You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

658 lines
20 KiB
C++

// SPDX-License-Identifier: GPL-2.0-or-later
// included header files
#include "types.hpp"
#include "enforce.hpp"
#include "futils.hpp"
#include "i18n.h" // for _exvGettext
#include "safe_op.hpp"
// + standard includes
#include <array>
#include <cctype>
#include <climits>
#include <cmath>
#include <cstring>
#include <iomanip>
#include <sstream>
#include <utility>
// *****************************************************************************
namespace {
//! Information pertaining to the defined %Exiv2 value type identifiers.
struct TypeInfoTable {
Exiv2::TypeId typeId_; //!< Type id
const char* name_; //!< Name of the type
size_t size_; //!< Bytes per data entry
//! Comparison operator for \em typeId
bool operator==(Exiv2::TypeId typeId) const {
return typeId_ == typeId;
}
//! Comparison operator for \em name
bool operator==(const std::string& name) const {
return name == name_;
}
}; // struct TypeInfoTable
//! Lookup list with information of Exiv2 types
constexpr auto typeInfoTable = std::array{
TypeInfoTable{Exiv2::invalidTypeId, "Invalid", 0},
TypeInfoTable{Exiv2::unsignedByte, "Byte", 1},
TypeInfoTable{Exiv2::asciiString, "Ascii", 1},
TypeInfoTable{Exiv2::unsignedShort, "Short", 2},
TypeInfoTable{Exiv2::unsignedLong, "Long", 4},
TypeInfoTable{Exiv2::unsignedRational, "Rational", 8},
TypeInfoTable{Exiv2::signedByte, "SByte", 1},
TypeInfoTable{Exiv2::undefined, "Undefined", 1},
TypeInfoTable{Exiv2::signedShort, "SShort", 2},
TypeInfoTable{Exiv2::signedLong, "SLong", 4},
TypeInfoTable{Exiv2::signedRational, "SRational", 8},
TypeInfoTable{Exiv2::tiffFloat, "Float", 4},
TypeInfoTable{Exiv2::tiffDouble, "Double", 8},
TypeInfoTable{Exiv2::tiffIfd, "Ifd", 4},
TypeInfoTable{Exiv2::string, "String", 1},
TypeInfoTable{Exiv2::date, "Date", 8},
TypeInfoTable{Exiv2::time, "Time", 11},
TypeInfoTable{Exiv2::comment, "Comment", 1},
TypeInfoTable{Exiv2::directory, "Directory", 1},
TypeInfoTable{Exiv2::xmpText, "XmpText", 1},
TypeInfoTable{Exiv2::xmpAlt, "XmpAlt", 1},
TypeInfoTable{Exiv2::xmpBag, "XmpBag", 1},
TypeInfoTable{Exiv2::xmpSeq, "XmpSeq", 1},
TypeInfoTable{Exiv2::langAlt, "LangAlt", 1},
};
} // namespace
// *****************************************************************************
// class member definitions
namespace Exiv2 {
const char* TypeInfo::typeName(TypeId typeId) {
auto tit = std::find(typeInfoTable.begin(), typeInfoTable.end(), typeId);
if (tit == typeInfoTable.end())
return nullptr;
return tit->name_;
}
TypeId TypeInfo::typeId(const std::string& typeName) {
auto tit = std::find(typeInfoTable.begin(), typeInfoTable.end(), typeName);
if (tit == typeInfoTable.end())
return invalidTypeId;
return tit->typeId_;
}
size_t TypeInfo::typeSize(TypeId typeId) {
auto tit = std::find(typeInfoTable.begin(), typeInfoTable.end(), typeId);
if (tit == typeInfoTable.end())
return 0;
return tit->size_;
}
DataBuf::DataBuf(size_t size) : pData_(size) {
}
DataBuf::DataBuf(const byte* pData, size_t size) : pData_(size) {
std::copy_n(pData, size, pData_.begin());
}
void DataBuf::alloc(size_t size) {
pData_.resize(size);
}
void DataBuf::resize(size_t size) {
pData_.resize(size);
}
void DataBuf::reset() {
pData_.clear();
}
