summaryrefslogtreecommitdiff
path: root/Build/source/libs/icu-xetex/test/intltest/strcase.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'Build/source/libs/icu-xetex/test/intltest/strcase.cpp')
-rw-r--r--Build/source/libs/icu-xetex/test/intltest/strcase.cpp542
1 files changed, 0 insertions, 542 deletions
diff --git a/Build/source/libs/icu-xetex/test/intltest/strcase.cpp b/Build/source/libs/icu-xetex/test/intltest/strcase.cpp
deleted file mode 100644
index 0645f1269ce..00000000000
--- a/Build/source/libs/icu-xetex/test/intltest/strcase.cpp
+++ /dev/null
@@ -1,542 +0,0 @@
-/*
-*******************************************************************************
-*
-* Copyright (C) 2002-2007, International Business Machines
-* Corporation and others. All Rights Reserved.
-*
-*******************************************************************************
-* file name: strcase.cpp
-* encoding: US-ASCII
-* tab size: 8 (not used)
-* indentation:4
-*
-* created on: 2002mar12
-* created by: Markus W. Scherer
-*
-* Test file for string casing C++ API functions.
-*/
-
-#include "unicode/uchar.h"
-#include "unicode/ures.h"
-#include "unicode/uloc.h"
-#include "unicode/locid.h"
-#include "unicode/ubrk.h"
-#include "unicode/unistr.h"
-#include "unicode/ucasemap.h"
-#include "ustrtest.h"
-#include "unicode/tstdtmod.h"
-
-#define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
-
-StringCaseTest::~StringCaseTest() {}
-
-void
-StringCaseTest::runIndexedTest(int32_t index, UBool exec, const char *&name, char * /*par*/) {
- if (exec) logln("TestSuite StringCaseTest: ");
- switch (index) {
- case 0: name = "TestCaseConversion"; if (exec) TestCaseConversion(); break;
- case 1:
- name = "TestCasing";
-#if !UCONFIG_NO_BREAK_ITERATION
- if(exec) TestCasing();
-#endif
- break;
-
- default: name = ""; break; //needed to end loop
- }
-}
-
-void
-StringCaseTest::TestCaseConversion()
-{
- static const UChar uppercaseGreek[] =
- { 0x399, 0x395, 0x3a3, 0x3a5, 0x3a3, 0x20, 0x03a7, 0x3a1, 0x399, 0x3a3, 0x3a4,
- 0x39f, 0x3a3, 0 };
- // "IESUS CHRISTOS"
-
- static const UChar lowercaseGreek[] =
- { 0x3b9, 0x3b5, 0x3c3, 0x3c5, 0x3c2, 0x20, 0x03c7, 0x3c1, 0x3b9, 0x3c3, 0x3c4,
- 0x3bf, 0x3c2, 0 };
- // "iesus christos"
-
- static const UChar lowercaseTurkish[] =
- { 0x69, 0x73, 0x74, 0x61, 0x6e, 0x62, 0x75, 0x6c, 0x2c, 0x20, 0x6e, 0x6f, 0x74, 0x20, 0x63, 0x6f,
- 0x6e, 0x73, 0x74, 0x61, 0x6e, 0x74, 0x0131, 0x6e, 0x6f, 0x70, 0x6c, 0x65, 0x21, 0 };
-
- static const UChar uppercaseTurkish[] =
- { 0x54, 0x4f, 0x50, 0x4b, 0x41, 0x50, 0x49, 0x20, 0x50, 0x41, 0x4c, 0x41, 0x43, 0x45, 0x2c, 0x20,
- 0x0130, 0x53, 0x54, 0x41, 0x4e, 0x42, 0x55, 0x4c, 0 };
-
- UnicodeString expectedResult;
- UnicodeString test3;
-
- test3 += (UChar32)0x0130;
- test3 += "STANBUL, NOT CONSTANTINOPLE!";
-
- UnicodeString test4(test3);
- test4.toLower(Locale(""));
- expectedResult = UnicodeString("i\\u0307stanbul, not constantinople!", "").unescape();
- if (test4 != expectedResult)
- errln("1. toLower failed: expected \"" + expectedResult + "\", got \"" + test4 + "\".");
-
- test4 = test3;
- test4.toLower(Locale("tr", "TR"));
