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 // Android-changed: On Android, use our patched version of openCommonData() to load the data,
115 // and do not try to load ICU data from other files.
116 #if U_PLATFORM == U_PF_ANDROID
117 static UDataFileAccess gDataFileAccess = UDATA_NO_FILES;
118 #elif !defined(ICU_DATA_DIR_WINDOWS)
119 static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; // Access not synchronized.
120 // Modifying is documented as thread-unsafe.
121 #else
122 // If we are using the Windows data directory, then look in one spot only.
123 static UDataFileAccess gDataFileAccess = UDATA_NO_FILES;
124 #endif
125
126 // BEGIN Android-added: Include android/host-linux-specific headers and variables.
127 #ifdef ANDROID // if using the AOSP build system, e.g. Soong, but not the normal GNU make used by ./updateicudata.py
128 #if U_PLATFORM == U_PF_ANDROID || U_PLATFORM == U_PF_LINUX // if targeting Android or host linux
129 #define AOSP_ICU_INIT 1
130 #endif
131 #endif
132
133 #ifdef AOSP_ICU_INIT
134 #include "androidicuinit/android_icu_init.h"
135 static icu::UInitOnce gAospInitOnce {};
136 #endif
137 // END Android-added: Include android/host-linux-specific headers and variables.
138
139
140 static UBool U_CALLCONV
udata_cleanup()141 udata_cleanup()
142 {
143 int32_t i;
144
145 if (gCommonDataCache) { /* Delete the cache of user data mappings. */
146 uhash_close(gCommonDataCache); /* Table owns the contents, and will delete them. */
147 gCommonDataCache = nullptr; /* Cleanup is not thread safe. */
148 }
149 gCommonDataCacheInitOnce.reset();
150
151 for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != nullptr; ++i) {
152 udata_close(gCommonICUDataArray[i]);
153 gCommonICUDataArray[i] = nullptr;
154 }
155 gHaveTriedToLoadCommonData = 0;
156
157 // BEGIN Android-added: Use specialized libandroidicuinit to unload the data on Android/ART host.
158 #ifdef AOSP_ICU_INIT
159 android_icu_cleanup();
160 gAospInitOnce.reset();
161 #endif
162 // END Android-added: Use specialized libandroidicuinit to unload the data on Android/ART host.
163
164 return true; /* Everything was cleaned up */
165 }
166
167 static UBool U_CALLCONV
findCommonICUDataByName(const char * inBasename,UErrorCode & err)168 findCommonICUDataByName(const char *inBasename, UErrorCode &err)
169 {
170 UBool found = false;
171 int32_t i;
172
173 UDataMemory *pData = udata_findCachedData(inBasename, err);
174 if (U_FAILURE(err) || pData == nullptr)
175 return false;
176
177 {
178 Mutex lock;
179 for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) {
180 if ((gCommonICUDataArray[i] != nullptr) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) {
181 /* The data pointer is already in the array. */
182 found = true;
183 break;
184 }
185 }
186 }
187 return found;
188 }
189
190
191 /*
192 * setCommonICUData. Set a UDataMemory to be the global ICU Data
193 */
194 static UBool
setCommonICUData(UDataMemory * pData,UBool warn,UErrorCode * pErr)195 setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to caller, we copy it. */
196 UBool warn, /* If true, set USING_DEFAULT warning if ICUData was */
197 /* changed by another thread before we got to it. */
198 UErrorCode *pErr)
199 {
200 UDataMemory *newCommonData = UDataMemory_createNewInstance(pErr);
201 int32_t i;
202 UBool didUpdate = false;
203 if (U_FAILURE(*pErr)) {
204 return false;
205 }
206
207 /* For the assignment, other threads must cleanly see either the old */
208 /* or the new, not some partially initialized new. The old can not be */
209 /* deleted - someone may still have a pointer to it lying around in */
210 /* their locals. */
211 UDatamemory_assign(newCommonData, pData);
212 umtx_lock(nullptr);
213 for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) {
214 if (gCommonICUDataArray[i] == nullptr) {
215 gCommonICUDataArray[i] = newCommonData;
216 didUpdate = true;
217 break;
218 } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) {
219 /* The same data pointer is already in the array. */
220 break;
221 }
222 }
223 umtx_unlock(nullptr);
224
225 if (i == UPRV_LENGTHOF(gCommonICUDataArray) && warn) {
226 *pErr = U_USING_DEFAULT_WARNING;
227 }
228 if (didUpdate) {
229 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup);
230 } else {
231 uprv_free(newCommonData);
232 }
233 return didUpdate;
234 }
235
236 #if !defined(ICU_DATA_DIR_WINDOWS)
237
238 static UBool
setCommonICUDataPointer(const void * pData,UBool,UErrorCode * pErrorCode)239 setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) {
240 UDataMemory tData;
241 UDataMemory_init(&tData);
242 UDataMemory_setData(&tData, pData);
243 udata_checkCommonData(&tData, pErrorCode);
244 return setCommonICUData(&tData, false, pErrorCode);
245 }
246
247 #endif
248
249 static const char *
findBasename(const char * path)250 findBasename(const char *path) {
251 const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR);
252 if(basename==nullptr) {
253 return path;
254 } else {
255 return basename+1;
256 }
257 }
258
259 #ifdef UDATA_DEBUG
260 static const char *
packageNameFromPath(const char * path)261 packageNameFromPath(const char *path)
262 {
263 if((path == nullptr) || (*path == 0)) {
264 return U_ICUDATA_NAME;
265 }
266
267 path = findBasename(path);
268
269 if((path == nullptr) || (*path == 0)) {
270 return U_ICUDATA_NAME;
271 }
272
273 return path;
274 }
275 #endif
276
277 /*----------------------------------------------------------------------*
278 * *
279 * Cache for common data *
280 * Functions for looking up or adding entries to a cache of *
281 * data that has been previously opened. Avoids a potentially *
282 * expensive operation of re-opening the data for subsequent *
283 * uses. *
284 * *
285 * Data remains cached for the duration of the process. *
286 * *
287 *----------------------------------------------------------------------*/
288
289 typedef struct DataCacheElement {
290 char *name;
291 UDataMemory *item;
292 } DataCacheElement;
293
294
295
296 /*
297 * Deleter function for DataCacheElements.
298 * udata cleanup function closes the hash table; hash table in turn calls back to
299 * here for each entry.
