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 "uassert.h"
37 #include "ucln_cmn.h"
38 #include "ucmndata.h"
39 #include "udatamem.h"
40 #include "uhash.h"
41 #include "umapfile.h"
42 #include "umutex.h"
43 
44 /***********************************************************************
45 *
46 *   Notes on the organization of the ICU data implementation
47 *
48 *      All of the public API is defined in udata.h
49 *
50 *      The implementation is split into several files...
51 *
52 *         - udata.c  (this file) contains higher level code that knows about
53 *                     the search paths for locating data, caching opened data, etc.
54 *
55 *         - umapfile.c  contains the low level platform-specific code for actually loading
56 *                     (memory mapping, file reading, whatever) data into memory.
57 *
58 *         - ucmndata.c  deals with the tables of contents of ICU data items within
59 *                     an ICU common format data file.  The implementation includes
60 *                     an abstract interface and support for multiple TOC formats.
61 *                     All knowledge of any specific TOC format is encapsulated here.
62 *
63 *         - udatamem.c has code for managing UDataMemory structs.  These are little
64 *                     descriptor objects for blocks of memory holding ICU data of
65 *                     various types.
66 */
67 
68 /* configuration ---------------------------------------------------------- */
69 
70 /* If you are excruciatingly bored turn this on .. */
71 /* #define UDATA_DEBUG 1 */
72 
73 #if defined(UDATA_DEBUG)
74 #   include <stdio.h>
75 #endif
76 
77 U_NAMESPACE_USE
78 
79 /*
80  *  Forward declarations
81  */
82 static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err);
83 
84 /***********************************************************************
85 *
86 *    static (Global) data
87 *
88 ************************************************************************/
89 
90 /*
91  * Pointers to the common ICU data.
92  *
93  * We store multiple pointers to ICU data packages and iterate through them
94  * when looking for a data item.
95  *
96  * It is possible to combine this with dependency inversion:
97  * One or more data package libraries may export
98  * functions that each return a pointer to their piece of the ICU data,
99  * and this file would import them as weak functions, without a
100  * strong linker dependency from the common library on the data library.
101  *
102  * Then we can have applications depend on only that part of ICU's data
103  * that they really need, reducing the size of binaries that take advantage
104  * of this.
105  */
106 static UDataMemory *gCommonICUDataArray[10] = { NULL };   // Access protected by icu global mutex.
107 
108 static u_atomic_int32_t gHaveTriedToLoadCommonData = ATOMIC_INT32_T_INITIALIZER(0);  //  See extendICUData().
109 
110 static UHashtable  *gCommonDataCache = NULL;  /* Global hash table of opened ICU data files.  */
111 static icu::UInitOnce gCommonDataCacheInitOnce = U_INITONCE_INITIALIZER;
112 
113 #if U_PLATFORM_HAS_WINUWP_API == 0
114 static UDataFileAccess  gDataFileAccess = UDATA_DEFAULT_ACCESS;  // Access not synchronized.
115                                                                  // Modifying is documented as thread-unsafe.
116 #else
117 static UDataFileAccess  gDataFileAccess = UDATA_NO_FILES;        // Windows UWP looks in one spot explicitly
118 #endif
119 
120 static UBool U_CALLCONV
udata_cleanup(void)121 udata_cleanup(void)
122 {
123     int32_t i;
124 
125     if (gCommonDataCache) {             /* Delete the cache of user data mappings.  */
126         uhash_close(gCommonDataCache);  /*   Table owns the contents, and will delete them. */
127         gCommonDataCache = NULL;        /*   Cleanup is not thread safe.                */
128     }
129     gCommonDataCacheInitOnce.reset();
130 
131     for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != NULL; ++i) {
132         udata_close(gCommonICUDataArray[i]);
133         gCommonICUDataArray[i] = NULL;
134     }
135     gHaveTriedToLoadCommonData = 0;
136 
137     return TRUE;                   /* Everything was cleaned up */
138 }
139 
140 static UBool U_CALLCONV
findCommonICUDataByName(const char * inBasename,UErrorCode & err)141 findCommonICUDataByName(const char *inBasename, UErrorCode &err)
142 {
143     UBool found = FALSE;
144     int32_t i;
145 
146     UDataMemory  *pData = udata_findCachedData(inBasename, err);
147     if (U_FAILURE(err) || pData == NULL)
148         return FALSE;
149 
150     {
151         Mutex lock;
152         for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) {
153             if ((gCommonICUDataArray[i] != NULL) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) {
154                 /* The data pointer is already in the array. */
155                 found = TRUE;
156                 break;
157             }
158         }
159     }
160     return found;
161 }
162 
163 
164 /*
165  * setCommonICUData.   Set a UDataMemory to be the global ICU Data
166  */
167 static UBool
setCommonICUData(UDataMemory * pData,UBool warn,UErrorCode * pErr)168 setCommonICUData(UDataMemory *pData,     /*  The new common data.  Belongs to caller, we copy it. */
169                  UBool       warn,       /*  If true, set USING_DEFAULT warning if ICUData was    */
170                                          /*    changed by another thread before we got to it.     */
171                  UErrorCode *pErr)
172 {
173     UDataMemory  *newCommonData = UDataMemory_createNewInstance(pErr);
174     int32_t i;
175     UBool didUpdate = FALSE;
176     if (U_FAILURE(*pErr)) {
177         return FALSE;
178     }
179 
180     /*  For the assignment, other threads must cleanly see either the old            */
181     /*    or the new, not some partially initialized new.  The old can not be        */
182     /*    deleted - someone may still have a pointer to it lying around in           */
183     /*    their locals.                                                              */
184     UDatamemory_assign(newCommonData, pData);
185     umtx_lock(NULL);
186     for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) {
187         if (gCommonICUDataArray[i] == NULL) {
188             gCommonICUDataArray[i] = newCommonData;
189             didUpdate = TRUE;
190             break;
191         } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) {
192             /* The same data pointer is already in the array. */
193             break;
194         }
195     }
196     umtx_unlock(NULL);
197 
198     if (i == UPRV_LENGTHOF(gCommonICUDataArray) && warn) {
199         *pErr = U_USING_DEFAULT_WARNING;
200     }
201     if (didUpdate) {
202         ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup);
203     } else {
204         uprv_free(newCommonData);
205     }
206     return didUpdate;
207 }
208 
209 #if U_PLATFORM_HAS_WINUWP_API == 0
210 
211 static UBool
setCommonICUDataPointer(const void * pData,UBool,UErrorCode * pErrorCode)212 setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) {
213     UDataMemory tData;
214     UDataMemory_init(&tData);
215     UDataMemory_setData(&tData, pData);
216     udata_checkCommonData(&tData, pErrorCode);
217     return setCommonICUData(&tData, FALSE, pErrorCode);
218 }
219 
220 #endif
221 
222 static const char *
findBasename(const char * path)223 findBasename(const char *path) {
224     const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR);
225     if(basename==NULL) {
226         return path;
227     } else {
228         return basename+1;
229     }
230 }
231 
232 #ifdef UDATA_DEBUG
233 static const char *
packageNameFromPath(const char * path)234 packageNameFromPath(const char *path)
235 {
236     if((path == NULL) || (*path == 0)) {
237         return U_ICUDATA_NAME;
238     }
239 
240     path = findBasename(path);
241 
242     if((path == NULL) || (*path == 0)) {
243         return U_ICUDATA_NAME;
244     }
245 
246     return path;
247 }
248 #endif
249 
250 /*----------------------------------------------------------------------*
251  *                                                                      *
252  *   Cache for common data                                              *
253  *      Functions for looking up or adding entries to a cache of        *
254  *      data that has been previously opened.  Avoids a potentially     *
255  *      expensive operation of re-opening the data for subsequent       *
256  *      uses.                                                           *
257  *                                                                      *
258  *      Data remains cached for the duration of the process.            *
259  *                                                                      *
260  *----------------------------------------------------------------------*/
261 
262 typedef struct DataCacheElement {
263     char          *name;
264     UDataMemory   *item;
265 } DataCacheElement;
266 
267 
268 
269 /*
270  * Deleter function for DataCacheElements.
