summaryrefslogtreecommitdiffstats
path: root/tests/auto
diff options
context:
space:
mode:
authorThiago Macieira <thiago.macieira@nokia.com>2008-11-07 21:50:31 (GMT)
committerThiago Macieira <thiago.macieira@nokia.com>2009-05-11 11:27:55 (GMT)
commit61ba9b2a920fd561b292886b3b75bfde4847e26f (patch)
tree23e8b41a90bba73a99641046d72be96892bb3159 /tests/auto
parent60ff41273a58835466022f91d4f6eb195953c739 (diff)
downloadQt-61ba9b2a920fd561b292886b3b75bfde4847e26f.zip
Qt-61ba9b2a920fd561b292886b3b75bfde4847e26f.tar.gz
Qt-61ba9b2a920fd561b292886b3b75bfde4847e26f.tar.bz2
Add a test for UTF-8 conditions
Diffstat (limited to 'tests/auto')
-rw-r--r--tests/auto/utf8/tst_utf8.cpp309
-rw-r--r--tests/auto/utf8/utf8.pro3
2 files changed, 312 insertions, 0 deletions
diff --git a/tests/auto/utf8/tst_utf8.cpp b/tests/auto/utf8/tst_utf8.cpp
new file mode 100644
index 0000000..3aef69f
--- /dev/null
+++ b/tests/auto/utf8/tst_utf8.cpp
@@ -0,0 +1,309 @@
+/****************************************************************************
+**
+** Copyright (C) 2008 Nokia Corporation and/or its subsidiary(-ies).
+** Contact: Qt Software Information (qt-info@nokia.com)
+**
+** This file is part of the test suite of the Qt Toolkit.
+**
+** $QT_BEGIN_LICENSE:LGPL$
+** No Commercial Usage
+** This file contains pre-release code and may not be distributed.
+** You may use this file in accordance with the terms and conditions
+** contained in the either Technology Preview License Agreement or the
+** Beta Release License Agreement.
+**
+** GNU Lesser General Public License Usage
+** Alternatively, this file may be used under the terms of the GNU Lesser
+** General Public License version 2.1 as published by the Free Software
+** Foundation and appearing in the file LICENSE.LGPL included in the
+** packaging of this file. Please review the following information to
+** ensure the GNU Lesser General Public License version 2.1 requirements
+** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.
+**
+** In addition, as a special exception, Nokia gives you certain
+** additional rights. These rights are described in the Nokia Qt LGPL
+** Exception version 1.0, included in the file LGPL_EXCEPTION.txt in this
+** package.
+**
+** GNU General Public License Usage
+** Alternatively, this file may be used under the terms of the GNU
+** General Public License version 3.0 as published by the Free Software
+** Foundation and appearing in the file LICENSE.GPL included in the
+** packaging of this file. Please review the following information to
+** ensure the GNU General Public License version 3.0 requirements will be
+** met: http://www.gnu.org/copyleft/gpl.html.
+**
+** If you are unsure which license is appropriate for your use, please
+** contact the sales department at qt-sales@nokia.com.
+** $QT_END_LICENSE$
+**
+****************************************************************************/
+#include <QtTest/QtTest>
+
+#include <qtextcodec.h>
+#include <qsharedpointer.h>
+
+//TESTED_CLASS=
+//TESTED_FILES=
+
+static const char utf8bom[] = "\xEF\xBB\xBF";
+
+class tst_Utf8 : public QObject
+{
+ Q_OBJECT
+
+public:
+ // test data:
+ QTextCodec *codec;
+ QString (*from8BitPtr)(const char *, int);
+ QByteArray (QString:: *to8Bit)() const;
+
+ inline QString from8Bit(const QByteArray &ba)
+ { return from8BitPtr(ba.constData(), ba.length()); }
+public slots:
+ void initTestCase();
+ void init();
+
+private slots:
+ void roundTrip_data();
+ void roundTrip();
+
+ void charByChar_data();
+ void charByChar();
+
+ void invalidUtf8_data();
+ void invalidUtf8();
+};
+
+void tst_Utf8::initTestCase()
+{
+ QTest::addColumn<bool>("useLocale");
+ QTest::newRow("utf8codec") << false;
+
+ // is the locale UTF-8?
