X-Git-Url: https://git.saurik.com/apple/icu.git/blobdiff_plain/4388f060552cc537e71e957d32f35e9d75a61233..38fbf2fd31f5cd99b500914d6037b1d06b608645:/icuSources/common/udata.cpp diff --git a/icuSources/common/udata.cpp b/icuSources/common/udata.cpp index c4f8752a..80c10c04 100644 --- a/icuSources/common/udata.cpp +++ b/icuSources/common/udata.cpp @@ -1,12 +1,14 @@ +// © 2016 and later: Unicode, Inc. and others. +// License & terms of use: http://www.unicode.org/copyright.html /* ****************************************************************************** * -* Copyright (C) 1999-2012, International Business Machines +* Copyright (C) 1999-2016, International Business Machines * Corporation and others. All Rights Reserved. * ****************************************************************************** * file name: udata.cpp -* encoding: US-ASCII +* encoding: UTF-8 * tab size: 8 (not used) * indentation:4 * @@ -29,7 +31,9 @@ might have to #include some other header #include "charstr.h" #include "cmemory.h" #include "cstring.h" +#include "mutex.h" #include "putilimp.h" +#include "uassert.h" #include "ucln_cmn.h" #include "ucmndata.h" #include "udatamem.h" @@ -65,19 +69,19 @@ might have to #include some other header /* If you are excruciatingly bored turn this on .. */ /* #define UDATA_DEBUG 1 */ +/* For debugging use of timezone data in a separate file */ +/* #define UDATA_TZFILES_DEBUG 1 */ -#if defined(UDATA_DEBUG) +#if defined(UDATA_DEBUG) || defined(UDATA_TZFILES_DEBUG) # include #endif -#define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0])) - U_NAMESPACE_USE /* * Forward declarations */ -static UDataMemory *udata_findCachedData(const char *path); +static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err); /*********************************************************************** * @@ -101,13 +105,19 @@ static UDataMemory *udata_findCachedData(const char *path); * that they really need, reducing the size of binaries that take advantage * of this. */ -static UDataMemory *gCommonICUDataArray[10] = { NULL }; +static UDataMemory *gCommonICUDataArray[10] = { NULL }; // Access protected by icu global mutex. -static UBool gHaveTriedToLoadCommonData = FALSE; /* See extendICUData(). */ +static u_atomic_int32_t gHaveTriedToLoadCommonData = ATOMIC_INT32_T_INITIALIZER(0); // See extendICUData(). static UHashtable *gCommonDataCache = NULL; /* Global hash table of opened ICU data files. */ +static icu::UInitOnce gCommonDataCacheInitOnce = U_INITONCE_INITIALIZER; -static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; +#if U_PLATFORM_HAS_WINUWP_API == 0 +static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; // Access not synchronized. + // Modifying is documented as thread-unsafe. +#else +static UDataFileAccess gDataFileAccess = UDATA_NO_FILES; // Windows UWP looks in one spot explicitly +#endif static UBool U_CALLCONV udata_cleanup(void) @@ -118,34 +128,37 @@ udata_cleanup(void) uhash_close(gCommonDataCache); /* Table owns the contents, and will delete them. */ gCommonDataCache = NULL; /* Cleanup is not thread safe. */ } + gCommonDataCacheInitOnce.reset(); - for (i = 0; i < LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != NULL; ++i) { + for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != NULL; ++i) { udata_close(gCommonICUDataArray[i]); gCommonICUDataArray[i] = NULL; } - gHaveTriedToLoadCommonData = FALSE; + gHaveTriedToLoadCommonData = 0; return TRUE; /* Everything was cleaned up */ } static UBool U_CALLCONV -findCommonICUDataByName(const char *inBasename) +findCommonICUDataByName(const char *inBasename, UErrorCode &err) { UBool found = FALSE; int32_t i; - UDataMemory *pData = udata_findCachedData(inBasename); - if (pData == NULL) + UDataMemory *pData = udata_findCachedData(inBasename, err); + if (U_FAILURE(err) || pData == NULL) return FALSE; - for (i = 0; i < LENGTHOF(gCommonICUDataArray); ++i) { - if ((gCommonICUDataArray[i] != NULL) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) { - /* The