- expectedResult = lowercaseTurkish;
- if (test4 != expectedResult)
- errln("2. toLower failed: expected \"" + expectedResult + "\", got \"" + test4 + "\".");
-
- test3 = "topkap";
- test3 += (UChar32)0x0131;
- test3 += " palace, istanbul";
- test4 = test3;
-
- test4.toUpper(Locale(""));
- expectedResult = "TOPKAPI PALACE, ISTANBUL";
- if (test4 != expectedResult)
- errln("toUpper failed: expected \"" + expectedResult + "\", got \"" + test4 + "\".");
-
- test4 = test3;
- test4.toUpper(Locale("tr", "TR"));
- expectedResult = uppercaseTurkish;
- if (test4 != expectedResult)
- errln("toUpper failed: expected \"" + expectedResult + "\", got \"" + test4 + "\".");
-
- test3 = CharsToUnicodeString("S\\u00FC\\u00DFmayrstra\\u00DFe");
-
- test3.toUpper(Locale("de", "DE"));
- expectedResult = CharsToUnicodeString("S\\u00DCSSMAYRSTRASSE");
- if (test3 != expectedResult)
- errln("toUpper failed: expected \"" + expectedResult + "\", got \"" + test3 + "\".");
-
- test4.replace(0, test4.length(), uppercaseGreek);
-
- test4.toLower(Locale("el", "GR"));
- expectedResult = lowercaseGreek;
- if (test4 != expectedResult)
- errln("toLower failed: expected \"" + expectedResult + "\", got \"" + test4 + "\".");
-
- test4.replace(0, test4.length(), lowercaseGreek);
-
- test4.toUpper();
- expectedResult = uppercaseGreek;
- if (test4 != expectedResult)
- errln("toUpper failed: expected \"" + expectedResult + "\", got \"" + test4 + "\".");
-
- // more string case mapping tests with the new implementation
- {
- static const UChar
-
- beforeLower[]= { 0x61, 0x42, 0x49, 0x3a3, 0xdf, 0x3a3, 0x2f, 0xd93f, 0xdfff },
- lowerRoot[]= { 0x61, 0x62, 0x69, 0x3c3, 0xdf, 0x3c2, 0x2f, 0xd93f, 0xdfff },
- lowerTurkish[]={ 0x61, 0x62, 0x131, 0x3c3, 0xdf, 0x3c2, 0x2f, 0xd93f, 0xdfff },
-
- beforeUpper[]= { 0x61, 0x42, 0x69, 0x3c2, 0xdf, 0x3c3, 0x2f, 0xfb03, 0xfb03, 0xfb03, 0xd93f, 0xdfff },
- upperRoot[]= { 0x41, 0x42, 0x49, 0x3a3, 0x53, 0x53, 0x3a3, 0x2f, 0x46, 0x46, 0x49, 0x46, 0x46, 0x49, 0x46, 0x46, 0x49, 0xd93f, 0xdfff },
- upperTurkish[]={ 0x41, 0x42, 0x130, 0x3a3, 0x53, 0x53, 0x3a3, 0x2f, 0x46, 0x46, 0x49, 0x46, 0x46, 0x49, 0x46, 0x46, 0x49, 0xd93f, 0xdfff },
-
- beforeMiniUpper[]= { 0xdf, 0x61 },
- miniUpper[]= { 0x53, 0x53, 0x41 };
-
- UnicodeString s;
-
- /* lowercase with root locale */
- s=UnicodeString(FALSE, beforeLower, (int32_t)(sizeof(beforeLower)/U_SIZEOF_UCHAR));
- s.toLower("");
- if( s.length()!=(sizeof(lowerRoot)/U_SIZEOF_UCHAR) ||
- s!=UnicodeString(FALSE, lowerRoot, s.length())
- ) {
- errln("error in toLower(root locale)=\"" + s + "\" expected \"" + UnicodeString(FALSE, lowerRoot, (int32_t)(sizeof(lowerRoot)/U_SIZEOF_UCHAR)) + "\"");
- }
-
- /* lowercase with turkish locale */
- s=UnicodeString(FALSE, beforeLower, (int32_t)(sizeof(beforeLower)/U_SIZEOF_UCHAR));
- s.setCharAt(0, beforeLower[0]).toLower(Locale("tr"));
- if( s.length()!=(sizeof(lowerTurkish)/U_SIZEOF_UCHAR) ||
- s!=UnicodeString(FALSE, lowerTurkish, s.length())
- ) {
- errln("error in toLower(turkish locale)=\"" + s + "\" expected \"" + UnicodeString(FALSE, lowerTurkish, (int32_t)(sizeof(lowerTurkish)/U_SIZEOF_UCHAR)) + "\"");
- }
-
