Vendor integration dependencies.

integration/vendor/github.com/vbatts/tar-split/tar/asm/assemble.go

@@ -0,0 +1,130 @@
+package asm
+
+import (
+	"bytes"
+	"fmt"
+	"hash"
+	"hash/crc64"
+	"io"
+	"sync"
+
+	"github.com/vbatts/tar-split/tar/storage"
+)
+
+// NewOutputTarStream returns an io.ReadCloser that is an assembled tar archive
+// stream.
+//
+// It takes a storage.FileGetter, for mapping the file payloads that are to be read in,
+// and a storage.Unpacker, which has access to the rawbytes and file order
+// metadata. With the combination of these two items, a precise assembled Tar
+// archive is possible.
+func NewOutputTarStream(fg storage.FileGetter, up storage.Unpacker) io.ReadCloser {
+	// ... Since these are interfaces, this is possible, so let's not have a nil pointer
+	if fg == nil || up == nil {
+		return nil
+	}
+	pr, pw := io.Pipe()
+	go func() {
+		err := WriteOutputTarStream(fg, up, pw)
+		if err != nil {
+			pw.CloseWithError(err)
+		} else {
+			pw.Close()
+		}
+	}()
+	return pr
+}
+
+// WriteOutputTarStream writes assembled tar archive to a writer.
+func WriteOutputTarStream(fg storage.FileGetter, up storage.Unpacker, w io.Writer) error {
+	// ... Since these are interfaces, this is possible, so let's not have a nil pointer
+	if fg == nil || up == nil {
+		return nil
+	}
+	var copyBuffer []byte
+	var crcHash hash.Hash
+	var crcSum []byte
+	var multiWriter io.Writer
+	for {
+		entry, err := up.Next()
+		if err != nil {
+			if err == io.EOF {
+				return nil
+			}
+			return err
+		}
+		switch entry.Type {
+		case storage.SegmentType:
+			if _, err := w.Write(entry.Payload); err != nil {
+				return err
+			}
+		case storage.FileType:
+			if entry.Size == 0 {
+				continue
+			}
+			fh, err := fg.Get(entry.GetName())
+			if err != nil {
+				return err
+			}
+			if crcHash == nil {
+				crcHash = crc64.New(storage.CRCTable)
+				crcSum = make([]byte, 8)
+				multiWriter = io.MultiWriter(w, crcHash)
+				copyBuffer = byteBufferPool.Get().([]byte)
+				defer byteBufferPool.Put(copyBuffer)
+			} else {
+				crcHash.Reset()
+			}
+
+			if _, err := copyWithBuffer(multiWriter, fh, copyBuffer); err != nil {
+				fh.Close()
+				return err
+			}
+
+			if !bytes.Equal(crcHash.Sum(crcSum[:0]), entry.Payload) {
+				// I would rather this be a comparable ErrInvalidChecksum or such,
+				// but since it's coming through the PipeReader, the context of
+				// _which_ file would be lost...
+				fh.Close()
+				return fmt.Errorf("file integrity checksum failed for %q", entry.GetName())
+			}
+			fh.Close()
+		}
+	}
+}
+
+var byteBufferPool = &sync.Pool{
+	New: func() interface{} {
+		return make([]byte, 32*1024)
+	},
+}
+
+// copyWithBuffer is taken from stdlib io.Copy implementation
+// https://github.com/golang/go/blob/go1.5.1/src/io/io.go#L367
+func copyWithBuffer(dst io.Writer, src io.Reader, buf []byte) (written int64, err error) {
+	for {
+		nr, er := src.Read(buf)
+		if nr > 0 {
+			nw, ew := dst.Write(buf[0:nr])
+			if nw > 0 {
+				written += int64(nw)
+			}
+			if ew != nil {
+				err = ew
+				break
+			}
+			if nr != nw {
+				err = io.ErrShortWrite
+				break
+			}
+		}
+		if er == io.EOF {
+			break
+		}
+		if er != nil {
+			err = er
+			break
+		}
+	}
+	return written, err
+}