data pointer is already in the array. */ - found = TRUE; - break; + { + Mutex lock; + for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) { + if ((gCommonICUDataArray[i] != NULL) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) { + /* The data pointer is already in the array. */ + found = TRUE; + break; + } } } - return found; } @@ -172,10 +185,9 @@ setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to ca /* their locals. */ UDatamemory_assign(newCommonData, pData); umtx_lock(NULL); - for (i = 0; i < LENGTHOF(gCommonICUDataArray); ++i) { + for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) { if (gCommonICUDataArray[i] == NULL) { gCommonICUDataArray[i] = newCommonData; - ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); didUpdate = TRUE; break; } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) { @@ -185,10 +197,12 @@ setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to ca } umtx_unlock(NULL); - if (i == LENGTHOF(gCommonICUDataArray) && warn) { + if (i == UPRV_LENGTHOF(gCommonICUDataArray) && warn) { *pErr = U_USING_DEFAULT_WARNING; } - if (!didUpdate) { + if (didUpdate) { + ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); + } else { uprv_free(newCommonData); } return didUpdate; @@ -198,7 +212,7 @@ static UBool setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) { UDataMemory tData; UDataMemory_init(&tData); - tData.pHeader = (const DataHeader *)pData; + UDataMemory_setData(&tData, pData); udata_checkCommonData(&tData, pErrorCode); return setCommonICUData(&tData, FALSE, pErrorCode); } @@ -262,56 +276,41 @@ static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) { uprv_free(pDCEl); /* delete 'this' */ } - /* udata_getCacheHashTable() - * Get the hash table used to store the data cache entries. - * Lazy create it if it doesn't yet exist. - */ -static UHashtable *udata_getHashTable() { - UErrorCode err = U_ZERO_ERROR; - UBool cacheIsInitialized; - UHashtable *tHT = NULL; - - UMTX_CHECK(NULL, (gCommonDataCache != NULL), cacheIsInitialized); - - if (cacheIsInitialized) { - return gCommonDataCache; - } - - tHT = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err); - /* Check for null pointer. */ - if (tHT == NULL) { - return NULL; /* TODO: Handle this error better. */ - } - uhash_setValueDeleter(tHT, DataCacheElement_deleter); - - umtx_lock(NULL); - if (gCommonDataCache == NULL) { - gCommonDataCache = tHT; - tHT = NULL; - ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); - } - umtx_unlock(NULL); - if (tHT != NULL) { - uhash_close(tHT); - } - +static void U_CALLCONV udata_initHashTable(UErrorCode &err) { + U_ASSERT(gCommonDataCache == NULL); + gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err); if (U_FAILURE(err)) { - return NULL; /* TODO: handle this error better. */ + return; } + U_ASSERT(gCommonDataCache != NULL); + uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter); + ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); +} + + /* udata_getCacheHashTable() + * Get the hash table used to store the data cache entries. + * Lazy create it if it doesn't yet exist. + */ +static UHashtable *udata_getHashTable(UErrorCode &err) { + umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable, err); return gCommonDataCache; } -static UDataMemory *udata_findCachedData(const char *path) +static UDataMemory *udata_findCachedData(const char *path, UErrorCode &err) { UHashtable *htable; UDataMemory *retVal = NULL; DataCacheElement *el; const char *baseName; + htable = udata_getHashTable(err); + if (U_FAILURE(err)) { + return NULL; + } + baseName = findBasename(path); /* Cache remembers only the base name, not the full path. */ - htable = udata_getHashTable(); umtx_lock(NULL); el = (DataCacheElement *)uhash_get(htable, baseName); umtx_unlock(NULL); @@ -333,6 +332,7 @@ static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UEr DataCacheElement *oldValue = NULL; UErrorCode subErr = U_ZERO_ERROR; + htable = udata_getHashTable(*pErr); if (U_FAILURE(*pErr)) { return NULL; } @@ -365,7 +365,6 @@ static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UEr /* Stick the new DataCacheElement into the hash table. */ - htable = udata_getHashTable(); umtx_lock(NULL); oldValue = (DataCacheElement *)uhash_get(htable, path); if (oldValue != NULL) { @@ -403,9 +402,6 @@ static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UEr * * *----------------------------------------------------------------------*/ -#define U_DATA_PATHITER_BUFSIZ 128 /* Size of local buffer for paths */ - /* Overflow causes malloc of larger buf */ - U_NAMESPACE_BEGIN class UDataPathIterator @@ -629,12 +625,14 @@ U_NAMESPACE_END /*----------------------------------------------------------------------* * * - * Add a static reference to the common data library * + * Add a static reference to the common data library * * Unless overridden by an explicit udata_setCommonData, this will be * * our common data. * * * *----------------------------------------------------------------------*/ +#if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time extern "C" const DataHeader U_DATA_API U_ICUDATA_ENTRY_POINT; +#endif /* * This would be a good place for weak-linkage declarations of @@ -674,10 +672,15 @@ openCommonData(const char *path, /* Path from OpenChoice? */ /* ??????? TODO revisit this */ if (commonDataIndex >= 0) { /* "mini-cache" for common ICU data */ - if(commonDataIndex >= LENGTHOF(gCommonICUDataArray)) { + if(commonDataIndex >= UPRV_LENGTHOF(gCommonICUDataArray)) { return NULL; } - if(gCommonICUDataArray[commonDataIndex] == NULL) { + { + Mutex lock; + if(gCommonICUDataArray[commonDataIndex] != NULL) { + return gCommonICUDataArray[commonDataIndex]; + } +#if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time int32_t i; for(i = 0; i < commonDataIndex; ++i) { if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT) { @@ -685,23 +688,29 @@ openCommonData(const char *path, /* Path from OpenChoice? */ return NULL; } } +#endif + } - /* Add the linked-in data to the list. */ - /* - * This is where we would check and call weakly linked partial-data-library - * access functions. - */ - /* - if (uprv_getICUData_collation) { - setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErrorCode); - } - if (uprv_getICUData_conversion) { - setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pErrorCode); - } - */ - setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, FALSE, pErrorCode); + /* Add the linked-in data to the list. */ + /* + * This is where we would check and call weakly linked partial-data-library + * access functions. + */ + /* + if (uprv_getICUData_collation) { + setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErrorCode); } - return gCommonICUDataArray[commonDataIndex]; + if (uprv_getICUData_conversion) { + setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pErrorCode); + } + */ +#if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time + setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, FALSE, pErrorCode); + { + Mutex lock; + return gCommonICUDataArray[commonDataIndex]; + } +#endif } @@ -720,18 +729,18 @@ openCommonData(const char *path, /* Path from OpenChoice? */ #ifdef UDATA_DEBUG fprintf(stderr, "ocd: no basename in %s, bailing.\n", path); #endif - *pErrorCode=U_FILE_ACCESS_ERROR; + if (U_SUCCESS(*pErrorCode)) { + *pErrorCode=U_FILE_ACCESS_ERROR; + } return NULL; } /* Is the requested common data file already open and cached? */ /* Note that the cache is keyed by the base name only. The rest of the path, */ /* if any, is not considered. */ - { - UDataMemory *dataToReturn = udata_findCachedData(inBasename); - if (dataToReturn != NULL) { - return dataToReturn; - } + UDataMemory *dataToReturn = udata_findCachedData(inBasename, *pErrorCode); + if (dataToReturn != NULL || U_FAILURE(*pErrorCode)) { + return dataToReturn; } /* Requested item is not in the cache. @@ -762,6 +771,9 @@ openCommonData(const char *path, /* Path from OpenChoice? */ } #endif + if (U_FAILURE(*pErrorCode)) { + return NULL; + } if (!UDataMemory_isLoaded(&tData)) { /* no common data */ *pErrorCode=U_FILE_ACCESS_ERROR; @@ -806,10 +818,10 @@ static UBool extendICUData(UErrorCode *pErr) * Use a specific mutex to avoid nested locks of the global mutex. */ #if MAP_IMPLEMENTATION==MAP_STDIO - static UMTX extendICUDataMutex = NULL; + static UMutex extendICUDataMutex = U_MUTEX_INITIALIZER; umtx_lock(&extendICUDataMutex); #endif - if(!gHaveTriedToLoadCommonData) { + if(!umtx_loadAcquire(gHaveTriedToLoadCommonData)) { /* See if we can explicitly open a .dat file for the ICUData. */ pData = openCommonData( U_ICUDATA_NAME, /* "icudt20l" , for example. */ @@ -834,10 +846,10 @@ static UBool extendICUData(UErrorCode *pErr) pErr); /* setCommonICUData honors errors; NOP if error set */ } - gHaveTriedToLoadCommonData = TRUE; + umtx_storeRelease(gHaveTriedToLoadCommonData, 1); } - didUpdate = findCommonICUDataByName(U_ICUDATA_NAME); /* Return 'true' when a racing writes out the extended */ + didUpdate = findCommonICUDataByName(U_ICUDATA_NAME, *pErr); /* Return 'true' when a racing writes out the extended */ /* data after another thread has failed to see it (in openCommonData), so */ /* extended data can be examined. */ /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */ @@ -1078,6 +1090,17 @@ static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName } } +/* + * Identify the Time Zone resources that are subject to special override data loading. + */ +static UBool isTimeZoneFile(const char *name, const char *type) { + return ((uprv_strcmp(type, "res") == 0) && + (uprv_strcmp(name, "zoneinfo64") == 0 || + uprv_strcmp(name, "timezoneTypes") == 0 || + uprv_strcmp(name, "windowsZones") == 0 || + uprv_strcmp(name, "metaZones") == 0)); +} + /* * A note on the ownership of Mapped Memory * @@ -1234,9 +1257,14 @@ doOpenChoice(const char *path, const char *type, const char *name, fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data()); #endif +#if U_PLATFORM_HAS_WINUWP_API == 0 // Windows UWP Platform does not support dll icu data at this time if(path == NULL) { path = COMMON_DATA_NAME; /* "icudt26e" */ } +#else + // Windows UWP expects only a single data file. + path = COMMON_DATA_NAME; /* "icudt26e" */ +#endif /************************ Begin loop looking for ind. files ***************/ #ifdef UDATA_DEBUG @@ -1246,15 +1274,83 @@ doOpenChoice(const char *path, const char *type, const char *name, /* End of dealing with a null basename */ dataPath = u_getDataDirectory(); + /**** Time zone individual files override */ + if (isICUData && isTimeZoneFile(name, type)) { + const char *tzFilesDir = u_getTimeZoneFilesDirectory(pErrorCode); + if (tzFilesDir[0] != 0) { +#ifdef UDATA_DEBUG + fprintf(stderr, "Trying Time Zone Files directory = %s\n", tzFilesDir); +#endif +#ifdef UDATA_TZFILES_DEBUG + fprintf(stderr, "# dOC U_TIMEZONE_FILES_DIR: %s\n", U_TIMEZONE_FILES_DIR); +#endif + +#if defined(U_TIMEZONE_PACKAGE) + // make tztocEntryName, like tocEntryName but with our package name + UErrorCode tzpkgErrorCode = U_ZERO_ERROR; + CharString tztocPkgPath; + tztocPkgPath.append(tzFilesDir, tzpkgErrorCode); + tztocPkgPath.append(U_FILE_SEP_CHAR, tzpkgErrorCode).append(U_TIMEZONE_PACKAGE, tzpkgErrorCode); + CharString tztocEntryName; + tztocEntryName.append(U_TIMEZONE_PACKAGE, tzpkgErrorCode); + if(!treeName.isEmpty()) { + tztocEntryName.append(U_TREE_ENTRY_SEP_CHAR, tzpkgErrorCode).append(treeName, tzpkgErrorCode); + } + tztocEntryName.append(U_TREE_ENTRY_SEP_CHAR, tzpkgErrorCode).append(name, tzpkgErrorCode); + if(type!=NULL && *type!