mirror of
https://github.com/superseriousbusiness/gotosocial.git
synced 2024-11-22 19:56:39 +00:00
94e87610c4
* add back exif-terminator and use only for jpeg,png,webp * fix arguments passed to terminateExif() * pull in latest exif-terminator * fix test * update processed img --------- Co-authored-by: tobi <tobi.smethurst@protonmail.com>
387 lines
7.8 KiB
Go
387 lines
7.8 KiB
Go
package pngstructure
|
|
|
|
import (
|
|
"bytes"
|
|
"errors"
|
|
"fmt"
|
|
"io"
|
|
|
|
"encoding/binary"
|
|
"hash/crc32"
|
|
|
|
"github.com/dsoprea/go-exif/v3"
|
|
exifcommon "github.com/dsoprea/go-exif/v3/common"
|
|
riimage "github.com/dsoprea/go-utility/v2/image"
|
|
)
|
|
|
|
var (
|
|
PngSignature = [8]byte{137, 'P', 'N', 'G', '\r', '\n', 26, '\n'}
|
|
EXifChunkType = "eXIf"
|
|
IHDRChunkType = "IHDR"
|
|
)
|
|
|
|
var (
|
|
ErrNotPng = errors.New("not png data")
|
|
ErrCrcFailure = errors.New("crc failure")
|
|
)
|
|
|
|
// ChunkSlice encapsulates a slice of chunks.
|
|
type ChunkSlice struct {
|
|
chunks []*Chunk
|
|
}
|
|
|
|
func NewChunkSlice(chunks []*Chunk) (*ChunkSlice, error) {
|
|
if len(chunks) == 0 {
|
|
err := errors.New("ChunkSlice must be initialized with at least one chunk (IHDR)")
|
|
return nil, err
|
|
} else if chunks[0].Type != IHDRChunkType {
|
|
err := errors.New("first chunk in any ChunkSlice must be an IHDR")
|
|
return nil, err
|
|
}
|
|
|
|
return &ChunkSlice{chunks}, nil
|
|
}
|
|
|
|
func NewPngChunkSlice() (*ChunkSlice, error) {
|
|
ihdrChunk := &Chunk{
|
|
Type: IHDRChunkType,
|
|
}
|
|
|
|
ihdrChunk.UpdateCrc32()
|
|
|
|
return NewChunkSlice([]*Chunk{ihdrChunk})
|
|
}
|
|
|
|
func (cs *ChunkSlice) String() string {
|
|
return fmt.Sprintf("ChunkSlize<LEN=(%d)>", len(cs.chunks))
|
|
}
|
|
|
|
// Chunks exposes the actual slice.
|
|
func (cs *ChunkSlice) Chunks() []*Chunk {
|
|
return cs.chunks
|
|
}
|
|
|
|
// Write encodes and writes all chunks.
|
|
func (cs *ChunkSlice) WriteTo(w io.Writer) error {
|
|
if _, err := w.Write(PngSignature[:]); err != nil {
|
|
return err
|
|
}
|
|
|
|
// TODO(dustin): !! This should respect
|
|
// the safe-to-copy characteristic.
|
|
for _, c := range cs.chunks {
|
|
if _, err := c.WriteTo(w); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
// Index returns a map of chunk types to chunk slices, grouping all like chunks.
|
|
func (cs *ChunkSlice) Index() (index map[string][]*Chunk) {
|
|
index = make(map[string][]*Chunk)
|
|
for _, c := range cs.chunks {
|
|
if grouped, found := index[c.Type]; found {
|
|
index[c.Type] = append(grouped, c)
|
|
} else {
|
|
index[c.Type] = []*Chunk{c}
|
|
}
|
|
}
|
|
|
|
return index
|
|
}
|
|
|
|
// FindExif returns the the segment that hosts the EXIF data.
|
|
func (cs *ChunkSlice) FindExif() (chunk *Chunk, err error) {
|
|
index := cs.Index()
|
|
if chunks, found := index[EXifChunkType]; found {
|
|
return chunks[0], nil
|
|
}
|
|
|
|
return nil, exif.ErrNoExif
|
|
}
|
|
|
|
// Exif returns an `exif.Ifd` instance with the existing tags.