+ if (QString(QChar(QChar::ReplacementCharacter)).toLocal8Bit() == "\xEF\xBF\xBD") {
+ QTest::newRow("localecodec") << true;
+ qDebug() << "locale is utf8";
+ }
+}
+
+void tst_Utf8::init()
+{
+ QFETCH_GLOBAL(bool, useLocale);
+ if (useLocale) {
+ codec = QTextCodec::codecForLocale();
+ from8BitPtr = &QString::fromLocal8Bit;
+ to8Bit = &QString::toLocal8Bit;
+ } else {
+ codec = QTextCodec::codecForMib(106);
+ from8BitPtr = &QString::fromUtf8;
+ to8Bit = &QString::toUtf8;
+ }
+}
+
+void tst_Utf8::roundTrip_data()
+{
+ QTest::addColumn<QByteArray>("utf8");
+ QTest::addColumn<QString>("utf16");
+
+ QTest::newRow("empty") << QByteArray() << QString();
+ QTest::newRow("nul") << QByteArray("", 1) << QString(QChar(QChar::Null));
+
+ static const char ascii[] = "This is a standard US-ASCII message";
+ QTest::newRow("ascii") << QByteArray(ascii) << ascii;
+
+ static const char ascii2[] = "\1This\2is\3an\4US-ASCII\020 message interspersed with control chars";
+ QTest::newRow("ascii2") << QByteArray(ascii2) << ascii2;
+
+ static const char utf8_1[] = "\302\240"; // NBSP
+ QTest::newRow("utf8_1") << QByteArray(utf8_1) << QString(QChar(QChar::Nbsp));
+
+ static const char utf8_2[] = "\342\202\254"; // Euro symbol
+ QTest::newRow("utf8_2") << QByteArray(utf8_2) << QString(QChar(0x20AC));
+
+#if 0
+ // Can't test this because QString::fromUtf8 consumes it
+ static const char utf8_3[] = "\357\273\277"; // byte order mark
+ QTest::newRow("utf8_3") << QByteArray(utf8_3) << QString(QChar(QChar::ByteOrderMark));
+#endif
+
+ static const char utf8_4[] = "\357\277\275"; // replacement char
+ QTest::newRow("utf8_4") << QByteArray(utf8_4) << QString(QChar(QChar::ReplacementCharacter));
+
+ static const char utf8_5[] = "\360\220\210\203"; // U+010203
+ static const uint utf32_5[] = { 0x010203 };
+ QTest::newRow("utf8_5") << QByteArray(utf8_5) << QString::fromUcs4(utf32_5, 1);
+
+ static const char utf8_6[] = "\364\217\277\277"; // U+10FFFF
+ static const uint utf32_6[] = { 0x10FFFF };
+ QTest::newRow("utf8_6") << QByteArray(utf8_6) << QString::fromUcs4(utf32_6, 1);
+
+ static const char utf8_7[] = "abc\302\240\303\241\303\251\307\275 \342\202\254def";
+ static const ushort utf16_7[] = { 'a', 'b', 'c', 0x00A0,
+ 0x00E1, 0x00E9, 0x01FD,
+ ' ', 0x20AC, 'd', 'e', 'f', 0 };
+ QTest::newRow("utf8_7") << QByteArray(utf8_7) << QString::fromUtf16(utf16_7);
+
+ static const char utf8_8[] = "abc\302\240\303\241\303\251\307\275 \364\217\277\277 \342\202\254def";
+ static const uint utf32_8[] = { 'a', 'b', 'c', 0x00A0,
+ 0x00E1, 0x00E9, 0x01FD,
+ ' ', 0x10FFFF, ' ',
+ 0x20AC, 'd', 'e', 'f', 0 };
+ QTest::newRow("utf8_8") << QByteArray(utf8_8) << QString::fromUcs4(utf32_8);
+}
+
+void tst_Utf8::roundTrip()
+{
+ QFETCH(QByteArray, utf8);
+ QFETCH(QString, utf16);
+
+ QCOMPARE((utf16.*to8Bit)(), utf8);
+ QCOMPARE(from8Bit(utf8), utf16);
+
+ QCOMPARE((from8Bit(utf8).*to8Bit)(), utf8);
+ QCOMPARE(from8Bit((utf16.*to8Bit)()), utf16);
+}
+
+void tst_Utf8::charByChar_data()
+{
+ roundTrip_data();