300 */
DataCacheElement_deleter(void * pDCEl)301 static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) {
302 DataCacheElement* p = static_cast<DataCacheElement*>(pDCEl);
303 udata_close(p->item); /* unmaps storage */
304 uprv_free(p->name); /* delete the hash key string. */
305 uprv_free(pDCEl); /* delete 'this' */
306 }
307
udata_initHashTable(UErrorCode & err)308 static void U_CALLCONV udata_initHashTable(UErrorCode &err) {
309 U_ASSERT(gCommonDataCache == nullptr);
310 gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, nullptr, &err);
311 if (U_FAILURE(err)) {
312 return;
313 }
314 U_ASSERT(gCommonDataCache != nullptr);
315 uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter);
316 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup);
317 }
318
319 /* udata_getCacheHashTable()
320 * Get the hash table used to store the data cache entries.
321 * Lazy create it if it doesn't yet exist.
322 */
udata_getHashTable(UErrorCode & err)323 static UHashtable *udata_getHashTable(UErrorCode &err) {
324 umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable, err);
325 return gCommonDataCache;
326 }
327
328
329
udata_findCachedData(const char * path,UErrorCode & err)330 static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err)
331 {
332 UHashtable *htable;
333 UDataMemory *retVal = nullptr;
334 DataCacheElement *el;
335 const char *baseName;
336
337 htable = udata_getHashTable(err);
338 if (U_FAILURE(err)) {
339 return nullptr;
340 }
341
342 baseName = findBasename(path); /* Cache remembers only the base name, not the full path. */
343 umtx_lock(nullptr);
344 el = static_cast<DataCacheElement*>(uhash_get(htable, baseName));
345 umtx_unlock(nullptr);
346 if (el != nullptr) {
347 retVal = el->item;
348 }
349 #ifdef UDATA_DEBUG
350 fprintf(stderr, "Cache: [%s] -> %p\n", baseName, (void*) retVal);
351 #endif
352 return retVal;
353 }
354
355
udata_cacheDataItem(const char * path,UDataMemory * item,UErrorCode * pErr)356 static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UErrorCode *pErr) {
357 DataCacheElement *newElement;
358 const char *baseName;
359 int32_t nameLen;
360 UHashtable *htable;
361 DataCacheElement *oldValue = nullptr;
362 UErrorCode subErr = U_ZERO_ERROR;
363
364 htable = udata_getHashTable(*pErr);
365 if (U_FAILURE(*pErr)) {
366 return nullptr;
367 }
368
369 /* Create a new DataCacheElement - the thingy we store in the hash table -
370 * and copy the supplied path and UDataMemoryItems into it.
371 */
372 newElement = static_cast<DataCacheElement*>(uprv_malloc(sizeof(DataCacheElement)));
373 if (newElement == nullptr) {
374 *pErr = U_MEMORY_ALLOCATION_ERROR;
375 return nullptr;
376 }
377 newElement->item = UDataMemory_createNewInstance(pErr);
378 if (U_FAILURE(*pErr)) {
379 uprv_free(newElement);
380 return nullptr;
381 }
382 UDatamemory_assign(newElement->item, item);
383
384 baseName = findBasename(path);
385 nameLen = static_cast<int32_t>(uprv_strlen(baseName));
386 newElement->name = static_cast<char*>(uprv_malloc(nameLen + 1));
387 if (newElement->name == nullptr) {
388 *pErr = U_MEMORY_ALLOCATION_ERROR;
389 uprv_free(newElement->item);
390 uprv_free(newElement);
391 return nullptr;
392 }
393 uprv_strcpy(newElement->name, baseName);
394
395 /* Stick the new DataCacheElement into the hash table.
396 */
397 umtx_lock(nullptr);
398 oldValue = static_cast<DataCacheElement*>(uhash_get(htable, path));
399 if (oldValue != nullptr) {
400 subErr = U_USING_DEFAULT_WARNING;
401 }
402 else {
403 uhash_put(
404 htable,
405 newElement->name, /* Key */
406 newElement, /* Value */
407 &subErr);
408 }
409 umtx_unlock(nullptr);
410
411 #ifdef UDATA_DEBUG
412 fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n", newElement->name,
413 (void*) newElement->item, u_errorName(subErr), (void*) newElement->item->vFuncs);
414 #endif
415
416 if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) {
417 *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. */
418 uprv_free(newElement->name);
419 uprv_free(newElement->item);
420 uprv_free(newElement);
421 return oldValue ? oldValue->item : nullptr;
422 }
423
424 return newElement->item;
425 }
426
427 /*----------------------------------------------------------------------*==============
428 * *
429 * Path management. Could be shared with other tools/etc if need be *
430 * later on. *
431 * *
432 *----------------------------------------------------------------------*/
433
434 U_NAMESPACE_BEGIN
435
436 class UDataPathIterator
437 {
438 public:
439 UDataPathIterator(const char *path, const char *pkg,
440 const char *item, const char *suffix, UBool doCheckLastFour,
441 UErrorCode *pErrorCode);
442 const char *next(UErrorCode *pErrorCode);
443
444 private:
445 const char *path; /* working path (u_icudata_Dir) */
446 const char *nextPath; /* path following this one */
447 const char *basename; /* item's basename (icudt22e_mt.res)*/
448
449 StringPiece suffix; /* item suffix (can be null) */
450
451 uint32_t basenameLen; /* length of basename */
452
453 CharString itemPath; /* path passed in with item name */
454 CharString pathBuffer; /* output path for this it'ion */
455 CharString packageStub; /* example: "/icudt28b". Will ignore that leaf in set paths. */
456
457 UBool checkLastFour; /* if true then allow paths such as '/foo/myapp.dat'
458 * to match, checks last 4 chars of suffix with
459 * last 4 of path, then previous chars. */
460 };
461
462 /**
463 * @param iter The iterator to be initialized. Its current state does not matter.
464 * @param inPath The full pathname to be iterated over. If nullptr, defaults to U_ICUDATA_NAME
465 * @param pkg Package which is being searched for, ex "icudt28l". Will ignore leaf directories such as /icudt28l
466 * @param item Item to be searched for. Can include full path, such as /a/b/foo.dat
467 * @param inSuffix Optional item suffix, if not-null (ex. ".dat") then 'path' can contain 'item' explicitly.
468 * Ex: 'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar/baz' as item #2.
469 * '/blarg/stuff.dat' would also be found.
470 * Note: inSuffix may also be the 'item' being searched for as well, (ex: "ibm-5348_P100-1997.cnv"), in which case
471 * the 'item' parameter is often the same as pkg. (Though sometimes might have a tree part as well, ex: "icudt62l-curr").
