1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ******************************************************************** |
5 | * COPYRIGHT: |
6 | * Copyright (c) 1996-2016, International Business Machines Corporation and |
7 | * others. All Rights Reserved. |
8 | ******************************************************************** |
9 | * |
10 | * ucnv_bld.cpp: |
11 | * |
12 | * Defines functions that are used in the creation/initialization/deletion |
13 | * of converters and related structures. |
14 | * uses uconv_io.h routines to access disk information |
15 | * is used by ucnv.h to implement public API create/delete/flushCache routines |
16 | * Modification History: |
17 | * |
18 | * Date Name Description |
19 | * |
20 | * 06/20/2000 helena OS/400 port changes; mostly typecast. |
21 | * 06/29/2000 helena Major rewrite of the callback interface. |
22 | */ |
23 | |
24 | #include "unicode/utypes.h" |
25 | |
26 | #if !UCONFIG_NO_CONVERSION |
27 | |
28 | #include "unicode/putil.h" |
29 | #include "unicode/udata.h" |
30 | #include "unicode/ucnv.h" |
31 | #include "unicode/uloc.h" |
32 | #include "mutex.h" |
33 | #include "putilimp.h" |
34 | #include "uassert.h" |
35 | #include "utracimp.h" |
36 | #include "ucnv_io.h" |
37 | #include "ucnv_bld.h" |
38 | #include "ucnvmbcs.h" |
39 | #include "ucnv_ext.h" |
40 | #include "ucnv_cnv.h" |
41 | #include "ucnv_imp.h" |
42 | #include "uhash.h" |
43 | #include "umutex.h" |
44 | #include "cstring.h" |
45 | #include "cmemory.h" |
46 | #include "ucln_cmn.h" |
47 | #include "ustr_cnv.h" |
48 | |
49 | |
50 | #if 0 |
51 | #include <stdio.h> |
52 | extern void UCNV_DEBUG_LOG(char *what, char *who, void *p, int l); |
53 | #define UCNV_DEBUG_LOG(x,y,z) UCNV_DEBUG_LOG(x,y,z,__LINE__) |
54 | #else |
55 | # define UCNV_DEBUG_LOG(x,y,z) |
56 | #endif |
57 | |
58 | static const UConverterSharedData * const |
59 | converterData[UCNV_NUMBER_OF_SUPPORTED_CONVERTER_TYPES]={ |
60 | nullptr, nullptr, |
61 | |
62 | #if UCONFIG_NO_LEGACY_CONVERSION |
63 | nullptr, |
64 | #else |
65 | &_MBCSData, |
66 | #endif |
67 | |
68 | &_Latin1Data, |
69 | &_UTF8Data, &_UTF16BEData, &_UTF16LEData, |
70 | #if UCONFIG_ONLY_HTML_CONVERSION |
71 | nullptr, nullptr, |
72 | #else |
73 | &_UTF32BEData, &_UTF32LEData, |
74 | #endif |
75 | nullptr, |
76 | |
77 | #if UCONFIG_NO_LEGACY_CONVERSION |
78 | nullptr, |
79 | #else |
80 | &_ISO2022Data, |
81 | #endif |
82 | |
83 | #if UCONFIG_NO_LEGACY_CONVERSION || UCONFIG_ONLY_HTML_CONVERSION |
84 | nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, |
85 | nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, |
86 | nullptr, |
87 | #else |
88 | &_LMBCSData1,&_LMBCSData2, &_LMBCSData3, &_LMBCSData4, &_LMBCSData5, &_LMBCSData6, |
89 | &_LMBCSData8,&_LMBCSData11,&_LMBCSData16,&_LMBCSData17,&_LMBCSData18,&_LMBCSData19, |
90 | &_HZData, |
91 | #endif |
92 | |
93 | #if UCONFIG_ONLY_HTML_CONVERSION |
94 | nullptr, |
95 | #else |
96 | &_SCSUData, |
97 | #endif |
98 | |
99 | |
100 | #if UCONFIG_NO_LEGACY_CONVERSION || UCONFIG_ONLY_HTML_CONVERSION |
101 | nullptr, |
102 | #else |
103 | &_ISCIIData, |
104 | #endif |
105 | |
106 | &_ASCIIData, |
107 | #if UCONFIG_ONLY_HTML_CONVERSION |
108 | nullptr, nullptr, &_UTF16Data, nullptr, nullptr, nullptr, |
109 | #else |
110 | &_UTF7Data, &_Bocu1Data, &_UTF16Data, &_UTF32Data, &_CESU8Data, &_IMAPData, |
111 | #endif |
112 | |
113 | #if UCONFIG_NO_LEGACY_CONVERSION || UCONFIG_ONLY_HTML_CONVERSION |
114 | nullptr, |
115 | #else |
116 | &_CompoundTextData |
117 | #endif |
118 | }; |
119 | |
120 | /* Please keep this in binary sorted order for getAlgorithmicTypeFromName. |
121 | Also the name should be in lower case and all spaces, dashes and underscores |
122 | removed |
123 | */ |
124 | static struct { |
125 | const char *name; |
126 | const UConverterType type; |
127 | } const cnvNameType[] = { |
128 | #if !UCONFIG_ONLY_HTML_CONVERSION |
129 | { "bocu1" , UCNV_BOCU1 }, |
130 | { "cesu8" , UCNV_CESU8 }, |
131 | #endif |
132 | #if !UCONFIG_NO_LEGACY_CONVERSION && !UCONFIG_ONLY_HTML_CONVERSION |
133 | { "hz" ,UCNV_HZ }, |
134 | #endif |
135 | #if !UCONFIG_ONLY_HTML_CONVERSION |
136 | { "imapmailboxname" , UCNV_IMAP_MAILBOX }, |
137 | #endif |
138 | #if !UCONFIG_NO_LEGACY_CONVERSION && !UCONFIG_ONLY_HTML_CONVERSION |
139 | { "iscii" , UCNV_ISCII }, |
140 | #endif |
141 | #if !UCONFIG_NO_LEGACY_CONVERSION |
142 | { "iso2022" , UCNV_ISO_2022 }, |
143 | #endif |
144 | { "iso88591" , UCNV_LATIN_1 }, |
145 | #if !UCONFIG_NO_LEGACY_CONVERSION && !UCONFIG_ONLY_HTML_CONVERSION |
146 | { "lmbcs1" , UCNV_LMBCS_1 }, |
147 | { "lmbcs11" ,UCNV_LMBCS_11 }, |
148 | { "lmbcs16" ,UCNV_LMBCS_16 }, |
149 | { "lmbcs17" ,UCNV_LMBCS_17 }, |
150 | { "lmbcs18" ,UCNV_LMBCS_18 }, |
151 | { "lmbcs19" ,UCNV_LMBCS_19 }, |
152 | { "lmbcs2" , UCNV_LMBCS_2 }, |
153 | { "lmbcs3" , UCNV_LMBCS_3 }, |
154 | { "lmbcs4" , UCNV_LMBCS_4 }, |
155 | { "lmbcs5" , UCNV_LMBCS_5 }, |
156 | { "lmbcs6" , UCNV_LMBCS_6 }, |
157 | { "lmbcs8" , UCNV_LMBCS_8 }, |
158 | #endif |
159 | #if !UCONFIG_ONLY_HTML_CONVERSION |
160 | { "scsu" , UCNV_SCSU }, |
161 | #endif |
162 | { "usascii" , UCNV_US_ASCII }, |
163 | { "utf16" , UCNV_UTF16 }, |
164 | { "utf16be" , UCNV_UTF16_BigEndian }, |
165 | { "utf16le" , UCNV_UTF16_LittleEndian }, |
166 | #if U_IS_BIG_ENDIAN |
167 | { "utf16oppositeendian" , UCNV_UTF16_LittleEndian }, |
168 | { "utf16platformendian" , UCNV_UTF16_BigEndian }, |
169 | #else |
170 | { "utf16oppositeendian" , UCNV_UTF16_BigEndian}, |
171 | { "utf16platformendian" , UCNV_UTF16_LittleEndian }, |
172 | #endif |
173 | #if !UCONFIG_ONLY_HTML_CONVERSION |
174 | { "utf32" , UCNV_UTF32 }, |
175 | { "utf32be" , UCNV_UTF32_BigEndian }, |
176 | { "utf32le" , UCNV_UTF32_LittleEndian }, |
177 | #if U_IS_BIG_ENDIAN |
178 | { "utf32oppositeendian" , UCNV_UTF32_LittleEndian }, |
179 | { "utf32platformendian" , UCNV_UTF32_BigEndian }, |
180 | #else |
181 | { "utf32oppositeendian" , UCNV_UTF32_BigEndian }, |
182 | { "utf32platformendian" , UCNV_UTF32_LittleEndian }, |
183 | #endif |
184 | #endif |
185 | #if !UCONFIG_ONLY_HTML_CONVERSION |
186 | { "utf7" , UCNV_UTF7 }, |
187 | #endif |
188 | { "utf8" , UCNV_UTF8 }, |
189 | #if !UCONFIG_ONLY_HTML_CONVERSION |
190 | { "x11compoundtext" , UCNV_COMPOUND_TEXT} |
191 | #endif |
192 | }; |
193 | |
194 | |
195 | /*initializes some global variables */ |
196 | static UHashtable *SHARED_DATA_HASHTABLE = nullptr; |
197 | static icu::UMutex cnvCacheMutex; |
198 | /* Note: the global mutex is used for */ |
199 | /* reference count updates. */ |
200 | |
201 | static const char **gAvailableConverters = nullptr; |
202 | static uint16_t gAvailableConverterCount = 0; |
203 | static icu::UInitOnce gAvailableConvertersInitOnce {}; |
204 | |
205 | #if !U_CHARSET_IS_UTF8 |
206 | |
207 | /* This contains the resolved converter name. So no further alias lookup is needed again. */ |
208 | static char gDefaultConverterNameBuffer[UCNV_MAX_CONVERTER_NAME_LENGTH + 1]; /* +1 for nullptr */ |
209 | static const char *gDefaultConverterName = nullptr; |
210 | |
211 | /* |
212 | If the default converter is an algorithmic converter, this is the cached value. |
213 | We don't cache a full UConverter and clone it because ucnv_clone doesn't have |
214 | less overhead than an algorithmic open. We don't cache non-algorithmic converters |
215 | because ucnv_flushCache must be able to unload the default converter and its table. |
216 | */ |
217 | static const UConverterSharedData *gDefaultAlgorithmicSharedData = nullptr; |
218 | |
219 | /* Does gDefaultConverterName have a converter option and require extra parsing? */ |
220 | static UBool gDefaultConverterContainsOption; |