- /* uppercase with root locale */
- s=UnicodeString(FALSE, beforeUpper, (int32_t)(sizeof(beforeUpper)/U_SIZEOF_UCHAR));
- s.setCharAt(0, beforeUpper[0]).toUpper(Locale(""));
- if( s.length()!=(sizeof(upperRoot)/U_SIZEOF_UCHAR) ||
- s!=UnicodeString(FALSE, upperRoot, s.length())
- ) {
- errln("error in toUpper(root locale)=\"" + s + "\" expected \"" + UnicodeString(FALSE, upperRoot, (int32_t)(sizeof(upperRoot)/U_SIZEOF_UCHAR)) + "\"");
- }
-
- /* uppercase with turkish locale */
- s=UnicodeString(FALSE, beforeUpper, (int32_t)(sizeof(beforeUpper)/U_SIZEOF_UCHAR));
- s.toUpper(Locale("tr"));
- if( s.length()!=(sizeof(upperTurkish)/U_SIZEOF_UCHAR) ||
- s!=UnicodeString(FALSE, upperTurkish, s.length())
- ) {
- errln("error in toUpper(turkish locale)=\"" + s + "\" expected \"" + UnicodeString(FALSE, upperTurkish, (int32_t)(sizeof(upperTurkish)/U_SIZEOF_UCHAR)) + "\"");
- }
-
- /* uppercase a short string with root locale */
- s=UnicodeString(FALSE, beforeMiniUpper, (int32_t)(sizeof(beforeMiniUpper)/U_SIZEOF_UCHAR));
- s.setCharAt(0, beforeMiniUpper[0]).toUpper("");
- if( s.length()!=(sizeof(miniUpper)/U_SIZEOF_UCHAR) ||
- s!=UnicodeString(FALSE, miniUpper, s.length())
- ) {
- errln("error in toUpper(root locale)=\"" + s + "\" expected \"" + UnicodeString(FALSE, miniUpper, (int32_t)(sizeof(miniUpper)/U_SIZEOF_UCHAR)) + "\"");
- }
- }
-
- // test some supplementary characters (>= Unicode 3.1)
- {
- UnicodeString t;
-
- UnicodeString
- deseretInput=UnicodeString("\\U0001043C\\U00010414", "").unescape(),
- deseretLower=UnicodeString("\\U0001043C\\U0001043C", "").unescape(),
- deseretUpper=UnicodeString("\\U00010414\\U00010414", "").unescape();
- (t=deseretInput).toLower();
- if(t!=deseretLower) {
- errln("error lowercasing Deseret (plane 1) characters");
- }
- (t=deseretInput).toUpper();
- if(t!=deseretUpper) {
- errln("error uppercasing Deseret (plane 1) characters");
- }
- }
-
- // test some more cases that looked like problems
- {
- UnicodeString t;
-
- UnicodeString
- ljInput=UnicodeString("ab'cD \\uFB00i\\u0131I\\u0130 \\u01C7\\u01C8\\u01C9 \\U0001043C\\U00010414", "").unescape(),
- ljLower=UnicodeString("ab'cd \\uFB00i\\u0131ii\\u0307 \\u01C9\\u01C9\\u01C9 \\U0001043C\\U0001043C", "").unescape(),
- ljUpper=UnicodeString("AB'CD FFIII\\u0130 \\u01C7\\u01C7\\u01C7 \\U00010414\\U00010414", "").unescape();
- (t=ljInput).toLower("en");
- if(t!=ljLower) {
- errln("error lowercasing LJ characters");
- }
- (t=ljInput).toUpper("en");
- if(t!=ljUpper) {
- errln("error uppercasing LJ characters");
- }
- }
-
-#if !UCONFIG_NO_NORMALIZATION
- // some context-sensitive casing depends on normalization data being present
-
- // Unicode 3.1.1 SpecialCasing tests
- {
- UnicodeString t;
-
- // sigmas preceded and/or followed by cased letters
- UnicodeString
- sigmas=UnicodeString("i\\u0307\\u03a3\\u0308j \\u0307\\u03a3\\u0308j i\\u00ad\\u03a3\\u0308 \\u0307\\u03a3\\u0308 ", "").unescape(),
- sigmasLower=UnicodeString("i\\u0307\\u03c3\\u0308j \\u0307\\u03c3\\u0308j i\\u00ad\\u03c2\\u0308 \\u0307\\u03c3\\u0308 ", "").unescape(),
- sigmasUpper=UnicodeString("I\\u0307\\u03a3\\u0308J \\u0307\\u03a3\\u0308J I\\u00ad\\u03a3\\u0308 \\u0307\\u03a3\\u0308 ", "").unescape();
