|
1 /* |
|
2 ******************************************************************** |
|
3 * COPYRIGHT: |
|
4 * Copyright (c) 1996-2013, International Business Machines Corporation and |
|
5 * others. All Rights Reserved. |
|
6 ******************************************************************** |
|
7 * |
|
8 * uconv_bld.cpp: |
|
9 * |
|
10 * Defines functions that are used in the creation/initialization/deletion |
|
11 * of converters and related structures. |
|
12 * uses uconv_io.h routines to access disk information |
|
13 * is used by ucnv.h to implement public API create/delete/flushCache routines |
|
14 * Modification History: |
|
15 * |
|
16 * Date Name Description |
|
17 * |
|
18 * 06/20/2000 helena OS/400 port changes; mostly typecast. |
|
19 * 06/29/2000 helena Major rewrite of the callback interface. |
|
20 */ |
|
21 |
|
22 #include "unicode/utypes.h" |
|
23 |
|
24 #if !UCONFIG_NO_CONVERSION |
|
25 |
|
26 #include "unicode/putil.h" |
|
27 #include "unicode/udata.h" |
|
28 #include "unicode/ucnv.h" |
|
29 #include "unicode/uloc.h" |
|
30 #include "mutex.h" |
|
31 #include "putilimp.h" |
|
32 #include "uassert.h" |
|
33 #include "utracimp.h" |
|
34 #include "ucnv_io.h" |
|
35 #include "ucnv_bld.h" |
|
36 #include "ucnvmbcs.h" |
|
37 #include "ucnv_ext.h" |
|
38 #include "ucnv_cnv.h" |
|
39 #include "ucnv_imp.h" |
|
40 #include "uhash.h" |
|
41 #include "umutex.h" |
|
42 #include "cstring.h" |
|
43 #include "cmemory.h" |
|
44 #include "ucln_cmn.h" |
|
45 #include "ustr_cnv.h" |
|
46 |
|
47 |
|
48 #if 0 |
|
49 #include <stdio.h> |
|
50 extern void UCNV_DEBUG_LOG(char *what, char *who, void *p, int l); |
|
51 #define UCNV_DEBUG_LOG(x,y,z) UCNV_DEBUG_LOG(x,y,z,__LINE__) |
|
52 #else |
|
53 # define UCNV_DEBUG_LOG(x,y,z) |
|
54 #endif |
|
55 |
|
56 static const UConverterSharedData * const |
|
57 converterData[UCNV_NUMBER_OF_SUPPORTED_CONVERTER_TYPES]={ |
|
58 NULL, NULL, |
|
59 |
|
60 #if UCONFIG_NO_LEGACY_CONVERSION |
|
61 NULL, |
|
62 #else |
|
63 &_MBCSData, |
|
64 #endif |
|
65 |
|
66 &_Latin1Data, |
|
67 &_UTF8Data, &_UTF16BEData, &_UTF16LEData, &_UTF32BEData, &_UTF32LEData, |
|
68 NULL, |
|
69 |
|
70 #if UCONFIG_NO_LEGACY_CONVERSION |
|
71 NULL, |
|
72 NULL, NULL, NULL, NULL, NULL, NULL, |
|
73 NULL, NULL, NULL, NULL, NULL, NULL, |
|
74 NULL, |
|
75 #else |
|
76 &_ISO2022Data, |
|
77 &_LMBCSData1,&_LMBCSData2, &_LMBCSData3, &_LMBCSData4, &_LMBCSData5, &_LMBCSData6, |
|
78 &_LMBCSData8,&_LMBCSData11,&_LMBCSData16,&_LMBCSData17,&_LMBCSData18,&_LMBCSData19, |
|
79 &_HZData, |
|
80 #endif |
|
81 |
|
82 &_SCSUData, |
|
83 |
|
84 #if UCONFIG_NO_LEGACY_CONVERSION |
|
85 NULL, |
|
86 #else |
|
87 &_ISCIIData, |
|
88 #endif |
|
89 |
|
90 &_ASCIIData, |
|
91 &_UTF7Data, &_Bocu1Data, &_UTF16Data, &_UTF32Data, &_CESU8Data, &_IMAPData, |
|
92 |
|
93 #if UCONFIG_NO_LEGACY_CONVERSION |
|
94 NULL, |
|
95 #else |
|
96 &_CompoundTextData |
|
97 #endif |
|
98 }; |
|
99 |
|
100 /* Please keep this in binary sorted order for getAlgorithmicTypeFromName. |
|
101 Also the name should be in lower case and all spaces, dashes and underscores |
|
102 removed |
|
103 */ |
|
104 static struct { |
|
105 const char *name; |
|
106 const UConverterType type; |
|
107 } const cnvNameType[] = { |
|
108 { "bocu1", UCNV_BOCU1 }, |
|
109 { "cesu8", UCNV_CESU8 }, |
|
110 #if !UCONFIG_NO_LEGACY_CONVERSION |
|
111 { "hz",UCNV_HZ }, |
|
112 #endif |
|
113 { "imapmailboxname", UCNV_IMAP_MAILBOX }, |
|
114 #if !UCONFIG_NO_LEGACY_CONVERSION |
|
115 { "iscii", UCNV_ISCII }, |
|
116 { "iso2022", UCNV_ISO_2022 }, |
|
117 #endif |
|
118 { "iso88591", UCNV_LATIN_1 }, |
|
119 #if !UCONFIG_NO_LEGACY_CONVERSION |
|
120 { "lmbcs1", UCNV_LMBCS_1 }, |
|
121 { "lmbcs11",UCNV_LMBCS_11 }, |
|
122 { "lmbcs16",UCNV_LMBCS_16 }, |
|
123 { "lmbcs17",UCNV_LMBCS_17 }, |
|
124 { "lmbcs18",UCNV_LMBCS_18 }, |
|
125 { "lmbcs19",UCNV_LMBCS_19 }, |
|
126 { "lmbcs2", UCNV_LMBCS_2 }, |
|
127 { "lmbcs3", UCNV_LMBCS_3 }, |
|
128 { "lmbcs4", UCNV_LMBCS_4 }, |
|
129 { "lmbcs5", UCNV_LMBCS_5 }, |
|
130 { "lmbcs6", UCNV_LMBCS_6 }, |
|
131 { "lmbcs8", UCNV_LMBCS_8 }, |
|
132 #endif |
|
133 { "scsu", UCNV_SCSU }, |
|
134 { "usascii", UCNV_US_ASCII }, |
|
135 { "utf16", UCNV_UTF16 }, |
|
136 { "utf16be", UCNV_UTF16_BigEndian }, |
|
137 { "utf16le", UCNV_UTF16_LittleEndian }, |
|
138 #if U_IS_BIG_ENDIAN |
|
139 { "utf16oppositeendian", UCNV_UTF16_LittleEndian }, |
|
140 { "utf16platformendian", UCNV_UTF16_BigEndian }, |
|
141 #else |
|
142 { "utf16oppositeendian", UCNV_UTF16_BigEndian}, |
|
143 { "utf16platformendian", UCNV_UTF16_LittleEndian }, |
|
144 #endif |
|
145 { "utf32", UCNV_UTF32 }, |
|
146 { "utf32be", UCNV_UTF32_BigEndian }, |
|
147 { "utf32le", UCNV_UTF32_LittleEndian }, |
|
148 #if U_IS_BIG_ENDIAN |
|
149 { "utf32oppositeendian", UCNV_UTF32_LittleEndian }, |
|
150 { "utf32platformendian", UCNV_UTF32_BigEndian }, |
|
151 #else |
|
152 { "utf32oppositeendian", UCNV_UTF32_BigEndian }, |
|
153 { "utf32platformendian", UCNV_UTF32_LittleEndian }, |
|
154 #endif |
|
155 { "utf7", UCNV_UTF7 }, |
|
156 { "utf8", UCNV_UTF8 }, |
|
157 { "x11compoundtext", UCNV_COMPOUND_TEXT} |
|
158 }; |
|
159 |
|
160 |
|
161 /*initializes some global variables */ |
|
162 static UHashtable *SHARED_DATA_HASHTABLE = NULL; |
|
163 static UMutex cnvCacheMutex = U_MUTEX_INITIALIZER; /* Mutex for synchronizing cnv cache access. */ |
|
164 /* Note: the global mutex is used for */ |
|
165 /* reference count updates. */ |
|
166 |
|
167 static const char **gAvailableConverters = NULL; |
|
168 static uint16_t gAvailableConverterCount = 0; |
|
169 static icu::UInitOnce gAvailableConvertersInitOnce = U_INITONCE_INITIALIZER; |
|
170 |
|
171 #if !U_CHARSET_IS_UTF8 |
|
172 |
|
173 /* This contains the resolved converter name. So no further alias lookup is needed again. */ |
|
174 static char gDefaultConverterNameBuffer[UCNV_MAX_CONVERTER_NAME_LENGTH + 1]; /* +1 for NULL */ |
|
175 static const char *gDefaultConverterName = NULL; |
|
176 |
|
177 /* |
|
178 If the default converter is an algorithmic converter, this is the cached value. |
|
179 We don't cache a full UConverter and clone it because ucnv_clone doesn't have |
|
180 less overhead than an algorithmic open. We don't cache non-algorithmic converters |
|
181 because ucnv_flushCache must be able to unload the default converter and its table. |
|
182 */ |
|
183 static const UConverterSharedData *gDefaultAlgorithmicSharedData = NULL; |
|
184 |
|
185 /* Does gDefaultConverterName have a converter option and require extra parsing? */ |
|
186 static UBool gDefaultConverterContainsOption; |
|
187 |
|
188 #endif /* !U_CHARSET_IS_UTF8 */ |
|
189 |
|
190 static const char DATA_TYPE[] = "cnv"; |
|
191 |
|
192 /* ucnv_flushAvailableConverterCache. This is only called from ucnv_cleanup(). |
|
193 * If it is ever to be called from elsewhere, synchronization |
|
194 * will need to be considered. |
|
195 */ |
|
196 static void |
|
197 ucnv_flushAvailableConverterCache() { |
|
198 gAvailableConverterCount = 0; |
|
199 if (gAvailableConverters) { |
|
200 uprv_free((char **)gAvailableConverters); |
|
201 gAvailableConverters = NULL; |
|
202 } |
|
203 gAvailableConvertersInitOnce.reset(); |
|
204 } |
|
205 |
|
206 /* ucnv_cleanup - delete all storage held by the converter cache, except any */ |
|
207 /* in use by open converters. */ |
|
208 /* Not thread safe. */ |
|
209 /* Not supported API. */ |
|
210 static UBool U_CALLCONV ucnv_cleanup(void) { |
|
211 ucnv_flushCache(); |
|
212 if (SHARED_DATA_HASHTABLE != NULL && uhash_count(SHARED_DATA_HASHTABLE) == 0) { |
|
213 uhash_close(SHARED_DATA_HASHTABLE); |
|
214 SHARED_DATA_HASHTABLE = NULL; |
|
215 } |
|
216 |
|
217 /* Isn't called from flushCache because other threads may have preexisting references to the table. */ |
|
218 ucnv_flushAvailableConverterCache(); |
|
219 |
|
220 #if !U_CHARSET_IS_UTF8 |
|
221 gDefaultConverterName = NULL; |
|
222 gDefaultConverterNameBuffer[0] = 0; |