221 | |
222 | #endif /* !U_CHARSET_IS_UTF8 */ |
223 | |
224 | static const char DATA_TYPE[] = "cnv" ; |
225 | |
226 | /* ucnv_flushAvailableConverterCache. This is only called from ucnv_cleanup(). |
227 | * If it is ever to be called from elsewhere, synchronization |
228 | * will need to be considered. |
229 | */ |
230 | static void |
231 | ucnv_flushAvailableConverterCache() { |
232 | gAvailableConverterCount = 0; |
233 | if (gAvailableConverters) { |
234 | uprv_free((char **)gAvailableConverters); |
235 | gAvailableConverters = nullptr; |
236 | } |
237 | gAvailableConvertersInitOnce.reset(); |
238 | } |
239 | |
240 | /* ucnv_cleanup - delete all storage held by the converter cache, except any */ |
241 | /* in use by open converters. */ |
242 | /* Not thread safe. */ |
243 | /* Not supported API. */ |
244 | static UBool U_CALLCONV ucnv_cleanup() { |
245 | ucnv_flushCache(); |
246 | if (SHARED_DATA_HASHTABLE != nullptr && uhash_count(SHARED_DATA_HASHTABLE) == 0) { |
247 | uhash_close(SHARED_DATA_HASHTABLE); |
248 | SHARED_DATA_HASHTABLE = nullptr; |
249 | } |
250 | |
251 | /* Isn't called from flushCache because other threads may have preexisting references to the table. */ |
252 | ucnv_flushAvailableConverterCache(); |
253 | |
254 | #if !U_CHARSET_IS_UTF8 |
255 | gDefaultConverterName = nullptr; |
256 | gDefaultConverterNameBuffer[0] = 0; |
257 | gDefaultConverterContainsOption = false; |
258 | gDefaultAlgorithmicSharedData = nullptr; |
259 | #endif |
260 | |
261 | return (SHARED_DATA_HASHTABLE == nullptr); |
262 | } |
263 | |
264 | U_CAPI void U_EXPORT2 |
265 | ucnv_enableCleanup() { |
266 | ucln_common_registerCleanup(UCLN_COMMON_UCNV, ucnv_cleanup); |
267 | } |
268 | |
269 | static UBool U_CALLCONV |
270 | isCnvAcceptable(void * /*context*/, |
271 | const char * /*type*/, const char * /*name*/, |
272 | const UDataInfo *pInfo) { |
273 | return (UBool)( |
274 | pInfo->size>=20 && |
275 | pInfo->isBigEndian==U_IS_BIG_ENDIAN && |
276 | pInfo->charsetFamily==U_CHARSET_FAMILY && |
277 | pInfo->sizeofUChar==U_SIZEOF_UCHAR && |
278 | pInfo->dataFormat[0]==0x63 && /* dataFormat="cnvt" */ |
279 | pInfo->dataFormat[1]==0x6e && |
280 | pInfo->dataFormat[2]==0x76 && |
281 | pInfo->dataFormat[3]==0x74 && |
282 | pInfo->formatVersion[0]==6); /* Everything will be version 6 */ |
283 | } |
284 | |
285 | /** |
286 | * Un flatten shared data from a UDATA.. |
287 | */ |
288 | static UConverterSharedData* |
289 | ucnv_data_unFlattenClone(UConverterLoadArgs *pArgs, UDataMemory *pData, UErrorCode *status) |
290 | { |
291 | /* UDataInfo info; -- necessary only if some converters have different formatVersion */ |
292 | const uint8_t *raw = (const uint8_t *)udata_getMemory(pData); |
293 | const UConverterStaticData *source = (const UConverterStaticData *) raw; |
294 | UConverterSharedData *data; |
295 | UConverterType type = (UConverterType)source->conversionType; |
296 | |
297 | if(U_FAILURE(*status)) |
298 | return nullptr; |
299 | |
300 | if( (uint16_t)type >= UCNV_NUMBER_OF_SUPPORTED_CONVERTER_TYPES || |
301 | converterData[type] == nullptr || |
302 | !converterData[type]->isReferenceCounted || |
303 | converterData[type]->referenceCounter != 1 || |
304 | source->structSize != sizeof(UConverterStaticData)) |
305 | { |
306 | *status = U_INVALID_TABLE_FORMAT; |
307 | return nullptr; |
308 | } |
309 | |
310 | data = (UConverterSharedData *)uprv_malloc(sizeof(UConverterSharedData)); |
311 | if(data == nullptr) { |
312 | *status = U_MEMORY_ALLOCATION_ERROR; |
313 | return nullptr; |
314 | } |
315 | |
316 | /* copy initial values from the static structure for this type */ |
317 | uprv_memcpy(data, converterData[type], sizeof(UConverterSharedData)); |
318 | |
319 | data->staticData = source; |
320 | |
321 | data->sharedDataCached = false; |
322 | |
323 | /* fill in fields from the loaded data */ |
324 | data->dataMemory = (void*)pData; /* for future use */ |
325 | |
326 | if(data->impl->load != nullptr) { |
327 | data->impl->load(data, pArgs, raw + source->structSize, status); |
328 | if(U_FAILURE(*status)) { |
329 | uprv_free(data); |
330 | return nullptr; |
331 | } |
332 | } |
333 | return data; |
334 | } |
335 | |
336 | /*Takes an alias name gets an actual converter file name |
337 | *goes to disk and opens it. |
338 | *allocates the memory and returns a new UConverter object |
339 | */ |
340 | static UConverterSharedData *createConverterFromFile(UConverterLoadArgs *pArgs, UErrorCode * err) |
341 | { |
342 | UDataMemory *data; |
343 | UConverterSharedData *sharedData; |
344 | |
345 | UTRACE_ENTRY_OC(UTRACE_UCNV_LOAD); |
346 | |
347 | if (U_FAILURE (*err)) { |
348 | UTRACE_EXIT_STATUS(*err); |
349 | return nullptr; |
350 | } |
351 | |
352 | UTRACE_DATA2(UTRACE_OPEN_CLOSE, "load converter %s from package %s" , pArgs->name, pArgs->pkg); |
353 | |
354 | data = udata_openChoice(pArgs->pkg, DATA_TYPE, pArgs->name, isCnvAcceptable, nullptr, err); |
355 | if(U_FAILURE(*err)) |
356 | { |
357 | UTRACE_EXIT_STATUS(*err); |
358 | return nullptr; |
359 | } |
360 | |
361 | sharedData = ucnv_data_unFlattenClone(pArgs, data, err); |
362 | if(U_FAILURE(*err)) |
363 | { |
364 | udata_close(data); |
365 | UTRACE_EXIT_STATUS(*err); |
366 | return nullptr; |
367 | } |
368 | |
369 | /* |
370 | * TODO Store pkg in a field in the shared data so that delta-only converters |
371 | * can load base converters from the same package. |
372 | * If the pkg name is longer than the field, then either do not load the converter |
373 | * in the first place, or just set the pkg field to "". |
374 | */ |
375 | |
376 | UTRACE_EXIT_PTR_STATUS(sharedData, *err); |
377 | return sharedData; |
378 | } |
379 | |
380 | /*returns a converter type from a string |
381 | */ |
382 | static const UConverterSharedData * |
383 | getAlgorithmicTypeFromName(const char *realName) |
384 | { |
385 | uint32_t mid, start, limit; |
386 | uint32_t lastMid; |
387 | int result; |
388 | char strippedName[UCNV_MAX_CONVERTER_NAME_LENGTH]; |
389 | |
390 | /* Lower case and remove ignoreable characters. */ |
391 | ucnv_io_stripForCompare(strippedName, realName); |
392 | |
393 | /* do a binary search for the alias */ |
394 | start = 0; |
395 | limit = UPRV_LENGTHOF(cnvNameType); |
396 | mid = limit; |
397 | lastMid = UINT32_MAX; |
398 | |
399 | for (;;) { |
400 | mid = (uint32_t)((start + limit) / 2); |
401 | if (lastMid == mid) { /* Have we moved? */ |
402 | break; /* We haven't moved, and it wasn't found. */ |
403 | } |
404 | lastMid = mid; |
405 | result = uprv_strcmp(strippedName, cnvNameType[mid].name); |
406 | |
407 | if (result < 0) { |
408 | limit = mid; |
409 | } else if (result > 0) { |
410 | start = mid; |
411 | } else { |
412 | return converterData[cnvNameType[mid].type]; |
413 | } |
414 | } |
415 | |
416 | return nullptr; |
417 | } |
418 | |
419 | /* |
420 | * Based on the number of known converters, this determines how many times larger |
421 | * the shared data hash table should be. When on small platforms, or just a couple |
422 | * of converters are used, this number should be 2. When memory is plentiful, or |
423 | * when ucnv_countAvailable is ever used with a lot of available converters, |
424 | * this should be 4. |
425 | * Larger numbers reduce the number of hash collisions, but use more memory. |
426 | */ |
427 | #define UCNV_CACHE_LOAD_FACTOR 2 |
428 | |
429 | /* Puts the shared data in the static hashtable SHARED_DATA_HASHTABLE */ |
430 | /* Will always be called with the cnvCacheMutex already being held */ |
431 | /* by the calling function. */ |
432 | /* Stores the shared data in the SHARED_DATA_HASHTABLE |
433 | * @param data The shared data |
