Extract some full-size raw previews without LibRaw

Not all image/x-nikon-nef will work like this,
so don't claim their MIME type.
This commit is contained in:
2023-05-23 02:09:15 +02:00
parent 0f1c61ae33
commit bb4b895cb5
4 changed files with 719 additions and 404 deletions

View File

@@ -21,348 +21,10 @@
#include <stdlib.h>
#include <string.h>
// --- Utilities ---------------------------------------------------------------
static char *
binhex(const uint8_t *data, size_t len)
{
static const char *alphabet = "0123456789abcdef";
char *buf = calloc(1, len * 2 + 1), *p = buf;
for (size_t i = 0; i < len; i++) {
*p++ = alphabet[data[i] >> 4];
*p++ = alphabet[data[i] & 0xF];
}
return buf;
}
static uint64_t
u64be(const uint8_t *p)
{
return (uint64_t) p[0] << 56 | (uint64_t) p[1] << 48 |
(uint64_t) p[2] << 40 | (uint64_t) p[3] << 32 |
(uint64_t) p[4] << 24 | p[5] << 16 | p[6] << 8 | p[7];
}
static uint32_t
u32be(const uint8_t *p)
{
return (uint32_t) p[0] << 24 | p[1] << 16 | p[2] << 8 | p[3];
}
static uint16_t
u16be(const uint8_t *p)
{
return (uint16_t) p[0] << 8 | p[1];
}
static uint64_t
u64le(const uint8_t *p)
{
return (uint64_t) p[7] << 56 | (uint64_t) p[6] << 48 |
(uint64_t) p[5] << 40 | (uint64_t) p[4] << 32 |
(uint64_t) p[3] << 24 | p[2] << 16 | p[1] << 8 | p[0];
}
static uint32_t
u32le(const uint8_t *p)
{
return (uint32_t) p[3] << 24 | p[2] << 16 | p[1] << 8 | p[0];
}
static uint16_t
u16le(const uint8_t *p)
{
return (uint16_t) p[1] << 8 | p[0];
}
// --- TIFF --------------------------------------------------------------------
// libtiff is a mess, and the format is not particularly complicated.
// Exiv2 is senselessly copylefted, and cannot do much.
// libexif is only marginally better.
// ExifTool is too user-oriented.
static struct un {
uint64_t (*u64) (const uint8_t *);
uint32_t (*u32) (const uint8_t *);
uint16_t (*u16) (const uint8_t *);
} unbe = {u64be, u32be, u16be}, unle = {u64le, u32le, u16le};
struct tiffer {
struct un *un;
const uint8_t *begin, *p, *end;
uint16_t remaining_fields;
};
static bool
tiffer_u32(struct tiffer *self, uint32_t *u)
{
if (self->p < self->begin || self->p + 4 > self->end)
return false;
*u = self->un->u32(self->p);
self->p += 4;
return true;
}
static bool
tiffer_u16(struct tiffer *self, uint16_t *u)
{
if (self->p < self->begin || self->p + 2 > self->end)
return false;
*u = self->un->u16(self->p);
self->p += 2;
return true;
}
// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
static bool
tiffer_init(struct tiffer *self, const uint8_t *tiff, size_t len)
{
self->un = NULL;
self->begin = self->p = tiff;
self->end = tiff + len;
self->remaining_fields = 0;
const uint8_t
le[4] = {'I', 'I', 42, 0},
be[4] = {'M', 'M', 0, 42};
if (tiff + 8 > self->end)
return false;
else if (!memcmp(tiff, le, sizeof le))
self->un = &unle;
else if (!memcmp(tiff, be, sizeof be))
self->un = &unbe;
else
return false;
self->p = tiff + 4;
// The first IFD needs to be read by caller explicitly,
// even though it's required to be present by TIFF 6.0.
return true;
}
/// Read the next IFD in a sequence.
static bool
tiffer_next_ifd(struct tiffer *self)
{
// All fields from any previous IFD need to be read first.
if (self->remaining_fields)
return false;
uint32_t ifd_offset = 0;
if (!tiffer_u32(self, &ifd_offset))
return false;
// There is nothing more to read, this chain has terminated.
if (!ifd_offset)
return false;
// Note that TIFF 6.0 requires there to be at least one entry,
// but there is no need for us to check it.
self->p = self->begin + ifd_offset;
return tiffer_u16(self, &self->remaining_fields);
}
/// Initialize a derived TIFF reader for a subIFD at the given location.
static bool
tiffer_subifd(struct tiffer *self, uint32_t offset, struct tiffer *subreader)
{
*subreader = *self;
subreader->p = subreader->begin + offset;
return tiffer_u16(subreader, &subreader->remaining_fields);
}
enum tiffer_type {
BYTE = 1, ASCII, SHORT, LONG, RATIONAL,
SBYTE, UNDEFINED, SSHORT, SLONG, SRATIONAL, FLOAT, DOUBLE,
IFD // This last type from TIFF Technical Note 1 isn't really used much.
};
static size_t
tiffer_value_size(enum tiffer_type type)
{
switch (type) {
case BYTE:
case SBYTE:
case ASCII:
case UNDEFINED:
return 1;
case SHORT:
case SSHORT:
return 2;
case LONG:
case SLONG:
case FLOAT:
case IFD:
return 4;
case RATIONAL:
case SRATIONAL:
case DOUBLE:
return 8;
default:
return 0;
}
}
/// A lean iterator for values within entries.
struct tiffer_entry {
uint16_t tag;
enum tiffer_type type;
// For {S,}BYTE, ASCII, UNDEFINED, use these fields directly.
const uint8_t *p;
uint32_t remaining_count;
};
static bool
tiffer_next_value(struct tiffer_entry *entry)
{
if (!entry->remaining_count)
return false;
entry->p += tiffer_value_size(entry->type);
entry->remaining_count--;
return true;
}
static bool
tiffer_integer(
const struct tiffer *self, const struct tiffer_entry *entry, int64_t *out)
{
if (!entry->remaining_count)
return false;
// Somewhat excessively lenient, intended for display.
// TIFF 6.0 only directly suggests that a reader is should accept
// any of BYTE/SHORT/LONG for unsigned integers.
switch (entry->type) {
case BYTE:
case ASCII:
case UNDEFINED:
*out = *entry->p;
return true;
case SBYTE:
*out = (int8_t) *entry->p;
return true;
case SHORT:
*out = self->un->u16(entry->p);
return true;
case SSHORT:
*out = (int16_t) self->un->u16(entry->p);
return true;
case LONG:
case IFD:
*out = self->un->u32(entry->p);
return true;
case SLONG:
*out = (int32_t) self->un->u32(entry->p);
return true;
default:
return false;
}
}
static bool
tiffer_rational(const struct tiffer *self, const struct tiffer_entry *entry,
int64_t *numerator, int64_t *denominator)
{
if (!entry->remaining_count)
return false;
// Somewhat excessively lenient, intended for display.
switch (entry->type) {
case RATIONAL:
*numerator = self->un->u32(entry->p);
*denominator = self->un->u32(entry->p + 4);
return true;
case SRATIONAL:
*numerator = (int32_t) self->un->u32(entry->p);
*denominator = (int32_t) self->un->u32(entry->p + 4);
return true;
default:
if (tiffer_integer(self, entry, numerator)) {
*denominator = 1;
return true;
}
return false;
}
}
static bool
tiffer_real(
const struct tiffer *self, const struct tiffer_entry *entry, double *out)
{
if (!entry->remaining_count)
return false;
// Somewhat excessively lenient, intended for display.
// Assuming the host architecture uses IEEE 754.
switch (entry->type) {
int64_t numerator, denominator;
case FLOAT:
*out = *(float *) entry->p;
return true;
case DOUBLE:
*out = *(double *) entry->p;
return true;
default:
if (tiffer_rational(self, entry, &numerator, &denominator)) {
*out = (double) numerator / denominator;
return true;
}
return false;
}
}
static bool
tiffer_next_entry(struct tiffer *self, struct tiffer_entry *entry)
{
if (!self->remaining_fields)
return false;
uint16_t type = entry->type = 0xFFFF;
if (!tiffer_u16(self, &entry->tag) || !tiffer_u16(self, &type) ||
!tiffer_u32(self, &entry->remaining_count))
return false;
// Short values may and will be inlined, rather than pointed to.
size_t values_size = tiffer_value_size(type) * entry->remaining_count;
uint32_t offset = 0;
if (values_size <= sizeof offset) {
entry->p = self->p;
self->p += sizeof offset;
} else if (tiffer_u32(self, &offset)) {
entry->p = self->begin + offset;
} else {
return false;
}
// All entries are pre-checked not to overflow.
if (entry->p + values_size > self->end)
return false;
// Setting it at the end may provide an indication while debugging.
entry->type = type;
self->remaining_fields--;
return true;
}
// --- TIFF/Exif tags ----------------------------------------------------------
struct tiff_value {
const char *name;
uint16_t value;
};
struct tiff_entry {
const char *name;
uint16_t tag;
struct tiff_value *values;
};
// --- TIFF/Exif ---------------------------------------------------------------
#include "tiff-tables.h"
#include "tiffer.h"
// TODO(p): Consider if these can't be inlined into `tiff_entries`.
static struct {
@@ -376,6 +38,27 @@ static struct {
{}
};
// --- Utilities ---------------------------------------------------------------
#define u64be tiffer_u64be
#define u32be tiffer_u32be
#define u16be tiffer_u16be
#define u64le tiffer_u64le
#define u32le tiffer_u32le
#define u16le tiffer_u16le
static char *
binhex(const uint8_t *data, size_t len)
{
static const char *alphabet = "0123456789abcdef";
char *buf = calloc(1, len * 2 + 1), *p = buf;
for (size_t i = 0; i < len; i++) {
*p++ = alphabet[data[i] >> 4];
*p++ = alphabet[data[i] & 0xF];
}
return buf;
}
// --- Analysis ----------------------------------------------------------------
static jv