271  *         udata cleanup function closes the hash table; hash table in turn calls back to
272  *         here for each entry.
273  */
DataCacheElement_deleter(void * pDCEl)274 static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) {
275     DataCacheElement *p = (DataCacheElement *)pDCEl;
276     udata_close(p->item);              /* unmaps storage */
277     uprv_free(p->name);                /* delete the hash key string. */
278     uprv_free(pDCEl);                  /* delete 'this'          */
279 }
280 
udata_initHashTable(UErrorCode & err)281 static void U_CALLCONV udata_initHashTable(UErrorCode &err) {
282     U_ASSERT(gCommonDataCache == NULL);
283     gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err);
284     if (U_FAILURE(err)) {
285        return;
286     }
287     U_ASSERT(gCommonDataCache != NULL);
288     uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter);
289     ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup);
290 }
291 
292  /*   udata_getCacheHashTable()
293   *     Get the hash table used to store the data cache entries.
294   *     Lazy create it if it doesn't yet exist.
295   */
udata_getHashTable(UErrorCode & err)296 static UHashtable *udata_getHashTable(UErrorCode &err) {
297     umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable, err);
298     return gCommonDataCache;
299 }
300 
301 
302 
udata_findCachedData(const char * path,UErrorCode & err)303 static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err)
304 {
305     UHashtable        *htable;
306     UDataMemory       *retVal = NULL;
307     DataCacheElement  *el;
308     const char        *baseName;
309 
310     htable = udata_getHashTable(err);
311     if (U_FAILURE(err)) {
312         return NULL;
313     }
314 
315     baseName = findBasename(path);   /* Cache remembers only the base name, not the full path. */
316     umtx_lock(NULL);
317     el = (DataCacheElement *)uhash_get(htable, baseName);
318     umtx_unlock(NULL);
319     if (el != NULL) {
320         retVal = el->item;
321     }
322 #ifdef UDATA_DEBUG
323     fprintf(stderr, "Cache: [%s] -> %p\n", baseName, retVal);
324 #endif
325     return retVal;
326 }
327 
328 
udata_cacheDataItem(const char * path,UDataMemory * item,UErrorCode * pErr)329 static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UErrorCode *pErr) {
330     DataCacheElement *newElement;
331     const char       *baseName;
332     int32_t           nameLen;
333     UHashtable       *htable;
334     DataCacheElement *oldValue = NULL;
335     UErrorCode        subErr = U_ZERO_ERROR;
336 
337     htable = udata_getHashTable(*pErr);
338     if (U_FAILURE(*pErr)) {
339         return NULL;
340     }
341 
342     /* Create a new DataCacheElement - the thingy we store in the hash table -
343      * and copy the supplied path and UDataMemoryItems into it.
344      */
345     newElement = (DataCacheElement *)uprv_malloc(sizeof(DataCacheElement));
346     if (newElement == NULL) {
347         *pErr = U_MEMORY_ALLOCATION_ERROR;
348         return NULL;
349     }
350     newElement->item = UDataMemory_createNewInstance(pErr);
351     if (U_FAILURE(*pErr)) {
352         uprv_free(newElement);
353         return NULL;
354     }
355     UDatamemory_assign(newElement->item, item);
356 
357     baseName = findBasename(path);
358     nameLen = (int32_t)uprv_strlen(baseName);
359     newElement->name = (char *)uprv_malloc(nameLen+1);
360     if (newElement->name == NULL) {
361         *pErr = U_MEMORY_ALLOCATION_ERROR;
362         uprv_free(newElement->item);
363         uprv_free(newElement);
364         return NULL;
365     }
366     uprv_strcpy(newElement->name, baseName);
367 
368     /* Stick the new DataCacheElement into the hash table.
369     */
370     umtx_lock(NULL);
371     oldValue = (DataCacheElement *)uhash_get(htable, path);
372     if (oldValue != NULL) {
373         subErr = U_USING_DEFAULT_WARNING;
374     }
375     else {
376         uhash_put(
377             htable,
378             newElement->name,               /* Key   */
379             newElement,                     /* Value */
380             &subErr);
381     }
382     umtx_unlock(NULL);
383 
384 #ifdef UDATA_DEBUG
385     fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n", newElement->name,
386     newElement->item, u_errorName(subErr), newElement->item->vFuncs);
387 #endif
388 
389     if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) {
390         *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. */
391         uprv_free(newElement->name);
392         uprv_free(newElement->item);
393         uprv_free(newElement);
394         return oldValue ? oldValue->item : NULL;
395     }
396 
397     return newElement->item;
398 }
399 
400 /*----------------------------------------------------------------------*==============
401  *                                                                      *
402  *  Path management.  Could be shared with other tools/etc if need be   *
403  * later on.                                                            *
404  *                                                                      *
405  *----------------------------------------------------------------------*/
406 
407 U_NAMESPACE_BEGIN
408 
409 class UDataPathIterator
410 {
411 public:
412     UDataPathIterator(const char *path, const char *pkg,
413                       const char *item, const char *suffix, UBool doCheckLastFour,
414                       UErrorCode *pErrorCode);
415     const char *next(UErrorCode *pErrorCode);
416 
417 private:
418     const char *path;                              /* working path (u_icudata_Dir) */
419     const char *nextPath;                          /* path following this one */
420     const char *basename;                          /* item's basename (icudt22e_mt.res)*/
421     const char *suffix;                            /* item suffix (can be null) */
422 
423     uint32_t    basenameLen;                       /* length of basename */
424 
425     CharString  itemPath;                          /* path passed in with item name */
426     CharString  pathBuffer;                        /* output path for this it'ion */
427     CharString  packageStub;                       /* example:  "/icudt28b". Will ignore that leaf in set paths. */
428 
429     UBool       checkLastFour;                     /* if TRUE then allow paths such as '/foo/myapp.dat'
430                                                     * to match, checks last 4 chars of suffix with
431                                                     * last 4 of path, then previous chars. */
432 };
433 
434 /**
435  * @param iter  The iterator to be initialized. Its current state does not matter.