|
223 gDefaultConverterContainsOption = FALSE; |
|
224 gDefaultAlgorithmicSharedData = NULL; |
|
225 #endif |
|
226 |
|
227 return (SHARED_DATA_HASHTABLE == NULL); |
|
228 } |
|
229 |
|
230 static UBool U_CALLCONV |
|
231 isCnvAcceptable(void * /*context*/, |
|
232 const char * /*type*/, const char * /*name*/, |
|
233 const UDataInfo *pInfo) { |
|
234 return (UBool)( |
|
235 pInfo->size>=20 && |
|
236 pInfo->isBigEndian==U_IS_BIG_ENDIAN && |
|
237 pInfo->charsetFamily==U_CHARSET_FAMILY && |
|
238 pInfo->sizeofUChar==U_SIZEOF_UCHAR && |
|
239 pInfo->dataFormat[0]==0x63 && /* dataFormat="cnvt" */ |
|
240 pInfo->dataFormat[1]==0x6e && |
|
241 pInfo->dataFormat[2]==0x76 && |
|
242 pInfo->dataFormat[3]==0x74 && |
|
243 pInfo->formatVersion[0]==6); /* Everything will be version 6 */ |
|
244 } |
|
245 |
|
246 /** |
|
247 * Un flatten shared data from a UDATA.. |
|
248 */ |
|
249 static UConverterSharedData* |
|
250 ucnv_data_unFlattenClone(UConverterLoadArgs *pArgs, UDataMemory *pData, UErrorCode *status) |
|
251 { |
|
252 /* UDataInfo info; -- necessary only if some converters have different formatVersion */ |
|
253 const uint8_t *raw = (const uint8_t *)udata_getMemory(pData); |
|
254 const UConverterStaticData *source = (const UConverterStaticData *) raw; |
|
255 UConverterSharedData *data; |
|
256 UConverterType type = (UConverterType)source->conversionType; |
|
257 |
|
258 if(U_FAILURE(*status)) |
|
259 return NULL; |
|
260 |
|
261 if( (uint16_t)type >= UCNV_NUMBER_OF_SUPPORTED_CONVERTER_TYPES || |
|
262 converterData[type] == NULL || |
|
263 converterData[type]->referenceCounter != 1 || |
|
264 source->structSize != sizeof(UConverterStaticData)) |
|
265 { |
|
266 *status = U_INVALID_TABLE_FORMAT; |
|
267 return NULL; |
|
268 } |
|
269 |
|
270 data = (UConverterSharedData *)uprv_malloc(sizeof(UConverterSharedData)); |
|
271 if(data == NULL) { |
|
272 *status = U_MEMORY_ALLOCATION_ERROR; |
|
273 return NULL; |
|
274 } |
|
275 |
|
276 /* copy initial values from the static structure for this type */ |
|
277 uprv_memcpy(data, converterData[type], sizeof(UConverterSharedData)); |
|
278 |
|
279 #if 0 /* made UConverterMBCSTable part of UConverterSharedData -- markus 20031107 */ |
|
280 /* |
|
281 * It would be much more efficient if the table were a direct member, not a pointer. |
|
282 * However, that would add to the size of all UConverterSharedData objects |
|
283 * even if they do not use this table (especially algorithmic ones). |
|
284 * If this changes, then the static templates from converterData[type] |
|
285 * need more entries. |
|
286 * |
|
287 * In principle, it would be cleaner if the load() function below |
|
288 * allocated the table. |
|
289 */ |
|
290 data->table = (UConverterTable *)uprv_malloc(sizeof(UConverterTable)); |
|
291 if(data->table == NULL) { |
|
292 uprv_free(data); |
|
293 *status = U_MEMORY_ALLOCATION_ERROR; |
|
294 return NULL; |
|
295 } |
|
296 uprv_memset(data->table, 0, sizeof(UConverterTable)); |
|
297 #endif |
|
298 |
|
299 data->staticData = source; |
|
300 |
|
301 data->sharedDataCached = FALSE; |
|
302 |
|
303 /* fill in fields from the loaded data */ |
|
304 data->dataMemory = (void*)pData; /* for future use */ |
|
305 |
|
306 if(data->impl->load != NULL) { |
|
307 data->impl->load(data, pArgs, raw + source->structSize, status); |
|
308 if(U_FAILURE(*status)) { |
|
309 uprv_free(data->table); |
|
310 uprv_free(data); |
|
311 return NULL; |
|
312 } |
|
313 } |
|
314 return data; |
|
315 } |
|
316 |
|
317 /*Takes an alias name gets an actual converter file name |
|
318 *goes to disk and opens it. |
|
319 *allocates the memory and returns a new UConverter object |
|
320 */ |
|
321 static UConverterSharedData *createConverterFromFile(UConverterLoadArgs *pArgs, UErrorCode * err) |
|
322 { |
|
323 UDataMemory *data; |
|
324 UConverterSharedData *sharedData; |
|
325 |
|
326 UTRACE_ENTRY_OC(UTRACE_UCNV_LOAD); |
|
327 |
|
328 if (U_FAILURE (*err)) { |
|
329 UTRACE_EXIT_STATUS(*err); |
|
330 return NULL; |
|
331 } |
|
332 |
|
333 UTRACE_DATA2(UTRACE_OPEN_CLOSE, "load converter %s from package %s", pArgs->name, pArgs->pkg); |
|
334 |
|
335 data = udata_openChoice(pArgs->pkg, DATA_TYPE, pArgs->name, isCnvAcceptable, NULL, err); |
|
336 if(U_FAILURE(*err)) |
|
337 { |
|
338 UTRACE_EXIT_STATUS(*err); |
|
339 return NULL; |
|
340 } |
|
341 |
|
342 sharedData = ucnv_data_unFlattenClone(pArgs, data, err); |
|
343 if(U_FAILURE(*err)) |
|
344 { |
|
345 udata_close(data); |
|
346 UTRACE_EXIT_STATUS(*err); |
|
347 return NULL; |
|
348 } |
|
349 |
|
350 /* |
|
351 * TODO Store pkg in a field in the shared data so that delta-only converters |
|
352 * can load base converters from the same package. |
|
353 * If the pkg name is longer than the field, then either do not load the converter |
|
354 * in the first place, or just set the pkg field to "". |
|
355 */ |
|
356 |
|
357 UTRACE_EXIT_PTR_STATUS(sharedData, *err); |
|
358 return sharedData; |
|
359 } |
|
360 |
|
361 /*returns a converter type from a string |
|
362 */ |
|
363 static const UConverterSharedData * |
|
364 getAlgorithmicTypeFromName(const char *realName) |
|
365 { |
|
366 uint32_t mid, start, limit; |
|
367 uint32_t lastMid; |
|
368 int result; |
|
369 char strippedName[UCNV_MAX_CONVERTER_NAME_LENGTH]; |
|
370 |
|
371 /* Lower case and remove ignoreable characters. */ |
|
372 ucnv_io_stripForCompare(strippedName, realName); |
|
373 |
|
374 /* do a binary search for the alias */ |
|
375 start = 0; |
|
376 limit = sizeof(cnvNameType)/sizeof(cnvNameType[0]); |
|
377 mid = limit; |
|
378 lastMid = UINT32_MAX; |
|
379 |
|
380 for (;;) { |
|
381 mid = (uint32_t)((start + limit) / 2); |
|
382 if (lastMid == mid) { /* Have we moved? */ |
|
383 break; /* We haven't moved, and it wasn't found. */ |
|
384 } |
|
385 lastMid = mid; |
|
386 result = uprv_strcmp(strippedName, cnvNameType[mid].name); |
|
387 |
|
388 if (result < 0) { |
|
389 limit = mid; |
|
390 } else if (result > 0) { |
|
391 start = mid; |
|
392 } else { |
|
393 return converterData[cnvNameType[mid].type]; |
|
394 } |
|
395 } |
|
396 |
|
397 return NULL; |
|
398 } |
|
399 |
|
400 /* |
|
401 * Based on the number of known converters, this determines how many times larger |
|
402 * the shared data hash table should be. When on small platforms, or just a couple |
|
403 * of converters are used, this number should be 2. When memory is plentiful, or |
|
404 * when ucnv_countAvailable is ever used with a lot of available converters, |
|
405 * this should be 4. |
|
406 * Larger numbers reduce the number of hash collisions, but use more memory. |
|
407 */ |
|
408 #define UCNV_CACHE_LOAD_FACTOR 2 |
|
409 |
|
410 /* Puts the shared data in the static hashtable SHARED_DATA_HASHTABLE */ |
|
411 /* Will always be called with the cnvCacheMutex alrady being held */ |
|
412 /* by the calling function. */ |
|
413 /* Stores the shared data in the SHARED_DATA_HASHTABLE |
|
414 * @param data The shared data |
|
415 */ |
|
416 static void |
|
417 ucnv_shareConverterData(UConverterSharedData * data) |
|
418 { |
|
419 UErrorCode err = U_ZERO_ERROR; |
|
420 /*Lazy evaluates the Hashtable itself */ |
|
421 /*void *sanity = NULL;*/ |
|
422 |
|
423 if (SHARED_DATA_HASHTABLE == NULL) |
|
424 { |
|
425 SHARED_DATA_HASHTABLE = uhash_openSize(uhash_hashChars, uhash_compareChars, NULL, |
|
426 ucnv_io_countKnownConverters(&err)*UCNV_CACHE_LOAD_FACTOR, |
|
427 &err); |
|
428 ucln_common_registerCleanup(UCLN_COMMON_UCNV, ucnv_cleanup); |
|
429 |
|
430 if (U_FAILURE(err)) |
|
431 return; |
|
432 } |
|
433 |
|
434 /* ### check to see if the element is not already there! */ |
|
435 |
|
436 /* |
|
437 sanity = ucnv_getSharedConverterData (data->staticData->name); |
|
438 if(sanity != NULL) |
|
439 { |
|
440 UCNV_DEBUG_LOG("put:overwrite!",data->staticData->name,sanity); |
|
441 } |
|
442 UCNV_DEBUG_LOG("put:chk",data->staticData->name,sanity); |