+}
+
+void tst_Utf8::charByChar()
+{
+ QFETCH(QByteArray, utf8);
+ QFETCH(QString, utf16);
+
+ {
+ // from utf16 to utf8 char by char:
+ QSharedPointer<QTextEncoder> encoder = QSharedPointer<QTextEncoder>(codec->makeEncoder());
+ QByteArray encoded;
+
+ for (int i = 0; i < utf16.length(); ++i) {
+ encoded += encoder->fromUnicode(utf16.constData() + i, 1);
+ QVERIFY(!encoder->hasFailure());
+ }
+
+ if (encoded.startsWith(utf8bom))
+ encoded = encoded.mid(strlen(utf8bom));
+ QCOMPARE(encoded, utf8);
+ }
+ {
+ // from utf8 to utf16 char by char:
+ QSharedPointer<QTextDecoder> decoder = QSharedPointer<QTextDecoder>(codec->makeDecoder());
+ QString decoded;
+
+ for (int i = 0; i < utf8.length(); ++i) {
+ decoded += decoder->toUnicode(utf8.constData() + i, 1);
+ QVERIFY(!decoder->hasFailure());
+ }
+
+ QCOMPARE(decoded, utf16);
+ }
+}
+
+void tst_Utf8::invalidUtf8_data()
+{
+ QTest::addColumn<QByteArray>("utf8");
+
+ QTest::newRow("1char") << QByteArray("\x80");
+ QTest::newRow("2chars") << QByteArray("\xC2\xC0");
+ QTest::newRow("3chars-1") << QByteArray("\xE0\xA0\xC0");
+ QTest::newRow("3chars-2") << QByteArray("\xE0\xC0\xA0");
+ QTest::newRow("4chars-1") << QByteArray("\xF0\x90\x80\xC0");
+ QTest::newRow("4chars-2") << QByteArray("\xF0\x90\xC0\x80");
+ QTest::newRow("4chars-3") << QByteArray("\xF0\xC0\x80\x80");
+
+ // U+FFFE and U+FFFF are non-characters and must not be present
+ // U+FFFE: 1111 11 1111 11 1110
+ // encoding: xxxz:1111 xz11:1111 xz11:1110
+ QTest::newRow("fffe") << QByteArray("\xEF\xBF\xBE");
+ // U+FFFF: 1111 11 1111 11 1111
+ // encoding: xxxz:1111 xz11:1111 xz11:1111
+ QTest::newRow("ffff") << QByteArray("\xEF\xBF\xBF");
+
+ // Surrogate pairs must now be present either
+ // U+D800: 1101 10 0000 00 0000
+ // encoding: xxxz:1101 xz10:0000 xz00:0000
+ QTest::newRow("hi-surrogate") << QByteArray("\xED\xA0\x80");
+ // U+DC00: 1101 11 0000 00 0000
+ // encoding: xxxz:1101 xz11:0000 xz00:0000
+ QTest::newRow("lo-surrogate") << QByteArray("\xED\xB0\x80");
+
+ // not even in pair:
+ QTest::newRow("surrogate-pair") << QByteArray("\xED\xA0\x80\xED\xB0\x80");
+
+ // Characters outside the Unicode range:
+ // 0x110000: 00 0100 01 0000 00 0000 00 0000
+ // encoding: xxxx:z100 xz01:0000 xz00:0000 xz00:0000
+ QTest::newRow("non-unicode-1") << QByteArray("\xF4\x90\x80\x80");
+ // 0x200000: 00 1000 00 0000 00 0000 00 0000
+ // encoding: xxxx:xz00 xz00:1000 xz00:0000 xz00:0000 xz00:0000
+ QTest::newRow("non-unicode-2") << QByteArray("\xF8\x88\x80\x80\x80");
+ // 0x04000000: 0100 00 0000 00 0000 00 0000 00 0000
+ // encoding: xxxx:xxz0 xz00:0100 xz00:0000 xz00:0000 xz00:0001 xz00:0001
+ QTest::newRow("non-unicode-3") << QByteArray("\xFC\x84\x80\x80\x80\x80");
+ // 0x7fffffff: 1 11 1111 11 1111 11 1111 11 1111 11 1111
+ // encoding: xxxx:xxz0 xz00:0100 xz00:0000 xz00:0000 xz00:0001 xz00:0001
+ QTest::newRow("non-unicode-3") << QByteArray("\xFD\xBF\xBF\xBF\xBF\xBF");
+
+ // As seen above, 0xFE and 0xFF never appear:
+ QTest::newRow("fe") << QByteArray("\xFE");