434 | */ |
435 | static void |
436 | ucnv_shareConverterData(UConverterSharedData * data) |
437 | { |
438 | UErrorCode err = U_ZERO_ERROR; |
439 | /*Lazy evaluates the Hashtable itself */ |
440 | /*void *sanity = nullptr;*/ |
441 | |
442 | if (SHARED_DATA_HASHTABLE == nullptr) |
443 | { |
444 | SHARED_DATA_HASHTABLE = uhash_openSize(uhash_hashChars, uhash_compareChars, nullptr, |
445 | ucnv_io_countKnownConverters(&err)*UCNV_CACHE_LOAD_FACTOR, |
446 | &err); |
447 | ucnv_enableCleanup(); |
448 | |
449 | if (U_FAILURE(err)) |
450 | return; |
451 | } |
452 | |
453 | /* ### check to see if the element is not already there! */ |
454 | |
455 | /* |
456 | sanity = ucnv_getSharedConverterData (data->staticData->name); |
457 | if(sanity != nullptr) |
458 | { |
459 | UCNV_DEBUG_LOG("put:overwrite!",data->staticData->name,sanity); |
460 | } |
461 | UCNV_DEBUG_LOG("put:chk",data->staticData->name,sanity); |
462 | */ |
463 | |
464 | /* Mark it shared */ |
465 | data->sharedDataCached = true; |
466 | |
467 | uhash_put(SHARED_DATA_HASHTABLE, |
468 | (void*) data->staticData->name, /* Okay to cast away const as long as |
469 | keyDeleter == nullptr */ |
470 | data, |
471 | &err); |
472 | UCNV_DEBUG_LOG("put" , data->staticData->name,data); |
473 | |
474 | } |
475 | |
476 | /* Look up a converter name in the shared data cache. */ |
477 | /* cnvCacheMutex must be held by the caller to protect the hash table. */ |
478 | /* gets the shared data from the SHARED_DATA_HASHTABLE (might return nullptr if it isn't there) |
479 | * @param name The name of the shared data |
480 | * @return the shared data from the SHARED_DATA_HASHTABLE |
481 | */ |
482 | static UConverterSharedData * |
483 | ucnv_getSharedConverterData(const char *name) |
484 | { |
485 | /*special case when no Table has yet been created we return nullptr */ |
486 | if (SHARED_DATA_HASHTABLE == nullptr) |
487 | { |
488 | return nullptr; |
489 | } |
490 | else |
491 | { |
492 | UConverterSharedData *rc; |
493 | |
494 | rc = (UConverterSharedData*)uhash_get(SHARED_DATA_HASHTABLE, name); |
495 | UCNV_DEBUG_LOG("get" ,name,rc); |
496 | return rc; |
497 | } |
498 | } |
499 | |
500 | /*frees the string of memory blocks associates with a sharedConverter |
501 | *if and only if the referenceCounter == 0 |
502 | */ |
503 | /* Deletes (frees) the Shared data it's passed. first it checks the referenceCounter to |
504 | * see if anyone is using it, if not it frees all the memory stemming from sharedConverterData and |
505 | * returns true, |
506 | * otherwise returns false |
507 | * @param sharedConverterData The shared data |
508 | * @return if not it frees all the memory stemming from sharedConverterData and |
509 | * returns true, otherwise returns false |
510 | */ |
511 | static UBool |
512 | ucnv_deleteSharedConverterData(UConverterSharedData * deadSharedData) |
513 | { |
514 | UTRACE_ENTRY_OC(UTRACE_UCNV_UNLOAD); |
515 | UTRACE_DATA2(UTRACE_OPEN_CLOSE, "unload converter %s shared data %p" , deadSharedData->staticData->name, deadSharedData); |
516 | |
517 | if (deadSharedData->referenceCounter > 0) { |
518 | UTRACE_EXIT_VALUE((int32_t)false); |
519 | return false; |
520 | } |
521 | |
522 | if (deadSharedData->impl->unload != nullptr) { |
523 | deadSharedData->impl->unload(deadSharedData); |
524 | } |
525 | |
526 | if(deadSharedData->dataMemory != nullptr) |
527 | { |
528 | UDataMemory *data = (UDataMemory*)deadSharedData->dataMemory; |
529 | udata_close(data); |
530 | } |
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==nullptr, 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 == nullptr || U_FAILURE(*err)) { |
547 | return nullptr; |
548 | } |
549 | |
550 | if(pArgs->pkg != nullptr && *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 == nullptr) |
557 | { |
558 | /*Not cached, we need to stream it in from file */ |
559 | mySharedConverterData = createConverterFromFile(pArgs, err); |
560 | if (U_FAILURE (*err) || (mySharedConverterData == nullptr)) |
561 | { |
562 | return nullptr; |
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->isReferenceCounted |
583 | * and this function must be called inside umtx_lock(&cnvCacheMutex). |
584 | */ |
585 | U_CAPI void |
586 | ucnv_unload(UConverterSharedData *sharedData) { |
587 | if(sharedData != nullptr) { |
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 | if(sharedData != nullptr && sharedData->isReferenceCounted) { |
602 | umtx_lock(&cnvCacheMutex); |
603 | ucnv_unload(sharedData); |
604 | umtx_unlock(&cnvCacheMutex); |
605 | } |
606 | } |
607 | |
608 | U_CFUNC void |
609 | ucnv_incrementRefCount(UConverterSharedData *sharedData) |
610 | { |
611 | if(sharedData != nullptr && sharedData->isReferenceCounted) { |
612 | umtx_lock(&cnvCacheMutex); |
613 | sharedData->referenceCounter++; |
614 | umtx_unlock(&cnvCacheMutex); |
615 | } |
616 | } |
617 | |
618 | /* |
619 | * *pPieces must be initialized. |
620 | * The name without options will be copied to pPieces->cnvName. |
621 | * The locale and options will be copied to pPieces only if present in inName, |
622 | * otherwise the existing values in pPieces remain. |
623 | * *pArgs will be set to the pPieces values. |
624 | */ |
625 | static void |
626 | parseConverterOptions(const char *inName, |
627 | UConverterNamePieces *pPieces, |
628 | UConverterLoadArgs *pArgs, |
629 | UErrorCode *err) |
630 | { |
631 | char *cnvName = pPieces->cnvName; |
632 | char c; |
633 | int32_t len = 0; |
634 | |
635 | pArgs->name=inName; |
636 | pArgs->locale=pPieces->locale; |
637 | pArgs->options=pPieces->options; |
638 | |
639 | /* copy the converter name itself to cnvName */ |
640 | while((c=*inName)!=0 && c!=UCNV_OPTION_SEP_CHAR) { |
641 | if (++len>=UCNV_MAX_CONVERTER_NAME_LENGTH) { |
642 | *err = U_ILLEGAL_ARGUMENT_ERROR; /* bad name */ |
643 | pPieces->cnvName[0]=0; |
644 | return; |
645 | } |
646 | *cnvName++=c; |
647 | inName++; |
648 | } |
649 | *cnvName=0; |
650 | pArgs->name=pPieces->cnvName; |
651 | |
652 | /* parse options. No more name copying should occur. */ |
653 | while((c=*inName)!=0) { |
654 | if(c==UCNV_OPTION_SEP_CHAR) { |
655 | ++inName; |
656 | } |
657 | |
658 | /* inName is behind an option separator */ |
659 | if(uprv_strncmp(inName, "locale=" , 7)==0) { |
660 | /* do not modify locale itself in case we have multiple locale options */ |
661 | char *dest=pPieces->locale; |
662 | |
663 | /* copy the locale option value */ |
664 | inName+=7; |
665 | len=0; |
666 | while((c=*inName)!=0 && c!=UCNV_OPTION_SEP_CHAR) { |
667 | ++inName; |
668 | |
669 | if(++len>=ULOC_FULLNAME_CAPACITY) { |
670 | *err=U_ILLEGAL_ARGUMENT_ERROR; /* bad name */ |
671 | pPieces->locale[0]=0; |
672 | return; |
673 | } |
674 | |
675 | *dest++=c; |
676 | } |
677 | *dest=0; |
678 | } else if(uprv_strncmp(inName, "version=" , 8)==0) { |
679 | /* copy the version option value into bits 3..0 of pPieces->options */ |
680 | inName+=8; |
681 | c=*inName; |
682 | if(c==0) { |
683 | pArgs->options=(pPieces->options&=~UCNV_OPTION_VERSION); |
684 | return; |
685 | } else if((uint8_t)(c-'0')<10) { |
686 | pArgs->options=pPieces->options=(pPieces->options&~UCNV_OPTION_VERSION)|(uint32_t)(c-'0'); |
687 | ++inName; |
688 | } |
689 | } else if(uprv_strncmp(inName, "swaplfnl" , 8)==0) { |
690 | inName+=8; |
691 | pArgs->options=(pPieces->options|=UCNV_OPTION_SWAP_LFNL); |
692 | /* add processing for new options here with another } else if(uprv_strncmp(inName, "option-name=", XX)==0) { */ |
693 | } else { |
694 | /* ignore any other options until we define some */ |
695 | while(((c = *inName++) != 0) && (c != UCNV_OPTION_SEP_CHAR)) { |
696 | } |
697 | if(c==0) { |
698 | return; |