|
443 */ |
|
444 |
|
445 /* Mark it shared */ |
|
446 data->sharedDataCached = TRUE; |
|
447 |
|
448 uhash_put(SHARED_DATA_HASHTABLE, |
|
449 (void*) data->staticData->name, /* Okay to cast away const as long as |
|
450 keyDeleter == NULL */ |
|
451 data, |
|
452 &err); |
|
453 UCNV_DEBUG_LOG("put", data->staticData->name,data); |
|
454 |
|
455 } |
|
456 |
|
457 /* Look up a converter name in the shared data cache. */ |
|
458 /* cnvCacheMutex must be held by the caller to protect the hash table. */ |
|
459 /* gets the shared data from the SHARED_DATA_HASHTABLE (might return NULL if it isn't there) |
|
460 * @param name The name of the shared data |
|
461 * @return the shared data from the SHARED_DATA_HASHTABLE |
|
462 */ |
|
463 static UConverterSharedData * |
|
464 ucnv_getSharedConverterData(const char *name) |
|
465 { |
|
466 /*special case when no Table has yet been created we return NULL */ |
|
467 if (SHARED_DATA_HASHTABLE == NULL) |
|
468 { |
|
469 return NULL; |
|
470 } |
|
471 else |
|
472 { |
|
473 UConverterSharedData *rc; |
|
474 |
|
475 rc = (UConverterSharedData*)uhash_get(SHARED_DATA_HASHTABLE, name); |
|
476 UCNV_DEBUG_LOG("get",name,rc); |
|
477 return rc; |
|
478 } |
|
479 } |
|
480 |
|
481 /*frees the string of memory blocks associates with a sharedConverter |
|
482 *if and only if the referenceCounter == 0 |
|
483 */ |
|
484 /* Deletes (frees) the Shared data it's passed. first it checks the referenceCounter to |
|
485 * see if anyone is using it, if not it frees all the memory stemming from sharedConverterData and |
|
486 * returns TRUE, |
|
487 * otherwise returns FALSE |
|
488 * @param sharedConverterData The shared data |
|
489 * @return if not it frees all the memory stemming from sharedConverterData and |
|
490 * returns TRUE, otherwise returns FALSE |
|
491 */ |
|
492 static UBool |
|
493 ucnv_deleteSharedConverterData(UConverterSharedData * deadSharedData) |
|
494 { |
|
495 UTRACE_ENTRY_OC(UTRACE_UCNV_UNLOAD); |
|
496 UTRACE_DATA2(UTRACE_OPEN_CLOSE, "unload converter %s shared data %p", deadSharedData->staticData->name, deadSharedData); |
|
497 |
|
498 if (deadSharedData->referenceCounter > 0) { |
|
499 UTRACE_EXIT_VALUE((int32_t)FALSE); |
|
500 return FALSE; |
|
501 } |
|
502 |
|
503 if (deadSharedData->impl->unload != NULL) { |
|
504 deadSharedData->impl->unload(deadSharedData); |
|
505 } |
|
506 |
|
507 if(deadSharedData->dataMemory != NULL) |
|
508 { |
|
509 UDataMemory *data = (UDataMemory*)deadSharedData->dataMemory; |
|
510 udata_close(data); |
|
511 } |
|
512 |
|
513 if(deadSharedData->table != NULL) |
|
514 { |
|
515 uprv_free(deadSharedData->table); |
|
516 } |
|
517 |
|
518 #if 0 |
|
519 /* if the static data is actually owned by the shared data */ |
|
520 /* enable if we ever have this situation. */ |
|
521 if(deadSharedData->staticDataOwned == TRUE) /* see ucnv_bld.h */ |
|
522 { |
|
523 uprv_free((void*)deadSharedData->staticData); |
|
524 } |
|
525 #endif |
|
526 |
|
527 #if 0 |
|
528 /* Zap it ! */ |
|
529 uprv_memset(deadSharedData->0, sizeof(*deadSharedData)); |
|
530 #endif |
|
531 |
|
532 uprv_free(deadSharedData); |
|
533 |
|
534 UTRACE_EXIT_VALUE((int32_t)TRUE); |
|
535 return TRUE; |
|
536 } |
|
537 |
|
538 /** |
|
539 * Load a non-algorithmic converter. |
|
540 * If pkg==NULL, then this function must be called inside umtx_lock(&cnvCacheMutex). |
|
541 */ |
|
542 UConverterSharedData * |
|
543 ucnv_load(UConverterLoadArgs *pArgs, UErrorCode *err) { |
|
544 UConverterSharedData *mySharedConverterData; |
|
545 |
|
546 if(err == NULL || U_FAILURE(*err)) { |
|
547 return NULL; |
|
548 } |
|
549 |
|
550 if(pArgs->pkg != NULL && *pArgs->pkg != 0) { |
|
551 /* application-provided converters are not currently cached */ |
|
552 return createConverterFromFile(pArgs, err); |
|
553 } |
|
554 |
|
555 mySharedConverterData = ucnv_getSharedConverterData(pArgs->name); |
|
556 if (mySharedConverterData == NULL) |
|
557 { |
|
558 /*Not cached, we need to stream it in from file */ |
|
559 mySharedConverterData = createConverterFromFile(pArgs, err); |
|
560 if (U_FAILURE (*err) || (mySharedConverterData == NULL)) |
|
561 { |
|
562 return NULL; |
|
563 } |
|
564 else if (!pArgs->onlyTestIsLoadable) |
|
565 { |
|
566 /* share it with other library clients */ |
|
567 ucnv_shareConverterData(mySharedConverterData); |
|
568 } |
|
569 } |
|
570 else |
|
571 { |
|
572 /* The data for this converter was already in the cache. */ |
|
573 /* Update the reference counter on the shared data: one more client */ |
|
574 mySharedConverterData->referenceCounter++; |
|
575 } |
|
576 |
|
577 return mySharedConverterData; |
|
578 } |
|
579 |
|
580 /** |
|
581 * Unload a non-algorithmic converter. |
|
582 * It must be sharedData->referenceCounter != ~0 |
|
583 * and this function must be called inside umtx_lock(&cnvCacheMutex). |
|
584 */ |
|
585 U_CAPI void |
|
586 ucnv_unload(UConverterSharedData *sharedData) { |
|
587 if(sharedData != NULL) { |
|
588 if (sharedData->referenceCounter > 0) { |
|
589 sharedData->referenceCounter--; |
|
590 } |
|
591 |
|
592 if((sharedData->referenceCounter <= 0)&&(sharedData->sharedDataCached == FALSE)) { |
|
593 ucnv_deleteSharedConverterData(sharedData); |
|
594 } |
|
595 } |
|
596 } |
|
597 |
|
598 U_CFUNC void |
|
599 ucnv_unloadSharedDataIfReady(UConverterSharedData *sharedData) |
|
600 { |
|
601 /* |
|
602 Checking whether it's an algorithic converter is okay |
|
603 in multithreaded applications because the value never changes. |
|
604 Don't check referenceCounter for any other value. |
|
605 */ |
|
606 if(sharedData != NULL && sharedData->referenceCounter != (uint32_t)~0) { |
|
607 umtx_lock(&cnvCacheMutex); |
|
608 ucnv_unload(sharedData); |
|
609 umtx_unlock(&cnvCacheMutex); |
|
610 } |
|
611 } |
|
612 |
|
613 U_CFUNC void |
|
614 ucnv_incrementRefCount(UConverterSharedData *sharedData) |
|
615 { |
|
616 /* |
|
617 Checking whether it's an algorithic converter is okay |
|
618 in multithreaded applications because the value never changes. |
|
619 Don't check referenceCounter for any other value. |
|
620 */ |
|
621 if(sharedData != NULL && sharedData->referenceCounter != (uint32_t)~0) { |
|
622 umtx_lock(&cnvCacheMutex); |
|
623 sharedData->referenceCounter++; |
|
624 umtx_unlock(&cnvCacheMutex); |
|
625 } |
|
626 } |
|
627 |
|
628 /* |
|
629 * *pPieces must be initialized. |
|
630 * The name without options will be copied to pPieces->cnvName. |
|
631 * The locale and options will be copied to pPieces only if present in inName, |
|
632 * otherwise the existing values in pPieces remain. |
|
633 * *pArgs will be set to the pPieces values. |
|
634 */ |
|
635 static void |
|
636 parseConverterOptions(const char *inName, |
|
637 UConverterNamePieces *pPieces, |
|
638 UConverterLoadArgs *pArgs, |
|
639 UErrorCode *err) |
|
640 { |
|
641 char *cnvName = pPieces->cnvName; |
|
642 char c; |
|
643 int32_t len = 0; |
|
644 |
|
645 pArgs->name=inName; |
|
646 pArgs->locale=pPieces->locale; |
|
647 pArgs->options=pPieces->options; |
|
648 |
|
649 /* copy the converter name itself to cnvName */ |
|
650 while((c=*inName)!=0 && c!=UCNV_OPTION_SEP_CHAR) { |
|
651 if (++len>=UCNV_MAX_CONVERTER_NAME_LENGTH) { |
|
652 *err = U_ILLEGAL_ARGUMENT_ERROR; /* bad name */ |
|
653 pPieces->cnvName[0]=0; |
|
654 return; |
|
655 } |
|
656 *cnvName++=c; |
|
657 inName++; |
|
658 } |
|
659 *cnvName=0; |
|
660 pArgs->name=pPieces->cnvName; |
|
661 |
|
662 /* parse options. No more name copying should occur. */ |
|
663 while((c=*inName)!=0) { |
|
664 if(c==UCNV_OPTION_SEP_CHAR) { |
|
665 ++inName; |
|
666 } |
|
667 |
|
668 /* inName is behind an option separator */ |
|
669 if(uprv_strncmp(inName, "locale=", 7)==0) { |
|
670 /* do not modify locale itself in case we have multiple locale options */ |
|
671 char *dest=pPieces->locale; |
|
672 |
|
673 /* copy the locale option value */ |
|
674 inName+=7; |
|
675 len=0; |
|