472 */
UDataPathIterator(const char * inPath,const char * pkg,const char * item,const char * inSuffix,UBool doCheckLastFour,UErrorCode * pErrorCode)473 UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg,
474 const char *item, const char *inSuffix, UBool doCheckLastFour,
475 UErrorCode *pErrorCode)
476 {
477 #ifdef UDATA_DEBUG
478 fprintf(stderr, "SUFFIX1=%s PATH=%s\n", inSuffix, inPath);
479 #endif
480 /** Path **/
481 if(inPath == nullptr) {
482 path = u_getDataDirectory();
483 } else {
484 path = inPath;
485 }
486
487 /** Package **/
488 if(pkg != nullptr) {
489 packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode);
490 #ifdef UDATA_DEBUG
491 fprintf(stderr, "STUB=%s [%d]\n", packageStub.data(), packageStub.length());
492 #endif
493 }
494
495 /** Item **/
496 basename = findBasename(item);
497 basenameLen = static_cast<int32_t>(uprv_strlen(basename));
498
499 /** Item path **/
500 if(basename == item) {
501 nextPath = path;
502 } else {
503 itemPath.append(item, static_cast<int32_t>(basename - item), *pErrorCode);
504 nextPath = itemPath.data();
505 }
506 #ifdef UDATA_DEBUG
507 fprintf(stderr, "SUFFIX=%s [%p]\n", inSuffix, (void*) inSuffix);
508 #endif
509
510 /** Suffix **/
511 if(inSuffix != nullptr) {
512 suffix = inSuffix;
513 } else {
514 suffix = "";
515 }
516
517 checkLastFour = doCheckLastFour;
518
519 /* pathBuffer will hold the output path strings returned by this iterator */
520
521 #ifdef UDATA_DEBUG
522 fprintf(stderr, "0: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%s], [nextpath=%s], [checklast4=%s]\n",
523 item,
524 path,
525 basename,
526 suffix.data(),
527 itemPath.data(),
528 nextPath,
529 checkLastFour?"true":"false");
530 #endif
531 }
532
533 /**
534 * Get the next path on the list.
535 *
536 * @param iter The Iter to be used
537 * @param len If set, pointer to the length of the returned path, for convenience.
538 * @return Pointer to the next path segment, or nullptr if there are no more.
539 */
next(UErrorCode * pErrorCode)540 const char *UDataPathIterator::next(UErrorCode *pErrorCode)
541 {
542 if(U_FAILURE(*pErrorCode)) {
543 return nullptr;
544 }
545
546 const char *currentPath = nullptr;
547 int32_t pathLen = 0;
548 const char *pathBasename;
549
550 do
551 {
552 if( nextPath == nullptr ) {
553 break;
554 }
555 currentPath = nextPath;
556
557 if(nextPath == itemPath.data()) { /* we were processing item's path. */
558 nextPath = path; /* start with regular path next tm. */
559 pathLen = static_cast<int32_t>(uprv_strlen(currentPath));
560 } else {
561 /* fix up next for next time */
562 nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR);
563 if(nextPath == nullptr) {
564 /* segment: entire path */
565 pathLen = static_cast<int32_t>(uprv_strlen(currentPath));
566 } else {
567 /* segment: until next segment */
568 pathLen = static_cast<int32_t>(nextPath - currentPath);
569 /* skip divider */
570 nextPath ++;
571 }
572 }
573
574 if(pathLen == 0) {
575 continue;
576 }
577
578 #ifdef UDATA_DEBUG
579 fprintf(stderr, "rest of path (IDD) = %s\n", currentPath);
580 fprintf(stderr, " ");
581 {
582 int32_t qqq;
583 for(qqq=0;qqq<pathLen;qqq++)
584 {
585 fprintf(stderr, " ");
586 }
587
588 fprintf(stderr, "^\n");
589 }
590 #endif
591 pathBuffer.clear().append(currentPath, pathLen, *pErrorCode);
592
593 /* check for .dat files */
594 pathBasename = findBasename(pathBuffer.data());
595
596 if(checkLastFour &&
597 (pathLen>=4) &&
598 uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix.data(), 4)==0 && /* suffix matches */
599 uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)==0 && /* base matches */
600 uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full len */
601
602 #ifdef UDATA_DEBUG
603 fprintf(stderr, "Have %s file on the path: %s\n", suffix.data(), pathBuffer.data());
604 #endif
605 /* do nothing */
606 }
607 else
608 { /* regular dir path */
609 if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) {
610 if((pathLen>=4) &&
611 uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat", 4) == 0)
612 {
613 #ifdef UDATA_DEBUG
614 fprintf(stderr, "skipping non-directory .dat file %s\n", pathBuffer.data());
615 #endif
616 continue;
617 }
618
619 /* Check if it is a directory with the same name as our package */
620 if(!packageStub.isEmpty() &&
621 (pathLen > packageStub.length()) &&
622 !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length(), packageStub.data())) {
623 #ifdef UDATA_DEBUG
624 fprintf(stderr, "Found stub %s (will add package %s of len %d)\n", packageStub.data(), basename, basenameLen);
625 #endif
626 pathBuffer.truncate(pathLen - packageStub.length());
627 }
628 pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode);
629 }
630
631 /* + basename */
632 pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pErrorCode);
633
634 if (!suffix.empty()) /* tack on suffix */
635 {
636 if (suffix.length() > 4) {
637 // If the suffix is actually an item ("ibm-5348_P100-1997.cnv") and not an extension (".res")
638 // then we need to ensure that the path ends with a separator.
639 pathBuffer.ensureEndsWithFileSeparator(*pErrorCode);
640 }
641 pathBuffer.append(suffix, *pErrorCode);
642 }
643 }
644
645 #ifdef UDATA_DEBUG
646 fprintf(stderr, " --> %s\n", pathBuffer.data());
647 #endif
648
649 return pathBuffer.data();
650
651 } while(path);
652
653 /* fell way off the end */
654 return nullptr;
655 }
656
657 U_NAMESPACE_END
658
659 /* ==================================================================================*/
660
661
662 /*----------------------------------------------------------------------*
663 * *
664 * Add a static reference to the common data library *
665 * Unless overridden by an explicit udata_setCommonData, this will be *
666 * our common data. *
667 * *
668 *----------------------------------------------------------------------*/
669 #if !defined(ICU_DATA_DIR_WINDOWS)
670 // When using the Windows system data, we expect only a single data file.