=0) { + tztocEntryName.append(".", tzpkgErrorCode).append(type, tzpkgErrorCode); + } +#ifdef UDATA_TZFILES_DEBUG + fprintf(stderr, "# dOC tz pkg, doLoadFromCommonData start; U_TIMEZONE_PACKAGE: %s, tztocPkgPath.data(): %s, tztocEntryName.data(): %s, name: %s\n", + U_TIMEZONE_PACKAGE, tztocPkgPath.data(), tztocEntryName.data(), name); +#endif + retVal = doLoadFromCommonData(FALSE, "" /*ignored*/, "" /*ignored*/, "" /*ignored*/, + tztocEntryName.data(), // tocEntryName, like icutz44/zoneinfo64.res + tztocPkgPath.data(), // path = path to pkg, like /usr/share/icu/icutz44l + type, name, isAcceptable, context, &subErrorCode, &tzpkgErrorCode); +#ifdef UDATA_TZFILES_DEBUG + fprintf(stderr, "# dOC tz pkg, doLoadFromCommonData end; status %d, retVal %p\n", tzpkgErrorCode, retVal); +#endif + if(U_SUCCESS(tzpkgErrorCode) && retVal != NULL) { + return retVal; + } +#endif /* defined(U_TIMEZONE_PACKAGE) */ + // The following assumes any timezone resources in tzFilesDir are in individual .res files +#ifdef UDATA_TZFILES_DEBUG + fprintf(stderr, "# dOC tz files, doLoadFromIndividualFiles start; tzFilesDir: %s, tocEntryPathSuffix: %s, name: %s\n", + tzFilesDir, tocEntryPathSuffix, name); +#endif + retVal = doLoadFromIndividualFiles(/* pkgName.data() */ "", tzFilesDir, tocEntryPathSuffix, + /* path */ "", type, name, isAcceptable, context, &subErrorCode, pErrorCode); +#ifdef UDATA_TZFILES_DEBUG + fprintf(stderr, "# dOC tz files, doLoadFromIndividualFiles end; status %d, retVal %p\n", *pErrorCode, retVal); +#endif + if((retVal != NULL) || U_FAILURE(*pErrorCode)) { + return retVal; + } + } + } + /**** COMMON PACKAGE - only if packages are first. */ if(gDataFileAccess == UDATA_PACKAGES_FIRST) { #ifdef UDATA_DEBUG fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n"); #endif /* #2 */ +#ifdef UDATA_TZFILES_DEBUG + if (isTimeZoneFile(name, type)) { + fprintf(stderr, "# dOC std common 1, doLoadFromCommonData start; U_TIMEZONE_PACKAGE: path: %s, tocEntryName.data(): %s, name: %s\n", + path, tocEntryName.data(), name); + } +#endif retVal = doLoadFromCommonData(isICUData, pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); +#ifdef UDATA_TZFILES_DEBUG + if (isTimeZoneFile(name, type)) { + fprintf(stderr, "# dOC std common 1, doLoadFromCommonData end; status %d, retVal %p\n", *pErrorCode, retVal); + } +#endif if((retVal != NULL) || U_FAILURE(*pErrorCode)) { return retVal; } @@ -1268,8 +1364,19 @@ doOpenChoice(const char *path, const char *type, const char *name, #endif /* Check to make sure that there is a dataPath to iterate over */ if ((dataPath && *dataPath) || !isICUData) { +#ifdef UDATA_TZFILES_DEBUG + if (isTimeZoneFile(name, type)) { + fprintf(stderr, "# dOC std indiv files, doLoadFromIndividualFiles start; dataPath: %s, tocEntryPathSuffix: %s, name: %s\n", + dataPath, tocEntryPathSuffix, name); + } +#endif retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix, path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); +#ifdef UDATA_TZFILES_DEBUG + if (isTimeZoneFile(name, type)) { + fprintf(stderr, "# dOC std indiv files, doLoadFromIndividualFiles end; status %d, retVal %p\n", *pErrorCode, retVal); + } +#endif if((retVal != NULL) || U_FAILURE(*pErrorCode)) { return retVal; } @@ -1281,10 +1388,21 @@ doOpenChoice(const char *path, const char *type, const char *name, (gDataFileAccess==UDATA_FILES_FIRST)) { #ifdef UDATA_DEBUG fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n"); +#endif +#ifdef UDATA_TZFILES_DEBUG + if (isTimeZoneFile(name, type)) { + fprintf(stderr, "# dOC std common 2, doLoadFromCommonData start; U_TIMEZONE_PACKAGE: path: %s, tocEntryName.data(): %s, name: %s\n", + path, tocEntryName.data(), name); + } #endif retVal = doLoadFromCommonData(isICUData, pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); +#ifdef UDATA_TZFILES_DEBUG + if (isTimeZoneFile(name, type)) { + fprintf(stderr, "# dOC std common 2, doLoadFromCommonData end; status %d, retVal %p\n", *pErrorCode, retVal); + } +#endif if((retVal != NULL) || U_FAILURE(*pErrorCode)) { return retVal; } @@ -1386,5 +1504,6 @@ udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) { U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/) { + // Note: this function is documented as not thread safe. gDataFileAccess = access; }