676 while((c=*inName)!=0 && c!=UCNV_OPTION_SEP_CHAR) { |
|
677 ++inName; |
|
678 |
|
679 if(++len>=ULOC_FULLNAME_CAPACITY) { |
|
680 *err=U_ILLEGAL_ARGUMENT_ERROR; /* bad name */ |
|
681 pPieces->locale[0]=0; |
|
682 return; |
|
683 } |
|
684 |
|
685 *dest++=c; |
|
686 } |
|
687 *dest=0; |
|
688 } else if(uprv_strncmp(inName, "version=", 8)==0) { |
|
689 /* copy the version option value into bits 3..0 of pPieces->options */ |
|
690 inName+=8; |
|
691 c=*inName; |
|
692 if(c==0) { |
|
693 pArgs->options=(pPieces->options&=~UCNV_OPTION_VERSION); |
|
694 return; |
|
695 } else if((uint8_t)(c-'0')<10) { |
|
696 pArgs->options=pPieces->options=(pPieces->options&~UCNV_OPTION_VERSION)|(uint32_t)(c-'0'); |
|
697 ++inName; |
|
698 } |
|
699 } else if(uprv_strncmp(inName, "swaplfnl", 8)==0) { |
|
700 inName+=8; |
|
701 pArgs->options=(pPieces->options|=UCNV_OPTION_SWAP_LFNL); |
|
702 /* add processing for new options here with another } else if(uprv_strncmp(inName, "option-name=", XX)==0) { */ |
|
703 } else { |
|
704 /* ignore any other options until we define some */ |
|
705 while(((c = *inName++) != 0) && (c != UCNV_OPTION_SEP_CHAR)) { |
|
706 } |
|
707 if(c==0) { |
|
708 return; |
|
709 } |
|
710 } |
|
711 } |
|
712 } |
|
713 |
|
714 /*Logic determines if the converter is Algorithmic AND/OR cached |
|
715 *depending on that: |
|
716 * -we either go to get data from disk and cache it (Data=TRUE, Cached=False) |
|
717 * -Get it from a Hashtable (Data=X, Cached=TRUE) |
|
718 * -Call dataConverter initializer (Data=TRUE, Cached=TRUE) |
|
719 * -Call AlgorithmicConverter initializer (Data=FALSE, Cached=TRUE) |
|
720 */ |
|
721 U_CFUNC UConverterSharedData * |
|
722 ucnv_loadSharedData(const char *converterName, |
|
723 UConverterNamePieces *pPieces, |
|
724 UConverterLoadArgs *pArgs, |
|
725 UErrorCode * err) { |
|
726 UConverterNamePieces stackPieces; |
|
727 UConverterLoadArgs stackArgs; |
|
728 UConverterSharedData *mySharedConverterData = NULL; |
|
729 UErrorCode internalErrorCode = U_ZERO_ERROR; |
|
730 UBool mayContainOption = TRUE; |
|
731 UBool checkForAlgorithmic = TRUE; |
|
732 |
|
733 if (U_FAILURE (*err)) { |
|
734 return NULL; |
|
735 } |
|
736 |
|
737 if(pPieces == NULL) { |
|
738 if(pArgs != NULL) { |
|
739 /* |
|
740 * Bad: We may set pArgs pointers to stackPieces fields |
|
741 * which will be invalid after this function returns. |
|
742 */ |
|
743 *err = U_INTERNAL_PROGRAM_ERROR; |
|
744 return NULL; |
|
745 } |
|
746 pPieces = &stackPieces; |
|
747 } |
|
748 if(pArgs == NULL) { |
|
749 uprv_memset(&stackArgs, 0, sizeof(stackArgs)); |
|
750 stackArgs.size = (int32_t)sizeof(stackArgs); |
|
751 pArgs = &stackArgs; |
|
752 } |
|
753 |
|
754 pPieces->cnvName[0] = 0; |
|
755 pPieces->locale[0] = 0; |
|
756 pPieces->options = 0; |
|
757 |
|
758 pArgs->name = converterName; |
|
759 pArgs->locale = pPieces->locale; |
|
760 pArgs->options = pPieces->options; |
|
761 |
|
762 /* In case "name" is NULL we want to open the default converter. */ |
|
763 if (converterName == NULL) { |
|
764 #if U_CHARSET_IS_UTF8 |
|
765 pArgs->name = "UTF-8"; |
|
766 return (UConverterSharedData *)converterData[UCNV_UTF8]; |
|
767 #else |
|
768 /* Call ucnv_getDefaultName first to query the name from the OS. */ |
|
769 pArgs->name = ucnv_getDefaultName(); |
|
770 if (pArgs->name == NULL) { |
|
771 *err = U_MISSING_RESOURCE_ERROR; |
|
772 return NULL; |
|
773 } |
|
774 mySharedConverterData = (UConverterSharedData *)gDefaultAlgorithmicSharedData; |
|
775 checkForAlgorithmic = FALSE; |
|
776 mayContainOption = gDefaultConverterContainsOption; |
|
777 /* the default converter name is already canonical */ |
|
778 #endif |
|
779 } |
|
780 else if(UCNV_FAST_IS_UTF8(converterName)) { |
|
781 /* fastpath for UTF-8 */ |
|
782 pArgs->name = "UTF-8"; |
|
783 return (UConverterSharedData *)converterData[UCNV_UTF8]; |
|
784 } |
|
785 else { |
|
786 /* separate the converter name from the options */ |
|
787 parseConverterOptions(converterName, pPieces, pArgs, err); |
|
788 if (U_FAILURE(*err)) { |
|
789 /* Very bad name used. */ |
|
790 return NULL; |
|
791 } |
|
792 |
|
793 /* get the canonical converter name */ |
|
794 pArgs->name = ucnv_io_getConverterName(pArgs->name, &mayContainOption, &internalErrorCode); |
|
795 if (U_FAILURE(internalErrorCode) || pArgs->name == NULL) { |
|
796 /* |
|
797 * set the input name in case the converter was added |
|
798 * without updating the alias table, or when there is no alias table |
|
799 */ |
|
800 pArgs->name = pPieces->cnvName; |
|
801 } else if (internalErrorCode == U_AMBIGUOUS_ALIAS_WARNING) { |
|
802 *err = U_AMBIGUOUS_ALIAS_WARNING; |
|
803 } |
|
804 } |
|
805 |
|
806 /* separate the converter name from the options */ |
|
807 if(mayContainOption && pArgs->name != pPieces->cnvName) { |
|
808 parseConverterOptions(pArgs->name, pPieces, pArgs, err); |
|
809 } |
|
810 |
|
811 /* get the shared data for an algorithmic converter, if it is one */ |
|
812 if (checkForAlgorithmic) { |
|
813 mySharedConverterData = (UConverterSharedData *)getAlgorithmicTypeFromName(pArgs->name); |
|
814 } |
|
815 if (mySharedConverterData == NULL) |
|
816 { |
|
817 /* it is a data-based converter, get its shared data. */ |
|
818 /* Hold the cnvCacheMutex through the whole process of checking the */ |
|
819 /* converter data cache, and adding new entries to the cache */ |
|
820 /* to prevent other threads from modifying the cache during the */ |
|
821 /* process. */ |
|
822 pArgs->nestedLoads=1; |
|
823 pArgs->pkg=NULL; |
|
824 |
|
825 umtx_lock(&cnvCacheMutex); |
|
826 mySharedConverterData = ucnv_load(pArgs, err); |
|
827 umtx_unlock(&cnvCacheMutex); |
|
828 if (U_FAILURE (*err) || (mySharedConverterData == NULL)) |
|
829 { |
|
830 return NULL; |
|
831 } |
|
832 } |
|
833 |
|
834 return mySharedConverterData; |
|
835 } |
|
836 |
|
837 U_CAPI UConverter * |
|
838 ucnv_createConverter(UConverter *myUConverter, const char *converterName, UErrorCode * err) |
|
839 { |
|
840 UConverterNamePieces stackPieces; |
|
841 UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
|
842 UConverterSharedData *mySharedConverterData; |
|
843 |
|
844 UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN); |
|
845 |
|
846 if(U_SUCCESS(*err)) { |
|
847 UTRACE_DATA1(UTRACE_OPEN_CLOSE, "open converter %s", converterName); |
|
848 |
|
849 mySharedConverterData = ucnv_loadSharedData(converterName, &stackPieces, &stackArgs, err); |
|
850 |
|
851 myUConverter = ucnv_createConverterFromSharedData( |
|
852 myUConverter, mySharedConverterData, |
|
853 &stackArgs, |
|
854 err); |
|
855 |
|
856 if(U_SUCCESS(*err)) { |
|
857 UTRACE_EXIT_PTR_STATUS(myUConverter, *err); |
|
858 return myUConverter; |
|
859 } |
|
860 } |
|
861 |
|
862 /* exit with error */ |
|
863 UTRACE_EXIT_STATUS(*err); |
|
864 return NULL; |
|
865 } |
|
866 |
|
867 U_CFUNC UBool |
|
868 ucnv_canCreateConverter(const char *converterName, UErrorCode *err) { |
|
869 UConverter myUConverter; |
|
870 UConverterNamePieces stackPieces; |
|
871 UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
|
872 UConverterSharedData *mySharedConverterData; |
|
873 |
|
874 UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN); |
|
875 |
|
876 if(U_SUCCESS(*err)) { |
|
877 UTRACE_DATA1(UTRACE_OPEN_CLOSE, "test if can open converter %s", converterName); |
|
878 |
|
879 stackArgs.onlyTestIsLoadable=TRUE; |
|
880 mySharedConverterData = ucnv_loadSharedData(converterName, &stackPieces, &stackArgs, err); |
|
881 ucnv_createConverterFromSharedData( |
|
882 &myUConverter, mySharedConverterData, |
|
883 &stackArgs, |
|
884 err); |
|
885 ucnv_unloadSharedDataIfReady(mySharedConverterData); |
|
886 } |
|
887 |
|
888 UTRACE_EXIT_STATUS(*err); |
|
889 return U_SUCCESS(*err); |
|
890 } |
|
891 |
|
892 UConverter * |
|
893 ucnv_createAlgorithmicConverter(UConverter *myUConverter, |
|
894 UConverterType type, |
|
895 const char *locale, uint32_t options, |
|
