From e9041c7fc1052167f1ec2df0ea9623059e55d00f Mon Sep 17 00:00:00 2001 From: Thiago Macieira Date: Thu, 28 Apr 2016 22:09:01 -0700 Subject: [PATCH] Fix parsing of tzfile(5) POSIX rule zone names with bracket quotes POSIX.1-2001 allows quoting a zone name so that it can contain other characters besides letters, by enclosing it in angle brackets ('<' and '>'). This hadn't been used until recently (tzdata2016b), when the Asia/Barnaul rule started using a zone name "+07" (the name variable contained the value "<+07>-7"). Thanks to Paul Eggert for reporting and investigating the root cause. Task-number: QTBUG-53071 Change-Id: Id5480807d25e49e78b79ffff1449bc410776cb66 Reviewed-by: Edward Welbourne Reviewed-by: Lars Knoll --- src/corelib/tools/qtimezoneprivate_tz.cpp | 176 ++++++++++++++------- .../auto/corelib/tools/qtimezone/tst_qtimezone.cpp | 10 ++ 2 files changed, 130 insertions(+), 56 deletions(-) diff --git a/src/corelib/tools/qtimezoneprivate_tz.cpp b/src/corelib/tools/qtimezoneprivate_tz.cpp index 85ed345..cb9581a 100644 --- a/src/corelib/tools/qtimezoneprivate_tz.cpp +++ b/src/corelib/tools/qtimezoneprivate_tz.cpp @@ -41,6 +41,8 @@ #include +#include "qlocale_tools_p.h" + #include QT_BEGIN_NAMESPACE @@ -384,25 +386,100 @@ static QTime parsePosixTime(const QByteArray &timeRule) return QTime(2, 0, 0); } -static int parsePosixOffset(const QByteArray &timeRule) +static int parsePosixOffset(const char *begin, const char *end) { // Format "[+|-]hh[:mm[:ss]]" - QList parts = timeRule.split(':'); - int count = parts.count(); - if (count == 3) { - int hour = parts.at(0).toInt(); - int sign = hour >= 0 ? -1 : 1; - return sign * ((qAbs(hour) * 60 * 60) + (parts.at(1).toInt() * 60) + parts.at(2).toInt()); - } else if (count == 2) { - int hour = parts.at(0).toInt(); - int sign = hour >= 0 ? -1 : 1; - return sign * ((qAbs(hour) * 60 * 60) + (parts.at(1).toInt() * 60)); - } else if (count == 1) { - int hour = parts.at(0).toInt(); - int sign = hour >= 0 ? -1 : 1; - return sign * (qAbs(hour) * 60 * 60); - } - return 0; + int hour, min = 0, sec = 0; + + // note that the sign is inverted because POSIX counts in hours West of GMT + bool negate = true; + if (*begin == '+') { + ++begin; + } else if (*begin == '-') { + negate = false; + ++begin; + } + + bool ok = false; + hour = qstrtoll(begin, &begin, 10, &ok); + if (!ok) + return INT_MIN; + if (begin < end && *begin == ':') { + // minutes + ++begin; + min = qstrtoll(begin, &begin, 10, &ok); + if (!ok || min < 0) + return INT_MIN; + + if (begin < end && *begin == ':') { + // seconds + ++begin; + sec = qstrtoll(begin, &begin, 10, &ok); + if (!ok || sec < 0) + return INT_MIN; + } + } + + // we must have consumed everything + if (begin != end) + return INT_MIN; + + int value = (hour * 60 + min) * 60 + sec; + return negate ? -value : value; +} + +static inline bool asciiIsLetter(char ch) +{ + ch |= 0x20; // lowercases if it is a letter, otherwise just corrupts ch + return ch >= 'a' && ch <= 'z'; +} + +// Returns the zone name, the offset (in seconds) and advances \a begin to +// where the parsing ended. Returns a zone of INT_MIN in case an offset +// couldn't be read. +static QPair parsePosixZoneNameAndOffset(const char *&pos, const char *end) +{ + static const char offsetChars[] = "0123456789:"; + QPair result = qMakePair(QString(), INT_MIN); + + const char *nameBegin = pos; + const char *nameEnd; + Q_ASSERT(pos < end); + + if (*pos == '<') { + nameBegin = pos + 1; // skip the '<' + nameEnd = nameBegin; + while (nameEnd < end && *nameEnd != '>') { + // POSIX says only alphanumeric, but we allow anything + ++nameEnd; + } + pos = nameEnd + 1; // skip the '>' + } else { + nameBegin = pos; + nameEnd = nameBegin; + while (nameEnd < end && asciiIsLetter(*nameEnd)) + ++nameEnd; + pos = nameEnd; + } + if (nameEnd - nameBegin < 3) + return result; // name must be at least 3 characters long + + // zone offset, form [+-]hh:mm:ss + const char *zoneBegin = pos; + const char *zoneEnd = pos; + if (zoneEnd < end && (zoneEnd[0] == '+' || zoneEnd[0] == '-')) + ++zoneEnd; + while (zoneEnd < end) { + if (strchr(offsetChars, char(*zoneEnd)) == NULL) + break; + ++zoneEnd; + } + + result.first = QString::fromUtf8(nameBegin, nameEnd - nameBegin); + if (zoneEnd > zoneBegin) + result.second = parsePosixOffset(zoneBegin, zoneEnd); + pos = zoneEnd; + return result; } static QVector calculatePosixTransitions(const QByteArray &posixRule, @@ -419,51 +496,38 @@ static QVector calculatePosixTransitions(const QByteArra // POSIX Format is like "TZ=CST6CDT,M3.2.0/2:00:00,M11.1.0/2:00:00" // i.e. "std offset dst [offset],start[/time],end[/time]" - // See http://www.gnu.org/software/libc/manual/html_node/TZ-Variable.html + // See the section about TZ at http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap08.html QList parts = posixRule.split(','); - QString name = QString::fromUtf8(parts.at(0)); - QString stdName; - QString stdOffsetString; - QString dstName; - QString dstOffsetString; - bool parsedStdName = false; - bool parsedStdOffset = false; - for (int i = 0; i < name.size(); ++i) { - if (name.at(i).isLetter()) { - if (parsedStdName) { - parsedStdOffset = true; - dstName.append(name.at(i)); - } else { - stdName.append(name.at(i)); + QPair stdZone, dstZone; + { + const QByteArray &zoneinfo = parts.at(0); + const char *begin = zoneinfo.constBegin(); + + stdZone = parsePosixZoneNameAndOffset(begin, zoneinfo.constEnd()); + if (stdZone.second == INT_MIN) { + stdZone.second = 0; // reset to UTC if we failed to parse + } else if (begin < zoneinfo.constEnd()) { + dstZone = parsePosixZoneNameAndOffset(begin, zoneinfo.constEnd()); + if (dstZone.second == INT_MIN) { + // if the dst offset isn't provided, it is 1 hour ahead of the standard offset + dstZone.second = stdZone.second + (60 * 60); } - } else { - parsedStdName = true; - if (parsedStdOffset) - dstOffsetString.append(name.at(i)); - else - stdOffsetString.append(name.at(i)); } } - int utcOffset = parsePosixOffset(stdOffsetString.toUtf8()); - // If only the name part then no transitions if (parts.count() == 1) { QTimeZonePrivate::Data data; data.atMSecsSinceEpoch = lastTranMSecs; - data.offsetFromUtc = utcOffset; - data.standardTimeOffset = utcOffset; + data.offsetFromUtc = stdZone.second; + data.standardTimeOffset = stdZone.second; data.daylightTimeOffset = 0; - data.abbreviation = stdName; + data.abbreviation = stdZone.first; result << data; return result; } - // If not populated the total dst offset is 1 hour - int dstOffset = utcOffset + (60 * 60); - if (!dstOffsetString.isEmpty()) - dstOffset = parsePosixOffset(dstOffsetString.toUtf8()); // Get the std to dst transtion details QList dstParts = parts.at(1).split('/'); @@ -486,18 +550,18 @@ static QVector calculatePosixTransitions(const QByteArra for (int year = startYear; year <= endYear; ++year) { QTimeZonePrivate::Data dstData; QDateTime dst(calculatePosixDate(dstDateRule, year), dstTime, Qt::UTC); - dstData.atMSecsSinceEpoch = dst.toMSecsSinceEpoch() - (utcOffset * 1000); - dstData.offsetFromUtc = dstOffset; - dstData.standardTimeOffset = utcOffset; - dstData.daylightTimeOffset = dstOffset - utcOffset; - dstData.abbreviation = dstName; + dstData.atMSecsSinceEpoch = dst.toMSecsSinceEpoch() - (stdZone.second * 1000); + dstData.offsetFromUtc = dstZone.second; + dstData.standardTimeOffset = stdZone.second; + dstData.daylightTimeOffset = dstZone.second - stdZone.second; + dstData.abbreviation = dstZone.first; QTimeZonePrivate::Data stdData; QDateTime std(calculatePosixDate(stdDateRule, year), stdTime, Qt::UTC); - stdData.atMSecsSinceEpoch = std.toMSecsSinceEpoch() - (dstOffset * 1000); - stdData.offsetFromUtc = utcOffset; - stdData.standardTimeOffset = utcOffset; + stdData.atMSecsSinceEpoch = std.toMSecsSinceEpoch() - (dstZone.second * 1000); + stdData.offsetFromUtc = stdZone.second; + stdData.standardTimeOffset = stdZone.second; stdData.daylightTimeOffset = 0; - stdData.abbreviation = stdName; + stdData.abbreviation = stdZone.first; // Part of the high year will overflow if (year == 292278994 && (dstData.atMSecsSinceEpoch < 0 || stdData.atMSecsSinceEpoch < 0)) { if (dstData.atMSecsSinceEpoch > 0) { diff --git a/tests/auto/corelib/tools/qtimezone/tst_qtimezone.cpp b/tests/auto/corelib/tools/qtimezone/tst_qtimezone.cpp index ea83510..ce72e7c 100644 --- a/tests/auto/corelib/tools/qtimezone/tst_qtimezone.cpp +++ b/tests/auto/corelib/tools/qtimezone/tst_qtimezone.cpp @@ -847,6 +847,16 @@ void tst_QTimeZone::tzTest() QTzTimeZonePrivate::Data datatz2 = tztz2.data(std); QTzTimeZonePrivate::Data datautc2 = tzutc2.data(std); QCOMPARE(datatz2.offsetFromUtc, datautc2.offsetFromUtc); + + // Test a timezone with a name that isn't all letters + QTzTimeZonePrivate tzBarnaul("Asia/Barnaul"); + if (tzBarnaul.isValid()) { + QCOMPARE(tzBarnaul.data(std).abbreviation, QString("+07")); + + // first full day of the new rule (tzdata2016b) + QDateTime dt(QDate(2016, 3, 28), QTime(0, 0, 0), Qt::UTC); + QCOMPARE(tzBarnaul.data(dt.toMSecsSinceEpoch()).abbreviation, QString("+07")); + } #endif // Q_OS_UNIX }