/*
*****************************************************************
-* Copyright (c) 2002-2011, International Business Machines Corporation
+* Copyright (c) 2002-2014, International Business Machines Corporation
* and others. All Rights Reserved.
*****************************************************************
* Date Name Description
#include "unicode/uobject.h"
#include "unicode/uscript.h"
-#include "nultrans.h"
+
#include "anytrans.h"
-#include "uvector.h"
-#include "tridpars.h"
#include "hash.h"
+#include "mutex.h"
+#include "nultrans.h"
#include "putilimp.h"
+#include "tridpars.h"
#include "uinvchar.h"
+#include "uvector.h"
//------------------------------------------------------------
// Constants
*/
static void U_CALLCONV
_deleteTransliterator(void *obj) {
- delete (icu::Transliterator*) obj;
+ delete (icu::Transliterator*) obj;
}
U_CDECL_END
* The end of the run, exclusive, valid after next() returns.
*/
int32_t limit;
-
+
/**
* Constructs a run iterator over the given text from start
* (inclusive) to limit (exclusive).
UScriptCode theTargetScript,
UErrorCode& ec) :
Transliterator(id, NULL),
- targetScript(theTargetScript)
+ targetScript(theTargetScript)
{
cache = uhash_open(uhash_hashLong, uhash_compareLong, NULL, &ec);
if (U_FAILURE(ec)) {
// Try to instantiate transliterator from it.scriptCode to
// our target or target/variant
Transliterator* t = getTransliterator(it.scriptCode);
-
+
if (t == NULL) {
// We have no transliterator. Do nothing, but keep
// pos.start up to date.
// a non-incremental transliteration. Otherwise do an
// incremental one.
UBool incremental = isIncremental && (it.limit >= allLimit);
-
+
pos.start = uprv_max(allStart, it.start);
pos.limit = uprv_min(allLimit, it.limit);
int32_t limit = pos.limit;
return NULL;
}
- Transliterator* t = (Transliterator*) uhash_iget(cache, (int32_t) source);
+ Transliterator* t = NULL;
+ {
+ Mutex m(NULL);
+ t = (Transliterator*) uhash_iget(cache, (int32_t) source);
+ }
if (t == NULL) {
UErrorCode ec = U_ZERO_ERROR;
UnicodeString sourceName(uscript_getName(source), -1, US_INV);
UnicodeString id(sourceName);
id.append(TARGET_SEP).append(target);
-
+
t = Transliterator::createInstance(id, UTRANS_FORWARD, ec);
if (U_FAILURE(ec) || t == NULL) {
delete t;
-
+
// Try to pivot around Latin, our most common script
id = sourceName;
id.append(LATIN_PIVOT, -1).append(target);
}
if (t != NULL) {
- uhash_iput(cache, (int32_t) source, t, &ec);
+ Transliterator *rt = NULL;
+ {
+ Mutex m(NULL);
+ rt = static_cast<Transliterator *> (uhash_iget(cache, (int32_t) source));
+ if (rt == NULL) {
+ // Common case, no race to cache this new transliterator.
+ uhash_iput(cache, (int32_t) source, t, &ec);
+ } else {
+ // Race case, some other thread beat us to caching this transliterator.
+ Transliterator *temp = rt;
+ rt = t; // Our newly created transliterator that lost the race & now needs deleting.
+ t = temp; // The transliterator from the cache that we will return.
+ }
+ }
+ delete rt; // will be non-null only in case of races.
}
}
-
return t;
}
UErrorCode ec = U_ZERO_ERROR;
int32_t nameLen = name.length();
UBool isInvariant = uprv_isInvariantUString(name.getBuffer(), nameLen);
-
+
if (isInvariant) {
name.extract(0, nameLen, buf, (int32_t)sizeof(buf), US_INV);
buf[127] = 0; // Make sure that we NULL terminate the string.
if (seen.geti(target) != 0) continue;
ec = U_ZERO_ERROR;
seen.puti(target, 1, ec);
-
+
// Get the script code for the target. If not a script, ignore.