896 UErrorCode *err) { |
|
897 UConverter *cnv; |
|
898 const UConverterSharedData *sharedData; |
|
899 UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
|
900 |
|
901 UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN_ALGORITHMIC); |
|
902 UTRACE_DATA1(UTRACE_OPEN_CLOSE, "open algorithmic converter type %d", (int32_t)type); |
|
903 |
|
904 if(type<0 || UCNV_NUMBER_OF_SUPPORTED_CONVERTER_TYPES<=type) { |
|
905 *err = U_ILLEGAL_ARGUMENT_ERROR; |
|
906 UTRACE_EXIT_STATUS(U_ILLEGAL_ARGUMENT_ERROR); |
|
907 return NULL; |
|
908 } |
|
909 |
|
910 sharedData = converterData[type]; |
|
911 /* |
|
912 Checking whether it's an algorithic converter is okay |
|
913 in multithreaded applications because the value never changes. |
|
914 Don't check referenceCounter for any other value. |
|
915 */ |
|
916 if(sharedData == NULL || sharedData->referenceCounter != (uint32_t)~0) { |
|
917 /* not a valid type, or not an algorithmic converter */ |
|
918 *err = U_ILLEGAL_ARGUMENT_ERROR; |
|
919 UTRACE_EXIT_STATUS(U_ILLEGAL_ARGUMENT_ERROR); |
|
920 return NULL; |
|
921 } |
|
922 |
|
923 stackArgs.name = ""; |
|
924 stackArgs.options = options; |
|
925 stackArgs.locale=locale; |
|
926 cnv = ucnv_createConverterFromSharedData( |
|
927 myUConverter, (UConverterSharedData *)sharedData, |
|
928 &stackArgs, err); |
|
929 |
|
930 UTRACE_EXIT_PTR_STATUS(cnv, *err); |
|
931 return cnv; |
|
932 } |
|
933 |
|
934 U_CFUNC UConverter* |
|
935 ucnv_createConverterFromPackage(const char *packageName, const char *converterName, UErrorCode * err) |
|
936 { |
|
937 UConverter *myUConverter; |
|
938 UConverterSharedData *mySharedConverterData; |
|
939 UConverterNamePieces stackPieces; |
|
940 UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
|
941 |
|
942 UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN_PACKAGE); |
|
943 |
|
944 if(U_FAILURE(*err)) { |
|
945 UTRACE_EXIT_STATUS(*err); |
|
946 return NULL; |
|
947 } |
|
948 |
|
949 UTRACE_DATA2(UTRACE_OPEN_CLOSE, "open converter %s from package %s", converterName, packageName); |
|
950 |
|
951 /* first, get the options out of the converterName string */ |
|
952 stackPieces.cnvName[0] = 0; |
|
953 stackPieces.locale[0] = 0; |
|
954 stackPieces.options = 0; |
|
955 parseConverterOptions(converterName, &stackPieces, &stackArgs, err); |
|
956 if (U_FAILURE(*err)) { |
|
957 /* Very bad name used. */ |
|
958 UTRACE_EXIT_STATUS(*err); |
|
959 return NULL; |
|
960 } |
|
961 stackArgs.nestedLoads=1; |
|
962 stackArgs.pkg=packageName; |
|
963 |
|
964 /* open the data, unflatten the shared structure */ |
|
965 mySharedConverterData = createConverterFromFile(&stackArgs, err); |
|
966 |
|
967 if (U_FAILURE(*err)) { |
|
968 UTRACE_EXIT_STATUS(*err); |
|
969 return NULL; |
|
970 } |
|
971 |
|
972 /* create the actual converter */ |
|
973 myUConverter = ucnv_createConverterFromSharedData(NULL, mySharedConverterData, &stackArgs, err); |
|
974 |
|
975 if (U_FAILURE(*err)) { |
|
976 ucnv_close(myUConverter); |
|
977 UTRACE_EXIT_STATUS(*err); |
|
978 return NULL; |
|
979 } |
|
980 |
|
981 UTRACE_EXIT_PTR_STATUS(myUConverter, *err); |
|
982 return myUConverter; |
|
983 } |
|
984 |
|
985 |
|
986 U_CFUNC UConverter* |
|
987 ucnv_createConverterFromSharedData(UConverter *myUConverter, |
|
988 UConverterSharedData *mySharedConverterData, |
|
989 UConverterLoadArgs *pArgs, |
|
990 UErrorCode *err) |
|
991 { |
|
992 UBool isCopyLocal; |
|
993 |
|
994 if(U_FAILURE(*err)) { |
|
995 ucnv_unloadSharedDataIfReady(mySharedConverterData); |
|
996 return myUConverter; |
|
997 } |
|
998 if(myUConverter == NULL) |
|
999 { |
|
1000 myUConverter = (UConverter *) uprv_malloc (sizeof (UConverter)); |
|
1001 if(myUConverter == NULL) |
|
1002 { |
|
1003 *err = U_MEMORY_ALLOCATION_ERROR; |
|
1004 ucnv_unloadSharedDataIfReady(mySharedConverterData); |
|
1005 return NULL; |
|
1006 } |
|
1007 isCopyLocal = FALSE; |
|
1008 } else { |
|
1009 isCopyLocal = TRUE; |
|
1010 } |
|
1011 |
|
1012 /* initialize the converter */ |
|
1013 uprv_memset(myUConverter, 0, sizeof(UConverter)); |
|
1014 myUConverter->isCopyLocal = isCopyLocal; |
|
1015 /*myUConverter->isExtraLocal = FALSE;*/ /* Set by the memset call */ |
|
1016 myUConverter->sharedData = mySharedConverterData; |
|
1017 myUConverter->options = pArgs->options; |
|
1018 if(!pArgs->onlyTestIsLoadable) { |
|
1019 myUConverter->preFromUFirstCP = U_SENTINEL; |
|
1020 myUConverter->fromCharErrorBehaviour = UCNV_TO_U_DEFAULT_CALLBACK; |
|
1021 myUConverter->fromUCharErrorBehaviour = UCNV_FROM_U_DEFAULT_CALLBACK; |
|
1022 myUConverter->toUnicodeStatus = mySharedConverterData->toUnicodeStatus; |
|
1023 myUConverter->maxBytesPerUChar = mySharedConverterData->staticData->maxBytesPerChar; |
|
1024 myUConverter->subChar1 = mySharedConverterData->staticData->subChar1; |
|
1025 myUConverter->subCharLen = mySharedConverterData->staticData->subCharLen; |
|
1026 myUConverter->subChars = (uint8_t *)myUConverter->subUChars; |
|
1027 uprv_memcpy(myUConverter->subChars, mySharedConverterData->staticData->subChar, myUConverter->subCharLen); |
|
1028 myUConverter->toUCallbackReason = UCNV_ILLEGAL; /* default reason to invoke (*fromCharErrorBehaviour) */ |
|
1029 } |
|
1030 |
|
1031 if(mySharedConverterData->impl->open != NULL) { |
|
1032 mySharedConverterData->impl->open(myUConverter, pArgs, err); |
|
1033 if(U_FAILURE(*err) && !pArgs->onlyTestIsLoadable) { |
|
1034 /* don't ucnv_close() if onlyTestIsLoadable because not fully initialized */ |
|
1035 ucnv_close(myUConverter); |
|
1036 return NULL; |
|
1037 } |
|
1038 } |
|
1039 |
|
1040 return myUConverter; |
|
1041 } |
|
1042 |
|
1043 /*Frees all shared immutable objects that aren't referred to (reference count = 0) |
|
1044 */ |
|
1045 U_CAPI int32_t U_EXPORT2 |
|
1046 ucnv_flushCache () |
|
1047 { |
|
1048 UConverterSharedData *mySharedData = NULL; |
|
1049 int32_t pos; |
|
1050 int32_t tableDeletedNum = 0; |
|
1051 const UHashElement *e; |
|
1052 /*UErrorCode status = U_ILLEGAL_ARGUMENT_ERROR;*/ |
|
1053 int32_t i, remaining; |
|
1054 |
|
1055 UTRACE_ENTRY_OC(UTRACE_UCNV_FLUSH_CACHE); |
|
1056 |
|
1057 /* Close the default converter without creating a new one so that everything will be flushed. */ |
|
1058 u_flushDefaultConverter(); |
|
1059 |
|
1060 /*if shared data hasn't even been lazy evaluated yet |
|
1061 * return 0 |
|
1062 */ |
|
1063 if (SHARED_DATA_HASHTABLE == NULL) { |
|
1064 UTRACE_EXIT_VALUE((int32_t)0); |
|
1065 return 0; |
|
1066 } |
|
1067 |
|
1068 /*creates an enumeration to iterate through every element in the |
|
1069 * table |
|
1070 * |
|
1071 * Synchronization: holding cnvCacheMutex will prevent any other thread from |
|
1072 * accessing or modifying the hash table during the iteration. |
|
1073 * The reference count of an entry may be decremented by |
|
1074 * ucnv_close while the iteration is in process, but this is |
|
1075 * benign. It can't be incremented (in ucnv_createConverter()) |
|
1076 * because the sequence of looking up in the cache + incrementing |
|
1077 * is protected by cnvCacheMutex. |
|
1078 */ |
|
1079 umtx_lock(&cnvCacheMutex); |
|
1080 /* |
|
1081 * double loop: A delta/extension-only converter has a pointer to its base table's |
|
1082 * shared data; the first iteration of the outer loop may see the delta converter |
|
1083 * before the base converter, and unloading the delta converter may get the base |
|
1084 * converter's reference counter down to 0. |
|
1085 */ |
|
1086 i = 0; |
|
1087 do { |
|
1088 remaining = 0; |
|
1089 pos = -1; |
|
1090 while ((e = uhash_nextElement (SHARED_DATA_HASHTABLE, &pos)) != NULL) |
|
1091 { |
|
1092 mySharedData = (UConverterSharedData *) e->value.pointer; |
|
1093 /*deletes only if reference counter == 0 */ |
|
1094 if (mySharedData->referenceCounter == 0) |
|
1095 { |
|
1096 tableDeletedNum++; |
|
1097 |
|
1098 UCNV_DEBUG_LOG("del",mySharedData->staticData->name,mySharedData); |
|
1099 |
|
1100 uhash_removeElement(SHARED_DATA_HASHTABLE, e); |
