...
Run Format

Source file src/archive/tar/writer.go

Documentation: archive/tar

  // Copyright 2009 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  
  package tar
  
  // TODO(dsymonds):
  // - catch more errors (no first header, etc.)
  
  import (
  	"bytes"
  	"errors"
  	"fmt"
  	"io"
  	"path"
  	"sort"
  	"strconv"
  	"strings"
  	"time"
  )
  
  var (
  	ErrWriteTooLong    = errors.New("archive/tar: write too long")
  	ErrFieldTooLong    = errors.New("archive/tar: header field too long")
  	ErrWriteAfterClose = errors.New("archive/tar: write after close")
  	errInvalidHeader   = errors.New("archive/tar: header field too long or contains invalid values")
  )
  
  // A Writer provides sequential writing of a tar archive in POSIX.1 format.
  // A tar archive consists of a sequence of files.
  // Call WriteHeader to begin a new file, and then call Write to supply that file's data,
  // writing at most hdr.Size bytes in total.
  type Writer struct {
  	w          io.Writer
  	err        error
  	nb         int64 // number of unwritten bytes for current file entry
  	pad        int64 // amount of padding to write after current file entry
  	closed     bool
  	usedBinary bool  // whether the binary numeric field extension was used
  	preferPax  bool  // use PAX header instead of binary numeric header
  	hdrBuff    block // buffer to use in writeHeader when writing a regular header
  	paxHdrBuff block // buffer to use in writeHeader when writing a PAX header
  }
  
  // NewWriter creates a new Writer writing to w.
  func NewWriter(w io.Writer) *Writer { return &Writer{w: w} }
  
  // Flush finishes writing the current file (optional).
  func (tw *Writer) Flush() error {
  	if tw.nb > 0 {
  		tw.err = fmt.Errorf("archive/tar: missed writing %d bytes", tw.nb)
  		return tw.err
  	}
  
  	n := tw.nb + tw.pad
  	for n > 0 && tw.err == nil {
  		nr := n
  		if nr > blockSize {
  			nr = blockSize
  		}
  		var nw int
  		nw, tw.err = tw.w.Write(zeroBlock[0:nr])
  		n -= int64(nw)
  	}
  	tw.nb = 0
  	tw.pad = 0
  	return tw.err
  }
  
  var (
  	minTime = time.Unix(0, 0)
  	// There is room for 11 octal digits (33 bits) of mtime.
  	maxTime = minTime.Add((1<<33 - 1) * time.Second)
  )
  
  // WriteHeader writes hdr and prepares to accept the file's contents.
  // WriteHeader calls Flush if it is not the first header.
  // Calling after a Close will return ErrWriteAfterClose.
  func (tw *Writer) WriteHeader(hdr *Header) error {
  	return tw.writeHeader(hdr, true)
  }
  
  // WriteHeader writes hdr and prepares to accept the file's contents.
  // WriteHeader calls Flush if it is not the first header.
  // Calling after a Close will return ErrWriteAfterClose.
  // As this method is called internally by writePax header to allow it to
  // suppress writing the pax header.
  func (tw *Writer) writeHeader(hdr *Header, allowPax bool) error {
  	if tw.closed {
  		return ErrWriteAfterClose
  	}
  	if tw.err == nil {
  		tw.Flush()
  	}
  	if tw.err != nil {
  		return tw.err
  	}
  
  	// a map to hold pax header records, if any are needed
  	paxHeaders := make(map[string]string)
  
  	// TODO(dsnet): we might want to use PAX headers for
  	// subsecond time resolution, but for now let's just capture
  	// too long fields or non ascii characters
  
  	// We need to select which scratch buffer to use carefully,
  	// since this method is called recursively to write PAX headers.
  	// If allowPax is true, this is the non-recursive call, and we will use hdrBuff.
  	// If allowPax is false, we are being called by writePAXHeader, and hdrBuff is
  	// already being used by the non-recursive call, so we must use paxHdrBuff.
  	header := &tw.hdrBuff
  	if !allowPax {
  		header = &tw.paxHdrBuff
  	}
  	copy(header[:], zeroBlock[:])
  
  	// Wrappers around formatter that automatically sets paxHeaders if the
  	// argument extends beyond the capacity of the input byte slice.
  	var f formatter
  	var formatString = func(b []byte, s string, paxKeyword string) {
  		needsPaxHeader := paxKeyword != paxNone && len(s) > len(b) || !isASCII(s)
  		if needsPaxHeader {
  			paxHeaders[paxKeyword] = s
  		}
  