436  * @param path  The full pathname to be iterated over.  If NULL, defaults to U_ICUDATA_NAME
437  * @param pkg   Package which is being searched for, ex "icudt28l".  Will ignore leave directories such as /icudt28l
438  * @param item  Item to be searched for.  Can include full path, such as /a/b/foo.dat
439  * @param suffix  Optional item suffix, if not-null (ex. ".dat") then 'path' can contain 'item' explicitly.
440  *               Ex:   'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar/baz' as item #2.
441  *                     '/blarg/stuff.dat' would also be found.
442  */
UDataPathIterator(const char * inPath,const char * pkg,const char * item,const char * inSuffix,UBool doCheckLastFour,UErrorCode * pErrorCode)443 UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg,
444                                      const char *item, const char *inSuffix, UBool doCheckLastFour,
445                                      UErrorCode *pErrorCode)
446 {
447 #ifdef UDATA_DEBUG
448         fprintf(stderr, "SUFFIX1=%s PATH=%s\n", inSuffix, inPath);
449 #endif
450     /** Path **/
451     if(inPath == NULL) {
452         path = u_getDataDirectory();
453     } else {
454         path = inPath;
455     }
456 
457     /** Package **/
458     if(pkg != NULL) {
459       packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode);
460 #ifdef UDATA_DEBUG
461       fprintf(stderr, "STUB=%s [%d]\n", packageStub.data(), packageStub.length());
462 #endif
463     }
464 
465     /** Item **/
466     basename = findBasename(item);
467     basenameLen = (int32_t)uprv_strlen(basename);
468 
469     /** Item path **/
470     if(basename == item) {
471         nextPath = path;
472     } else {
473         itemPath.append(item, (int32_t)(basename-item), *pErrorCode);
474         nextPath = itemPath.data();
475     }
476 #ifdef UDATA_DEBUG
477     fprintf(stderr, "SUFFIX=%s [%p]\n", inSuffix, inSuffix);
478 #endif
479 
480     /** Suffix  **/
481     if(inSuffix != NULL) {
482         suffix = inSuffix;
483     } else {
484         suffix = "";
485     }
486 
487     checkLastFour = doCheckLastFour;
488 
489     /* pathBuffer will hold the output path strings returned by this iterator */
490 
491 #ifdef UDATA_DEBUG
492     fprintf(stderr, "%p: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%s], [nextpath=%s], [checklast4=%s]\n",
493             iter,
494             item,
495             path,
496             basename,
497             suffix,
498             itemPath.data(),
499             nextPath,
500             checkLastFour?"TRUE":"false");
501 #endif
502 }
503 
504 /**
505  * Get the next path on the list.
506  *
507  * @param iter The Iter to be used
508  * @param len  If set, pointer to the length of the returned path, for convenience.
509  * @return Pointer to the next path segment, or NULL if there are no more.
510  */
next(UErrorCode * pErrorCode)511 const char *UDataPathIterator::next(UErrorCode *pErrorCode)
512 {
513     if(U_FAILURE(*pErrorCode)) {
514         return NULL;
515     }
516 
517     const char *currentPath = NULL;
518     int32_t     pathLen = 0;
519     const char *pathBasename;
520 
521     do
522     {
523         if( nextPath == NULL ) {
524             break;
525         }
526         currentPath = nextPath;
527 
528         if(nextPath == itemPath.data()) { /* we were processing item's path. */
529             nextPath = path; /* start with regular path next tm. */
530             pathLen = (int32_t)uprv_strlen(currentPath);
531         } else {
532             /* fix up next for next time */
533             nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR);
534             if(nextPath == NULL) {
535                 /* segment: entire path */
536                 pathLen = (int32_t)uprv_strlen(currentPath);
537             } else {
538                 /* segment: until next segment */
539                 pathLen = (int32_t)(nextPath - currentPath);
540                 /* skip divider */
541                 nextPath ++;
542             }
543         }
544 
545         if(pathLen == 0) {
546             continue;
547         }
548 
549 #ifdef UDATA_DEBUG
550         fprintf(stderr, "rest of path (IDD) = %s\n", currentPath);
551         fprintf(stderr, "                     ");
552         {
553             uint32_t qqq;
554             for(qqq=0;qqq<pathLen;qqq++)
555             {
556                 fprintf(stderr, " ");
557             }
558 
559             fprintf(stderr, "^\n");
560         }
561 #endif
562         pathBuffer.clear().append(currentPath, pathLen, *pErrorCode);
563 
564         /* check for .dat files */
565         pathBasename = findBasename(pathBuffer.data());
566 
567         if(checkLastFour == TRUE &&
568            (pathLen>=4) &&
569            uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix, 4)==0 && /* suffix matches */
570            uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)==0  && /* base matches */
571            uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full len */
572 
573 #ifdef UDATA_DEBUG
574             fprintf(stderr, "Have %s file on the path: %s\n", suffix, pathBuffer.data());
575 #endif
576             /* do nothing */
577         }
578         else
579         {       /* regular dir path */
580             if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) {
581                 if((pathLen>=4) &&
582                    uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat", 4) == 0)
583                 {
584 #ifdef UDATA_DEBUG
585                     fprintf(stderr, "skipping non-directory .dat file %s\n", pathBuffer.data());
586 #endif
587                     continue;
588                 }
589 
590                 /* Check if it is a directory with the same name as our package */
591                 if(!packageStub.isEmpty() &&
592                    (pathLen > packageStub.length()) &&
593                    !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length(), packageStub.data())) {
594 #ifdef UDATA_DEBUG
595                   fprintf(stderr, "Found stub %s (will add package %s of len %d)\n", packageStub.data(), basename, basenameLen);
596 #endif
597                   pathBuffer.truncate(pathLen - packageStub.length());
598                 }
599                 pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode);
600             }
601 
602             /* + basename */
603             pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pErrorCode);
604 
605             if(*suffix)  /* tack on suffix */
606             {
607                 pathBuffer.append(suffix, *pErrorCode);
608             }
609         }
610 
611 #ifdef UDATA_DEBUG
612         fprintf(stderr, " -->  %s\n", pathBuffer.data());
613 #endif
614 
615         return pathBuffer.data();
616 
617     } while(path);
618 
619     /* fell way off the end */
620     return NULL;
621 }
622 
623 U_NAMESPACE_END
624 
625 /* ==================================================================================*/
626 
627 
628 /*----------------------------------------------------------------------*
629  *                                                                      *
630  *  Add a static reference to the common data library                   *
631  *   Unless overridden by an explicit udata_setCommonData, this will be *
632  *      our common data.                                                *
633  *                                                                      *
634  *----------------------------------------------------------------------*/
635 #if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time
636 extern "C" const DataHeader U_DATA_API U_ICUDATA_ENTRY_POINT;
637 #endif
638 
639 /*
640  * This would be a good place for weak-linkage declarations of
641  * partial-data-library access functions where each returns a pointer
642  * to its data package, if it is linked in.
