16_t dataInfoSize=udata_getInfoSize(info); |
1440 | if(pInfo->size>dataInfoSize) { |
1441 | pInfo->size=dataInfoSize; |
1442 | } |
1443 | uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->size-2); |
1444 | if(info->isBigEndian!=U_IS_BIG_ENDIAN) { |
1445 | /* opposite endianness */ |
1446 | uint16_t x=info->reservedWord; |
1447 | pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8)); |
1448 | } |
1449 | } else { |
1450 | pInfo->size=0; |
1451 | } |
1452 | } |
1453 | } |
1454 | |
1455 | |
1456 | U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/) |
1457 | { |
1458 | // Note: this function is documented as not thread safe. |
1459 | gDataFileAccess = access; |
1460 | } |
1461 | |