671 extern "C" const DataHeader U_DATA_API U_ICUDATA_ENTRY_POINT;
672 #endif
673
674 /*
675 * This would be a good place for weak-linkage declarations of
676 * partial-data-library access functions where each returns a pointer
677 * to its data package, if it is linked in.
678 */
679 /*
680 extern const void *uprv_getICUData_collation() ATTRIBUTE_WEAK;
681 extern const void *uprv_getICUData_conversion() ATTRIBUTE_WEAK;
682 */
683
684 /*----------------------------------------------------------------------*
685 * *
686 * openCommonData Attempt to open a common format (.dat) file *
687 * Map it into memory (if it's not there already) *
688 * and return a UDataMemory object for it. *
689 * *
690 * If the requested data is already open and cached *
691 * just return the cached UDataMem object. *
692 * *
693 *----------------------------------------------------------------------*/
694 static UDataMemory *
openCommonData(const char * path,int32_t commonDataIndex,UErrorCode * pErrorCode)695 openCommonData(const char *path, /* Path from OpenChoice? */
696 int32_t commonDataIndex, /* ICU Data (index >= 0) if path == nullptr */
697 UErrorCode *pErrorCode)
698 {
699 UDataMemory tData;
700 const char *pathBuffer;
701 const char *inBasename;
702
703 if (U_FAILURE(*pErrorCode)) {
704 return nullptr;
705 }
706
707 UDataMemory_init(&tData);
708
709 /* ??????? TODO revisit this */
710 if (commonDataIndex >= 0) {
711 /* "mini-cache" for common ICU data */
712 if(commonDataIndex >= UPRV_LENGTHOF(gCommonICUDataArray)) {
713 return nullptr;
714 }
715 {
716 Mutex lock;
717 if(gCommonICUDataArray[commonDataIndex] != nullptr) {
718 return gCommonICUDataArray[commonDataIndex];
719 }
720 #if !defined(ICU_DATA_DIR_WINDOWS)
721 // When using the Windows system data, we expect only a single data file.
722 int32_t i;
723 for(i = 0; i < commonDataIndex; ++i) {
724 if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT) {
725 /* The linked-in data is already in the list. */
726 return nullptr;
727 }
728 }
729 #endif
730 }
731
732 // BEGIN Android-added: Use specialized libandroidicuinit to load the data on Android/ART host.
733 #ifdef AOSP_ICU_INIT // Do nothing on other platforms, e.g. Windows
734 // android_icu_init() is only called once.
735 umtx_initOnce(gAospInitOnce, &android_icu_init);
736 #endif // AOSP_ICU_INIT
737 // END Android-added: Use specialized libandroidicuinit to load the data on Android/ART host.
738
739 /* Add the linked-in data to the list. */
740 /*
741 * This is where we would check and call weakly linked partial-data-library
742 * access functions.
743 */
744 /*
745 if (uprv_getICUData_collation) {
746 setCommonICUDataPointer(uprv_getICUData_collation(), false, pErrorCode);
747 }
748 if (uprv_getICUData_conversion) {
749 setCommonICUDataPointer(uprv_getICUData_conversion(), false, pErrorCode);
750 }
751 */
752 #if !defined(ICU_DATA_DIR_WINDOWS)
753 // When using the Windows system data, we expect only a single data file.
754 setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, false, pErrorCode);
755 {
756 Mutex lock;
757 return gCommonICUDataArray[commonDataIndex];
758 }
759 #endif
760 }
761
762
763 /* request is NOT for ICU Data. */
764
765 /* Find the base name portion of the supplied path. */
766 /* inBasename will be left pointing somewhere within the original path string. */
767 inBasename = findBasename(path);
768 #ifdef UDATA_DEBUG
769 fprintf(stderr, "inBasename = %s\n", inBasename);
770 #endif
771
772 if(*inBasename==0) {
773 /* no basename. This will happen if the original path was a directory name, */
774 /* like "a/b/c/". (Fallback to separate files will still work.) */
775 #ifdef UDATA_DEBUG
776 fprintf(stderr, "ocd: no basename in %s, bailing.\n", path);
777 #endif
778 if (U_SUCCESS(*pErrorCode)) {
779 *pErrorCode=U_FILE_ACCESS_ERROR;
780 }
781 return nullptr;
782 }
783
784 /* Is the requested common data file already open and cached? */
785 /* Note that the cache is keyed by the base name only. The rest of the path, */
786 /* if any, is not considered. */
787 UDataMemory *dataToReturn = udata_findCachedData(inBasename, *pErrorCode);
788 if (dataToReturn != nullptr || U_FAILURE(*pErrorCode)) {
789 return dataToReturn;
790 }
791
792 /* Requested item is not in the cache.
793 * Hunt it down, trying all the path locations
794 */
795
796 UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat", true, pErrorCode);
797
798 while ((UDataMemory_isLoaded(&tData)==false) && (pathBuffer = iter.next(pErrorCode)) != nullptr)
799 {
800 #ifdef UDATA_DEBUG
801 fprintf(stderr, "ocd: trying path %s - ", pathBuffer);
802 #endif
803 uprv_mapFile(&tData, pathBuffer, pErrorCode);
804 #ifdef UDATA_DEBUG
805 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&tData)?"LOADED":"not loaded");
806 #endif
807 }
808 if (U_FAILURE(*pErrorCode)) {
809 return nullptr;
810 }
811
812 if (U_FAILURE(*pErrorCode)) {
813 return nullptr;
814 }
815 if (!UDataMemory_isLoaded(&tData)) {
816 /* no common data */
817 *pErrorCode=U_FILE_ACCESS_ERROR;
818 return nullptr;
819 }
820
821 /* we have mapped a file, check its header */
822 udata_checkCommonData(&tData, pErrorCode);
823
824
825 /* Cache the UDataMemory struct for this .dat file,
826 * so we won't need to hunt it down and map it again next time
827 * something is needed from it. */
828 return udata_cacheDataItem(inBasename, &tData, pErrorCode);
829 }
830
831
832 /*----------------------------------------------------------------------*
833 * *
834 * extendICUData If the full set of ICU data was not loaded at *
835 * program startup, load it now. This function will *
836 * be called when the lookup of an ICU data item in *
837 * the common ICU data fails. *
838 * *
839 * return true if new data is loaded, false otherwise.*
840 * *
841 *----------------------------------------------------------------------*/
extendICUData(UErrorCode * pErr)842 static UBool extendICUData(UErrorCode *pErr)
843 {
844 UDataMemory *pData;
845 UDataMemory copyPData;
846 UBool didUpdate = false;
847
848 /*
849 * There is a chance for a race condition here.
