archive/tar: re-implement USTAR path splitting

The logic for USTAR was disabled because a previous implementation of
Writer had a wrong understanding of the differences between USTAR and GNU,
causing the prefix field is incorrectly be populated in GNU files.

Now that this issue has been fixed, we can re-enable the logic for USTAR
path splitting, which allows Writer to use the USTAR for a wider range
of possible inputs.

Updates #9683
Updates #12594
Updates #17630

Change-Id: I9fe34e5df63f99c6dd56fee3a7e7e4d6ec3995c9
Reviewed-on: https://go-review.googlesource.com/55574
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

src/archive/tar/common.go

@@ -86,6 +86,7 @@ func (h *Header) allowedFormats() (format int, paxHdrs map[string]string) {
 	paxHdrs = make(map[string]string)
 
 	verifyString := func(s string, size int, gnuLong bool, paxKey string) {
+
 		// NUL-terminator is optional for path and linkpath.
 		// Technically, it is required for uname and gname,
 		// but neither GNU nor BSD tar checks for it.
@@ -95,9 +96,10 @@ func (h *Header) allowedFormats() (format int, paxHdrs map[string]string) {
 			format &^= formatGNU // No GNU
 		}
 		if !isASCII(s) || tooLong {
-			// TODO(dsnet): If the path is splittable, it is possible to still
-			// use the USTAR format.
+			canSplitUSTAR := paxKey == paxPath
+			if _, _, ok := splitUSTARPath(s); !canSplitUSTAR || !ok {
 				format &^= formatUSTAR // No USTAR
+			}
 			if paxKey == paxNone {
 				format &^= formatPAX // No PAX
 			} else {

src/archive/tar/writer.go

@@ -28,6 +28,7 @@ type Writer struct {
 	pad    int64 // amount of padding to write after current file entry
 	closed bool
 
+	hdr Header // Shallow copy of Header that is safe for mutations
 	blk block  // Buffer to use as temporary local storage
 }
 
@@ -59,25 +60,30 @@ func (tw *Writer) WriteHeader(hdr *Header) error {
 		return err
 	}
 
-	switch allowedFormats, paxHdrs := hdr.allowedFormats(); {
+	tw.hdr = *hdr // Shallow copy of Header
+	switch allowedFormats, paxHdrs := tw.hdr.allowedFormats(); {
 	case allowedFormats&formatUSTAR != 0:
-		return tw.writeUSTARHeader(hdr)
+		return tw.writeUSTARHeader(&tw.hdr)
 	case allowedFormats&formatPAX != 0:
-		return tw.writePAXHeader(hdr, paxHdrs)
+		return tw.writePAXHeader(&tw.hdr, paxHdrs)
 	case allowedFormats&formatGNU != 0:
-		return tw.writeGNUHeader(hdr)
+		return tw.writeGNUHeader(&tw.hdr)
 	default:
 		return ErrHeader
 	}
 }
 
 func (tw *Writer) writeUSTARHeader(hdr *Header) error {
-	// TODO(dsnet): Support USTAR prefix/suffix path splitting.
-	// See https://golang.org/issue/12594
+	// Check if we can use USTAR prefix/suffix splitting.
+	var namePrefix string
+	if prefix, suffix, ok := splitUSTARPath(hdr.Name); ok {
+		namePrefix, hdr.Name = prefix, suffix
+	}
 
 	// Pack the main header.
 	var f formatter
 	blk := tw.templateV7Plus(hdr, f.formatString, f.formatOctal)
+	f.formatString(blk.USTAR().Prefix(), namePrefix)
 	blk.SetFormat(formatUSTAR)
 	if f.err != nil {
 		return f.err // Should never happen since header is validated

src/archive/tar/writer_test.go

@@ -151,15 +151,9 @@ func TestWriter(t *testing.T) {
 			contents: strings.Repeat("\x00", 4<<10),
 		}},
 	}, {
-		// TODO(dsnet): The Writer output should match the following file.
-		// To fix an issue (see https://golang.org/issue/12594), we disabled
-		// prefix support, which alters the generated output.
-		/*
-			// This file was produced using gnu tar 1.17
+		// This file was produced using GNU tar v1.17.
 		//	gnutar -b 4 --format=ustar (longname/)*15 + file.txt
-			file: "testdata/ustar.tar"
-		*/
-		file: "testdata/ustar.issue12594.tar", // This is a valid tar file, but not expected
+		file: "testdata/ustar.tar",
 		entries: []*entry{{
 			header: &Header{
 				Name:     strings.Repeat("longname/", 15) + "file.txt",