+ QTest::newRow("fe-bis") << QByteArray("\xFE\xBF\xBF\xBF\xBF\xBF\xBF");
+ QTest::newRow("ff") << QByteArray("\xFF");
+ QTest::newRow("ff-bis") << QByteArray("\xFF\xBF\xBF\xBF\xBF\xBF\xBF\xBF");
+
+ // some combinations in UTF-8 are invalid even though they have the proper bits set
+ // these are known as overlong sequences
+
+ // "A": U+0041: 01 00 0001
+ // overlong 2: xxz0:0001 xz00:0001
+ QTest::newRow("overlong-1-2") << QByteArray("\xC1\x81");
+ // overlong 3: xxxz:0000 xz00:0001 xz00:0001
+ QTest::newRow("overlong-1-3") << QByteArray("\xE0\x81\x81");
+ // overlong 4: xxxx:z000 xz00:0000 xz00:0001 xz00:0001
+ QTest::newRow("overlong-1-4") << QByteArray("\xF0\x80\x81\x81");
+ // overlong 5: xxxx:xz00 xz00:0000 xz00:0000 xz00:0001 xz00:0001
+ QTest::newRow("overlong-1-5") << QByteArray("\xF8\x80\x80\x81\x81");
+ // overlong 6: xxxx:xxz0 xz00:0000 xz00:0000 xz00:0000 xz00:0001 xz00:0001
+ QTest::newRow("overlong-1-6") << QByteArray("\xFC\x80\x80\x80\x81\x81");
+
+ // NBSP: U+00A0: 10 00 0000
+ // proper encoding: xxz0:0010 xz00:0000
+ // overlong 3: xxxz:0000 xz00:0010 xz00:0000
+ QTest::newRow("overlong-2-3") << QByteArray("\xC0\x82\x80");
+ // overlong 4: xxxx:z000 xz00:0000 xz00:0010 xz00:0000
+ QTest::newRow("overlong-2-4") << QByteArray("\xF0\x80\x82\x80");
+ // overlong 5: xxxx:xz00 xz00:0000 xz00:0000 xz00:0010 xz00:0000
+ QTest::newRow("overlong-2-4") << QByteArray("\xF8\x80\x80\x82\x80");
+ // overlong 6: xxxx:xxz0 xz00:0000 xz00:0000 xz00:0000 xz00:0010 xz00:0000
+ QTest::newRow("overlong-2-4") << QByteArray("\xFC\x80\x80\x80\x82\x80");
+
+ // U+0800: 10 0000 00 0000
+ // proper encoding: xxxz:0000 xz10:0000 xz00:0000
+ // overlong 4: xxxx:z000 xz00:0000 xz10:0000 xz00:0000
+ QTest::newRow("overlong-3-4") << QByteArray("\xF0\x80\xA0\x80");
+ // overlong 5: xxxx:xz00 xz00:0000 xz00:0000 xz10:0000 xz00:0000
+ QTest::newRow("overlong-3-5") << QByteArray("\xF8\x80\x80\xA0\x80");
+ // overlong 6: xxxx:xxz0 xz00:0000 xz00:0000 xz00:0000 xz10:0000 xz00:0000
+ QTest::newRow("overlong-3-6") << QByteArray("\xFC\x80\x80\x80\xA0\x80");
+
+ // U+010000: 00 0100 00 0000 00 0000
+ // proper encoding: xxxx:z000 xz00:0100 xz00:0000 xz00:0000
+ // overlong 5: xxxx:xz00 xz00:0000 xz00:0100 xz00:0000 xz00:0000
+ QTest::newRow("overlong-4-5") << QByteArray("\xF8\x80\x84\x80\x80");
+ // overlong 6: xxxx:xxz0 xz00:0000 xz00:0000 xz00:0100 xz00:0000 xz00:0000
+ QTest::newRow("overlong-4-6") << QByteArray("\xFC\x80\x80\x84\x80\x80");
+}
+
+void tst_Utf8::invalidUtf8()
+{
+ QFETCH(QByteArray, utf8);
+
+ QSharedPointer<QTextDecoder> decoder = QSharedPointer<QTextDecoder>(codec->makeDecoder());
+ QString decoded = decoder->toUnicode(utf8);
+ QVERIFY(decoder->hasFailure());
+}
+
+QTEST_MAIN(tst_Utf8)
+#include "tst_utf8.moc"
diff --git a/tests/auto/utf8/utf8.pro b/tests/auto/utf8/utf8.pro
new file mode 100644
index 0000000..4ec6851
--- /dev/null
+++ b/tests/auto/utf8/utf8.pro
@@ -0,0 +1,3 @@
+load(qttest_p4)
+QT -= gui
+SOURCES += tst_utf8.cpp