UScriptCode targetScript = scriptNameToCode(target);
if (targetScript == USCRIPT_INVALID_CODE) continue;
for (int32_t v=0; v<variantCount; ++v) {
UnicodeString variant;
Transliterator::_getAvailableVariant(v, source, target, variant);
-
+
UnicodeString id;
TransliteratorIDParser::STVtoID(UnicodeString(TRUE, ANY, 3), target, variant, id);
ec = U_ZERO_ERROR;
#include "intltest.h"
#include "tsmthred.h"
#include "unicode/ushape.h"
-
+#include "unicode/translit.h"
#if U_PLATFORM_USES_ONLY_WIN32_API
/* Prefer native Windows APIs even if POSIX is implemented (i.e., on Cygwin). */
}
break;
- case 5:
+ case 5:
name = "TestArabicShapingThreads";
if (exec) {
TestArabicShapingThreads();
}
break;
-
+ case 6:
+ name = "TestAnyTranslit";
+ if (exec) {
+ TestAnyTranslit();
+ }
+ break;
+
default:
name = "";
break; //needed to end loop
}
delete testString;
}
+};
+
+
+// Test for ticket #10673, race in cache code in AnyTransliterator.
+// It's difficult to make the original unsafe code actually fail, but
+// this test will fairly reliably take the code path for races in
+// populating the cache.
+
+#if !UCONFIG_NO_TRANSLITERATION
+class TxThread: public SimpleThread {
+ private:
+ Transliterator *fSharedTranslit;
+ public:
+ UBool fSuccess;
+ TxThread(Transliterator *tx) : fSharedTranslit(tx), fSuccess(FALSE) {};
+ ~TxThread();
+ void run();
+};
+
+TxThread::~TxThread() {};
+void TxThread::run() {
+ UnicodeString greekString("\\u03B4\\u03B9\\u03B1\\u03C6\\u03BF\\u03C1\\u03B5\\u03C4\\u03B9\\u03BA\\u03BF\\u03CD\\u03C2");
+ greekString = greekString.unescape();
+ fSharedTranslit->transliterate(greekString);
+ fSuccess = greekString[0] == 0x64; // 'd'. The whole transliterated string is "diaphoretikous" (accented u).
+}
+#endif
+
+
+void MultithreadTest::TestAnyTranslit() {
+#if !UCONFIG_NO_TRANSLITERATION
+ UErrorCode status = U_ZERO_ERROR;
+ LocalPointer<Transliterator> tx(Transliterator::createInstance("Any-Latin", UTRANS_FORWARD, status));
+ if (U_FAILURE(status)) {
+ errln("File %s, Line %d: Error, status = %s", __FILE__, __LINE__, u_errorName(status));
+ return;
+ }
+ TxThread * threads[4];
+ int32_t i;
+ for (i=0; i<4; i++) {
+ threads[i] = new TxThread(tx.getAlias());
+ }
+ for (i=0; i<4; i++) {
+ threads[i]->start();
+ }
+ int32_t patience = 100;
+ UBool success;
+ UBool someThreadRunning;
+ do {
+ someThreadRunning = FALSE;
+ success = TRUE;
+ for (i=0; i<4; i++) {
+ if (threads[i]->isRunning()) {
+ someThreadRunning = TRUE;
+ SimpleThread::sleep(10);
+ break;
+ } else {
+ if (threads[i]->fSuccess == FALSE) {
+ success = FALSE;
+ }
+ }
+ }
+ } while (someThreadRunning && --patience > 0);
+
+ if (patience <= 0) {
+ errln("File %s, Line %d: Error, one or more threads did not complete.", __FILE__, __LINE__);
+ }
+ if (success == FALSE) {
+ errln("File %s, Line %d: Error, transliteration result incorrect.", __FILE__, __LINE__);
+ }
+
+ for (i=0; i<4; i++) {
+ delete threads[i];
+ }
+#endif // !UCONFIG_NO_TRANSLITERATION
}
#endif // ICU_USE_THREADS
projects / icu / commitdiff