643  */
644 /*
645 extern const void *uprv_getICUData_collation(void) ATTRIBUTE_WEAK;
646 extern const void *uprv_getICUData_conversion(void) ATTRIBUTE_WEAK;
647 */
648 
649 /*----------------------------------------------------------------------*
650  *                                                                      *
651  *   openCommonData   Attempt to open a common format (.dat) file       *
652  *                    Map it into memory (if it's not there already)    *
653  *                    and return a UDataMemory object for it.           *
654  *                                                                      *
655  *                    If the requested data is already open and cached  *
656  *                       just return the cached UDataMem object.        *
657  *                                                                      *
658  *----------------------------------------------------------------------*/
659 static UDataMemory *
openCommonData(const char * path,int32_t commonDataIndex,UErrorCode * pErrorCode)660 openCommonData(const char *path,          /*  Path from OpenChoice?          */
661                int32_t commonDataIndex,   /*  ICU Data (index >= 0) if path == NULL */
662                UErrorCode *pErrorCode)
663 {
664     UDataMemory tData;
665     const char *pathBuffer;
666     const char *inBasename;
667 
668     if (U_FAILURE(*pErrorCode)) {
669         return NULL;
670     }
671 
672     UDataMemory_init(&tData);
673 
674     /* ??????? TODO revisit this */
675     if (commonDataIndex >= 0) {
676         /* "mini-cache" for common ICU data */
677         if(commonDataIndex >= UPRV_LENGTHOF(gCommonICUDataArray)) {
678             return NULL;
679         }
680         {
681             Mutex lock;
682             if(gCommonICUDataArray[commonDataIndex] != NULL) {
683                 return gCommonICUDataArray[commonDataIndex];
684             }
685 #if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time
686             int32_t i;
687             for(i = 0; i < commonDataIndex; ++i) {
688                 if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT) {
689                     /* The linked-in data is already in the list. */
690                     return NULL;
691                 }
692             }
693 #endif
694         }
695 
696         /* Add the linked-in data to the list. */
697         /*
698          * This is where we would check and call weakly linked partial-data-library
699          * access functions.
700          */
701         /*
702         if (uprv_getICUData_collation) {
703             setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErrorCode);
704         }
705         if (uprv_getICUData_conversion) {
706             setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pErrorCode);
707         }
708         */
709 #if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time
710         setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, FALSE, pErrorCode);
711         {
712             Mutex lock;
713             return gCommonICUDataArray[commonDataIndex];
714         }
715 #endif
716     }
717 
718 
719     /* request is NOT for ICU Data.  */
720 
721     /* Find the base name portion of the supplied path.   */
722     /*   inBasename will be left pointing somewhere within the original path string.      */
723     inBasename = findBasename(path);
724 #ifdef UDATA_DEBUG
725     fprintf(stderr, "inBasename = %s\n", inBasename);
726 #endif
727 
728     if(*inBasename==0) {
729         /* no basename.     This will happen if the original path was a directory name,   */
730         /*    like  "a/b/c/".   (Fallback to separate files will still work.)             */
731 #ifdef UDATA_DEBUG
732         fprintf(stderr, "ocd: no basename in %s, bailing.\n", path);
733 #endif
734         if (U_SUCCESS(*pErrorCode)) {
735             *pErrorCode=U_FILE_ACCESS_ERROR;
736         }
737         return NULL;
738     }
739 
740    /* Is the requested common data file already open and cached?                     */
741    /*   Note that the cache is keyed by the base name only.  The rest of the path,   */
742    /*     if any, is not considered.                                                 */
743     UDataMemory  *dataToReturn = udata_findCachedData(inBasename, *pErrorCode);
744     if (dataToReturn != NULL || U_FAILURE(*pErrorCode)) {
745         return dataToReturn;
746     }
747 
748     /* Requested item is not in the cache.
749      * Hunt it down, trying all the path locations
750      */
751 
752     UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat", TRUE, pErrorCode);
753 
754     while((UDataMemory_isLoaded(&tData)==FALSE) && (pathBuffer = iter.next(pErrorCode)) != NULL)
755     {
756 #ifdef UDATA_DEBUG
757         fprintf(stderr, "ocd: trying path %s - ", pathBuffer);
758 #endif
759         uprv_mapFile(&tData, pathBuffer);
760 #ifdef UDATA_DEBUG
761         fprintf(stderr, "%s\n", UDataMemory_isLoaded(&tData)?"LOADED":"not loaded");
762 #endif
763     }
764 
765 #if defined(OS390_STUBDATA) && defined(OS390BATCH)
766     if (!UDataMemory_isLoaded(&tData)) {
767         char ourPathBuffer[1024];
768         /* One more chance, for extendCommonData() */
769         uprv_strncpy(ourPathBuffer, path, 1019);
770         ourPathBuffer[1019]=0;
771         uprv_strcat(ourPathBuffer, ".dat");
772         uprv_mapFile(&tData, ourPathBuffer);
773     }
774 #endif
775 
776     if (U_FAILURE(*pErrorCode)) {
777         return NULL;
778     }
779     if (!UDataMemory_isLoaded(&tData)) {
780         /* no common data */
781         *pErrorCode=U_FILE_ACCESS_ERROR;
782         return NULL;
783     }
784 
785     /* we have mapped a file, check its header */
786     udata_checkCommonData(&tData, pErrorCode);
787 
788 
789     /* Cache the UDataMemory struct for this .dat file,
790      *   so we won't need to hunt it down and map it again next time
791      *   something is needed from it.                */
792     return udata_cacheDataItem(inBasename, &tData, pErrorCode);
793 }
794 
795 
796 /*----------------------------------------------------------------------*
797  *                                                                      *
798  *   extendICUData   If the full set of ICU data was not loaded at      *
799  *                   program startup, load it now.  This function will  *
800  *                   be called when the lookup of an ICU data item in   *
801  *                   the common ICU data fails.                         *
802  *                                                                      *
803  *                   return true if new data is loaded, false otherwise.*
804  *                                                                      *
805  *----------------------------------------------------------------------*/
extendICUData(UErrorCode * pErr)806 static UBool extendICUData(UErrorCode *pErr)
807 {
808     UDataMemory   *pData;
809     UDataMemory   copyPData;
810     UBool         didUpdate = FALSE;
811 
812     /*
813      * There is a chance for a race condition here.