-
- (t=sigmas).toLower();
- if(t!=sigmasLower) {
- errln("error in sigmas.toLower()=\"" + t + "\" expected \"" + sigmasLower + "\"");
- }
-
- (t=sigmas).toUpper(Locale(""));
- if(t!=sigmasUpper) {
- errln("error in sigmas.toUpper()=\"" + t + "\" expected \"" + sigmasUpper + "\"");
- }
-
- // turkish & azerbaijani dotless i & dotted I
- // remove dot above if there was a capital I before and there are no more accents above
- UnicodeString
- dots=UnicodeString("I \\u0130 I\\u0307 I\\u0327\\u0307 I\\u0301\\u0307 I\\u0327\\u0307\\u0301", "").unescape(),
- dotsTurkish=UnicodeString("\\u0131 i i i\\u0327 \\u0131\\u0301\\u0307 i\\u0327\\u0301", "").unescape(),
- dotsDefault=UnicodeString("i i\\u0307 i\\u0307 i\\u0327\\u0307 i\\u0301\\u0307 i\\u0327\\u0307\\u0301", "").unescape();
-
- (t=dots).toLower("tr");
- if(t!=dotsTurkish) {
- errln("error in dots.toLower(tr)=\"" + t + "\" expected \"" + dotsTurkish + "\"");
- }
-
- (t=dots).toLower("de");
- if(t!=dotsDefault) {
- errln("error in dots.toLower(de)=\"" + t + "\" expected \"" + dotsDefault + "\"");
- }
- }
-
- // more Unicode 3.1.1 tests
- {
- UnicodeString t;
-
- // lithuanian dot above in uppercasing
- UnicodeString
- dots=UnicodeString("a\\u0307 \\u0307 i\\u0307 j\\u0327\\u0307 j\\u0301\\u0307", "").unescape(),
- dotsLithuanian=UnicodeString("A\\u0307 \\u0307 I J\\u0327 J\\u0301\\u0307", "").unescape(),
- dotsDefault=UnicodeString("A\\u0307 \\u0307 I\\u0307 J\\u0327\\u0307 J\\u0301\\u0307", "").unescape();
-
- (t=dots).toUpper("lt");
- if(t!=dotsLithuanian) {
- errln("error in dots.toUpper(lt)=\"" + t + "\" expected \"" + dotsLithuanian + "\"");
- }
-
- (t=dots).toUpper("de");
- if(t!=dotsDefault) {
- errln("error in dots.toUpper(de)=\"" + t + "\" expected \"" + dotsDefault + "\"");
- }
-
- // lithuanian adds dot above to i in lowercasing if there are more above accents
- UnicodeString
- i=UnicodeString("I I\\u0301 J J\\u0301 \\u012e \\u012e\\u0301 \\u00cc\\u00cd\\u0128", "").unescape(),
- iLithuanian=UnicodeString("i i\\u0307\\u0301 j j\\u0307\\u0301 \\u012f \\u012f\\u0307\\u0301 i\\u0307\\u0300i\\u0307\\u0301i\\u0307\\u0303", "").unescape(),
- iDefault=UnicodeString("i i\\u0301 j j\\u0301 \\u012f \\u012f\\u0301 \\u00ec\\u00ed\\u0129", "").unescape();
-
- (t=i).toLower("lt");
- if(t!=iLithuanian) {
- errln("error in i.toLower(lt)=\"" + t + "\" expected \"" + iLithuanian + "\"");
- }
-
- (t=i).toLower("de");
- if(t!=iDefault) {
- errln("error in i.toLower(de)=\"" + t + "\" expected \"" + iDefault + "\"");
- }
- }
-
-#endif
-
- // test case folding
- {
- UnicodeString
- s=UnicodeString("A\\u00df\\u00b5\\ufb03\\U0001040c\\u0130\\u0131", "").unescape(),
- f=UnicodeString("ass\\u03bcffi\\U00010434i\\u0307\\u0131", "").unescape(),
- g=UnicodeString("ass\\u03bcffi\\U00010434i\\u0131", "").unescape(),
- t;
-
- (t=s).foldCase();
- if(f!=t) {
- errln("error in foldCase(\"" + s + "\", default)=\"" + t + "\" but expected \"" + f + "\"");
- }
-
- // alternate handling for dotted I/dotless i (U+0130, U+0131)
- (t=s).foldCase(U_FOLD_CASE_EXCLUDE_SPECIAL_I);
- if(g!=t) {