|
1101 mySharedData->sharedDataCached = FALSE; |
|
1102 ucnv_deleteSharedConverterData (mySharedData); |
|
1103 } else { |
|
1104 ++remaining; |
|
1105 } |
|
1106 } |
|
1107 } while(++i == 1 && remaining > 0); |
|
1108 umtx_unlock(&cnvCacheMutex); |
|
1109 |
|
1110 UTRACE_DATA1(UTRACE_INFO, "ucnv_flushCache() exits with %d converters remaining", remaining); |
|
1111 |
|
1112 UTRACE_EXIT_VALUE(tableDeletedNum); |
|
1113 return tableDeletedNum; |
|
1114 } |
|
1115 |
|
1116 /* available converters list --------------------------------------------------- */ |
|
1117 |
|
1118 static void U_CALLCONV initAvailableConvertersList(UErrorCode &errCode) { |
|
1119 U_ASSERT(gAvailableConverterCount == 0); |
|
1120 U_ASSERT(gAvailableConverters == NULL); |
|
1121 |
|
1122 ucln_common_registerCleanup(UCLN_COMMON_UCNV, ucnv_cleanup); |
|
1123 UEnumeration *allConvEnum = ucnv_openAllNames(&errCode); |
|
1124 int32_t allConverterCount = uenum_count(allConvEnum, &errCode); |
|
1125 if (U_FAILURE(errCode)) { |
|
1126 return; |
|
1127 } |
|
1128 |
|
1129 /* We can't have more than "*converterTable" converters to open */ |
|
1130 gAvailableConverters = (const char **) uprv_malloc(allConverterCount * sizeof(char*)); |
|
1131 if (!gAvailableConverters) { |
|
1132 errCode = U_MEMORY_ALLOCATION_ERROR; |
|
1133 return; |
|
1134 } |
|
1135 |
|
1136 /* Open the default converter to make sure that it has first dibs in the hash table. */ |
|
1137 UErrorCode localStatus = U_ZERO_ERROR; |
|
1138 UConverter tempConverter; |
|
1139 ucnv_close(ucnv_createConverter(&tempConverter, NULL, &localStatus)); |
|
1140 |
|
1141 gAvailableConverterCount = 0; |
|
1142 |
|
1143 for (int32_t idx = 0; idx < allConverterCount; idx++) { |
|
1144 localStatus = U_ZERO_ERROR; |
|
1145 const char *converterName = uenum_next(allConvEnum, NULL, &localStatus); |
|
1146 if (ucnv_canCreateConverter(converterName, &localStatus)) { |
|
1147 gAvailableConverters[gAvailableConverterCount++] = converterName; |
|
1148 } |
|
1149 } |
|
1150 |
|
1151 uenum_close(allConvEnum); |
|
1152 } |
|
1153 |
|
1154 |
|
1155 static UBool haveAvailableConverterList(UErrorCode *pErrorCode) { |
|
1156 umtx_initOnce(gAvailableConvertersInitOnce, &initAvailableConvertersList, *pErrorCode); |
|
1157 return U_SUCCESS(*pErrorCode); |
|
1158 } |
|
1159 |
|
1160 U_CFUNC uint16_t |
|
1161 ucnv_bld_countAvailableConverters(UErrorCode *pErrorCode) { |
|
1162 if (haveAvailableConverterList(pErrorCode)) { |
|
1163 return gAvailableConverterCount; |
|
1164 } |
|
1165 return 0; |
|
1166 } |
|
1167 |
|
1168 U_CFUNC const char * |
|
1169 ucnv_bld_getAvailableConverter(uint16_t n, UErrorCode *pErrorCode) { |
|
1170 if (haveAvailableConverterList(pErrorCode)) { |
|
1171 if (n < gAvailableConverterCount) { |
|
1172 return gAvailableConverters[n]; |
|
1173 } |
|
1174 *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR; |
|
1175 } |
|
1176 return NULL; |
|
1177 } |
|
1178 |
|
1179 /* default converter name --------------------------------------------------- */ |
|
1180 |
|
1181 #if !U_CHARSET_IS_UTF8 |
|
1182 /* |
|
1183 Copy the canonical converter name. |
|
1184 ucnv_getDefaultName must be thread safe, which can call this function. |
|
1185 |
|
1186 ucnv_setDefaultName calls this function and it doesn't have to be |
|
1187 thread safe because there is no reliable/safe way to reset the |
|
1188 converter in use in all threads. If you did reset the converter, you |
|
1189 would not be sure that retrieving a default converter for one string |
|
1190 would be the same type of default converter for a successive string. |
|
1191 Since the name is a returned via ucnv_getDefaultName without copying, |
|
1192 you shouldn't be modifying or deleting the string from a separate thread. |
|
1193 */ |
|
1194 static inline void |
|
1195 internalSetName(const char *name, UErrorCode *status) { |
|
1196 UConverterNamePieces stackPieces; |
|
1197 UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
|
1198 int32_t length=(int32_t)(uprv_strlen(name)); |
|
1199 UBool containsOption = (UBool)(uprv_strchr(name, UCNV_OPTION_SEP_CHAR) != NULL); |
|
1200 const UConverterSharedData *algorithmicSharedData; |
|
1201 |
|
1202 stackArgs.name = name; |
|
1203 if(containsOption) { |
|
1204 stackPieces.cnvName[0] = 0; |
|
1205 stackPieces.locale[0] = 0; |
|
1206 stackPieces.options = 0; |
|
1207 parseConverterOptions(name, &stackPieces, &stackArgs, status); |
|
1208 if(U_FAILURE(*status)) { |
|
1209 return; |
|
1210 } |
|
1211 } |
|
1212 algorithmicSharedData = getAlgorithmicTypeFromName(stackArgs.name); |
|
1213 |
|
1214 umtx_lock(&cnvCacheMutex); |
|
1215 |
|
1216 gDefaultAlgorithmicSharedData = algorithmicSharedData; |
|
1217 gDefaultConverterContainsOption = containsOption; |
|
1218 uprv_memcpy(gDefaultConverterNameBuffer, name, length); |
|
1219 gDefaultConverterNameBuffer[length]=0; |
|
1220 |
|
1221 /* gDefaultConverterName MUST be the last global var set by this function. */ |
|
1222 /* It is the variable checked in ucnv_getDefaultName() to see if initialization is required. */ |
|
1223 // But there is nothing here preventing that from being reordered, either by the compiler |
|
1224 // or hardware. I'm adding the mutex to ucnv_getDefaultName for now. UMTX_CHECK is not enough. |
|
1225 // -- Andy |
|
1226 gDefaultConverterName = gDefaultConverterNameBuffer; |
|
1227 |
|
1228 ucln_common_registerCleanup(UCLN_COMMON_UCNV, ucnv_cleanup); |
|
1229 |
|
1230 umtx_unlock(&cnvCacheMutex); |
|
1231 } |
|
1232 #endif |
|
1233 |
|
1234 /* |
|
1235 * In order to be really thread-safe, the get function would have to take |
|
1236 * a buffer parameter and copy the current string inside a mutex block. |
|
1237 * This implementation only tries to be really thread-safe while |
|
1238 * setting the name. |
|
1239 * It assumes that setting a pointer is atomic. |
|
1240 */ |
|
1241 |
|
1242 U_CAPI const char* U_EXPORT2 |
|
1243 ucnv_getDefaultName() { |
|
1244 #if U_CHARSET_IS_UTF8 |
|
1245 return "UTF-8"; |
|
1246 #else |
|
1247 /* local variable to be thread-safe */ |
|
1248 const char *name; |
|
1249 |
|
1250 /* |
|
1251 Concurrent calls to ucnv_getDefaultName must be thread safe, |
|
1252 but ucnv_setDefaultName is not thread safe. |
|
1253 */ |
|
1254 { |
|
1255 icu::Mutex lock(&cnvCacheMutex); |
|
1256 name = gDefaultConverterName; |
|
1257 } |
|
1258 if(name==NULL) { |
|
1259 UErrorCode errorCode = U_ZERO_ERROR; |
|
1260 UConverter *cnv = NULL; |
|
1261 |
|
1262 name = uprv_getDefaultCodepage(); |
|
1263 |
|
1264 /* if the name is there, test it out and get the canonical name with options */ |
|
1265 if(name != NULL) { |
|
1266 cnv = ucnv_open(name, &errorCode); |
|
1267 if(U_SUCCESS(errorCode) && cnv != NULL) { |
|
1268 name = ucnv_getName(cnv, &errorCode); |
|
1269 } |
|
1270 } |
|
1271 |
|
1272 if(name == NULL || name[0] == 0 |
|
1273 || U_FAILURE(errorCode) || cnv == NULL |
|
1274 || uprv_strlen(name)>=sizeof(gDefaultConverterNameBuffer)) |
|
1275 { |
|
1276 /* Panic time, let's use a fallback. */ |
|
1277 #if (U_CHARSET_FAMILY == U_ASCII_FAMILY) |
|
1278 name = "US-ASCII"; |
|
1279 /* there is no 'algorithmic' converter for EBCDIC */ |
|
1280 #elif U_PLATFORM == U_PF_OS390 |
|
1281 name = "ibm-1047_P100-1995" UCNV_SWAP_LFNL_OPTION_STRING; |
|
1282 #else |
|
1283 name = "ibm-37_P100-1995"; |
|
1284 #endif |
|
1285 } |
|
1286 |
|
1287 internalSetName(name, &errorCode); |
|
1288 |
|
1289 /* The close may make the current name go away. */ |
|
1290 ucnv_close(cnv); |
|
1291 } |
|
1292 |
|
1293 return name; |
|
1294 #endif |
|
1295 } |
|
1296 |
|
1297 #if U_CHARSET_IS_UTF8 |
|
1298 U_CAPI void U_EXPORT2 ucnv_setDefaultName(const char *) {} |
|
1299 #else |
|
1300 /* |
|
1301 This function is not thread safe, and it can't be thread safe. |
|
1302 See internalSetName or the API reference for details. |
|
1303 */ |
|
1304 U_CAPI void U_EXPORT2 |
|