uint8_t Exiv2::DataBuf::read_uint8(size_t offset) const {
if (offset >= pData_.size()) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::read_uint8");
}
return pData_[offset];
}
void Exiv2::DataBuf::write_uint8(size_t offset, uint8_t x) {
if (offset >= pData_.size()) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::write_uint8");
}
pData_[offset] = x;
}
uint16_t Exiv2::DataBuf::read_uint16(size_t offset, ByteOrder byteOrder) const {
if (pData_.size() < 2 || offset > (pData_.size() - 2)) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::read_uint16");
}
return getUShort(&pData_[offset], byteOrder);
}
void Exiv2::DataBuf::write_uint16(size_t offset, uint16_t x, ByteOrder byteOrder) {
if (pData_.size() < 2 || offset > (pData_.size() - 2)) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::write_uint16");
}
us2Data(&pData_[offset], x, byteOrder);
}
uint32_t Exiv2::DataBuf::read_uint32(size_t offset, ByteOrder byteOrder) const {
if (pData_.size() < 4 || offset > (pData_.size() - 4)) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::read_uint32");
}
return getULong(&pData_[offset], byteOrder);
}
void Exiv2::DataBuf::write_uint32(size_t offset, uint32_t x, ByteOrder byteOrder) {
if (pData_.size() < 4 || offset > (pData_.size() - 4)) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::write_uint32");
}
ul2Data(&pData_[offset], x, byteOrder);
}
uint64_t Exiv2::DataBuf::read_uint64(size_t offset, ByteOrder byteOrder) const {
if (pData_.size() < 8 || offset > (pData_.size() - 8)) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::read_uint64");
}
return getULongLong(&pData_[offset], byteOrder);
}
void Exiv2::DataBuf::write_uint64(size_t offset, uint64_t x, ByteOrder byteOrder) {
if (pData_.size() < 8 || offset > (pData_.size() - 8)) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::write_uint64");
}
ull2Data(&pData_[offset], x, byteOrder);
}
int Exiv2::DataBuf::cmpBytes(size_t offset, const void* buf, size_t bufsize) const {
if (pData_.size() < bufsize || offset > pData_.size() - bufsize) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::cmpBytes");
}
return memcmp(&pData_[offset], buf, bufsize);
}
byte* Exiv2::DataBuf::data(size_t offset) {
return const_cast<byte*>(c_data(offset));
}
const byte* Exiv2::DataBuf::c_data(size_t offset) const {
if (pData_.empty()) {
return nullptr;
}
if (offset >= pData_.size()) {
throw std::out_of_range("Overflow in Exiv2::DataBuf::c_data");
}
return &pData_[offset];
}
const char* Exiv2::DataBuf::c_str(size_t offset) const {
return reinterpret_cast<const char*>(c_data(offset));
}
// *************************************************************************
// free functions
static void checkDataBufBounds(const DataBuf& buf, size_t end) {
enforce<std::invalid_argument>(end <= static_cast<size_t>(std::numeric_limits<long>::max()),
"end of slice too large to be compared with DataBuf bounds.");
enforce<std::out_of_range>(end <= buf.size(), "Invalid slice bounds specified");
}
Slice<byte*> makeSlice(DataBuf& buf, size_t begin, size_t end) {
checkDataBufBounds(buf, end);
return {buf.data(), begin, end};
}
Slice<const byte*> makeSlice(const DataBuf& buf, size_t begin, size_t end) {
checkDataBufBounds(buf, end);
return {buf.c_data(), begin, end};