850 * Normally, ICU data is loaded from a DLL or via mmap() and
851 * setCommonICUData() will detect if the same address is set twice.
852 * If ICU is built with data loading via fread() then the address will
853 * be different each time the common data is loaded and we may add
854 * multiple copies of the data.
855 * In this case, use a mutex to prevent the race.
856 * Use a specific mutex to avoid nested locks of the global mutex.
857 */
858 #if MAP_IMPLEMENTATION==MAP_STDIO
859 static UMutex extendICUDataMutex;
860 umtx_lock(&extendICUDataMutex);
861 #endif
862 if(!umtx_loadAcquire(gHaveTriedToLoadCommonData)) {
863 /* See if we can explicitly open a .dat file for the ICUData. */
864 pData = openCommonData(
865 U_ICUDATA_NAME, /* "icudt20l" , for example. */
866 -1, /* Pretend we're not opening ICUData */
867 pErr);
868
869 /* How about if there is no pData, eh... */
870
871 UDataMemory_init(©PData);
872 if(pData != nullptr) {
873 UDatamemory_assign(©PData, pData);
874 copyPData.map = nullptr; /* The mapping for this data is owned by the hash table */
875 copyPData.mapAddr = nullptr; /* which will unmap it when ICU is shut down. */
876 /* CommonICUData is also unmapped when ICU is shut down.*/
877 /* To avoid unmapping the data twice, zero out the map */
878 /* fields in the UDataMemory that we're assigning */
879 /* to CommonICUData. */
880
881 didUpdate = /* no longer using this result */
882 setCommonICUData(©PData,/* The new common data. */
883 false, /* No warnings if write didn't happen */
884 pErr); /* setCommonICUData honors errors; NOP if error set */
885 }
886
887 umtx_storeRelease(gHaveTriedToLoadCommonData, 1);
888 }
889
890 didUpdate = findCommonICUDataByName(U_ICUDATA_NAME, *pErr); /* Return 'true' when a racing writes out the extended */
891 /* data after another thread has failed to see it (in openCommonData), so */
892 /* extended data can be examined. */
893 /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */
894
895 #if MAP_IMPLEMENTATION==MAP_STDIO
896 umtx_unlock(&extendICUDataMutex);
897 #endif
898 return didUpdate; /* Return true if ICUData pointer was updated. */
899 /* (Could potentially have been done by another thread racing */
900 /* us through here, but that's fine, we still return true */
901 /* so that current thread will also examine extended data. */
902 }
903
904 /*----------------------------------------------------------------------*
905 * *
906 * udata_setCommonData *
907 * *
908 *----------------------------------------------------------------------*/
909 U_CAPI void U_EXPORT2
udata_setCommonData(const void * data,UErrorCode * pErrorCode)910 udata_setCommonData(const void *data, UErrorCode *pErrorCode) {
911 UDataMemory dataMemory;
912
913 if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) {
914 return;
915 }
916
917 if(data==nullptr) {
918 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
919 return;
920 }
921
922 /* set the data pointer and test for validity */
923 UDataMemory_init(&dataMemory);
924 UDataMemory_setData(&dataMemory, data);
925 udata_checkCommonData(&dataMemory, pErrorCode);
926 if (U_FAILURE(*pErrorCode)) {return;}
927
928 /* we have good data */
929 /* Set it up as the ICU Common Data. */
930 setCommonICUData(&dataMemory, true, pErrorCode);
931 }
932
933 /*---------------------------------------------------------------------------
934 *
935 * udata_setAppData
936 *
937 *---------------------------------------------------------------------------- */
938 U_CAPI void U_EXPORT2
udata_setAppData(const char * path,const void * data,UErrorCode * err)939 udata_setAppData(const char *path, const void *data, UErrorCode *err)
940 {
941 UDataMemory udm;
942
943 if(err==nullptr || U_FAILURE(*err)) {
944 return;
945 }
946 if(data==nullptr) {
947 *err=U_ILLEGAL_ARGUMENT_ERROR;
948 return;
949 }
950
951 UDataMemory_init(&udm);
952 UDataMemory_setData(&udm, data);
953 udata_checkCommonData(&udm, err);
954 udata_cacheDataItem(path, &udm, err);
955 }
956
957 /*----------------------------------------------------------------------------*
958 * *
959 * checkDataItem Given a freshly located/loaded data item, either *
960 * an entry in a common file or a separately loaded file, *
961 * sanity check its header, and see if the data is *
962 * acceptable to the app. *
963 * If the data is good, create and return a UDataMemory *
964 * object that can be returned to the application. *
965 * Return nullptr on any sort of failure. *
966 * *
967 *----------------------------------------------------------------------------*/
968 static UDataMemory *
checkDataItem(const DataHeader * pHeader,UDataMemoryIsAcceptable * isAcceptable,void * context,const char * type,const char * name,UErrorCode * nonFatalErr,UErrorCode * fatalErr)969 checkDataItem
970 (
971 const DataHeader *pHeader, /* The data item to be checked. */
972 UDataMemoryIsAcceptable *isAcceptable, /* App's call-back function */
973 void *context, /* pass-thru param for above. */
974 const char *type, /* pass-thru param for above. */
975 const char *name, /* pass-thru param for above. */
976 UErrorCode *nonFatalErr, /* Error code if this data was not acceptable */
977 /* but openChoice should continue with */
978 /* trying to get data from fallback path. */
979 UErrorCode *fatalErr /* Bad error, caller should return immediately */
980 )
981 {
982 UDataMemory *rDataMem = nullptr; /* the new UDataMemory, to be returned. */
983
984 if (U_FAILURE(*fatalErr)) {
985 return nullptr;
986 }
987
988 if(pHeader->dataHeader.magic1==0xda &&
989 pHeader->dataHeader.magic2==0x27 &&
990 (isAcceptable==nullptr || isAcceptable(context, type, name, &pHeader->info))
991 ) {
992 rDataMem=UDataMemory_createNewInstance(fatalErr);
993 if (U_FAILURE(*fatalErr)) {
994 return nullptr;
995 }
996 rDataMem->pHeader = pHeader;
997 } else {
998 /* the data is not acceptable, look further */
999 /* If we eventually find something good, this errorcode will be */
1000 /* cleared out. */
1001 *nonFatalErr=U_INVALID_FORMAT_ERROR;
1002 }
1003 return rDataMem;
1004 }
1005
1006 /**
1007 * @return 0 if not loaded, 1 if loaded or err
1008 */
doLoadFromIndividualFiles(const char * pkgName,const char * dataPath,const char * tocEntryPathSuffix,const char * path,const char * type,const char * name,UDataMemoryIsAcceptable * isAcceptable,void * context,UErrorCode * subErrorCode,UErrorCode * pErrorCode)1009 static UDataMemory *doLoadFromIndividualFiles(const char *pkgName,
1010 const char *dataPath, const char *tocEntryPathSuffix,
1011 /* following arguments are the same as doOpenChoice itself */
1012 const char *path, const char *type, const char *name,
1013 UDataMemoryIsAcceptable *isAcceptable, void *context,
1014 UErrorCode *subErrorCode,
1015 UErrorCode *pErrorCode)
1016 {
1017 const char *pathBuffer;
1018 UDataMemory dataMemory;
1019 UDataMemory *pEntryData;
1020
1021 /* look in ind. files: package\nam.typ ========================= */
1022 /* init path iterator for individual files */
1023 UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, false, pErrorCode);
1024
1025 while ((pathBuffer = iter.next(pErrorCode)) != nullptr)
1026 {
1027 #ifdef UDATA_DEBUG
1028 fprintf(stderr, "UDATA: trying individual file %s\n", pathBuffer);
1029 #endif
1030 if (uprv_mapFile(&dataMemory, pathBuffer, pErrorCode))
1031 {
1032 pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode);
1033 if (pEntryData != nullptr) {
1034 /* Data is good.