  		// Write string in a best-effort manner to satisfy readers that expect
  		// the field to be non-empty.
  		s = toASCII(s)
  		if len(s) > len(b) {
  			s = s[:len(b)]
  		}
  		f.formatString(b, s) // Should never error
  	}
  	var formatNumeric = func(b []byte, x int64, paxKeyword string) {
  		// Try octal first.
  		s := strconv.FormatInt(x, 8)
  		if len(s) < len(b) {
  			f.formatOctal(b, x)
  			return
  		}
  
  		// If it is too long for octal, and PAX is preferred, use a PAX header.
  		if paxKeyword != paxNone && tw.preferPax {
  			f.formatOctal(b, 0)
  			s := strconv.FormatInt(x, 10)
  			paxHeaders[paxKeyword] = s
  			return
  		}
  
  		tw.usedBinary = true
  		f.formatNumeric(b, x)
  	}
  
  	// Handle out of range ModTime carefully.
  	var modTime int64
  	if !hdr.ModTime.Before(minTime) && !hdr.ModTime.After(maxTime) {
  		modTime = hdr.ModTime.Unix()
  	}
  
  	v7 := header.V7()
  	formatString(v7.Name(), hdr.Name, paxPath)
  	// TODO(dsnet): The GNU format permits the mode field to be encoded in
  	// base-256 format. Thus, we can use formatNumeric instead of formatOctal.
  	f.formatOctal(v7.Mode(), hdr.Mode)
  	formatNumeric(v7.UID(), int64(hdr.Uid), paxUid)
  	formatNumeric(v7.GID(), int64(hdr.Gid), paxGid)
  	formatNumeric(v7.Size(), hdr.Size, paxSize)
  	// TODO(dsnet): Consider using PAX for finer time granularity.
  	formatNumeric(v7.ModTime(), modTime, paxNone)
  	v7.TypeFlag()[0] = hdr.Typeflag
  	formatString(v7.LinkName(), hdr.Linkname, paxLinkpath)
  
  	ustar := header.USTAR()
  	formatString(ustar.UserName(), hdr.Uname, paxUname)
  	formatString(ustar.GroupName(), hdr.Gname, paxGname)
  	formatNumeric(ustar.DevMajor(), hdr.Devmajor, paxNone)
  	formatNumeric(ustar.DevMinor(), hdr.Devminor, paxNone)
  
  	// TODO(dsnet): The logic surrounding the prefix field is broken when trying
  	// to encode the header as GNU format. The challenge with the current logic
  	// is that we are unsure what format we are using at any given moment until
  	// we have processed *all* of the fields. The problem is that by the time
  	// all fields have been processed, some work has already been done to handle
  	// each field under the assumption that it is for one given format or
  	// another. In some situations, this causes the Writer to be confused and
  	// encode a prefix field when the format being used is GNU. Thus, producing
  	// an invalid tar file.
  	//
  	// As a short-term fix, we disable the logic to use the prefix field, which
  	// will force the badly generated GNU files to become encoded as being
  	// the PAX format.
  	//
  	// As an alternative fix, we could hard-code preferPax to be true. However,
  	// this is problematic for the following reasons:
  	//	* The preferPax functionality is not tested at all.
  	//	* This can result in headers that try to use both the GNU and PAX
  	//	features at the same time, which is also wrong.
  	//
  	// The proper fix for this is to use a two-pass method:
  	//	* The first pass simply determines what set of formats can possibly
  	//	encode the given header.
  	//	* The second pass actually encodes the header as that given format
  	//	without worrying about violating the format.
  	//
  	// See the following:
  	//	https://golang.org/issue/12594
  	//	https://golang.org/issue/17630
  	//	https://golang.org/issue/9683
  	const usePrefix = false
  
  	// try to use a ustar header when only the name is too long
  	_, paxPathUsed := paxHeaders[paxPath]
  	if usePrefix && !tw.preferPax && len(paxHeaders) == 1 && paxPathUsed {
  		prefix, suffix, ok := splitUSTARPath(hdr.Name)
  		if ok {
  			// Since we can encode in USTAR format, disable PAX header.
  			delete(paxHeaders, paxPath)
  
  			// Update the path fields
  			formatString(v7.Name(), suffix, paxNone)
  			formatString(ustar.Prefix(), prefix, paxNone)
  		}
  	}
  
  	if tw.usedBinary {
  		header.SetFormat(formatGNU)
  	} else {
  		header.SetFormat(formatUSTAR)
  	}
  