}
std::ostream& operator<<(std::ostream& os, const Rational& r) {
return os << r.first << "/" << r.second;
}
template <typename T>
std::istream& fromStreamToRational(std::istream& is, T& r) {
// http://dev.exiv2.org/boards/3/topics/1912?r=1915
if (std::tolower(is.peek()) == 'f') {
char F = 0;
float f = 0.F;
is >> F >> f;
f = 2.0F * std::log(f) / std::log(2.0F);
r = Exiv2::floatToRationalCast(f);
} else {
int32_t nominator = 0;
int32_t denominator = 0;
char c('\0');
is >> nominator >> c >> denominator;
if (c != '/')
is.setstate(std::ios::failbit);
if (is)
r = {nominator, denominator};
}
return is;
}
std::istream& operator>>(std::istream& is, Rational& r) {
return fromStreamToRational(is, r);
}
std::ostream& operator<<(std::ostream& os, const URational& r) {
return os << r.first << "/" << r.second;
}
std::istream& operator>>(std::istream& is, URational& r) {
return fromStreamToRational(is, r);
}
uint16_t getUShort(const byte* buf, ByteOrder byteOrder) {
return getUShort(makeSliceUntil(buf, 2), byteOrder);
}
uint32_t getULong(const byte* buf, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
return buf[3] << 24 | buf[2] << 16 | buf[1] << 8 | buf[0];
}
return buf[0] << 24 | buf[1] << 16 | buf[2] << 8 | buf[3];
}
uint64_t getULongLong(const byte* buf, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
return static_cast<uint64_t>(buf[7]) << 56 | static_cast<uint64_t>(buf[6]) << 48 |
static_cast<uint64_t>(buf[5]) << 40 | static_cast<uint64_t>(buf[4]) << 32 |
static_cast<uint64_t>(buf[3]) << 24 | static_cast<uint64_t>(buf[2]) << 16 |
static_cast<uint64_t>(buf[1]) << 8 | static_cast<uint64_t>(buf[0]);
}
return static_cast<uint64_t>(buf[0]) << 56 | static_cast<uint64_t>(buf[1]) << 48 |
static_cast<uint64_t>(buf[2]) << 40 | static_cast<uint64_t>(buf[3]) << 32 |
static_cast<uint64_t>(buf[4]) << 24 | static_cast<uint64_t>(buf[5]) << 16 |
static_cast<uint64_t>(buf[6]) << 8 | static_cast<uint64_t>(buf[7]);
}
URational getURational(const byte* buf, ByteOrder byteOrder) {
uint32_t nominator = getULong(buf, byteOrder);
uint32_t denominator = getULong(buf + 4, byteOrder);
return {nominator, denominator};
}
int16_t getShort(const byte* buf, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
return buf[1] << 8 | buf[0];
}
return buf[0] << 8 | buf[1];
}
int32_t getLong(const byte* buf, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
return buf[3] << 24 | buf[2] << 16 | buf[1] << 8 | buf[0];
}
return buf[0] << 24 | buf[1] << 16 | buf[2] << 8 | buf[3];
}
Rational getRational(const byte* buf, ByteOrder byteOrder) {
int32_t nominator = getLong(buf, byteOrder);
int32_t denominator = getLong(buf + 4, byteOrder);
return {nominator, denominator};
}
float getFloat(const byte* buf, ByteOrder byteOrder) {
// This algorithm assumes that the internal representation of the float
// type is the 4-byte IEEE 754 binary32 format, which is common but not
// required by the C++ standard.
union {
uint32_t ul_;
float f_;
} u;
u.ul_ = getULong(buf, byteOrder);
return u.f_;
}
double getDouble(const byte* buf, ByteOrder byteOrder) {
// This algorithm assumes that the internal representation of the double
// type is the 8-byte IEEE 754 binary64 format, which is common but not
// required by the C++ standard.