814      * Normally, ICU data is loaded from a DLL or via mmap() and
815      * setCommonICUData() will detect if the same address is set twice.
816      * If ICU is built with data loading via fread() then the address will
817      * be different each time the common data is loaded and we may add
818      * multiple copies of the data.
819      * In this case, use a mutex to prevent the race.
820      * Use a specific mutex to avoid nested locks of the global mutex.
821      */
822 #if MAP_IMPLEMENTATION==MAP_STDIO
823     static UMutex extendICUDataMutex = U_MUTEX_INITIALIZER;
824     umtx_lock(&extendICUDataMutex);
825 #endif
826     if(!umtx_loadAcquire(gHaveTriedToLoadCommonData)) {
827         /* See if we can explicitly open a .dat file for the ICUData. */
828         pData = openCommonData(
829                    U_ICUDATA_NAME,            /*  "icudt20l" , for example.          */
830                    -1,                        /*  Pretend we're not opening ICUData  */
831                    pErr);
832 
833         /* How about if there is no pData, eh... */
834 
835        UDataMemory_init(&copyPData);
836        if(pData != NULL) {
837           UDatamemory_assign(&copyPData, pData);
838           copyPData.map = 0;              /* The mapping for this data is owned by the hash table */
839           copyPData.mapAddr = 0;          /*   which will unmap it when ICU is shut down.         */
840                                           /* CommonICUData is also unmapped when ICU is shut down.*/
841                                           /* To avoid unmapping the data twice, zero out the map  */
842                                           /*   fields in the UDataMemory that we're assigning     */
843                                           /*   to CommonICUData.                                  */
844 
845           didUpdate = /* no longer using this result */
846               setCommonICUData(&copyPData,/*  The new common data.                                */
847                        FALSE,             /*  No warnings if write didn't happen                  */
848                        pErr);             /*  setCommonICUData honors errors; NOP if error set    */
849         }
850 
851         umtx_storeRelease(gHaveTriedToLoadCommonData, 1);
852     }
853 
854     didUpdate = findCommonICUDataByName(U_ICUDATA_NAME, *pErr);  /* Return 'true' when a racing writes out the extended                 */
855                                                           /* data after another thread has failed to see it (in openCommonData), so     */
856                                                           /* extended data can be examined.                                             */
857                                                           /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */
858 
859 #if MAP_IMPLEMENTATION==MAP_STDIO
860     umtx_unlock(&extendICUDataMutex);
861 #endif
862     return didUpdate;               /* Return true if ICUData pointer was updated.   */
863                                     /*   (Could potentialy have been done by another thread racing */
864                                     /*   us through here, but that's fine, we still return true    */
865                                     /*   so that current thread will also examine extended data.   */
866 }
867 
868 /*----------------------------------------------------------------------*
869  *                                                                      *
870  *   udata_setCommonData                                                *
871  *                                                                      *
872  *----------------------------------------------------------------------*/
873 U_CAPI void U_EXPORT2
udata_setCommonData(const void * data,UErrorCode * pErrorCode)874 udata_setCommonData(const void *data, UErrorCode *pErrorCode) {
875     UDataMemory dataMemory;
876 
877     if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
878         return;
879     }
880 
881     if(data==NULL) {
882         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
883         return;
884     }
885 
886     /* set the data pointer and test for validity */
887     UDataMemory_init(&dataMemory);
888     UDataMemory_setData(&dataMemory, data);
889     udata_checkCommonData(&dataMemory, pErrorCode);
890     if (U_FAILURE(*pErrorCode)) {return;}
891 
892     /* we have good data */
893     /* Set it up as the ICU Common Data.  */
894     setCommonICUData(&dataMemory, TRUE, pErrorCode);
895 }
896 
897 /*---------------------------------------------------------------------------
898  *
899  *  udata_setAppData
900  *
901  *---------------------------------------------------------------------------- */
902 U_CAPI void U_EXPORT2
udata_setAppData(const char * path,const void * data,UErrorCode * err)903 udata_setAppData(const char *path, const void *data, UErrorCode *err)
904 {
905     UDataMemory     udm;
906 
907     if(err==NULL || U_FAILURE(*err)) {
908         return;
909     }
910     if(data==NULL) {
911         *err=U_ILLEGAL_ARGUMENT_ERROR;
912         return;
913     }
914 
915     UDataMemory_init(&udm);
916     UDataMemory_setData(&udm, data);
917     udata_checkCommonData(&udm, err);
918     udata_cacheDataItem(path, &udm, err);
919 }
920 
921 /*----------------------------------------------------------------------------*
922  *                                                                            *
923  *  checkDataItem     Given a freshly located/loaded data item, either        *
924  *                    an entry in a common file or a separately loaded file,  *
925  *                    sanity check its header, and see if the data is         *
926  *                    acceptable to the app.                                  *
927  *                    If the data is good, create and return a UDataMemory    *
928  *                    object that can be returned to the application.         *
929  *                    Return NULL on any sort of failure.                     *
930  *                                                                            *
931  *----------------------------------------------------------------------------*/
932 static UDataMemory *
checkDataItem(const DataHeader * pHeader,UDataMemoryIsAcceptable * isAcceptable,void * context,const char * type,const char * name,UErrorCode * nonFatalErr,UErrorCode * fatalErr)933 checkDataItem
934 (
935  const DataHeader         *pHeader,         /* The data item to be checked.                */
936  UDataMemoryIsAcceptable  *isAcceptable,    /* App's call-back function                    */
937  void                     *context,         /*   pass-thru param for above.                */
938  const char               *type,            /*   pass-thru param for above.                */
939  const char               *name,            /*   pass-thru param for above.                */