1035 * Hand off ownership of the backing memory to the user's UDataMemory.
1036 * and return it. */
1037 pEntryData->mapAddr = dataMemory.mapAddr;
1038 pEntryData->map = dataMemory.map;
1039
1040 #ifdef UDATA_DEBUG
1041 fprintf(stderr, "** Mapped file: %s\n", pathBuffer);
1042 #endif
1043 return pEntryData;
1044 }
1045
1046 /* the data is not acceptable, or some error occurred. Either way, unmap the memory */
1047 udata_close(&dataMemory);
1048
1049 /* If we had a nasty error, bail out completely. */
1050 if (U_FAILURE(*pErrorCode)) {
1051 return nullptr;
1052 }
1053
1054 /* Otherwise remember that we found data but didn't like it for some reason */
1055 *subErrorCode=U_INVALID_FORMAT_ERROR;
1056 }
1057 #ifdef UDATA_DEBUG
1058 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&dataMemory)?"LOADED":"not loaded");
1059 #endif
1060 }
1061 return nullptr;
1062 }
1063
1064 /**
1065 * @return 0 if not loaded, 1 if loaded or err
1066 */
doLoadFromCommonData(UBool isICUData,const char *,const char *,const char *,const char * tocEntryName,const char * path,const char * type,const char * name,UDataMemoryIsAcceptable * isAcceptable,void * context,UErrorCode * subErrorCode,UErrorCode * pErrorCode)1067 static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName*/,
1068 const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const char *tocEntryName,
1069 /* following arguments are the same as doOpenChoice itself */
1070 const char *path, const char *type, const char *name,
1071 UDataMemoryIsAcceptable *isAcceptable, void *context,
1072 UErrorCode *subErrorCode,
1073 UErrorCode *pErrorCode)
1074 {
1075 UDataMemory *pEntryData;
1076 const DataHeader *pHeader;
1077 UDataMemory *pCommonData;
1078 int32_t commonDataIndex;
1079 UBool checkedExtendedICUData = false;
1080 /* try to get common data. The loop is for platforms such as the 390 that do
1081 * not initially load the full set of ICU data. If the lookup of an ICU data item
1082 * fails, the full (but slower to load) set is loaded, the and the loop repeats,
1083 * trying the lookup again. Once the full set of ICU data is loaded, the loop wont
1084 * repeat because the full set will be checked the first time through.
1085 *
1086 * The loop also handles the fallback to a .dat file if the application linked
1087 * to the stub data library rather than a real library.
1088 */
1089 for (commonDataIndex = isICUData ? 0 : -1;;) {
1090 pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** search for pkg **/
1091
1092 if(U_SUCCESS(*subErrorCode) && pCommonData!=nullptr) {
1093 int32_t length;
1094
1095 /* look up the data piece in the common data */
1096 pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &length, subErrorCode);
1097 #ifdef UDATA_DEBUG
1098 fprintf(stderr, "%s: pHeader=%p - %s\n", tocEntryName, (void*) pHeader, u_errorName(*subErrorCode));
1099 #endif
1100
1101 if(pHeader!=nullptr) {
1102 pEntryData = checkDataItem(pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode);
1103 #ifdef UDATA_DEBUG
1104 fprintf(stderr, "pEntryData=%p\n", (void*) pEntryData);
1105 #endif
1106 if (U_FAILURE(*pErrorCode)) {
1107 return nullptr;
1108 }
1109 if (pEntryData != nullptr) {
1110 pEntryData->length = length;
1111 return pEntryData;
1112 }
1113 }
1114 }
1115 // If we failed due to being out-of-memory, then stop early and report the error.
1116 if (*subErrorCode == U_MEMORY_ALLOCATION_ERROR) {
1117 *pErrorCode = *subErrorCode;
1118 return nullptr;
1119 }
1120 /* Data wasn't found. If we were looking for an ICUData item and there is
1121 * more data available, load it and try again,
1122 * otherwise break out of this loop. */
1123 if (!isICUData) {
1124 return nullptr;
1125 } else if (pCommonData != nullptr) {
1126 ++commonDataIndex; /* try the next data package */
1127 } else if ((!checkedExtendedICUData) && extendICUData(subErrorCode)) {
1128 checkedExtendedICUData = true;
1129 /* try this data package slot again: it changed from nullptr to non-nullptr */
1130 } else {
1131 return nullptr;
1132 }
1133 }
1134 }
1135
1136 /*
1137 * Identify the Time Zone resources that are subject to special override data loading.
1138 */
isTimeZoneFile(const char * name,const char * type)1139 static UBool isTimeZoneFile(const char *name, const char *type) {
1140 return ((uprv_strcmp(type, "res") == 0) &&
1141 (uprv_strcmp(name, "zoneinfo64") == 0 ||
1142 uprv_strcmp(name, "timezoneTypes") == 0 ||
1143 uprv_strcmp(name, "windowsZones") == 0 ||
1144 uprv_strcmp(name, "metaZones") == 0));
1145 }
1146
1147 /*
1148 * A note on the ownership of Mapped Memory
1149 *
1150 * For common format files, ownership resides with the UDataMemory object
1151 * that lives in the cache of opened common data. These UDataMemorys are private
1152 * to the udata implementation, and are never seen directly by users.