integration/vendor/github.com/vbatts/tar-split/tar/asm/disassemble.go

@@ -0,0 +1,141 @@
+package asm
+
+import (
+	"io"
+	"io/ioutil"
+
+	"github.com/vbatts/tar-split/archive/tar"
+	"github.com/vbatts/tar-split/tar/storage"
+)
+
+// NewInputTarStream wraps the Reader stream of a tar archive and provides a
+// Reader stream of the same.
+//
+// In the middle it will pack the segments and file metadata to storage.Packer
+// `p`.
+//
+// The the storage.FilePutter is where payload of files in the stream are
+// stashed. If this stashing is not needed, you can provide a nil
+// storage.FilePutter. Since the checksumming is still needed, then a default
+// of NewDiscardFilePutter will be used internally
+func NewInputTarStream(r io.Reader, p storage.Packer, fp storage.FilePutter) (io.Reader, error) {
+	// What to do here... folks will want their own access to the Reader that is
+	// their tar archive stream, but we'll need that same stream to use our
+	// forked 'archive/tar'.
+	// Perhaps do an io.TeeReader that hands back an io.Reader for them to read
+	// from, and we'll MITM the stream to store metadata.
+	// We'll need a storage.FilePutter too ...
+
+	// Another concern, whether to do any storage.FilePutter operations, such that we
+	// don't extract any amount of the archive. But then again, we're not making
+	// files/directories, hardlinks, etc. Just writing the io to the storage.FilePutter.
+	// Perhaps we have a DiscardFilePutter that is a bit bucket.
+
+	// we'll return the pipe reader, since TeeReader does not buffer and will
+	// only read what the outputRdr Read's. Since Tar archives have padding on
+	// the end, we want to be the one reading the padding, even if the user's
+	// `archive/tar` doesn't care.
+	pR, pW := io.Pipe()
+	outputRdr := io.TeeReader(r, pW)
+
+	// we need a putter that will generate the crc64 sums of file payloads
+	if fp == nil {
+		fp = storage.NewDiscardFilePutter()
+	}
+
+	go func() {
+		tr := tar.NewReader(outputRdr)
+		tr.RawAccounting = true
+		for {
+			hdr, err := tr.Next()
+			if err != nil {
+				if err != io.EOF {
+					pW.CloseWithError(err)
+					return
+				}
+				// even when an EOF is reached, there is often 1024 null bytes on
+				// the end of an archive. Collect them too.
+				if b := tr.RawBytes(); len(b) > 0 {
+					_, err := p.AddEntry(storage.Entry{
+						Type:    storage.SegmentType,
+						Payload: b,
+					})
+					if err != nil {
+						pW.CloseWithError(err)
+						return
+					}
+				}
+				break // not return. We need the end of the reader.
+			}
+			if hdr == nil {
+				break // not return. We need the end of the reader.
+			}
+
+			if b := tr.RawBytes(); len(b) > 0 {
+				_, err := p.AddEntry(storage.Entry{
+					Type:    storage.SegmentType,
+					Payload: b,
+				})
+				if err != nil {
+					pW.CloseWithError(err)
+					return
+				}
+			}
+
+			var csum []byte
+			if hdr.Size > 0 {
+				var err error
+				_, csum, err = fp.Put(hdr.Name, tr)
+				if err != nil {
+					pW.CloseWithError(err)
+					return
+				}
+			}
+
+			entry := storage.Entry{
+				Type:    storage.FileType,
+				Size:    hdr.Size,
+				Payload: csum,
+			}
+			// For proper marshalling of non-utf8 characters
+			entry.SetName(hdr.Name)
+
+			// File entries added, regardless of size
+			_, err = p.AddEntry(entry)
+			if err != nil {
+				pW.CloseWithError(err)
+				return
+			}
+
+			if b := tr.RawBytes(); len(b) > 0 {
+				_, err = p.AddEntry(storage.Entry{
+					Type:    storage.SegmentType,
+					Payload: b,
+				})
+				if err != nil {
+					pW.CloseWithError(err)
+					return
+				}
+			}
+		}
+
+		// it is allowable, and not uncommon that there is further padding on the
+		// end of an archive, apart from the expected 1024 null bytes.
+		remainder, err := ioutil.ReadAll(outputRdr)
+		if err != nil && err != io.EOF {
+			pW.CloseWithError(err)
+			return
+		}
+		_, err = p.AddEntry(storage.Entry{
+			Type:    storage.SegmentType,
+			Payload: remainder,
+		})
+		if err != nil {
+			pW.CloseWithError(err)
+			return
+		}
+		pW.Close()
+	}()
+
+	return pR, nil
+}

integration/vendor/github.com/vbatts/tar-split/tar/asm/doc.go

@@ -0,0 +1,9 @@
+/*
+Package asm provides the API for streaming assembly and disassembly of tar
+archives.
+
+Using the `github.com/vbatts/tar-split/tar/storage` for Packing/Unpacking the
+metadata for a stream, as well as an implementation of Getting/Putting the file
+entries' payload.
+*/
+package asm