union {
uint64_t ull_;
double d_;
} u;
u.ull_ = 0;
if (byteOrder == littleEndian) {
u.ull_ = static_cast<uint64_t>(buf[7]) << 56 | static_cast<uint64_t>(buf[6]) << 48 |
static_cast<uint64_t>(buf[5]) << 40 | static_cast<uint64_t>(buf[4]) << 32 |
static_cast<uint64_t>(buf[3]) << 24 | static_cast<uint64_t>(buf[2]) << 16 |
static_cast<uint64_t>(buf[1]) << 8 | static_cast<uint64_t>(buf[0]);
} else {
u.ull_ = static_cast<uint64_t>(buf[0]) << 56 | static_cast<uint64_t>(buf[1]) << 48 |
static_cast<uint64_t>(buf[2]) << 40 | static_cast<uint64_t>(buf[3]) << 32 |
static_cast<uint64_t>(buf[4]) << 24 | static_cast<uint64_t>(buf[5]) << 16 |
static_cast<uint64_t>(buf[6]) << 8 | static_cast<uint64_t>(buf[7]);
}
return u.d_;
}
size_t us2Data(byte* buf, uint16_t s, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
buf[0] = static_cast<byte>(s & 0x00ffU);
buf[1] = static_cast<byte>((s & 0xff00U) >> 8);
} else {
buf[0] = static_cast<byte>((s & 0xff00U) >> 8);
buf[1] = static_cast<byte>(s & 0x00ffU);
}
return 2;
}
size_t ul2Data(byte* buf, uint32_t l, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
buf[0] = static_cast<byte>(l & 0x000000ffU);
buf[1] = static_cast<byte>((l & 0x0000ff00U) >> 8);
buf[2] = static_cast<byte>((l & 0x00ff0000U) >> 16);
buf[3] = static_cast<byte>((l & 0xff000000U) >> 24);
} else {
buf[0] = static_cast<byte>((l & 0xff000000U) >> 24);
buf[1] = static_cast<byte>((l & 0x00ff0000U) >> 16);
buf[2] = static_cast<byte>((l & 0x0000ff00U) >> 8);
buf[3] = static_cast<byte>(l & 0x000000ffU);
}
return 4;
}
size_t ull2Data(byte* buf, uint64_t l, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
for (size_t i = 0; i < 8; i++) {
buf[i] = static_cast<byte>(l & 0xff);
l >>= 8;
}
} else {
for (size_t i = 0; i < 8; i++) {
buf[8 - i - 1] = static_cast<byte>(l & 0xff);
l >>= 8;
}
}
return 8;
}
size_t ur2Data(byte* buf, URational l, ByteOrder byteOrder) {
size_t o = ul2Data(buf, l.first, byteOrder);
o += ul2Data(buf + o, l.second, byteOrder);
return o;
}
size_t s2Data(byte* buf, int16_t s, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
buf[0] = static_cast<byte>(s & 0x00ffU);
buf[1] = static_cast<byte>((s & 0xff00U) >> 8);
} else {
buf[0] = static_cast<byte>((s & 0xff00U) >> 8);
buf[1] = static_cast<byte>(s & 0x00ffU);
}
return 2;
}
size_t l2Data(byte* buf, int32_t l, ByteOrder byteOrder) {
if (byteOrder == littleEndian) {
buf[0] = static_cast<byte>(l & 0x000000ffU);
buf[1] = static_cast<byte>((l & 0x0000ff00U) >> 8);
buf[2] = static_cast<byte>((l & 0x00ff0000U) >> 16);
buf[3] = static_cast<byte>((l & 0xff000000U) >> 24);
} else {
buf[0] = static_cast<byte>((l & 0xff000000U) >> 24);
buf[1] = static_cast<byte>((l & 0x00ff0000U) >> 16);
buf[2] = static_cast<byte>((l & 0x0000ff00U) >> 8);
buf[3] = static_cast<byte>(l & 0x000000ffU);
}
return 4;
}
size_t r2Data(byte* buf, Rational l, ByteOrder byteOrder) {
size_t o = l2Data(buf, l.first, byteOrder);
o += l2Data(buf + o, l.second, byteOrder);
return o;
}
size_t f2Data(byte* buf, float f, ByteOrder byteOrder) {
// This algorithm assumes that the internal representation of the float
// type is the 4-byte IEEE 754 binary32 format, which is common but not
// required by the C++ standard.
union {
uint32_t ul_;
float f_;
} u;
u.f_ = f;
return ul2Data(buf, u.ul_, byteOrder);
}
size_t d2Data(byte* buf, double d, ByteOrder byteOrder) {
// This algorithm assumes that the internal representation of the double
// type is the 8-byte IEEE 754 binary64 format, which is common but not
// required by the C++ standard.