|
|
func (cs *ChunkSlice) Exif() (*exif.Ifd, []byte, error) {
|
|
chunk, err := cs.FindExif()
|
|
if err != nil {
|
|
return nil, nil, err
|
|
}
|
|
|
|
im, err := exifcommon.NewIfdMappingWithStandard()
|
|
if err != nil {
|
|
return nil, nil, err
|
|
}
|
|
|
|
ti := exif.NewTagIndex()
|
|
|
|
_, index, err := exif.Collect(im, ti, chunk.Data)
|
|
if err != nil {
|
|
return nil, nil, err
|
|
}
|
|
|
|
return index.RootIfd, chunk.Data, nil
|
|
}
|
|
|
|
// ConstructExifBuilder returns an `exif.IfdBuilder` instance
|
|
// (needed for modifying) preloaded with all existing tags.
|
|
func (cs *ChunkSlice) ConstructExifBuilder() (*exif.IfdBuilder, error) {
|
|
rootIfd, _, err := cs.Exif()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
return exif.NewIfdBuilderFromExistingChain(rootIfd), nil
|
|
}
|
|
|
|
// SetExif encodes and sets EXIF data into this segment.
|
|
func (cs *ChunkSlice) SetExif(ib *exif.IfdBuilder) error {
|
|
// Encode.
|
|
|
|
ibe := exif.NewIfdByteEncoder()
|
|
|
|
exifData, err := ibe.EncodeToExif(ib)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
// Set.
|
|
|
|
exifChunk, err := cs.FindExif()
|
|
|
|
switch {
|
|
case err == nil:
|
|
// EXIF chunk already exists.
|
|
exifChunk.Data = exifData
|
|
exifChunk.Length = uint32(len(exifData))
|
|
|
|
case errors.Is(err, exif.ErrNoExif):
|
|
// Add a EXIF chunk for the first time.
|
|
exifChunk = &Chunk{
|
|
Type: EXifChunkType,
|
|
Data: exifData,
|
|
Length: uint32(len(exifData)),
|
|
}
|
|
|
|
// Insert exif after the IHDR chunk; it's
|
|
// a reliably appropriate place to put it.
|
|
cs.chunks = append(
|
|
cs.chunks[:1],
|
|
append(
|
|
[]*Chunk{exifChunk},
|
|
cs.chunks[1:]...,
|
|
)...,
|
|
)
|
|
|
|
default:
|
|
return err
|
|
}
|
|
|
|
exifChunk.UpdateCrc32()
|
|
return nil
|
|
}
|
|
|
|
// PngSplitter hosts the princpal `Split()`
|
|
// method uses by `bufio.Scanner`.
|
|
type PngSplitter struct {
|
|
chunks []*Chunk
|
|
currentOffset int
|
|
|
|
doCheckCrc bool
|
|
crcErrors []string
|
|
}
|
|
|
|
func (ps *PngSplitter) Chunks() (*ChunkSlice, error) {
|
|
return NewChunkSlice(ps.chunks)
|
|
}
|
|
|
|
func (ps *PngSplitter) DoCheckCrc(doCheck bool) {
|
|
ps.doCheckCrc = doCheck
|
|
}
|
|
|
|
func (ps *PngSplitter) CrcErrors() []string {
|
|
return ps.crcErrors
|
|
}
|
|
|
|
func NewPngSplitter() *PngSplitter {
|
|
return &PngSplitter{
|
|
chunks: make([]*Chunk, 0),
|
|
doCheckCrc: true,
|
|
crcErrors: make([]string, 0),
|
|
}
|
|
}
|
|
|
|
// Chunk describes a single chunk.
|
|
type Chunk struct {
|
|
Offset int
|
|
Length uint32
|
|
Type string
|
|
Data []byte
|
|
Crc uint32
|
|
}
|
|
|
|
func (c *Chunk) String() string {
|
|
return fmt.Sprintf("Chunk<OFFSET=(%d) LENGTH=(%d) TYPE=[%s] CRC=(%d)>", c.Offset, c.Length, c.Type, c.Crc)
|
|
}
|
|
|
|
func calculateCrc32(chunk *Chunk) uint32 {
|
|
c := crc32.NewIEEE()
|
|
|
|
c.Write([]byte(chunk.Type))
|
|
c.Write(chunk.Data)
|
|
|
|
return c.Sum32()
|
|
}
|
|
|
|
func (c *Chunk) UpdateCrc32() {
|
|
c.Crc = calculateCrc32(c)
|
|
}
|
|
|
|
func (c *Chunk) CheckCrc32() bool {
|
|
expected := calculateCrc32(c)
|
|
return c.Crc == expected
|
|
}
|
|
|
|
// Bytes encodes and returns the bytes for this chunk.