699 | } |
700 | } |
701 | } |
702 | } |
703 | |
704 | /*Logic determines if the converter is Algorithmic AND/OR cached |
705 | *depending on that: |
706 | * -we either go to get data from disk and cache it (Data=true, Cached=false) |
707 | * -Get it from a Hashtable (Data=X, Cached=true) |
708 | * -Call dataConverter initializer (Data=true, Cached=true) |
709 | * -Call AlgorithmicConverter initializer (Data=false, Cached=true) |
710 | */ |
711 | U_CFUNC UConverterSharedData * |
712 | ucnv_loadSharedData(const char *converterName, |
713 | UConverterNamePieces *pPieces, |
714 | UConverterLoadArgs *pArgs, |
715 | UErrorCode * err) { |
716 | UConverterNamePieces stackPieces; |
717 | UConverterLoadArgs stackArgs; |
718 | UConverterSharedData *mySharedConverterData = nullptr; |
719 | UErrorCode internalErrorCode = U_ZERO_ERROR; |
720 | UBool mayContainOption = true; |
721 | UBool checkForAlgorithmic = true; |
722 | |
723 | if (U_FAILURE (*err)) { |
724 | return nullptr; |
725 | } |
726 | |
727 | if(pPieces == nullptr) { |
728 | if(pArgs != nullptr) { |
729 | /* |
730 | * Bad: We may set pArgs pointers to stackPieces fields |
731 | * which will be invalid after this function returns. |
732 | */ |
733 | *err = U_INTERNAL_PROGRAM_ERROR; |
734 | return nullptr; |
735 | } |
736 | pPieces = &stackPieces; |
737 | } |
738 | if(pArgs == nullptr) { |
739 | uprv_memset(&stackArgs, 0, sizeof(stackArgs)); |
740 | stackArgs.size = (int32_t)sizeof(stackArgs); |
741 | pArgs = &stackArgs; |
742 | } |
743 | |
744 | pPieces->cnvName[0] = 0; |
745 | pPieces->locale[0] = 0; |
746 | pPieces->options = 0; |
747 | |
748 | pArgs->name = converterName; |
749 | pArgs->locale = pPieces->locale; |
750 | pArgs->options = pPieces->options; |
751 | |
752 | /* In case "name" is nullptr we want to open the default converter. */ |
753 | if (converterName == nullptr) { |
754 | #if U_CHARSET_IS_UTF8 |
755 | pArgs->name = "UTF-8" ; |
756 | return (UConverterSharedData *)converterData[UCNV_UTF8]; |
757 | #else |
758 | /* Call ucnv_getDefaultName first to query the name from the OS. */ |
759 | pArgs->name = ucnv_getDefaultName(); |
760 | if (pArgs->name == nullptr) { |
761 | *err = U_MISSING_RESOURCE_ERROR; |
762 | return nullptr; |
763 | } |
764 | mySharedConverterData = (UConverterSharedData *)gDefaultAlgorithmicSharedData; |
765 | checkForAlgorithmic = false; |
766 | mayContainOption = gDefaultConverterContainsOption; |
767 | /* the default converter name is already canonical */ |
768 | #endif |
769 | } |
770 | else if(UCNV_FAST_IS_UTF8(converterName)) { |
771 | /* fastpath for UTF-8 */ |
772 | pArgs->name = "UTF-8" ; |
773 | return (UConverterSharedData *)converterData[UCNV_UTF8]; |
774 | } |
775 | else { |
776 | /* separate the converter name from the options */ |
777 | parseConverterOptions(converterName, pPieces, pArgs, err); |
778 | if (U_FAILURE(*err)) { |
779 | /* Very bad name used. */ |
780 | return nullptr; |
781 | } |
782 | |
783 | /* get the canonical converter name */ |
784 | pArgs->name = ucnv_io_getConverterName(pArgs->name, &mayContainOption, &internalErrorCode); |
785 | if (U_FAILURE(internalErrorCode) || pArgs->name == nullptr) { |
786 | /* |
787 | * set the input name in case the converter was added |
788 | * without updating the alias table, or when there is no alias table |
789 | */ |
790 | pArgs->name = pPieces->cnvName; |
791 | } else if (internalErrorCode == U_AMBIGUOUS_ALIAS_WARNING) { |
792 | *err = U_AMBIGUOUS_ALIAS_WARNING; |
793 | } |
794 | } |
795 | |
796 | /* separate the converter name from the options */ |
797 | if(mayContainOption && pArgs->name != pPieces->cnvName) { |
798 | parseConverterOptions(pArgs->name, pPieces, pArgs, err); |
799 | } |
800 | |
801 | /* get the shared data for an algorithmic converter, if it is one */ |
802 | if (checkForAlgorithmic) { |
803 | mySharedConverterData = (UConverterSharedData *)getAlgorithmicTypeFromName(pArgs->name); |
804 | } |
805 | if (mySharedConverterData == nullptr) |
806 | { |
807 | /* it is a data-based converter, get its shared data. */ |
808 | /* Hold the cnvCacheMutex through the whole process of checking the */ |
809 | /* converter data cache, and adding new entries to the cache */ |
810 | /* to prevent other threads from modifying the cache during the */ |
811 | /* process. */ |
812 | pArgs->nestedLoads=1; |
813 | pArgs->pkg=nullptr; |
814 | |
815 | umtx_lock(&cnvCacheMutex); |
816 | mySharedConverterData = ucnv_load(pArgs, err); |
817 | umtx_unlock(&cnvCacheMutex); |
818 | if (U_FAILURE (*err) || (mySharedConverterData == nullptr)) |
819 | { |
820 | return nullptr; |
821 | } |
822 | } |
823 | |
824 | return mySharedConverterData; |
825 | } |
826 | |
827 | U_CAPI UConverter * |
828 | ucnv_createConverter(UConverter *myUConverter, const char *converterName, UErrorCode * err) |
829 | { |
830 | UConverterNamePieces stackPieces; |
831 | UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
832 | UConverterSharedData *mySharedConverterData; |
833 | |
834 | UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN); |
835 | |
836 | if(U_SUCCESS(*err)) { |
837 | UTRACE_DATA1(UTRACE_OPEN_CLOSE, "open converter %s" , converterName); |
838 | |
839 | mySharedConverterData = ucnv_loadSharedData(converterName, &stackPieces, &stackArgs, err); |
840 | |
841 | myUConverter = ucnv_createConverterFromSharedData( |
842 | myUConverter, mySharedConverterData, |
843 | &stackArgs, |
844 | err); |
845 | |
846 | if(U_SUCCESS(*err)) { |
847 | UTRACE_EXIT_PTR_STATUS(myUConverter, *err); |
848 | return myUConverter; |
849 | } |
850 | } |
851 | |
852 | /* exit with error */ |
853 | UTRACE_EXIT_STATUS(*err); |
854 | return nullptr; |
855 | } |
856 | |
857 | U_CFUNC UBool |
858 | ucnv_canCreateConverter(const char *converterName, UErrorCode *err) { |
859 | UConverter myUConverter; |
860 | UConverterNamePieces stackPieces; |
861 | UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
862 | UConverterSharedData *mySharedConverterData; |
863 | |
864 | UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN); |
865 | |
866 | if(U_SUCCESS(*err)) { |
867 | UTRACE_DATA1(UTRACE_OPEN_CLOSE, "test if can open converter %s" , converterName); |
868 | |
869 | stackArgs.onlyTestIsLoadable=true; |
870 | mySharedConverterData = ucnv_loadSharedData(converterName, &stackPieces, &stackArgs, err); |
871 | ucnv_createConverterFromSharedData( |
872 | &myUConverter, mySharedConverterData, |
873 | &stackArgs, |
874 | err); |
875 | ucnv_unloadSharedDataIfReady(mySharedConverterData); |
876 | } |
877 | |
878 | UTRACE_EXIT_STATUS(*err); |
879 | return U_SUCCESS(*err); |
880 | } |
881 | |
882 | UConverter * |
883 | ucnv_createAlgorithmicConverter(UConverter *myUConverter, |
884 | UConverterType type, |
885 | const char *locale, uint32_t options, |
886 | UErrorCode *err) { |
887 | UConverter *cnv; |
888 | const UConverterSharedData *sharedData; |
889 | UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
890 | |
891 | UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN_ALGORITHMIC); |
892 | UTRACE_DATA1(UTRACE_OPEN_CLOSE, "open algorithmic converter type %d" , (int32_t)type); |
893 | |
894 | if(type<0 || UCNV_NUMBER_OF_SUPPORTED_CONVERTER_TYPES<=type) { |
895 | *err = U_ILLEGAL_ARGUMENT_ERROR; |
896 | UTRACE_EXIT_STATUS(U_ILLEGAL_ARGUMENT_ERROR); |
897 | return nullptr; |
898 | } |
899 | |
900 | sharedData = converterData[type]; |
901 | if(sharedData == nullptr || sharedData->isReferenceCounted) { |
902 | /* not a valid type, or not an algorithmic converter */ |
903 | *err = U_ILLEGAL_ARGUMENT_ERROR; |
904 | UTRACE_EXIT_STATUS(U_ILLEGAL_ARGUMENT_ERROR); |
905 | return nullptr; |
906 | } |
907 | |