1153 *
1154 * The UDataMemory objects returned to users will have the address of some desired
1155 * data within the mapped region, but they wont have the mapping info itself, and thus
1156 * won't cause anything to be removed from memory when they are closed.
1157 *
1158 * For individual data files, the UDataMemory returned to the user holds the
1159 * information necessary to unmap the data on close. If the user independently
1160 * opens the same data file twice, two completely independent mappings will be made.
1161 * (There is no cache of opened data items from individual files, only a cache of
1162 * opened Common Data files, that is, files containing a collection of data items.)
1163 *
1164 * For common data passed in from the user via udata_setAppData() or
1165 * udata_setCommonData(), ownership remains with the user.
1166 *
1167 * UDataMemory objects themselves, as opposed to the memory they describe,
1168 * can be anywhere - heap, stack/local or global.
1169 * They have a flag to indicate when they're heap allocated and thus
1170 * must be deleted when closed.
1171 */
1172
1173
1174 /*----------------------------------------------------------------------------*
1175 * *
1176 * main data loading functions *
1177 * *
1178 *----------------------------------------------------------------------------*/
1179 static UDataMemory *
doOpenChoice(const char * path,const char * type,const char * name,UDataMemoryIsAcceptable * isAcceptable,void * context,UErrorCode * pErrorCode)1180 doOpenChoice(const char *path, const char *type, const char *name,
1181 UDataMemoryIsAcceptable *isAcceptable, void *context,
1182 UErrorCode *pErrorCode)
1183 {
1184 UDataMemory *retVal = nullptr;
1185
1186 const char *dataPath;
1187
1188 int32_t tocEntrySuffixIndex;
1189 const char *tocEntryPathSuffix;
1190 UErrorCode subErrorCode=U_ZERO_ERROR;
1191 const char *treeChar;
1192
1193 UBool isICUData = false;
1194
1195
1196 FileTracer::traceOpen(path, type, name);
1197
1198
1199 /* Is this path ICU data? */
1200 if(path == nullptr ||
1201 !strcmp(path, U_ICUDATA_ALIAS) || /* "ICUDATA" */
1202 !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-" */
1203 uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) ||
1204 !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-" */
1205 uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) {
1206 isICUData = true;
1207 }
1208
1209 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR) /* Windows: try "foo\bar" and "foo/bar" */
1210 /* remap from alternate path char to the main one */
1211 CharString altSepPath;
1212 if(path) {
1213 if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != nullptr) {
1214 altSepPath.append(path, *pErrorCode);
1215 char *p;
1216 while ((p = uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR)) != nullptr) {
1217 *p = U_FILE_SEP_CHAR;
1218 }
1219 #if defined (UDATA_DEBUG)
1220 fprintf(stderr, "Changed path from [%s] to [%s]\n", path, altSepPath.data());
1221 #endif
1222 path = altSepPath.data();
1223 }
1224 }
1225 #endif
1226
1227 CharString tocEntryName; /* entry name in tree format. ex: 'icudt28b/coll/ar.res' */
1228 CharString tocEntryPath; /* entry name in path format. ex: 'icudt28b\\coll\\ar.res' */
1229
1230 CharString pkgName;
1231 CharString treeName;
1232
1233 /* ======= Set up strings */
1234 if(path==nullptr) {
1235 pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1236 } else {
1237 const char *pkg;
1238 const char *first;
1239 pkg = uprv_strrchr(path, U_FILE_SEP_CHAR);
1240 first = uprv_strchr(path, U_FILE_SEP_CHAR);
1241 if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash in the path- not a tree name */
1242 /* see if this is an /absolute/path/to/package path */
1243 if(pkg) {
1244 pkgName.append(pkg+1, *pErrorCode);
1245 } else {
1246 pkgName.append(path, *pErrorCode);
1247 }
1248 } else {
1249 treeChar = uprv_strchr(path, U_TREE_SEPARATOR);
1250 if(treeChar) {
1251 treeName.append(treeChar+1, *pErrorCode); /* following '-' */
1252 if(isICUData) {
1253 pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1254 } else {
1255 pkgName.append(path, static_cast<int32_t>(treeChar - path), *pErrorCode);
1256 if (first == nullptr) {
1257 /*
1258 This user data has no path, but there is a tree name.
1259 Look up the correct path from the data cache later.
1260 */
1261 path = pkgName.data();
1262 }
1263 }
1264 } else {
1265 if(isICUData) {
1266 pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1267 } else {
1268 pkgName.append(path, *pErrorCode);
1269 }
1270 }
1271 }
1272 }
1273
1274 #ifdef UDATA_DEBUG
1275 fprintf(stderr, " P=%s T=%s\n", pkgName.data(), treeName.data());
1276 #endif
1277
1278 /* setting up the entry name and file name
1279 * Make up a full name by appending the type to the supplied
1280 * name, assuming that a type was supplied.
1281 */
1282
1283 /* prepend the package */
1284 tocEntryName.append(pkgName, *pErrorCode);
1285 tocEntryPath.append(pkgName, *pErrorCode);
1286 tocEntrySuffixIndex = tocEntryName.length();
1287
1288 if(!treeName.isEmpty()) {
1289 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode);
1290 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode);
1291 }
1292
1293 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pErrorCode);
1294 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode);
1295 if(type!=nullptr && *type!=0) {
1296 tocEntryName.append(".", *pErrorCode).append(type, *pErrorCode);
1297 tocEntryPath.append(".", *pErrorCode).append(type, *pErrorCode);
1298 }
1299 // The +1 is for the U_FILE_SEP_CHAR that is always appended above.
1300 tocEntryPathSuffix = tocEntryPath.data() + tocEntrySuffixIndex + 1; /* suffix starts here */
1301
1302 #ifdef UDATA_DEBUG
1303 fprintf(stderr, " tocEntryName = %s\n", tocEntryName.data());
1304 fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data());
1305 #endif
1306
1307 #if !defined(ICU_DATA_DIR_WINDOWS)
1308 if(path == nullptr) {
1309 path = COMMON_DATA_NAME; /* "icudt26e" */
1310 }
1311 #else
1312 // When using the Windows system data, we expects only a single data file.