1305 ucnv_setDefaultName(const char *converterName) { |
|
1306 if(converterName==NULL) { |
|
1307 /* reset to the default codepage */ |
|
1308 gDefaultConverterName=NULL; |
|
1309 } else { |
|
1310 UErrorCode errorCode = U_ZERO_ERROR; |
|
1311 UConverter *cnv = NULL; |
|
1312 const char *name = NULL; |
|
1313 |
|
1314 /* if the name is there, test it out and get the canonical name with options */ |
|
1315 cnv = ucnv_open(converterName, &errorCode); |
|
1316 if(U_SUCCESS(errorCode) && cnv != NULL) { |
|
1317 name = ucnv_getName(cnv, &errorCode); |
|
1318 } |
|
1319 |
|
1320 if(U_SUCCESS(errorCode) && name!=NULL) { |
|
1321 internalSetName(name, &errorCode); |
|
1322 } |
|
1323 /* else this converter is bad to use. Don't change it to a bad value. */ |
|
1324 |
|
1325 /* The close may make the current name go away. */ |
|
1326 ucnv_close(cnv); |
|
1327 |
|
1328 /* reset the converter cache */ |
|
1329 u_flushDefaultConverter(); |
|
1330 } |
|
1331 } |
|
1332 #endif |
|
1333 |
|
1334 /* data swapping ------------------------------------------------------------ */ |
|
1335 |
|
1336 /* most of this might belong more properly into ucnvmbcs.c, but that is so large */ |
|
1337 |
|
1338 #if !UCONFIG_NO_LEGACY_CONVERSION |
|
1339 |
|
1340 U_CAPI int32_t U_EXPORT2 |
|
1341 ucnv_swap(const UDataSwapper *ds, |
|
1342 const void *inData, int32_t length, void *outData, |
|
1343 UErrorCode *pErrorCode) { |
|
1344 const UDataInfo *pInfo; |
|
1345 int32_t headerSize; |
|
1346 |
|
1347 const uint8_t *inBytes; |
|
1348 uint8_t *outBytes; |
|
1349 |
|
1350 uint32_t offset, count, staticDataSize; |
|
1351 int32_t size; |
|
1352 |
|
1353 const UConverterStaticData *inStaticData; |
|
1354 UConverterStaticData *outStaticData; |
|
1355 |
|
1356 const _MBCSHeader *inMBCSHeader; |
|
1357 _MBCSHeader *outMBCSHeader; |
|
1358 _MBCSHeader mbcsHeader; |
|
1359 uint32_t mbcsHeaderLength; |
|
1360 UBool noFromU=FALSE; |
|
1361 |
|
1362 uint8_t outputType; |
|
1363 |
|
1364 int32_t maxFastUChar, mbcsIndexLength; |
|
1365 |
|
1366 const int32_t *inExtIndexes; |
|
1367 int32_t extOffset; |
|
1368 |
|
1369 /* udata_swapDataHeader checks the arguments */ |
|
1370 headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode); |
|
1371 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
|
1372 return 0; |
|
1373 } |
|
1374 |
|
1375 /* check data format and format version */ |
|
1376 pInfo=(const UDataInfo *)((const char *)inData+4); |
|
1377 if(!( |
|
1378 pInfo->dataFormat[0]==0x63 && /* dataFormat="cnvt" */ |
|
1379 pInfo->dataFormat[1]==0x6e && |
|
1380 pInfo->dataFormat[2]==0x76 && |
|
1381 pInfo->dataFormat[3]==0x74 && |
|
1382 pInfo->formatVersion[0]==6 && |
|
1383 pInfo->formatVersion[1]>=2 |
|
1384 )) { |
|
1385 udata_printError(ds, "ucnv_swap(): data format %02x.%02x.%02x.%02x (format version %02x.%02x) is not recognized as an ICU .cnv conversion table\n", |
|
1386 pInfo->dataFormat[0], pInfo->dataFormat[1], |
|
1387 pInfo->dataFormat[2], pInfo->dataFormat[3], |
|
1388 pInfo->formatVersion[0], pInfo->formatVersion[1]); |
|
1389 *pErrorCode=U_UNSUPPORTED_ERROR; |
|
1390 return 0; |
|
1391 } |
|
1392 |
|
1393 inBytes=(const uint8_t *)inData+headerSize; |
|
1394 outBytes=(uint8_t *)outData+headerSize; |
|
1395 |
|
1396 /* read the initial UConverterStaticData structure after the UDataInfo header */ |
|
1397 inStaticData=(const UConverterStaticData *)inBytes; |
|
1398 outStaticData=(UConverterStaticData *)outBytes; |
|
1399 |
|
1400 if(length<0) { |
|
1401 staticDataSize=ds->readUInt32(inStaticData->structSize); |
|
1402 } else { |
|
1403 length-=headerSize; |
|
1404 if( length<(int32_t)sizeof(UConverterStaticData) || |
|
1405 (uint32_t)length<(staticDataSize=ds->readUInt32(inStaticData->structSize)) |
|
1406 ) { |
|
1407 udata_printError(ds, "ucnv_swap(): too few bytes (%d after header) for an ICU .cnv conversion table\n", |
|
1408 length); |
|
1409 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
|
1410 return 0; |
|
1411 } |
|
1412 } |
|
1413 |
|
1414 if(length>=0) { |
|
1415 /* swap the static data */ |
|
1416 if(inStaticData!=outStaticData) { |
|
1417 uprv_memcpy(outStaticData, inStaticData, staticDataSize); |
|
1418 } |
|
1419 |
|
1420 ds->swapArray32(ds, &inStaticData->structSize, 4, |
|
1421 &outStaticData->structSize, pErrorCode); |
|
1422 ds->swapArray32(ds, &inStaticData->codepage, 4, |
|
1423 &outStaticData->codepage, pErrorCode); |
|
1424 |
|
1425 ds->swapInvChars(ds, inStaticData->name, (int32_t)uprv_strlen(inStaticData->name), |
|
1426 outStaticData->name, pErrorCode); |
|
1427 if(U_FAILURE(*pErrorCode)) { |
|
1428 udata_printError(ds, "ucnv_swap(): error swapping converter name\n"); |
|
1429 return 0; |
|
1430 } |
|
1431 } |
|
1432 |
|
1433 inBytes+=staticDataSize; |
|
1434 outBytes+=staticDataSize; |
|
1435 if(length>=0) { |
|
1436 length-=(int32_t)staticDataSize; |
|
1437 } |
|
1438 |
|
1439 /* check for supported conversionType values */ |
|
1440 if(inStaticData->conversionType==UCNV_MBCS) { |
|
1441 /* swap MBCS data */ |
|
1442 inMBCSHeader=(const _MBCSHeader *)inBytes; |
|
1443 outMBCSHeader=(_MBCSHeader *)outBytes; |
|
1444 |
|
1445 if(0<=length && length<(int32_t)sizeof(_MBCSHeader)) { |
|
1446 udata_printError(ds, "ucnv_swap(): too few bytes (%d after headers) for an ICU MBCS .cnv conversion table\n", |
|
1447 length); |
|
1448 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
|
1449 return 0; |
|
1450 } |
|
1451 if(inMBCSHeader->version[0]==4 && inMBCSHeader->version[1]>=1) { |
|
1452 mbcsHeaderLength=MBCS_HEADER_V4_LENGTH; |
|
1453 } else if(inMBCSHeader->version[0]==5 && inMBCSHeader->version[1]>=3 && |
|
1454 ((mbcsHeader.options=ds->readUInt32(inMBCSHeader->options))& |
|
1455 MBCS_OPT_UNKNOWN_INCOMPATIBLE_MASK)==0 |
|
1456 ) { |
|
1457 mbcsHeaderLength=mbcsHeader.options&MBCS_OPT_LENGTH_MASK; |
|
1458 noFromU=(UBool)((mbcsHeader.options&MBCS_OPT_NO_FROM_U)!=0); |
|
1459 } else { |
|
1460 udata_printError(ds, "ucnv_swap(): unsupported _MBCSHeader.version %d.%d\n", |
|
1461 inMBCSHeader->version[0], inMBCSHeader->version[1]); |
|
1462 *pErrorCode=U_UNSUPPORTED_ERROR; |
|
1463 return 0; |
|
1464 } |
|
1465 |
|
1466 uprv_memcpy(mbcsHeader.version, inMBCSHeader->version, 4); |
|
1467 mbcsHeader.countStates= ds->readUInt32(inMBCSHeader->countStates); |
|
1468 mbcsHeader.countToUFallbacks= ds->readUInt32(inMBCSHeader->countToUFallbacks); |
|
1469 mbcsHeader.offsetToUCodeUnits= ds->readUInt32(inMBCSHeader->offsetToUCodeUnits); |
|
1470 mbcsHeader.offsetFromUTable= ds->readUInt32(inMBCSHeader->offsetFromUTable); |
|
1471 mbcsHeader.offsetFromUBytes= ds->readUInt32(inMBCSHeader->offsetFromUBytes); |
|
1472 mbcsHeader.flags= ds->readUInt32(inMBCSHeader->flags); |
|
1473 mbcsHeader.fromUBytesLength= ds->readUInt32(inMBCSHeader->fromUBytesLength); |
|
1474 /* mbcsHeader.options have been read above */ |
|
1475 |
|
1476 extOffset=(int32_t)(mbcsHeader.flags>>8); |
|
1477 outputType=(uint8_t)mbcsHeader.flags; |
|
1478 if(noFromU && outputType==MBCS_OUTPUT_1) { |
|
1479 udata_printError(ds, "ucnv_swap(): unsupported combination of makeconv --small with SBCS\n"); |
|
1480 *pErrorCode=U_UNSUPPORTED_ERROR; |
|
1481 return 0; |
|
1482 } |
|
1483 |
|
1484 /* make sure that the output type is known */ |
|
1485 switch(outputType) { |
|
1486 case MBCS_OUTPUT_1: |
|
1487 case MBCS_OUTPUT_2: |
|
1488 case MBCS_OUTPUT_3: |
|
1489 case MBCS_OUTPUT_4: |
|
1490 case MBCS_OUTPUT_3_EUC: |
|
1491 case MBCS_OUTPUT_4_EUC: |
|
1492 case MBCS_OUTPUT_2_SISO: |
|
1493 case MBCS_OUTPUT_EXT_ONLY: |
|
1494 /* OK */ |
|
1495 break; |
|
1496 default: |
|
1497 udata_printError(ds, "ucnv_swap(): unsupported MBCS output type 0x%x\n", |
|
1498 outputType); |
|
1499 *pErrorCode=U_UNSUPPORTED_ERROR; |
|
1500 return 0; |
|
1501 } |
|
1502 |
|
1503 /* calculate the length of the MBCS data */ |
|
1504 |
|
1505 /* |
|
1506 * utf8Friendly MBCS files (mbcsHeader.version 4.3) |
|
1507 * contain an additional mbcsIndex table: |