- errln("error in foldCase(\"" + s + "\", U_FOLD_CASE_EXCLUDE_SPECIAL_I)=\"" + t + "\" but expected \"" + g + "\"");
- }
- }
-}
-
-// data-driven case mapping tests ------------------------------------------ ***
-
-enum {
- TEST_LOWER,
- TEST_UPPER,
- TEST_TITLE,
- TEST_FOLD,
- TEST_COUNT
-};
-
-// names of TestData children in casing.txt
-static const char *const dataNames[TEST_COUNT+1]={
- "lowercasing",
- "uppercasing",
- "titlecasing",
- "casefolding",
- ""
-};
-
-void
-StringCaseTest::TestCasingImpl(const UnicodeString &input,
- const UnicodeString &output,
- int32_t whichCase,
- void *iter, const char *localeID, uint32_t options) {
- // UnicodeString
- UnicodeString result;
- const char *name;
- Locale locale(localeID);
-
- result=input;
- switch(whichCase) {
- case TEST_LOWER:
- name="toLower";
- result.toLower(locale);
- break;
- case TEST_UPPER:
- name="toUpper";
- result.toUpper(locale);
- break;
-#if !UCONFIG_NO_BREAK_ITERATION
- case TEST_TITLE:
- name="toTitle";
- result.toTitle((BreakIterator *)iter, locale, options);
- break;
-#endif
- case TEST_FOLD:
- name="foldCase";
- result.foldCase(options);
- break;
- default:
- name="";
- break; // won't happen
- }
- if(result!=output) {
- errln("error: UnicodeString.%s() got a wrong result for a test case from casing.res", name);
- }
-#if !UCONFIG_NO_BREAK_ITERATION
- if(whichCase==TEST_TITLE && options==0) {
- result=input;
- result.toTitle((BreakIterator *)iter, locale);
- if(result!=output) {
- errln("error: UnicodeString.toTitle(options=0) got a wrong result for a test case from casing.res");
- }
- }
-#endif
-
- // UTF-8
- char utf8In[100], utf8Out[100];
- int32_t utf8InLength, utf8OutLength, resultLength;
- UChar *buffer;
-
- UCaseMap *csm;
- UErrorCode errorCode;
-
- errorCode=U_ZERO_ERROR;
- csm=ucasemap_open(localeID, options, &errorCode);
-#if !UCONFIG_NO_BREAK_ITERATION
- if(iter!=NULL) {
- // Clone the break iterator so that the UCaseMap can safely adopt it.
- int32_t size=1; // Not 0 because that only gives preflighting.
- UBreakIterator *clone=ubrk_safeClone((UBreakIterator *)iter, NULL, &size, &errorCode);
- ucasemap_setBreakIterator(csm, clone, &errorCode);
- }
-#endif
-
- u_strToUTF8(utf8In, (int32_t)sizeof(utf8In), &utf8InLength, input.getBuffer(), input.length(), &errorCode);
- switch(whichCase) {
- case TEST_LOWER:
- name="ucasemap_utf8ToLower";
- utf8OutLength=ucasemap_utf8ToLower(csm,
- utf8Out, (int32_t)sizeof(utf8Out),
- utf8In, utf8InLength, &errorCode);
- break;
- case TEST_UPPER:
- name="ucasemap_utf8ToUpper";
- utf8OutLength=ucasemap_utf8ToUpper(csm,
- utf8Out, (int32_t)sizeof(utf8Out),
- utf8In, utf8InLength, &errorCode);
- break;
-#if !UCONFIG_NO_BREAK_ITERATION
- case TEST_TITLE:
- name="ucasemap_utf8ToTitle";
- utf8OutLength=ucasemap_utf8ToTitle(csm,
- utf8Out, (int32_t)sizeof(utf8Out),
- utf8In, utf8InLength, &errorCode);
- break;
-#endif
- case TEST_FOLD:
- name="ucasemap_utf8FoldCase";
- utf8OutLength=ucasemap_utf8FoldCase(csm,
- utf8Out, (int32_t)sizeof(utf8Out),
- utf8In, utf8InLength, &errorCode);
- break;
- default:
- name="";
- utf8OutLength=0;
- break; // won't happen
- }