union {
uint64_t ull_;
double d_;
} u;
u.d_ = d;
uint64_t m = 0xff;
if (byteOrder == littleEndian) {
buf[0] = static_cast<byte>(u.ull_ & m);
buf[1] = static_cast<byte>((u.ull_ & (m << 8)) >> 8);
buf[2] = static_cast<byte>((u.ull_ & (m << 16)) >> 16);
buf[3] = static_cast<byte>((u.ull_ & (m << 24)) >> 24);
buf[4] = static_cast<byte>((u.ull_ & (m << 32)) >> 32);
buf[5] = static_cast<byte>((u.ull_ & (m << 40)) >> 40);
buf[6] = static_cast<byte>((u.ull_ & (m << 48)) >> 48);
buf[7] = static_cast<byte>((u.ull_ & (m << 56)) >> 56);
} else {
buf[0] = static_cast<byte>((u.ull_ & (m << 56)) >> 56);
buf[1] = static_cast<byte>((u.ull_ & (m << 48)) >> 48);
buf[2] = static_cast<byte>((u.ull_ & (m << 40)) >> 40);
buf[3] = static_cast<byte>((u.ull_ & (m << 32)) >> 32);
buf[4] = static_cast<byte>((u.ull_ & (m << 24)) >> 24);
buf[5] = static_cast<byte>((u.ull_ & (m << 16)) >> 16);
buf[6] = static_cast<byte>((u.ull_ & (m << 8)) >> 8);
buf[7] = static_cast<byte>(u.ull_ & m);
}
return 8;
}
void hexdump(std::ostream& os, const byte* buf, size_t len, size_t offset) {
const std::string::size_type pos = 8 + 16 * 3 + 2;
const std::string align(pos, ' ');
std::ios::fmtflags f(os.flags());
size_t i = 0;
while (i < len) {
os << " " << std::setw(4) << std::setfill('0') << std::hex << i + offset << " ";
std::ostringstream ss;
do {
byte c = buf[i];
os << std::setw(2) << std::setfill('0') << std::right << std::hex << static_cast<int>(c) << " ";
ss << (static_cast<int>(c) >= 31 && static_cast<int>(c) < 127 ? static_cast<char>(buf[i]) : '.');
} while (++i < len && i % 16 != 0);
std::string::size_type width = 9 + ((i - 1) % 16 + 1) * 3;
os << (width > pos ? "" : align.substr(width)) << ss.str() << "\n";
}
os << std::dec << std::setfill(' ');
os.flags(f);
}
bool isHex(const std::string& str, size_t size, const std::string& prefix) {
if (str.size() <= prefix.size() || str.substr(0, prefix.size()) != prefix)
return false;
if (size > 0 && str.size() != size + prefix.size())
return false;
for (size_t i = prefix.size(); i < str.size(); ++i) {
if (!isxdigit(str[i]))
return false;
}
return true;
} // isHex
int exifTime(const char* buf, struct tm* tm) {
int rc = 1;
int year = 0, mon = 0, mday = 0, hour = 0, min = 0, sec = 0;
int scanned = std::sscanf(buf, "%4d:%2d:%2d %2d:%2d:%2d", &year, &mon, &mday, &hour, &min, &sec);
if (scanned == 6) {
tm->tm_year = year - 1900;
tm->tm_mon = mon - 1;
tm->tm_mday = mday;
tm->tm_hour = hour;
tm->tm_min = min;
tm->tm_sec = sec;
rc = 0;
}
return rc;
} // exifTime
const char* exvGettext(const char* str) {
#ifdef EXV_ENABLE_NLS
return _exvGettext(str);
#else
return str;
#endif
}
template <>
bool stringTo<bool>(const std::string& s, bool& ok) {
std::string lcs(s); /* lowercase string */
for (size_t i = 0; i < lcs.length(); i++) {
lcs[i] = std::tolower(s[i]);
}
/* handle the same values as xmp sdk */
if (lcs == "false" || lcs == "f" || lcs == "0") {
ok = true;
return false;
}
if (lcs == "true" || lcs == "t" || lcs == "1") {