940  UErrorCode               *nonFatalErr,     /* Error code if this data was not acceptable  */
941                                             /*   but openChoice should continue with       */
942                                             /*   trying to get data from fallback path.    */
943  UErrorCode               *fatalErr         /* Bad error, caller should return immediately */
944  )
945 {
946     UDataMemory  *rDataMem = NULL;          /* the new UDataMemory, to be returned.        */
947 
948     if (U_FAILURE(*fatalErr)) {
949         return NULL;
950     }
951 
952     if(pHeader->dataHeader.magic1==0xda &&
953         pHeader->dataHeader.magic2==0x27 &&
954         (isAcceptable==NULL || isAcceptable(context, type, name, &pHeader->info))
955     ) {
956         rDataMem=UDataMemory_createNewInstance(fatalErr);
957         if (U_FAILURE(*fatalErr)) {
958             return NULL;
959         }
960         rDataMem->pHeader = pHeader;
961     } else {
962         /* the data is not acceptable, look further */
963         /* If we eventually find something good, this errorcode will be */
964         /*    cleared out.                                              */
965         *nonFatalErr=U_INVALID_FORMAT_ERROR;
966     }
967     return rDataMem;
968 }
969 
970 /**
971  * @return 0 if not loaded, 1 if loaded or err
972  */
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)973 static UDataMemory *doLoadFromIndividualFiles(const char *pkgName,
974         const char *dataPath, const char *tocEntryPathSuffix,
975             /* following arguments are the same as doOpenChoice itself */
976             const char *path, const char *type, const char *name,
977              UDataMemoryIsAcceptable *isAcceptable, void *context,
978              UErrorCode *subErrorCode,
979              UErrorCode *pErrorCode)
980 {
981     const char         *pathBuffer;
982     UDataMemory         dataMemory;
983     UDataMemory *pEntryData;
984 
985     /* look in ind. files: package\nam.typ  ========================= */
986     /* init path iterator for individual files */
987     UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, FALSE, pErrorCode);
988 
989     while((pathBuffer = iter.next(pErrorCode)) != NULL)
990     {
991 #ifdef UDATA_DEBUG
992         fprintf(stderr, "UDATA: trying individual file %s\n", pathBuffer);
993 #endif
994         if(uprv_mapFile(&dataMemory, pathBuffer))
995         {
996             pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode);
997             if (pEntryData != NULL) {
998                 /* Data is good.
999                 *  Hand off ownership of the backing memory to the user's UDataMemory.
1000                 *  and return it.   */
1001                 pEntryData->mapAddr = dataMemory.mapAddr;
1002                 pEntryData->map     = dataMemory.map;
1003 
1004 #ifdef UDATA_DEBUG
1005                 fprintf(stderr, "** Mapped file: %s\n", pathBuffer);
1006 #endif
1007                 return pEntryData;
1008             }
1009 
1010             /* the data is not acceptable, or some error occured.  Either way, unmap the memory */
1011             udata_close(&dataMemory);
1012 
1013             /* If we had a nasty error, bail out completely.  */
1014             if (U_FAILURE(*pErrorCode)) {
1015                 return NULL;
1016             }
1017 
1018             /* Otherwise remember that we found data but didn't like it for some reason  */
1019             *subErrorCode=U_INVALID_FORMAT_ERROR;
1020         }
1021 #ifdef UDATA_DEBUG
1022         fprintf(stderr, "%s\n", UDataMemory_isLoaded(&dataMemory)?"LOADED":"not loaded");
1023 #endif
1024     }
1025     return NULL;
1026 }
1027 
1028 /**
1029  * @return 0 if not loaded, 1 if loaded or err
1030  */
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)1031 static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName*/,
1032         const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const char *tocEntryName,
1033             /* following arguments are the same as doOpenChoice itself */
1034             const char *path, const char *type, const char *name,
1035              UDataMemoryIsAcceptable *isAcceptable, void *context,
1036              UErrorCode *subErrorCode,
1037              UErrorCode *pErrorCode)
1038 {
1039     UDataMemory        *pEntryData;
1040     const DataHeader   *pHeader;
1041     UDataMemory        *pCommonData;
1042     int32_t            commonDataIndex;
1043     UBool              checkedExtendedICUData = FALSE;
1044     /* try to get common data.  The loop is for platforms such as the 390 that do
1045      *  not initially load the full set of ICU data.  If the lookup of an ICU data item
1046      *  fails, the full (but slower to load) set is loaded, the and the loop repeats,
1047      *  trying the lookup again.  Once the full set of ICU data is loaded, the loop wont
1048      *  repeat because the full set will be checked the first time through.
1049      *
1050      *  The loop also handles the fallback to a .dat file if the application linked
1051      *   to the stub data library rather than a real library.
1052      */
1053     for (commonDataIndex = isICUData ? 0 : -1;;) {
1054         pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** search for pkg **/
1055 
1056         if(U_SUCCESS(*subErrorCode) && pCommonData!=NULL) {
1057             int32_t length;
1058 
1059             /* look up the data piece in the common data */
1060             pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &length, subErrorCode);
1061 #ifdef UDATA_DEBUG
1062             fprintf(stderr, "%s: pHeader=%p - %s\n", tocEntryName, pHeader, u_errorName(*subErrorCode));
1063 #endif
1064 
1065             if(pHeader!=NULL) {
1066                 pEntryData = checkDataItem(pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode);
1067 #ifdef UDATA_DEBUG
1068                 fprintf(stderr, "pEntryData=%p\n", pEntryData);
1069 #endif
1070                 if (U_FAILURE(*pErrorCode)) {
1071                     return NULL;
1072                 }
1073                 if (pEntryData != NULL) {
1074                     pEntryData->length = length;
1075                     return pEntryData;
1076                 }
1077             }
1078         }
1079         /* Data wasn't found.  If we were looking for an ICUData item and there is
1080          * more data available, load it and try again,
1081          * otherwise break out of this loop. */
1082         if (!isICUData) {
1083             return NULL;
1084         } else if (pCommonData != NULL) {
1085             ++commonDataIndex;  /* try the next data package */
1086         } else if ((!checkedExtendedICUData) && extendICUData(subErrorCode)) {
1087             checkedExtendedICUData = TRUE;
1088             /* try this data package slot again: it changed from NULL to non-NULL */
1089         } else {
1090             return NULL;
1091         }
1092     }
1093 }
1094 
1095 /*
1096  * Identify the Time Zone resources that are subject to special override data loading.