908 | stackArgs.name = "" ; |
909 | stackArgs.options = options; |
910 | stackArgs.locale=locale; |
911 | cnv = ucnv_createConverterFromSharedData( |
912 | myUConverter, (UConverterSharedData *)sharedData, |
913 | &stackArgs, err); |
914 | |
915 | UTRACE_EXIT_PTR_STATUS(cnv, *err); |
916 | return cnv; |
917 | } |
918 | |
919 | U_CFUNC UConverter* |
920 | ucnv_createConverterFromPackage(const char *packageName, const char *converterName, UErrorCode * err) |
921 | { |
922 | UConverter *myUConverter; |
923 | UConverterSharedData *mySharedConverterData; |
924 | UConverterNamePieces stackPieces; |
925 | UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
926 | |
927 | UTRACE_ENTRY_OC(UTRACE_UCNV_OPEN_PACKAGE); |
928 | |
929 | if(U_FAILURE(*err)) { |
930 | UTRACE_EXIT_STATUS(*err); |
931 | return nullptr; |
932 | } |
933 | |
934 | UTRACE_DATA2(UTRACE_OPEN_CLOSE, "open converter %s from package %s" , converterName, packageName); |
935 | |
936 | /* first, get the options out of the converterName string */ |
937 | stackPieces.cnvName[0] = 0; |
938 | stackPieces.locale[0] = 0; |
939 | stackPieces.options = 0; |
940 | parseConverterOptions(converterName, &stackPieces, &stackArgs, err); |
941 | if (U_FAILURE(*err)) { |
942 | /* Very bad name used. */ |
943 | UTRACE_EXIT_STATUS(*err); |
944 | return nullptr; |
945 | } |
946 | stackArgs.nestedLoads=1; |
947 | stackArgs.pkg=packageName; |
948 | |
949 | /* open the data, unflatten the shared structure */ |
950 | mySharedConverterData = createConverterFromFile(&stackArgs, err); |
951 | |
952 | if (U_FAILURE(*err)) { |
953 | UTRACE_EXIT_STATUS(*err); |
954 | return nullptr; |
955 | } |
956 | |
957 | /* create the actual converter */ |
958 | myUConverter = ucnv_createConverterFromSharedData(nullptr, mySharedConverterData, &stackArgs, err); |
959 | |
960 | if (U_FAILURE(*err)) { |
961 | ucnv_close(myUConverter); |
962 | UTRACE_EXIT_STATUS(*err); |
963 | return nullptr; |
964 | } |
965 | |
966 | UTRACE_EXIT_PTR_STATUS(myUConverter, *err); |
967 | return myUConverter; |
968 | } |
969 | |
970 | |
971 | U_CFUNC UConverter* |
972 | ucnv_createConverterFromSharedData(UConverter *myUConverter, |
973 | UConverterSharedData *mySharedConverterData, |
974 | UConverterLoadArgs *pArgs, |
975 | UErrorCode *err) |
976 | { |
977 | UBool isCopyLocal; |
978 | |
979 | if(U_FAILURE(*err)) { |
980 | ucnv_unloadSharedDataIfReady(mySharedConverterData); |
981 | return myUConverter; |
982 | } |
983 | if(myUConverter == nullptr) |
984 | { |
985 | myUConverter = (UConverter *) uprv_malloc (sizeof (UConverter)); |
986 | if(myUConverter == nullptr) |
987 | { |
988 | *err = U_MEMORY_ALLOCATION_ERROR; |
989 | ucnv_unloadSharedDataIfReady(mySharedConverterData); |
990 | return nullptr; |
991 | } |
992 | isCopyLocal = false; |
993 | } else { |
994 | isCopyLocal = true; |
995 | } |
996 | |
997 | /* initialize the converter */ |
998 | uprv_memset(myUConverter, 0, sizeof(UConverter)); |
999 | myUConverter->isCopyLocal = isCopyLocal; |
1000 | /*myUConverter->isExtraLocal = false;*/ /* Set by the memset call */ |
1001 | myUConverter->sharedData = mySharedConverterData; |
1002 | myUConverter->options = pArgs->options; |
1003 | if(!pArgs->onlyTestIsLoadable) { |
1004 | myUConverter->preFromUFirstCP = U_SENTINEL; |
1005 | myUConverter->fromCharErrorBehaviour = UCNV_TO_U_DEFAULT_CALLBACK; |
1006 | myUConverter->fromUCharErrorBehaviour = UCNV_FROM_U_DEFAULT_CALLBACK; |
1007 | myUConverter->toUnicodeStatus = mySharedConverterData->toUnicodeStatus; |
1008 | myUConverter->maxBytesPerUChar = mySharedConverterData->staticData->maxBytesPerChar; |
1009 | myUConverter->subChar1 = mySharedConverterData->staticData->subChar1; |
1010 | myUConverter->subCharLen = mySharedConverterData->staticData->subCharLen; |
1011 | myUConverter->subChars = (uint8_t *)myUConverter->subUChars; |
1012 | uprv_memcpy(myUConverter->subChars, mySharedConverterData->staticData->subChar, myUConverter->subCharLen); |
1013 | myUConverter->toUCallbackReason = UCNV_ILLEGAL; /* default reason to invoke (*fromCharErrorBehaviour) */ |
1014 | } |
1015 | |
1016 | if(mySharedConverterData->impl->open != nullptr) { |
1017 | mySharedConverterData->impl->open(myUConverter, pArgs, err); |
1018 | if(U_FAILURE(*err) && !pArgs->onlyTestIsLoadable) { |
1019 | /* don't ucnv_close() if onlyTestIsLoadable because not fully initialized */ |
1020 | ucnv_close(myUConverter); |
1021 | return nullptr; |
1022 | } |
1023 | } |
1024 | |
1025 | return myUConverter; |
1026 | } |
1027 | |
1028 | /*Frees all shared immutable objects that aren't referred to (reference count = 0) |
1029 | */ |
1030 | U_CAPI int32_t U_EXPORT2 |
1031 | ucnv_flushCache () |
1032 | { |
1033 | UConverterSharedData *mySharedData = nullptr; |
1034 | int32_t pos; |
1035 | int32_t tableDeletedNum = 0; |
1036 | const UHashElement *e; |
1037 | /*UErrorCode status = U_ILLEGAL_ARGUMENT_ERROR;*/ |
1038 | int32_t i, remaining; |
1039 | |
1040 | UTRACE_ENTRY_OC(UTRACE_UCNV_FLUSH_CACHE); |
1041 | |
1042 | /* Close the default converter without creating a new one so that everything will be flushed. */ |
1043 | u_flushDefaultConverter(); |
1044 | |
1045 | /*if shared data hasn't even been lazy evaluated yet |
1046 | * return 0 |
1047 | */ |
1048 | if (SHARED_DATA_HASHTABLE == nullptr) { |
1049 | UTRACE_EXIT_VALUE((int32_t)0); |
1050 | return 0; |
1051 | } |
1052 | |
1053 | /*creates an enumeration to iterate through every element in the |
1054 | * table |
1055 | * |
1056 | * Synchronization: holding cnvCacheMutex will prevent any other thread from |
1057 | * accessing or modifying the hash table during the iteration. |
1058 | * The reference count of an entry may be decremented by |
1059 | * ucnv_close while the iteration is in process, but this is |
1060 | * benign. It can't be incremented (in ucnv_createConverter()) |
1061 | * because the sequence of looking up in the cache + incrementing |
1062 | * is protected by cnvCacheMutex. |
1063 | */ |
1064 | umtx_lock(&cnvCacheMutex); |
1065 | /* |
1066 | * double loop: A delta/extension-only converter has a pointer to its base table's |
1067 | * shared data; the first iteration of the outer loop may see the delta converter |
1068 | * before the base converter, and unloading the delta converter may get the base |
1069 | * converter's reference counter down to 0. |
1070 | */ |
1071 | i = 0; |
1072 | do { |
1073 | remaining = 0; |
1074 | pos = UHASH_FIRST; |
1075 | while ((e = uhash_nextElement (SHARED_DATA_HASHTABLE, &pos)) != nullptr) |
1076 | { |
1077 | mySharedData = (UConverterSharedData *) e->value.pointer; |
1078 | /*deletes only if reference counter == 0 */ |
1079 | if (mySharedData->referenceCounter == 0) |
1080 | { |
1081 | tableDeletedNum++; |
1082 | |
1083 | UCNV_DEBUG_LOG("del" ,mySharedData->staticData->name,mySharedData); |
1084 | |
1085 | uhash_removeElement(SHARED_DATA_HASHTABLE, e); |
1086 | mySharedData->sharedDataCached = false; |
1087 | ucnv_deleteSharedConverterData (mySharedData); |
1088 | } else { |
1089 | ++remaining; |
1090 | } |
1091 | } |
1092 | } while(++i == 1 && remaining > 0); |
1093 | umtx_unlock(&cnvCacheMutex); |
1094 | |
1095 | UTRACE_DATA1(UTRACE_INFO, "ucnv_flushCache() exits with %d converters remaining" , remaining); |
1096 | |
1097 | UTRACE_EXIT_VALUE(tableDeletedNum); |
1098 | return tableDeletedNum; |
1099 | } |
1100 | |
1101 | /* available converters list --------------------------------------------------- */ |
1102 | |
1103 | static void U_CALLCONV initAvailableConvertersList(UErrorCode &errCode) { |
1104 | U_ASSERT(gAvailableConverterCount == 0); |
1105 | U_ASSERT(gAvailableConverters == nullptr); |
1106 | |