  	// Check if there were any formatting errors.
  	if f.err != nil {
  		tw.err = f.err
  		return tw.err
  	}
  
  	if allowPax {
  		for k, v := range hdr.Xattrs {
  			paxHeaders[paxXattr+k] = v
  		}
  	}
  
  	if len(paxHeaders) > 0 {
  		if !allowPax {
  			return errInvalidHeader
  		}
  		if err := tw.writePAXHeader(hdr, paxHeaders); err != nil {
  			return err
  		}
  	}
  	tw.nb = hdr.Size
  	tw.pad = (blockSize - (tw.nb % blockSize)) % blockSize
  
  	_, tw.err = tw.w.Write(header[:])
  	return tw.err
  }
  
  // splitUSTARPath splits a path according to USTAR prefix and suffix rules.
  // If the path is not splittable, then it will return ("", "", false).
  func splitUSTARPath(name string) (prefix, suffix string, ok bool) {
  	length := len(name)
  	if length <= nameSize || !isASCII(name) {
  		return "", "", false
  	} else if length > prefixSize+1 {
  		length = prefixSize + 1
  	} else if name[length-1] == '/' {
  		length--
  	}
  
  	i := strings.LastIndex(name[:length], "/")
  	nlen := len(name) - i - 1 // nlen is length of suffix
  	plen := i                 // plen is length of prefix
  	if i <= 0 || nlen > nameSize || nlen == 0 || plen > prefixSize {
  		return "", "", false
  	}
  	return name[:i], name[i+1:], true
  }
  
  // writePaxHeader writes an extended pax header to the
  // archive.
  func (tw *Writer) writePAXHeader(hdr *Header, paxHeaders map[string]string) error {
  	// Prepare extended header
  	ext := new(Header)
  	ext.Typeflag = TypeXHeader
  	// Setting ModTime is required for reader parsing to
  	// succeed, and seems harmless enough.
  	ext.ModTime = hdr.ModTime
  	// The spec asks that we namespace our pseudo files
  	// with the current pid. However, this results in differing outputs
  	// for identical inputs. As such, the constant 0 is now used instead.
  	// golang.org/issue/12358
  	dir, file := path.Split(hdr.Name)
  	fullName := path.Join(dir, "PaxHeaders.0", file)
  
  	ascii := toASCII(fullName)
  	if len(ascii) > nameSize {
  		ascii = ascii[:nameSize]
  	}
  	ext.Name = ascii
  	// Construct the body
  	var buf bytes.Buffer
  
  	// Keys are sorted before writing to body to allow deterministic output.
  	keys := make([]string, 0, len(paxHeaders))
  	for k := range paxHeaders {
  		keys = append(keys, k)
  	}
  	sort.Strings(keys)
  
  	for _, k := range keys {
  		fmt.Fprint(&buf, formatPAXRecord(k, paxHeaders[k]))
  	}
  
  	ext.Size = int64(len(buf.Bytes()))
  	if err := tw.writeHeader(ext, false); err != nil {
  		return err
  	}
  	if _, err := tw.Write(buf.Bytes()); err != nil {
  		return err
  	}
  	if err := tw.Flush(); err != nil {
  		return err
  	}
  	return nil
  }
  
  // Write writes to the current entry in the tar archive.
  // Write returns the error ErrWriteTooLong if more than
  // hdr.Size bytes are written after WriteHeader.
  func (tw *Writer) Write(b []byte) (n int, err error) {
  	if tw.closed {
  		err = ErrWriteAfterClose
  		return
  	}
  	overwrite := false
  	if int64(len(b)) > tw.nb {
  		b = b[0:tw.nb]
  		overwrite = true
  	}
  	n, err = tw.w.Write(b)
  	tw.nb -= int64(n)
  	if err == nil && overwrite {
  		err = ErrWriteTooLong
  		return
  	}
  	tw.err = err
  	return
  }
  
  // Close closes the tar archive, flushing any unwritten
  // data to the underlying writer.
  func (tw *Writer) Close() error {
  	if tw.err != nil || tw.closed {
  		return tw.err
  	}
  	tw.Flush()
  	tw.closed = true
  	if tw.err != nil {
  		return tw.err
  	}
  
  	// trailer: two zero blocks
  	for i := 0; i < 2; i++ {
  		_, tw.err = tw.w.Write(zeroBlock[:])
  		if tw.err != nil {
  			break
  		}
  	}
  	return tw.err
  }
  

View as plain text