- buffer=result.getBuffer(utf8OutLength);
- u_strFromUTF8(buffer, result.getCapacity(), &resultLength, utf8Out, utf8OutLength, &errorCode);
- result.releaseBuffer(U_SUCCESS(errorCode) ? resultLength : 0);
-
- if(U_FAILURE(errorCode)) {
- errln("error: %s() got an error for a test case from casing.res - %s", name, u_errorName(errorCode));
- } else if(result!=output) {
- errln("error: %s() got a wrong result for a test case from casing.res", name);
- }
- ucasemap_close(csm);
-}
-
-void
-StringCaseTest::TestCasing() {
- UErrorCode status = U_ZERO_ERROR;
- void *iter;
- char cLocaleID[100];
- UnicodeString locale, input, output, optionsString, result;
- uint32_t options;
- int32_t whichCase, type;
- TestDataModule *driver = TestDataModule::getTestDataModule("casing", *this, status);
- if(U_SUCCESS(status)) {
- for(whichCase=0; whichCase<TEST_COUNT; ++whichCase) {
-#if UCONFIG_NO_BREAK_ITERATION
- if(whichCase==TEST_TITLE) {
- continue;
- }
-#endif
- TestData *casingTest = driver->createTestData(dataNames[whichCase], status);
- if(U_FAILURE(status)) {
- errln("TestCasing failed to createTestData(%s) - %s", dataNames[whichCase], u_errorName(status));
- break;
- }
- const DataMap *myCase = NULL;
- while(casingTest->nextCase(myCase, status)) {
- input = myCase->getString("Input", status);
- output = myCase->getString("Output", status);
-
- if(whichCase!=TEST_FOLD) {
- locale = myCase->getString("Locale", status);
- }
- locale.extract(0, 0x7fffffff, cLocaleID, sizeof(cLocaleID), "");
-
- iter=NULL;
-#if !UCONFIG_NO_BREAK_ITERATION
- if(whichCase==TEST_TITLE) {
- type = myCase->getInt("Type", status);
- if(type>=0) {
- iter=ubrk_open((UBreakIteratorType)type, cLocaleID, NULL, 0, &status);
- } else if(type==-2) {
- // Open a trivial break iterator that only delivers { 0, length }
- // or even just { 0 } as boundaries.
- static const UChar rules[] = { 0x2e, 0x2a, 0x3b }; // ".*;"
- UParseError parseError;
- iter=ubrk_openRules(rules, LENGTHOF(rules), NULL, 0, &parseError, &status);
- }
- }
-#endif
- options = 0;
- if(whichCase==TEST_TITLE || whichCase==TEST_FOLD) {
- optionsString = myCase->getString("Options", status);
- if(optionsString.indexOf((UChar)0x54)>=0) { // T
- options|=U_FOLD_CASE_EXCLUDE_SPECIAL_I;
- }
- if(optionsString.indexOf((UChar)0x4c)>=0) { // L
- options|=U_TITLECASE_NO_LOWERCASE;
- }
- if(optionsString.indexOf((UChar)0x41)>=0) { // A
- options|=U_TITLECASE_NO_BREAK_ADJUSTMENT;
- }
- }
-
- if(U_FAILURE(status)) {
- errln("error: TestCasing() setup failed for %s test case from casing.res: %s", dataNames[whichCase], u_errorName(status));
- status = U_ZERO_ERROR;
- } else {
- TestCasingImpl(input, output, whichCase, iter, cLocaleID, options);
- }
-
-#if !UCONFIG_NO_BREAK_ITERATION
- if(iter!=NULL) {
- ubrk_close(iter);
- }
-#endif
- }
- delete casingTest;
- }
- }
- delete driver;
-
-#if !UCONFIG_NO_BREAK_ITERATION
- // more tests for API coverage
- status=U_ZERO_ERROR;
- input=UNICODE_STRING_SIMPLE("sTrA\\u00dfE").unescape();
- (result=input).toTitle(NULL);
- if(result!=UNICODE_STRING_SIMPLE("Stra\\u00dfe").unescape()) {
- errln("UnicodeString::toTitle(NULL) failed");
- }
-#endif
-}