1107 | ucnv_enableCleanup(); |
1108 | UEnumeration *allConvEnum = ucnv_openAllNames(&errCode); |
1109 | int32_t allConverterCount = uenum_count(allConvEnum, &errCode); |
1110 | if (U_FAILURE(errCode)) { |
1111 | return; |
1112 | } |
1113 | |
1114 | /* We can't have more than "*converterTable" converters to open */ |
1115 | gAvailableConverters = (const char **) uprv_malloc(allConverterCount * sizeof(char*)); |
1116 | if (!gAvailableConverters) { |
1117 | errCode = U_MEMORY_ALLOCATION_ERROR; |
1118 | return; |
1119 | } |
1120 | |
1121 | /* Open the default converter to make sure that it has first dibs in the hash table. */ |
1122 | UErrorCode localStatus = U_ZERO_ERROR; |
1123 | UConverter tempConverter; |
1124 | ucnv_close(ucnv_createConverter(&tempConverter, nullptr, &localStatus)); |
1125 | |
1126 | gAvailableConverterCount = 0; |
1127 | |
1128 | for (int32_t idx = 0; idx < allConverterCount; idx++) { |
1129 | localStatus = U_ZERO_ERROR; |
1130 | const char *converterName = uenum_next(allConvEnum, nullptr, &localStatus); |
1131 | if (ucnv_canCreateConverter(converterName, &localStatus)) { |
1132 | gAvailableConverters[gAvailableConverterCount++] = converterName; |
1133 | } |
1134 | } |
1135 | |
1136 | uenum_close(allConvEnum); |
1137 | } |
1138 | |
1139 | |
1140 | static UBool haveAvailableConverterList(UErrorCode *pErrorCode) { |
1141 | umtx_initOnce(gAvailableConvertersInitOnce, &initAvailableConvertersList, *pErrorCode); |
1142 | return U_SUCCESS(*pErrorCode); |
1143 | } |
1144 | |
1145 | U_CFUNC uint16_t |
1146 | ucnv_bld_countAvailableConverters(UErrorCode *pErrorCode) { |
1147 | if (haveAvailableConverterList(pErrorCode)) { |
1148 | return gAvailableConverterCount; |
1149 | } |
1150 | return 0; |
1151 | } |
1152 | |
1153 | U_CFUNC const char * |
1154 | ucnv_bld_getAvailableConverter(uint16_t n, UErrorCode *pErrorCode) { |
1155 | if (haveAvailableConverterList(pErrorCode)) { |
1156 | if (n < gAvailableConverterCount) { |
1157 | return gAvailableConverters[n]; |
1158 | } |
1159 | *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR; |
1160 | } |
1161 | return nullptr; |
1162 | } |
1163 | |
1164 | /* default converter name --------------------------------------------------- */ |
1165 | |
1166 | #if !U_CHARSET_IS_UTF8 |
1167 | /* |
1168 | Copy the canonical converter name. |
1169 | ucnv_getDefaultName must be thread safe, which can call this function. |
1170 | |
1171 | ucnv_setDefaultName calls this function and it doesn't have to be |
1172 | thread safe because there is no reliable/safe way to reset the |
1173 | converter in use in all threads. If you did reset the converter, you |
1174 | would not be sure that retrieving a default converter for one string |
1175 | would be the same type of default converter for a successive string. |
1176 | Since the name is a returned via ucnv_getDefaultName without copying, |
1177 | you shouldn't be modifying or deleting the string from a separate thread. |
1178 | */ |
1179 | static inline void |
1180 | internalSetName(const char *name, UErrorCode *status) { |
1181 | UConverterNamePieces stackPieces; |
1182 | UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER; |
1183 | int32_t length=(int32_t)(uprv_strlen(name)); |
1184 | UBool containsOption = (UBool)(uprv_strchr(name, UCNV_OPTION_SEP_CHAR) != nullptr); |
1185 | const UConverterSharedData *algorithmicSharedData; |
1186 | |
1187 | stackArgs.name = name; |
1188 | if(containsOption) { |
1189 | stackPieces.cnvName[0] = 0; |
1190 | stackPieces.locale[0] = 0; |
1191 | stackPieces.options = 0; |
1192 | parseConverterOptions(name, &stackPieces, &stackArgs, status); |
1193 | if(U_FAILURE(*status)) { |
1194 | return; |
1195 | } |
1196 | } |
1197 | algorithmicSharedData = getAlgorithmicTypeFromName(stackArgs.name); |
1198 | |
1199 | umtx_lock(&cnvCacheMutex); |
1200 | |
1201 | gDefaultAlgorithmicSharedData = algorithmicSharedData; |
1202 | gDefaultConverterContainsOption = containsOption; |
1203 | uprv_memcpy(gDefaultConverterNameBuffer, name, length); |
1204 | gDefaultConverterNameBuffer[length]=0; |
1205 | |
1206 | /* gDefaultConverterName MUST be the last global var set by this function. */ |
1207 | /* It is the variable checked in ucnv_getDefaultName() to see if initialization is required. */ |
1208 | // But there is nothing here preventing that from being reordered, either by the compiler |
1209 | // or hardware. I'm adding the mutex to ucnv_getDefaultName for now. UMTX_CHECK is not enough. |
1210 | // -- Andy |
1211 | gDefaultConverterName = gDefaultConverterNameBuffer; |
1212 | |
1213 | ucnv_enableCleanup(); |
1214 | |
1215 | umtx_unlock(&cnvCacheMutex); |
1216 | } |
1217 | #endif |
1218 | |
1219 | /* |
1220 | * In order to be really thread-safe, the get function would have to take |
1221 | * a buffer parameter and copy the current string inside a mutex block. |
1222 | * This implementation only tries to be really thread-safe while |
1223 | * setting the name. |
1224 | * It assumes that setting a pointer is atomic. |
1225 | */ |
1226 | |
1227 | U_CAPI const char* U_EXPORT2 |
1228 | ucnv_getDefaultName() { |
1229 | #if U_CHARSET_IS_UTF8 |
1230 | return "UTF-8" ; |
1231 | #else |
1232 | /* local variable to be thread-safe */ |
1233 | const char *name; |
1234 | |
1235 | /* |
1236 | Concurrent calls to ucnv_getDefaultName must be thread safe, |
1237 | but ucnv_setDefaultName is not thread safe. |
1238 | */ |
1239 | { |
1240 | icu::Mutex lock(&cnvCacheMutex); |
1241 | name = gDefaultConverterName; |
1242 | } |
1243 | if(name==nullptr) { |
1244 | UErrorCode errorCode = U_ZERO_ERROR; |
1245 | UConverter *cnv = nullptr; |
1246 | |
1247 | name = uprv_getDefaultCodepage(); |
1248 | |
1249 | /* if the name is there, test it out and get the canonical name with options */ |
1250 | if(name != nullptr) { |
1251 | cnv = ucnv_open(name, &errorCode); |
1252 | if(U_SUCCESS(errorCode) && cnv != nullptr) { |
1253 | name = ucnv_getName(cnv, &errorCode); |
1254 | } |
1255 | } |
1256 | |
1257 | if(name == nullptr || name[0] == 0 |
1258 | || U_FAILURE(errorCode) || cnv == nullptr |
1259 | || uprv_strlen(name)>=sizeof(gDefaultConverterNameBuffer)) |
1260 | { |
1261 | /* Panic time, let's use a fallback. */ |
1262 | #if (U_CHARSET_FAMILY == U_ASCII_FAMILY) |
1263 | name = "US-ASCII" ; |
1264 | /* there is no 'algorithmic' converter for EBCDIC */ |
1265 | #elif U_PLATFORM == U_PF_OS390 |
1266 | name = "ibm-1047_P100-1995" UCNV_SWAP_LFNL_OPTION_STRING; |
1267 | #else |
1268 | name = "ibm-37_P100-1995" ; |
1269 | #endif |
1270 | } |
1271 | |
1272 | internalSetName(name, &errorCode); |
1273 | |
1274 | /* The close may make the current name go away. */ |
1275 | ucnv_close(cnv); |
1276 | } |
1277 | |
1278 | return name; |
1279 | #endif |
1280 | } |
1281 | |
1282 | #if U_CHARSET_IS_UTF8 |
1283 | U_CAPI void U_EXPORT2 ucnv_setDefaultName(const char *) {} |
1284 | #else |
1285 | /* |
1286 | This function is not thread safe, and it can't be thread safe. |
1287 | See internalSetName or the API reference for details. |
1288 | */ |
1289 | U_CAPI void U_EXPORT2 |
1290 | ucnv_setDefaultName(const char *converterName) { |
1291 | if(converterName==nullptr) { |
1292 | /* reset to the default codepage */ |
1293 | gDefaultConverterName=nullptr; |
1294 | } else { |
1295 | UErrorCode errorCode = U_ZERO_ERROR; |
1296 | UConverter *cnv = nullptr; |
1297 | const char *name = nullptr; |
1298 | |
1299 | /* if the name is there, test it out and get the canonical name with options */ |
1300 | cnv = ucnv_open(converterName, &errorCode); |
1301 | if(U_SUCCESS(errorCode) && cnv != nullptr) { |
1302 | name = ucnv_getName(cnv, &errorCode); |
1303 | } |