1097  */
isTimeZoneFile(const char * name,const char * type)1098 static UBool isTimeZoneFile(const char *name, const char *type) {
1099     return ((uprv_strcmp(type, "res") == 0) &&
1100             (uprv_strcmp(name, "zoneinfo64") == 0 ||
1101              uprv_strcmp(name, "timezoneTypes") == 0 ||
1102              uprv_strcmp(name, "windowsZones") == 0 ||
1103              uprv_strcmp(name, "metaZones") == 0));
1104 }
1105 
1106 /*
1107  *  A note on the ownership of Mapped Memory
1108  *
1109  *  For common format files, ownership resides with the UDataMemory object
1110  *    that lives in the cache of opened common data.  These UDataMemorys are private
1111  *    to the udata implementation, and are never seen directly by users.
1112  *
1113  *    The UDataMemory objects returned to users will have the address of some desired
1114  *    data within the mapped region, but they wont have the mapping info itself, and thus
1115  *    won't cause anything to be removed from memory when they are closed.
1116  *
1117  *  For individual data files, the UDataMemory returned to the user holds the
1118  *  information necessary to unmap the data on close.  If the user independently
1119  *  opens the same data file twice, two completely independent mappings will be made.
1120  *  (There is no cache of opened data items from individual files, only a cache of
1121  *   opened Common Data files, that is, files containing a collection of data items.)
1122  *
1123  *  For common data passed in from the user via udata_setAppData() or
1124  *  udata_setCommonData(), ownership remains with the user.
1125  *
1126  *  UDataMemory objects themselves, as opposed to the memory they describe,
1127  *  can be anywhere - heap, stack/local or global.
1128  *  They have a flag to indicate when they're heap allocated and thus
1129  *  must be deleted when closed.
1130  */
1131 
1132 
1133 /*----------------------------------------------------------------------------*
1134  *                                                                            *
1135  * main data loading functions                                                *
1136  *                                                                            *
1137  *----------------------------------------------------------------------------*/
1138 static UDataMemory *
doOpenChoice(const char * path,const char * type,const char * name,UDataMemoryIsAcceptable * isAcceptable,void * context,UErrorCode * pErrorCode)1139 doOpenChoice(const char *path, const char *type, const char *name,
1140              UDataMemoryIsAcceptable *isAcceptable, void *context,
1141              UErrorCode *pErrorCode)
1142 {
1143     UDataMemory         *retVal = NULL;
1144 
1145     const char         *dataPath;
1146 
1147     int32_t             tocEntrySuffixIndex;
1148     const char         *tocEntryPathSuffix;
1149     UErrorCode          subErrorCode=U_ZERO_ERROR;
1150     const char         *treeChar;
1151 
1152     UBool               isICUData = FALSE;
1153 
1154 
1155     /* Is this path ICU data? */
1156     if(path == NULL ||
1157        !strcmp(path, U_ICUDATA_ALIAS) ||  /* "ICUDATA" */
1158        !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-" */
1159                      uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) ||
1160        !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-" */
1161                      uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) {
1162       isICUData = TRUE;
1163     }
1164 
1165 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR)  /* Windows:  try "foo\bar" and "foo/bar" */
1166     /* remap from alternate path char to the main one */
1167     CharString altSepPath;
1168     if(path) {
1169         if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != NULL) {
1170             altSepPath.append(path, *pErrorCode);
1171             char *p;
1172             while ((p = uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR)) != NULL) {
1173                 *p = U_FILE_SEP_CHAR;
1174             }
1175 #if defined (UDATA_DEBUG)
1176             fprintf(stderr, "Changed path from [%s] to [%s]\n", path, altSepPath.s);
1177 #endif
1178             path = altSepPath.data();
1179         }
1180     }
1181 #endif
1182 
1183     CharString tocEntryName; /* entry name in tree format. ex:  'icudt28b/coll/ar.res' */
1184     CharString tocEntryPath; /* entry name in path format. ex:  'icudt28b\\coll\\ar.res' */
1185 
1186     CharString pkgName;
1187     CharString treeName;
1188 
1189     /* ======= Set up strings */
1190     if(path==NULL) {
1191         pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1192     } else {
1193         const char *pkg;
1194         const char *first;
1195         pkg = uprv_strrchr(path, U_FILE_SEP_CHAR);
1196         first = uprv_strchr(path, U_FILE_SEP_CHAR);
1197         if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash in the path- not a tree name */
1198             /* see if this is an /absolute/path/to/package  path */
1199             if(pkg) {
1200                 pkgName.append(pkg+1, *pErrorCode);
1201             } else {
1202                 pkgName.append(path, *pErrorCode);
1203             }
1204         } else {
1205             treeChar = uprv_strchr(path, U_TREE_SEPARATOR);
1206             if(treeChar) {
1207                 treeName.append(treeChar+1, *pErrorCode); /* following '-' */
1208                 if(isICUData) {
1209                     pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1210                 } else {
1211                     pkgName.append(path, (int32_t)(treeChar-path), *pErrorCode);
1212                     if (first == NULL) {
1213                         /*
1214                         This user data has no path, but there is a tree name.
1215                         Look up the correct path from the data cache later.
1216                         */
1217                         path = pkgName.data();
1218                     }
1219                 }
1220             } else {
1221                 if(isICUData) {
1222                     pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1223                 } else {
1224                     pkgName.append(path, *pErrorCode);
1225                 }
1226             }
1227         }
1228     }
1229 
1230 #ifdef UDATA_DEBUG
1231     fprintf(stderr, " P=%s T=%s\n", pkgName.data(), treeName.data());
1232 #endif
1233 
1234     /* setting up the entry name and file name
1235      * Make up a full name by appending the type to the supplied
1236      *  name, assuming that a type was supplied.
1237      */
1238 
1239     /* prepend the package */
1240     tocEntryName.append(pkgName, *pErrorCode);
1241     tocEntryPath.append(pkgName, *pErrorCode);
1242     tocEntrySuffixIndex = tocEntryName.length();
1243 
1244     if(!treeName.isEmpty()) {
1245         tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode);
1246         tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode);
1247     }
1248 
1249     tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pErrorCode);
1250     tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode);
1251     if(type!=NULL && *type!=0) {
1252         tocEntryName.append(".", *pErrorCode).append(type, *pErrorCode);
1253         tocEntryPath.append(".", *pErrorCode).append(type, *pErrorCode);
1254     }
1255     tocEntryPathSuffix = tocEntryPath.data()+tocEntrySuffixIndex; /* suffix starts here */
1256 
1257 #ifdef UDATA_DEBUG
1258     fprintf(stderr, " tocEntryName = %s\n", tocEntryName.data());
1259     fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data());
1260 #endif
1261 
1262 #if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time
1263     if(path == NULL) {
1264         path = COMMON_DATA_NAME; /* "icudt26e" */
1265     }
1266 #else
1267     // Windows UWP expects only a single data file.