1313 path = COMMON_DATA_NAME; /* "icudt26e" */
1314 #endif
1315
1316 /************************ Begin loop looking for ind. files ***************/
1317 #ifdef UDATA_DEBUG
1318 fprintf(stderr, "IND: inBasename = %s, pkg=%s\n", "(n/a)", packageNameFromPath(path));
1319 #endif
1320
1321 /* End of dealing with a null basename */
1322 dataPath = u_getDataDirectory();
1323
1324 /**** Time zone individual files override */
1325 if (isICUData && isTimeZoneFile(name, type)) {
1326 const char *tzFilesDir = u_getTimeZoneFilesDirectory(pErrorCode);
1327 if (tzFilesDir[0] != 0) {
1328 #ifdef UDATA_DEBUG
1329 fprintf(stderr, "Trying Time Zone Files directory = %s\n", tzFilesDir);
1330 #endif
1331 retVal = doLoadFromIndividualFiles(/* pkgName.data() */ "", tzFilesDir, tocEntryPathSuffix,
1332 /* path */ "", type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1333 if((retVal != nullptr) || U_FAILURE(*pErrorCode)) {
1334 return retVal;
1335 }
1336 }
1337 }
1338
1339 /**** COMMON PACKAGE - only if packages are first. */
1340 if(gDataFileAccess == UDATA_PACKAGES_FIRST) {
1341 #ifdef UDATA_DEBUG
1342 fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n");
1343 #endif
1344 /* #2 */
1345 retVal = doLoadFromCommonData(isICUData,
1346 pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(),
1347 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1348 if((retVal != nullptr) || U_FAILURE(*pErrorCode)) {
1349 return retVal;
1350 }
1351 }
1352
1353 /**** INDIVIDUAL FILES */
1354 if((gDataFileAccess==UDATA_PACKAGES_FIRST) ||
1355 (gDataFileAccess==UDATA_FILES_FIRST)) {
1356 #ifdef UDATA_DEBUG
1357 fprintf(stderr, "Trying individual files\n");
1358 #endif
1359 /* Check to make sure that there is a dataPath to iterate over */
1360 if ((dataPath && *dataPath) || !isICUData) {
1361 retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix,
1362 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1363 if((retVal != nullptr) || U_FAILURE(*pErrorCode)) {
1364 return retVal;
1365 }
1366 }
1367 }
1368
1369 /**** COMMON PACKAGE */
1370 if((gDataFileAccess==UDATA_ONLY_PACKAGES) ||
1371 (gDataFileAccess==UDATA_FILES_FIRST)) {
1372 #ifdef UDATA_DEBUG
1373 fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n");
1374 #endif
1375 retVal = doLoadFromCommonData(isICUData,
1376 pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(),
1377 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1378 if((retVal != nullptr) || U_FAILURE(*pErrorCode)) {
1379 return retVal;
1380 }
1381 }
1382
1383 /* Load from DLL. If we haven't attempted package load, we also haven't had any chance to
1384 try a DLL (static or setCommonData/etc) load.
1385 If we ever have a "UDATA_ONLY_FILES", add it to the or list here. */
1386 if(gDataFileAccess==UDATA_NO_FILES) {
1387 #ifdef UDATA_DEBUG
1388 fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n");
1389 #endif
1390 retVal = doLoadFromCommonData(isICUData,
1391 pkgName.data(), "", tocEntryPathSuffix, tocEntryName.data(),
1392 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1393 if((retVal != nullptr) || U_FAILURE(*pErrorCode)) {
1394 return retVal;
1395 }
1396 }
1397
1398 /* data not found */
1399 if(U_SUCCESS(*pErrorCode)) {
1400 if(U_SUCCESS(subErrorCode)) {
1401 /* file not found */
1402 *pErrorCode=U_FILE_ACCESS_ERROR;
1403 } else {
1404 /* entry point not found or rejected */
1405 *pErrorCode=subErrorCode;
1406 }
1407 }
1408 return retVal;
1409 }
1410
1411
1412
1413 /* API ---------------------------------------------------------------------- */
1414
1415 U_CAPI UDataMemory * U_EXPORT2
udata_open(const char * path,const char * type,const char * name,UErrorCode * pErrorCode)1416 udata_open(const char *path, const char *type, const char *name,
1417 UErrorCode *pErrorCode) {
1418 #ifdef UDATA_DEBUG
1419 fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n", (path?path:"nullptr"), name, type);
1420 fflush(stderr);
1421 #endif
1422
1423 if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) {
1424 return nullptr;
1425 } else if(name==nullptr || *name==0) {
1426 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1427 return nullptr;
1428 } else {
1429 return doOpenChoice(path, type, name, nullptr, nullptr, pErrorCode);
1430 }
1431 }
1432
1433
1434
1435 U_CAPI UDataMemory * U_EXPORT2
udata_openChoice(const char * path,const char * type,const char * name,UDataMemoryIsAcceptable * isAcceptable,void * context,UErrorCode * pErrorCode)1436 udata_openChoice(const char *path, const char *type, const char *name,
1437 UDataMemoryIsAcceptable *isAcceptable, void *context,
1438 UErrorCode *pErrorCode) {
1439 #ifdef UDATA_DEBUG
1440 fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n", (path?path:"nullptr"), name, type);
1441 #endif
1442
1443 if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) {
1444 return nullptr;
1445 } else if(name==nullptr || *name==0 || isAcceptable==nullptr) {
1446 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1447 return nullptr;
1448 } else {
1449 return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode);
1450 }
1451 }
1452
1453
1454
1455 U_CAPI void U_EXPORT2
udata_getInfo(UDataMemory * pData,UDataInfo * pInfo)1456 udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) {
1457 if(pInfo!=nullptr) {
1458 if(pData!=nullptr && pData->pHeader!=nullptr) {
1459 const UDataInfo *info=&pData->pHeader->info;
1460 uint16_t dataInfoSize=udata_getInfoSize(info);
1461 if(pInfo->size>dataInfoSize) {
1462 pInfo->size=dataInfoSize;
1463 }
1464 uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->size-2);
1465 if(info->isBigEndian!=U_IS_BIG_ENDIAN) {
1466 /* opposite endianness */
1467 uint16_t x=info->reservedWord;
1468 pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8));
1469 }
1470 } else {
1471 pInfo->size=0;
1472 }
1473 }
1474 }
1475
1476
udata_setFileAccess(UDataFileAccess access,UErrorCode *)1477 U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/)
1478 {
1479 // Note: this function is documented as not thread safe.
1480 gDataFileAccess = access;
1481 }
1482