1304 | |
1305 | if(U_SUCCESS(errorCode) && name!=nullptr) { |
1306 | internalSetName(name, &errorCode); |
1307 | } |
1308 | /* else this converter is bad to use. Don't change it to a bad value. */ |
1309 | |
1310 | /* The close may make the current name go away. */ |
1311 | ucnv_close(cnv); |
1312 | |
1313 | /* reset the converter cache */ |
1314 | u_flushDefaultConverter(); |
1315 | } |
1316 | } |
1317 | #endif |
1318 | |
1319 | /* data swapping ------------------------------------------------------------ */ |
1320 | |
1321 | /* most of this might belong more properly into ucnvmbcs.c, but that is so large */ |
1322 | |
1323 | #if !UCONFIG_NO_LEGACY_CONVERSION |
1324 | |
1325 | U_CAPI int32_t U_EXPORT2 |
1326 | ucnv_swap(const UDataSwapper *ds, |
1327 | const void *inData, int32_t length, void *outData, |
1328 | UErrorCode *pErrorCode) { |
1329 | const UDataInfo *pInfo; |
1330 | int32_t headerSize; |
1331 | |
1332 | const uint8_t *inBytes; |
1333 | uint8_t *outBytes; |
1334 | |
1335 | uint32_t offset, count, staticDataSize; |
1336 | int32_t size; |
1337 | |
1338 | const UConverterStaticData *inStaticData; |
1339 | UConverterStaticData *outStaticData; |
1340 | |
1341 | const _MBCSHeader *inMBCSHeader; |
1342 | _MBCSHeader *outMBCSHeader; |
1343 | _MBCSHeader mbcsHeader; |
1344 | uint32_t mbcsHeaderLength; |
1345 | UBool noFromU=false; |
1346 | |
1347 | uint8_t outputType; |
1348 | |
1349 | int32_t maxFastUChar, mbcsIndexLength; |
1350 | |
1351 | const int32_t *inExtIndexes; |
1352 | int32_t extOffset; |
1353 | |
1354 | /* udata_swapDataHeader checks the arguments */ |
1355 | headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode); |
1356 | if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) { |
1357 | return 0; |
1358 | } |
1359 | |
1360 | /* check data format and format version */ |
1361 | pInfo=(const UDataInfo *)((const char *)inData+4); |
1362 | if(!( |
1363 | pInfo->dataFormat[0]==0x63 && /* dataFormat="cnvt" */ |
1364 | pInfo->dataFormat[1]==0x6e && |
1365 | pInfo->dataFormat[2]==0x76 && |
1366 | pInfo->dataFormat[3]==0x74 && |
1367 | pInfo->formatVersion[0]==6 && |
1368 | pInfo->formatVersion[1]>=2 |
1369 | )) { |
1370 | 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" , |
1371 | pInfo->dataFormat[0], pInfo->dataFormat[1], |
1372 | pInfo->dataFormat[2], pInfo->dataFormat[3], |
1373 | pInfo->formatVersion[0], pInfo->formatVersion[1]); |
1374 | *pErrorCode=U_UNSUPPORTED_ERROR; |
1375 | return 0; |
1376 | } |
1377 | |
1378 | inBytes=(const uint8_t *)inData+headerSize; |
1379 | outBytes=(outData == nullptr) ? nullptr : (uint8_t *)outData+headerSize; |
1380 | |
1381 | /* read the initial UConverterStaticData structure after the UDataInfo header */ |
1382 | inStaticData=(const UConverterStaticData *)inBytes; |
1383 | outStaticData=(UConverterStaticData *)outBytes; |
1384 | |
1385 | if(length<0) { |
1386 | staticDataSize=ds->readUInt32(inStaticData->structSize); |
1387 | } else { |
1388 | length-=headerSize; |
1389 | if( length<(int32_t)sizeof(UConverterStaticData) || |
1390 | (uint32_t)length<(staticDataSize=ds->readUInt32(inStaticData->structSize)) |
1391 | ) { |
1392 | udata_printError(ds, "ucnv_swap(): too few bytes (%d after header) for an ICU .cnv conversion table\n" , |
1393 | length); |
1394 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
1395 | return 0; |
1396 | } |
1397 | } |
1398 | |
1399 | if(length>=0) { |
1400 | /* swap the static data */ |
1401 | if(inStaticData!=outStaticData) { |
1402 | uprv_memcpy(outStaticData, inStaticData, staticDataSize); |
1403 | } |
1404 | |
1405 | ds->swapArray32(ds, &inStaticData->structSize, 4, |
1406 | &outStaticData->structSize, pErrorCode); |
1407 | ds->swapArray32(ds, &inStaticData->codepage, 4, |
1408 | &outStaticData->codepage, pErrorCode); |
1409 | |
1410 | ds->swapInvChars(ds, inStaticData->name, (int32_t)uprv_strlen(inStaticData->name), |
1411 | outStaticData->name, pErrorCode); |
1412 | if(U_FAILURE(*pErrorCode)) { |
1413 | udata_printError(ds, "ucnv_swap(): error swapping converter name\n" ); |
1414 | return 0; |
1415 | } |
1416 | } |
1417 | |
1418 | inBytes+=staticDataSize; |
1419 | if (outBytes != nullptr) outBytes+=staticDataSize; |
1420 | if(length>=0) { |
1421 | length-=(int32_t)staticDataSize; |
1422 | } |
1423 | |
1424 | /* check for supported conversionType values */ |
1425 | if(inStaticData->conversionType==UCNV_MBCS) { |
1426 | /* swap MBCS data */ |
1427 | inMBCSHeader=(const _MBCSHeader *)inBytes; |
1428 | outMBCSHeader=(_MBCSHeader *)outBytes; |
1429 | |
1430 | if(0<=length && length<(int32_t)sizeof(_MBCSHeader)) { |
1431 | udata_printError(ds, "ucnv_swap(): too few bytes (%d after headers) for an ICU MBCS .cnv conversion table\n" , |
1432 | length); |
1433 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
1434 | return 0; |
1435 | } |
1436 | if(inMBCSHeader->version[0]==4 && inMBCSHeader->version[1]>=1) { |
1437 | mbcsHeaderLength=MBCS_HEADER_V4_LENGTH; |
1438 | } else if(inMBCSHeader->version[0]==5 && inMBCSHeader->version[1]>=3 && |
1439 | ((mbcsHeader.options=ds->readUInt32(inMBCSHeader->options))& |
1440 | MBCS_OPT_UNKNOWN_INCOMPATIBLE_MASK)==0 |
1441 | ) { |
1442 | mbcsHeaderLength=mbcsHeader.options&MBCS_OPT_LENGTH_MASK; |
1443 | noFromU=(UBool)((mbcsHeader.options&MBCS_OPT_NO_FROM_U)!=0); |
1444 | } else { |
1445 | udata_printError(ds, "ucnv_swap(): unsupported _MBCSHeader.version %d.%d\n" , |
1446 | inMBCSHeader->version[0], inMBCSHeader->version[1]); |
1447 | *pErrorCode=U_UNSUPPORTED_ERROR; |
1448 | return 0; |
1449 | } |
1450 | |
1451 | uprv_memcpy(mbcsHeader.version, inMBCSHeader->version, 4); |
1452 | mbcsHeader.countStates= ds->readUInt32(inMBCSHeader->countStates); |
1453 | mbcsHeader.countToUFallbacks= ds->readUInt32(inMBCSHeader->countToUFallbacks); |
1454 | mbcsHeader.offsetToUCodeUnits= ds->readUInt32(inMBCSHeader->offsetToUCodeUnits); |
1455 | mbcsHeader.offsetFromUTable= ds->readUInt32(inMBCSHeader->offsetFromUTable); |
1456 | mbcsHeader.offsetFromUBytes= ds->readUInt32(inMBCSHeader->offsetFromUBytes); |
1457 | mbcsHeader.flags= ds->readUInt32(inMBCSHeader->flags); |
1458 | mbcsHeader.fromUBytesLength= ds->readUInt32(inMBCSHeader->fromUBytesLength); |
1459 | /* mbcsHeader.options have been read above */ |
1460 | |
1461 | extOffset=(int32_t)(mbcsHeader.flags>>8); |
1462 | outputType=(uint8_t)mbcsHeader.flags; |
1463 | if(noFromU && outputType==MBCS_OUTPUT_1) { |
1464 | udata_printError(ds, "ucnv_swap(): unsupported combination of makeconv --small with SBCS\n" ); |
1465 | *pErrorCode=U_UNSUPPORTED_ERROR; |
1466 | return 0; |
1467 | } |
1468 | |
1469 | /* make sure that the output type is known */ |
1470 | switch(outputType) { |
1471 | case MBCS_OUTPUT_1: |
1472 | case MBCS_OUTPUT_2: |
1473 | case MBCS_OUTPUT_3: |
1474 | case MBCS_OUTPUT_4: |
1475 | case MBCS_OUTPUT_3_EUC: |
1476 | case MBCS_OUTPUT_4_EUC: |
1477 | case MBCS_OUTPUT_2_SISO: |
1478 | case MBCS_OUTPUT_EXT_ONLY: |
1479 | /* OK */ |
1480 | break; |
1481 | default: |
1482 | udata_printError(ds, "ucnv_swap(): unsupported MBCS output type 0x%x\n" , |
1483 | outputType); |
1484 | *pErrorCode=U_UNSUPPORTED_ERROR; |
1485 | return 0; |
1486 | } |
1487 | |
1488 | /* calculate the length of the MBCS data */ |
1489 | |
1490 | /* |
1491 | * utf8Friendly MBCS files (mbcsHeader.version 4.3) |
1492 | * contain an additional mbcsIndex table: |
1493 | * uint16_t[(maxFastUChar+1)>>6]; |
1494 | * where maxFastUChar=((mbcsHeader.version[2]<<8)|0xff). |
1495 | */ |
1496 | maxFastUChar=0; |
1497 | mbcsIndexLength=0; |
1498 | if( outputType!=MBCS_OUTPUT_EXT_ONLY && outputType!=MBCS_OUTPUT_1 && |