|
1508 * uint16_t[(maxFastUChar+1)>>6]; |
|
1509 * where maxFastUChar=((mbcsHeader.version[2]<<8)|0xff). |
|
1510 */ |
|
1511 maxFastUChar=0; |
|
1512 mbcsIndexLength=0; |
|
1513 if( outputType!=MBCS_OUTPUT_EXT_ONLY && outputType!=MBCS_OUTPUT_1 && |
|
1514 mbcsHeader.version[1]>=3 && (maxFastUChar=mbcsHeader.version[2])!=0 |
|
1515 ) { |
|
1516 maxFastUChar=(maxFastUChar<<8)|0xff; |
|
1517 mbcsIndexLength=((maxFastUChar+1)>>6)*2; /* number of bytes */ |
|
1518 } |
|
1519 |
|
1520 if(extOffset==0) { |
|
1521 size=(int32_t)(mbcsHeader.offsetFromUBytes+mbcsIndexLength); |
|
1522 if(!noFromU) { |
|
1523 size+=(int32_t)mbcsHeader.fromUBytesLength; |
|
1524 } |
|
1525 |
|
1526 /* avoid compiler warnings - not otherwise necessary, and the value does not matter */ |
|
1527 inExtIndexes=NULL; |
|
1528 } else { |
|
1529 /* there is extension data after the base data, see ucnv_ext.h */ |
|
1530 if(length>=0 && length<(extOffset+UCNV_EXT_INDEXES_MIN_LENGTH*4)) { |
|
1531 udata_printError(ds, "ucnv_swap(): too few bytes (%d after headers) for an ICU MBCS .cnv conversion table with extension data\n", |
|
1532 length); |
|
1533 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
|
1534 return 0; |
|
1535 } |
|
1536 |
|
1537 inExtIndexes=(const int32_t *)(inBytes+extOffset); |
|
1538 size=extOffset+udata_readInt32(ds, inExtIndexes[UCNV_EXT_SIZE]); |
|
1539 } |
|
1540 |
|
1541 if(length>=0) { |
|
1542 if(length<size) { |
|
1543 udata_printError(ds, "ucnv_swap(): too few bytes (%d after headers) for an ICU MBCS .cnv conversion table\n", |
|
1544 length); |
|
1545 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
|
1546 return 0; |
|
1547 } |
|
1548 |
|
1549 /* copy the data for inaccessible bytes */ |
|
1550 if(inBytes!=outBytes) { |
|
1551 uprv_memcpy(outBytes, inBytes, size); |
|
1552 } |
|
1553 |
|
1554 /* swap the MBCSHeader, except for the version field */ |
|
1555 count=mbcsHeaderLength*4; |
|
1556 ds->swapArray32(ds, &inMBCSHeader->countStates, count-4, |
|
1557 &outMBCSHeader->countStates, pErrorCode); |
|
1558 |
|
1559 if(outputType==MBCS_OUTPUT_EXT_ONLY) { |
|
1560 /* |
|
1561 * extension-only file, |
|
1562 * contains a base name instead of normal base table data |
|
1563 */ |
|
1564 |
|
1565 /* swap the base name, between the header and the extension data */ |
|
1566 const char *inBaseName=(const char *)inBytes+count; |
|
1567 char *outBaseName=(char *)outBytes+count; |
|
1568 ds->swapInvChars(ds, inBaseName, (int32_t)uprv_strlen(inBaseName), |
|
1569 outBaseName, pErrorCode); |
|
1570 } else { |
|
1571 /* normal file with base table data */ |
|
1572 |
|
1573 /* swap the state table, 1kB per state */ |
|
1574 offset=count; |
|
1575 count=mbcsHeader.countStates*1024; |
|
1576 ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
|
1577 outBytes+offset, pErrorCode); |
|
1578 |
|
1579 /* swap the toUFallbacks[] */ |
|
1580 offset+=count; |
|
1581 count=mbcsHeader.countToUFallbacks*8; |
|
1582 ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
|
1583 outBytes+offset, pErrorCode); |
|
1584 |
|
1585 /* swap the unicodeCodeUnits[] */ |
|
1586 offset=mbcsHeader.offsetToUCodeUnits; |
|
1587 count=mbcsHeader.offsetFromUTable-offset; |
|
1588 ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
|
1589 outBytes+offset, pErrorCode); |
|
1590 |
|
1591 /* offset to the stage 1 table, independent of the outputType */ |
|
1592 offset=mbcsHeader.offsetFromUTable; |
|
1593 |
|
1594 if(outputType==MBCS_OUTPUT_1) { |
|
1595 /* SBCS: swap the fromU tables, all 16 bits wide */ |
|
1596 count=(mbcsHeader.offsetFromUBytes-offset)+mbcsHeader.fromUBytesLength; |
|
1597 ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
|
1598 outBytes+offset, pErrorCode); |
|
1599 } else { |
|
1600 /* otherwise: swap the stage tables separately */ |
|
1601 |
|
1602 /* stage 1 table: uint16_t[0x440 or 0x40] */ |
|
1603 if(inStaticData->unicodeMask&UCNV_HAS_SUPPLEMENTARY) { |
|
1604 count=0x440*2; /* for all of Unicode */ |
|
1605 } else { |
|
1606 count=0x40*2; /* only BMP */ |
|
1607 } |
|
1608 ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
|
1609 outBytes+offset, pErrorCode); |
|
1610 |
|
1611 /* stage 2 table: uint32_t[] */ |
|
1612 offset+=count; |
|
1613 count=mbcsHeader.offsetFromUBytes-offset; |
|
1614 ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
|
1615 outBytes+offset, pErrorCode); |
|
1616 |
|
1617 /* stage 3/result bytes: sometimes uint16_t[] or uint32_t[] */ |
|
1618 offset=mbcsHeader.offsetFromUBytes; |
|
1619 count= noFromU ? 0 : mbcsHeader.fromUBytesLength; |
|
1620 switch(outputType) { |
|
1621 case MBCS_OUTPUT_2: |
|
1622 case MBCS_OUTPUT_3_EUC: |
|
1623 case MBCS_OUTPUT_2_SISO: |
|
1624 ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
|
1625 outBytes+offset, pErrorCode); |
|
1626 break; |
|
1627 case MBCS_OUTPUT_4: |
|
1628 ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
|
1629 outBytes+offset, pErrorCode); |
|
1630 break; |
|
1631 default: |
|
1632 /* just uint8_t[], nothing to swap */ |
|
1633 break; |
|
1634 } |
|
1635 |
|
1636 if(mbcsIndexLength!=0) { |
|
1637 offset+=count; |
|
1638 count=mbcsIndexLength; |
|
1639 ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
|
1640 outBytes+offset, pErrorCode); |
|
1641 } |
|
1642 } |
|
1643 } |
|
1644 |
|
1645 if(extOffset!=0) { |
|
1646 /* swap the extension data */ |
|
1647 inBytes+=extOffset; |
|
1648 outBytes+=extOffset; |
|
1649 |
|
1650 /* swap toUTable[] */ |
|
1651 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_INDEX]); |
|
1652 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_LENGTH]); |
|
1653 ds->swapArray32(ds, inBytes+offset, length*4, outBytes+offset, pErrorCode); |
|
1654 |
|
1655 /* swap toUUChars[] */ |
|
1656 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_UCHARS_INDEX]); |
|
1657 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_UCHARS_LENGTH]); |
|
1658 ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
|
1659 |
|
1660 /* swap fromUTableUChars[] */ |
|
1661 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_UCHARS_INDEX]); |
|
1662 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_LENGTH]); |
|
1663 ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
|
1664 |
|
1665 /* swap fromUTableValues[] */ |
|
1666 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_VALUES_INDEX]); |
|
1667 /* same length as for fromUTableUChars[] */ |
|
1668 ds->swapArray32(ds, inBytes+offset, length*4, outBytes+offset, pErrorCode); |
|
1669 |
|
1670 /* no need to swap fromUBytes[] */ |
|
1671 |
|
1672 /* swap fromUStage12[] */ |
|
1673 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_12_INDEX]); |
|
1674 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_12_LENGTH]); |
|
1675 ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
|
1676 |
|
1677 /* swap fromUStage3[] */ |
|
1678 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3_INDEX]); |
|
1679 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3_LENGTH]); |
|
1680 ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
|
1681 |
|
1682 /* swap fromUStage3b[] */ |
|
1683 offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3B_INDEX]); |
|
1684 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3B_LENGTH]); |
|
1685 ds->swapArray32(ds, inBytes+offset, length*4, outBytes+offset, pErrorCode); |
|
1686 |
|
1687 /* swap indexes[] */ |
|
1688 length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_INDEXES_LENGTH]); |
|
1689 ds->swapArray32(ds, inBytes, length*4, outBytes, pErrorCode); |
|
1690 } |
|
1691 } |
|
1692 } else { |
|
1693 udata_printError(ds, "ucnv_swap(): unknown conversionType=%d!=UCNV_MBCS\n", |
|
1694 inStaticData->conversionType); |
|
1695 *pErrorCode=U_UNSUPPORTED_ERROR; |
|
1696 return 0; |
|
1697 } |
|
1698 |
|
1699 return headerSize+(int32_t)staticDataSize+size; |
|
1700 } |
|
1701 |
|
1702 #endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */ |
|
1703 |
|
1704 #endif |