1268     path = COMMON_DATA_NAME; /* "icudt26e" */
1269 #endif
1270 
1271     /************************ Begin loop looking for ind. files ***************/
1272 #ifdef UDATA_DEBUG
1273     fprintf(stderr, "IND: inBasename = %s, pkg=%s\n", "(n/a)", packageNameFromPath(path));
1274 #endif
1275 
1276     /* End of dealing with a null basename */
1277     dataPath = u_getDataDirectory();
1278 
1279     /****    Time zone individual files override  */
1280     if (isICUData && isTimeZoneFile(name, type)) {
1281         const char *tzFilesDir = u_getTimeZoneFilesDirectory(pErrorCode);
1282         if (tzFilesDir[0] != 0) {
1283 #ifdef UDATA_DEBUG
1284             fprintf(stderr, "Trying Time Zone Files directory = %s\n", tzFilesDir);
1285 #endif
1286             retVal = doLoadFromIndividualFiles(/* pkgName.data() */ "", tzFilesDir, tocEntryPathSuffix,
1287                             /* path */ "", type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1288             if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1289                 return retVal;
1290             }
1291         }
1292     }
1293 
1294     /****    COMMON PACKAGE  - only if packages are first. */
1295     if(gDataFileAccess == UDATA_PACKAGES_FIRST) {
1296 #ifdef UDATA_DEBUG
1297         fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n");
1298 #endif
1299         /* #2 */
1300         retVal = doLoadFromCommonData(isICUData,
1301                             pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(),
1302                             path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1303         if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1304             return retVal;
1305         }
1306     }
1307 
1308     /****    INDIVIDUAL FILES  */
1309     if((gDataFileAccess==UDATA_PACKAGES_FIRST) ||
1310        (gDataFileAccess==UDATA_FILES_FIRST)) {
1311 #ifdef UDATA_DEBUG
1312         fprintf(stderr, "Trying individual files\n");
1313 #endif
1314         /* Check to make sure that there is a dataPath to iterate over */
1315         if ((dataPath && *dataPath) || !isICUData) {
1316             retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix,
1317                             path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1318             if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1319                 return retVal;
1320             }
1321         }
1322     }
1323 
1324     /****    COMMON PACKAGE  */
1325     if((gDataFileAccess==UDATA_ONLY_PACKAGES) ||
1326        (gDataFileAccess==UDATA_FILES_FIRST)) {
1327 #ifdef UDATA_DEBUG
1328         fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n");
1329 #endif
1330         retVal = doLoadFromCommonData(isICUData,
1331                             pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(),
1332                             path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1333         if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1334             return retVal;
1335         }
1336     }
1337 
1338     /* Load from DLL.  If we haven't attempted package load, we also haven't had any chance to
1339         try a DLL (static or setCommonData/etc)  load.
1340          If we ever have a "UDATA_ONLY_FILES", add it to the or list here.  */
1341     if(gDataFileAccess==UDATA_NO_FILES) {
1342 #ifdef UDATA_DEBUG
1343         fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n");
1344 #endif
1345         retVal = doLoadFromCommonData(isICUData,
1346                             pkgName.data(), "", tocEntryPathSuffix, tocEntryName.data(),
1347                             path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1348         if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1349             return retVal;
1350         }
1351     }
1352 
1353     /* data not found */
1354     if(U_SUCCESS(*pErrorCode)) {
1355         if(U_SUCCESS(subErrorCode)) {
1356             /* file not found */
1357             *pErrorCode=U_FILE_ACCESS_ERROR;
1358         } else {
1359             /* entry point not found or rejected */
1360             *pErrorCode=subErrorCode;
1361         }
1362     }
1363     return retVal;
1364 }
1365 
1366 
1367 
1368 /* API ---------------------------------------------------------------------- */
1369 
1370 U_CAPI UDataMemory * U_EXPORT2
udata_open(const char * path,const char * type,const char * name,UErrorCode * pErrorCode)1371 udata_open(const char *path, const char *type, const char *name,
1372            UErrorCode *pErrorCode) {
1373 #ifdef UDATA_DEBUG
1374   fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type);
1375     fflush(stderr);
1376 #endif
1377 
1378     if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
1379         return NULL;
1380     } else if(name==NULL || *name==0) {
1381         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1382         return NULL;
1383     } else {
1384         return doOpenChoice(path, type, name, NULL, NULL, pErrorCode);
1385     }
1386 }
1387 
1388 
1389 
1390 U_CAPI UDataMemory * U_EXPORT2
udata_openChoice(const char * path,const char * type,const char * name,UDataMemoryIsAcceptable * isAcceptable,void * context,UErrorCode * pErrorCode)1391 udata_openChoice(const char *path, const char *type, const char *name,
1392                  UDataMemoryIsAcceptable *isAcceptable, void *context,
1393                  UErrorCode *pErrorCode) {
1394 #ifdef UDATA_DEBUG
1395   fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type);
1396 #endif
1397 
1398     if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
1399         return NULL;
1400     } else if(name==NULL || *name==0 || isAcceptable==NULL) {
1401         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1402         return NULL;
1403     } else {
1404         return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode);
1405     }
1406 }
1407 
1408 
1409 
1410 U_CAPI void U_EXPORT2
udata_getInfo(UDataMemory * pData,UDataInfo * pInfo)1411 udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) {
1412     if(pInfo!=NULL) {
1413         if(pData!=NULL && pData->pHeader!=NULL) {
1414             const UDataInfo *info=&pData->pHeader->info;
1415             uint16_t dataInfoSize=udata_getInfoSize(info);
1416             if(pInfo->size>dataInfoSize) {
1417                 pInfo->size=dataInfoSize;
1418             }
1419             uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->size-2);
1420             if(info->isBigEndian!=U_IS_BIG_ENDIAN) {
1421                 /* opposite endianness */
1422                 uint16_t x=info->reservedWord;
1423                 pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8));
1424             }
1425         } else {
1426             pInfo->size=0;
1427         }
1428     }
1429 }
1430 
1431 
udata_setFileAccess(UDataFileAccess access,UErrorCode *)1432 U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/)
1433 {
1434     // Note: this function is documented as not thread safe.
1435     gDataFileAccess = access;
1436 }
1437