1499 | mbcsHeader.version[1]>=3 && (maxFastUChar=mbcsHeader.version[2])!=0 |
1500 | ) { |
1501 | maxFastUChar=(maxFastUChar<<8)|0xff; |
1502 | mbcsIndexLength=((maxFastUChar+1)>>6)*2; /* number of bytes */ |
1503 | } |
1504 | |
1505 | if(extOffset==0) { |
1506 | size=(int32_t)(mbcsHeader.offsetFromUBytes+mbcsIndexLength); |
1507 | if(!noFromU) { |
1508 | size+=(int32_t)mbcsHeader.fromUBytesLength; |
1509 | } |
1510 | |
1511 | /* avoid compiler warnings - not otherwise necessary, and the value does not matter */ |
1512 | inExtIndexes=nullptr; |
1513 | } else { |
1514 | /* there is extension data after the base data, see ucnv_ext.h */ |
1515 | if(length>=0 && length<(extOffset+UCNV_EXT_INDEXES_MIN_LENGTH*4)) { |
1516 | udata_printError(ds, "ucnv_swap(): too few bytes (%d after headers) for an ICU MBCS .cnv conversion table with extension data\n" , |
1517 | length); |
1518 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
1519 | return 0; |
1520 | } |
1521 | |
1522 | inExtIndexes=(const int32_t *)(inBytes+extOffset); |
1523 | size=extOffset+udata_readInt32(ds, inExtIndexes[UCNV_EXT_SIZE]); |
1524 | } |
1525 | |
1526 | if(length>=0) { |
1527 | if(length<size) { |
1528 | udata_printError(ds, "ucnv_swap(): too few bytes (%d after headers) for an ICU MBCS .cnv conversion table\n" , |
1529 | length); |
1530 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
1531 | return 0; |
1532 | } |
1533 | |
1534 | /* copy the data for inaccessible bytes */ |
1535 | if(inBytes!=outBytes) { |
1536 | uprv_memcpy(outBytes, inBytes, size); |
1537 | } |
1538 | |
1539 | /* swap the MBCSHeader, except for the version field */ |
1540 | count=mbcsHeaderLength*4; |
1541 | ds->swapArray32(ds, &inMBCSHeader->countStates, count-4, |
1542 | &outMBCSHeader->countStates, pErrorCode); |
1543 | |
1544 | if(outputType==MBCS_OUTPUT_EXT_ONLY) { |
1545 | /* |
1546 | * extension-only file, |
1547 | * contains a base name instead of normal base table data |
1548 | */ |
1549 | |
1550 | /* swap the base name, between the header and the extension data */ |
1551 | const char *inBaseName=(const char *)inBytes+count; |
1552 | char *outBaseName=(char *)outBytes+count; |
1553 | ds->swapInvChars(ds, inBaseName, (int32_t)uprv_strlen(inBaseName), |
1554 | outBaseName, pErrorCode); |
1555 | } else { |
1556 | /* normal file with base table data */ |
1557 | |
1558 | /* swap the state table, 1kB per state */ |
1559 | offset=count; |
1560 | count=mbcsHeader.countStates*1024; |
1561 | ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
1562 | outBytes+offset, pErrorCode); |
1563 | |
1564 | /* swap the toUFallbacks[] */ |
1565 | offset+=count; |
1566 | count=mbcsHeader.countToUFallbacks*8; |
1567 | ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
1568 | outBytes+offset, pErrorCode); |
1569 | |
1570 | /* swap the unicodeCodeUnits[] */ |
1571 | offset=mbcsHeader.offsetToUCodeUnits; |
1572 | count=mbcsHeader.offsetFromUTable-offset; |
1573 | ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
1574 | outBytes+offset, pErrorCode); |
1575 | |
1576 | /* offset to the stage 1 table, independent of the outputType */ |
1577 | offset=mbcsHeader.offsetFromUTable; |
1578 | |
1579 | if(outputType==MBCS_OUTPUT_1) { |
1580 | /* SBCS: swap the fromU tables, all 16 bits wide */ |
1581 | count=(mbcsHeader.offsetFromUBytes-offset)+mbcsHeader.fromUBytesLength; |
1582 | ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
1583 | outBytes+offset, pErrorCode); |
1584 | } else { |
1585 | /* otherwise: swap the stage tables separately */ |
1586 | |
1587 | /* stage 1 table: uint16_t[0x440 or 0x40] */ |
1588 | if(inStaticData->unicodeMask&UCNV_HAS_SUPPLEMENTARY) { |
1589 | count=0x440*2; /* for all of Unicode */ |
1590 | } else { |
1591 | count=0x40*2; /* only BMP */ |
1592 | } |
1593 | ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
1594 | outBytes+offset, pErrorCode); |
1595 | |
1596 | /* stage 2 table: uint32_t[] */ |
1597 | offset+=count; |
1598 | count=mbcsHeader.offsetFromUBytes-offset; |
1599 | ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
1600 | outBytes+offset, pErrorCode); |
1601 | |
1602 | /* stage 3/result bytes: sometimes uint16_t[] or uint32_t[] */ |
1603 | offset=mbcsHeader.offsetFromUBytes; |
1604 | count= noFromU ? 0 : mbcsHeader.fromUBytesLength; |
1605 | switch(outputType) { |
1606 | case MBCS_OUTPUT_2: |
1607 | case MBCS_OUTPUT_3_EUC: |
1608 | case MBCS_OUTPUT_2_SISO: |
1609 | ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
1610 | outBytes+offset, pErrorCode); |
1611 | break; |
1612 | case MBCS_OUTPUT_4: |
1613 | ds->swapArray32(ds, inBytes+offset, (int32_t)count, |
1614 | outBytes+offset, pErrorCode); |
1615 | break; |
1616 | default: |
1617 | /* just uint8_t[], nothing to swap */ |
1618 | break; |
1619 | } |
1620 | |
1621 | if(mbcsIndexLength!=0) { |
1622 | offset+=count; |
1623 | count=mbcsIndexLength; |
1624 | ds->swapArray16(ds, inBytes+offset, (int32_t)count, |
1625 | outBytes+offset, pErrorCode); |
1626 | } |
1627 | } |
1628 | } |
1629 | |
1630 | if(extOffset!=0) { |
1631 | /* swap the extension data */ |
1632 | inBytes+=extOffset; |
1633 | outBytes+=extOffset; |
1634 | |
1635 | /* swap toUTable[] */ |
1636 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_INDEX]); |
1637 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_LENGTH]); |
1638 | ds->swapArray32(ds, inBytes+offset, length*4, outBytes+offset, pErrorCode); |
1639 | |
1640 | /* swap toUUChars[] */ |
1641 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_UCHARS_INDEX]); |
1642 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_TO_U_UCHARS_LENGTH]); |
1643 | ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
1644 | |
1645 | /* swap fromUTableUChars[] */ |
1646 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_UCHARS_INDEX]); |
1647 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_LENGTH]); |
1648 | ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
1649 | |
1650 | /* swap fromUTableValues[] */ |
1651 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_VALUES_INDEX]); |
1652 | /* same length as for fromUTableUChars[] */ |
1653 | ds->swapArray32(ds, inBytes+offset, length*4, outBytes+offset, pErrorCode); |
1654 | |
1655 | /* no need to swap fromUBytes[] */ |
1656 | |
1657 | /* swap fromUStage12[] */ |
1658 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_12_INDEX]); |
1659 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_12_LENGTH]); |
1660 | ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
1661 | |
1662 | /* swap fromUStage3[] */ |
1663 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3_INDEX]); |
1664 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3_LENGTH]); |
1665 | ds->swapArray16(ds, inBytes+offset, length*2, outBytes+offset, pErrorCode); |
1666 | |
1667 | /* swap fromUStage3b[] */ |
1668 | offset=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3B_INDEX]); |
1669 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_FROM_U_STAGE_3B_LENGTH]); |
1670 | ds->swapArray32(ds, inBytes+offset, length*4, outBytes+offset, pErrorCode); |
1671 | |
1672 | /* swap indexes[] */ |
1673 | length=udata_readInt32(ds, inExtIndexes[UCNV_EXT_INDEXES_LENGTH]); |
1674 | ds->swapArray32(ds, inBytes, length*4, outBytes, pErrorCode); |
1675 | } |
1676 | } |
1677 | } else { |
1678 | udata_printError(ds, "ucnv_swap(): unknown conversionType=%d!=UCNV_MBCS\n" , |
1679 | inStaticData->conversionType); |
1680 | *pErrorCode=U_UNSUPPORTED_ERROR; |
1681 | return 0; |
1682 | } |
1683 | |
1684 | return headerSize+(int32_t)staticDataSize+size; |
1685 | } |
1686 | |
1687 | #endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */ |
1688 | |
1689 | #endif |
1690 | |