ok = true;
return true;
}
ok = false;
return false;
}
int64_t parseInt64(const std::string& s, bool& ok) {
auto ret = stringTo<int64_t>(s, ok);
if (ok)
return ret;
auto f = stringTo<float>(s, ok);
if (ok)
return static_cast<int64_t>(f);
auto [r, st] = stringTo<Rational>(s, ok);
if (ok) {
if (st <= 0) {
ok = false;
return 0;
}
return static_cast<int64_t>(static_cast<float>(r) / st);
}
bool b = stringTo<bool>(s, ok);
if (ok)
return b ? 1 : 0;
// everything failed, return from stringTo<int64_t> is probably the best fit
return ret;
}
uint32_t parseUint32(const std::string& s, bool& ok) {
const int64_t x = parseInt64(s, ok);
if (ok && 0 <= x && x <= std::numeric_limits<uint32_t>::max()) {
return static_cast<uint32_t>(x);
}
ok = false;
return 0;
}
float parseFloat(const std::string& s, bool& ok) {
auto ret = stringTo<float>(s, ok);
if (ok)
return ret;
auto [r, st] = stringTo<Rational>(s, ok);
if (ok) {
if (st == 0) {
ok = false;
return 0.0;
}
return static_cast<float>(r) / st;
}
bool b = stringTo<bool>(s, ok);
if (ok)
return b ? 1.0F : 0.0F;
// everything failed, return from stringTo<float> is probably the best fit
return ret;
}
Rational parseRational(const std::string& s, bool& ok) {
auto ret = stringTo<Rational>(s, ok);
if (ok)
return ret;
auto l = stringTo<long>(s, ok);
if (ok)
return {l, 1};
auto f = stringTo<float>(s, ok);
if (ok)
return floatToRationalCast(f);
bool b = stringTo<bool>(s, ok);
if (ok)
return {b ? 1 : 0, 1};
// everything failed, return from stringTo<Rational> is probably the best fit
return ret;
}
Rational floatToRationalCast(float f) {
// Convert f to double because it simplifies the "in_range" check
// below. (INT_MAX can be represented accurately as a double, but
// gets rounded when it's converted to float.)
const double d = f;
const bool in_range = std::numeric_limits<int32_t>::min() <= d && d <= std::numeric_limits<int32_t>::max();
if (!in_range) {
return {d > 0 ? 1 : -1, 0};
}
// Beware: primitive conversion algorithm
int32_t den = 1000000;
const auto d_as_int32_t = static_cast<int32_t>(d);
if (Safe::abs(d_as_int32_t) > 21474836) {
den = 1;
} else if (Safe::abs(d_as_int32_t) > 214748) {
den = 100;
} else if (Safe::abs(d_as_int32_t) > 2147) {
den = 10000;
}
const auto nom = static_cast<int32_t>(std::round(d * den));
const int32_t g = gcd(nom, den);
return {nom / g, den / g};
}
} // namespace Exiv2
#ifdef EXV_ENABLE_NLS
// Declaration is in i18n.h
const char* _exvGettext(const char* str) {
static bool exvGettextInitialized = false;
if (!exvGettextInitialized) {
// bindtextdomain(EXV_PACKAGE_NAME, EXV_LOCALEDIR);
const std::string localeDir =
EXV_LOCALEDIR[0] == '/' ? EXV_LOCALEDIR : (Exiv2::getProcessPath() + EXV_SEPARATOR_STR + EXV_LOCALEDIR);
bindtextdomain(EXV_PACKAGE_NAME, localeDir.c_str());
#ifdef EXV_HAVE_BIND_TEXTDOMAIN_CODESET
bind_textdomain_codeset(EXV_PACKAGE_NAME, "UTF-8");
#endif
exvGettextInitialized = true;
}
return dgettext(EXV_PACKAGE_NAME, str);
}
#endif // EXV_ENABLE_NLS