|
|
func (c *Chunk) Bytes() ([]byte, error) {
|
|
if len(c.Data) != int(c.Length) {
|
|
return nil, errors.New("length of data not correct")
|
|
}
|
|
b := make([]byte, 0, 4+4+c.Length+4)
|
|
b = binary.BigEndian.AppendUint32(b, c.Length)
|
|
b = append(b, c.Type...)
|
|
b = append(b, c.Data...)
|
|
b = binary.BigEndian.AppendUint32(b, c.Crc)
|
|
return b, nil
|
|
}
|
|
|
|
// Write encodes and writes the bytes for this chunk.
|
|
func (c *Chunk) WriteTo(w io.Writer) (int, error) {
|
|
if len(c.Data) != int(c.Length) {
|
|
return 0, errors.New("length of data not correct")
|
|
}
|
|
|
|
var n int
|
|
|
|
b := make([]byte, 4) // uint32 buf
|
|
|
|
binary.BigEndian.PutUint32(b, c.Length)
|
|
if nn, err := w.Write(b); err != nil {
|
|
return n + nn, err
|
|
}
|
|
|
|
n += len(b)
|
|
|
|
if nn, err := io.WriteString(w, c.Type); err != nil {
|
|
return n + nn, err
|
|
}
|
|
|
|
n += len(c.Type)
|
|
|
|
if nn, err := w.Write(c.Data); err != nil {
|
|
return n + nn, err
|
|
}
|
|
|
|
n += len(c.Data)
|
|
|
|
binary.BigEndian.PutUint32(b, c.Crc)
|
|
if nn, err := w.Write(b); err != nil {
|
|
return n + nn, err
|
|
}
|
|
|
|
n += len(b)
|
|
|
|
return n, nil
|
|
}
|
|
|
|
// readHeader verifies that the PNG header bytes appear next.
|
|
func (ps *PngSplitter) readHeader(r io.Reader) error {
|
|
var (
|
|
sigLen = len(PngSignature)
|
|
header = make([]byte, sigLen)
|
|
)
|
|
|
|
if _, err := r.Read(header); err != nil {
|
|
return err
|
|
}
|
|
|
|
ps.currentOffset += sigLen
|
|
if !bytes.Equal(header, PngSignature[:]) {
|
|
return ErrNotPng
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
// Split fulfills the `bufio.SplitFunc`
|
|
// function definition for `bufio.Scanner`.
|
|
func (ps *PngSplitter) Split(
|
|
data []byte,
|
|
atEOF bool,
|
|
) (
|
|
advance int,
|
|
token []byte,
|
|
err error,
|
|
) {
|
|
// We might have more than one chunk's worth, and,
|
|
// if `atEOF` is true, we won't be called again.
|
|
// We'll repeatedly try to read additional chunks,
|
|
// but, when we run out of the data we were given
|
|
// then we'll return the number of bytes for the
|
|
// chunks we've already completely read. Then, we'll
|
|
// be called again from the end ofthose bytes, at
|
|
// which point we'll indicate that we don't yet have
|
|
// enough for another chunk, and we should be then
|
|
// called with more.
|
|
for {
|
|
len_ := len(data)
|
|
if len_ < 8 {
|
|
return advance, nil, nil
|
|
}
|
|
|
|
length := binary.BigEndian.Uint32(data[:4])
|
|
type_ := string(data[4:8])
|
|
chunkSize := (8 + int(length) + 4)
|
|
|
|
if len_ < chunkSize {
|
|
return advance, nil, nil
|
|
}
|
|
|
|
crcIndex := 8 + length
|
|
crc := binary.BigEndian.Uint32(data[crcIndex : crcIndex+4])
|
|
|
|
content := make([]byte, length)
|
|
copy(content, data[8:8+length])
|
|
|
|
c := &Chunk{
|
|
Length: length,
|
|
Type: type_,
|
|
Data: content,
|
|
Crc: crc,
|
|
Offset: ps.currentOffset,
|
|
}
|
|
|
|
ps.chunks = append(ps.chunks, c)
|
|
|
|
if !c.CheckCrc32() {
|
|
ps.crcErrors = append(ps.crcErrors, type_)
|
|
|
|
if ps.doCheckCrc {
|
|
err = ErrCrcFailure
|
|
return
|
|
}
|
|
}
|
|
|
|
advance += chunkSize
|
|
ps.currentOffset += chunkSize
|
|
|
|
data = data[chunkSize:]
|
|
}
|
|
}
|
|
|
|
var (
|
|
// Enforce interface conformance.
|
|
_ riimage.MediaContext = new(ChunkSlice)
|
|
)
|