1 ### the gnu tar specification:
2 ### http://www.gnu.org/software/tar/manual/tar.html
4 ### and the pax format spec, which tar derives from:
5 ### http://www.opengroup.org/onlinepubs/007904975/utilities/pax.html
13 use Carp qw(carp croak);
15 use File::Spec::Unix ();
18 use Archive::Tar::File;
19 use Archive::Tar::Constant;
24 use vars qw[$DEBUG $error $VERSION $WARN $FOLLOW_SYMLINK $CHOWN $CHMOD
25 $DO_NOT_USE_PREFIX $HAS_PERLIO $HAS_IO_STRING
26 $INSECURE_EXTRACT_MODE @ISA @EXPORT
30 @EXPORT = ( COMPRESS_GZIP, COMPRESS_BZIP );
37 $DO_NOT_USE_PREFIX = 0;
38 $INSECURE_EXTRACT_MODE = 0;
42 $HAS_PERLIO = $Config::Config{useperlio};
44 ### try and load IO::String anyway, so you can dynamically
45 ### switch between perlio and IO::String
46 $HAS_IO_STRING = eval {
55 Archive::Tar - module for manipulations of tar archives
60 my $tar = Archive::Tar->new;
62 $tar->read('origin.tgz');
65 $tar->add_files('file/foo.pl', 'docs/README');
66 $tar->add_data('file/baz.txt', 'This is the contents now');
68 $tar->rename('oldname', 'new/file/name');
70 $tar->write('files.tar'); # plain tar
71 $tar->write('files.tgz', COMPRESSED_GZIP); # gzip compressed
72 $tar->write('files.tbz', COMPRESSED_BZIP); # bzip2 compressed
76 Archive::Tar provides an object oriented mechanism for handling tar
77 files. It provides class methods for quick and easy files handling
78 while also allowing for the creation of tar file objects for custom
79 manipulation. If you have the IO::Zlib module installed,
80 Archive::Tar will also support compressed or gzipped tar files.
82 An object of class Archive::Tar represents a .tar(.gz) archive full
87 =head2 Archive::Tar->new( [$file, $compressed] )
89 Returns a new Tar object. If given any arguments, C<new()> calls the
90 C<read()> method automatically, passing on the arguments provided to
93 If C<new()> is invoked with arguments and the C<read()> method fails
94 for any reason, C<new()> returns undef.
103 ### install get/set accessors for this object.
104 for my $key ( keys %$tmpl ) {
106 *{__PACKAGE__."::$key"} = sub {
108 $self->{$key} = $_[0] if @_;
109 return $self->{$key};
115 $class = ref $class if ref $class;
117 ### copying $tmpl here since a shallow copy makes it use the
118 ### same aref, causing for files to remain in memory always.
119 my $obj = bless { _data => [ ], _file => 'Unknown' }, $class;
122 unless ( $obj->read( @_ ) ) {
123 $obj->_error(qq[No data could be read from file]);
131 =head2 $tar->read ( $filename|$handle, [$compressed, {opt => 'val'}] )
133 Read the given tar file into memory.
134 The first argument can either be the name of a file or a reference to
135 an already open filehandle (or an IO::Zlib object if it's compressed)
137 The C<read> will I<replace> any previous content in C<$tar>!
139 The second argument may be considered optional, but remains for
140 backwards compatibility. Archive::Tar now looks at the file
141 magic to determine what class should be used to open the file
142 and will transparently Do The Right Thing.
144 Archive::Tar will warn if you try to pass a bzip2 compressed file and the
145 IO::Zlib / IO::Uncompress::Bunzip2 modules are not available and simply return.
147 Note that you can currently B<not> pass a C<gzip> compressed
148 filehandle, which is not opened with C<IO::Zlib>, a C<bzip2> compressed
149 filehandle, which is not opened with C<IO::Uncompress::Bunzip2>, nor a string
150 containing the full archive information (either compressed or
151 uncompressed). These are worth while features, but not currently
152 implemented. See the C<TODO> section.
154 The third argument can be a hash reference with options. Note that
155 all options are case-sensitive.
161 Do not read more than C<limit> files. This is useful if you have
162 very big archives, and are only interested in the first few files.
166 Can be set to a regular expression. Only files with names that match
167 the expression will be read.
171 If set to true, immediately extract entries when reading them. This
172 gives you the same memory break as the C<extract_archive> function.
173 Note however that entries will not be read into memory, but written
174 straight to disk. This means no C<Archive::Tar::File> objects are
175 created for you to inspect.
179 All files are stored internally as C<Archive::Tar::File> objects.
180 Please consult the L<Archive::Tar::File> documentation for details.
182 Returns the number of files read in scalar context, and a list of
183 C<Archive::Tar::File> objects in list context.
190 my $gzip = shift || 0;
191 my $opts = shift || {};
193 unless( defined $file ) {
194 $self->_error( qq[No file to read from!] );
197 $self->_file( $file );
200 my $handle = $self->_get_handle($file, $gzip, READ_ONLY->( ZLIB ) )
203 my $data = $self->_read_tar( $handle, $opts ) or return;
205 $self->_data( $data );
207 return wantarray ? @$data : scalar @$data;
212 my $file = shift; return unless defined $file;
213 return $file if ref $file;
214 my $compress = shift || 0;
215 my $mode = shift || READ_ONLY->( ZLIB ); # default to read only
218 ### get a FH opened to the right class, so we can use it transparently
219 ### throughout the program
221 { ### reading magic only makes sense if we're opening a file for
222 ### reading. otherwise, just use what the user requested.
224 if( MODE_READ->($mode) ) {
225 open my $tmp, $file or do {
226 $self->_error( qq[Could not open '$file' for reading: $!] );
230 ### read the first 4 bites of the file to figure out which class to
231 ### use to open the file.
232 sysread( $tmp, $magic, 4 );
237 ### if you asked specifically for bzip compression, or if we're in
238 ### read mode and the magic numbers add up, use bzip
240 ($compress eq COMPRESS_BZIP) or
241 ( MODE_READ->($mode) and $magic =~ BZIP_MAGIC_NUM )
245 ### different reader/writer modules, different error vars... sigh
246 if( MODE_READ->($mode) ) {
247 $fh = IO::Uncompress::Bunzip2->new( $file ) or do {
248 $self->_error( qq[Could not read '$file': ] .
249 $IO::Uncompress::Bunzip2::Bunzip2Error
255 $fh = IO::Compress::Bzip2->new( $file ) or do {
256 $self->_error( qq[Could not write to '$file': ] .
257 $IO::Compress::Bzip2::Bzip2Error
264 ### if you asked for compression, if you wanted to read or the gzip
265 ### magic number is present (redundant with read)
267 $compress or MODE_READ->($mode) or $magic =~ GZIP_MAGIC_NUM
272 unless( $fh->open( $file, $mode ) ) {
273 $self->_error(qq[Could not create filehandle for '$file': $!]);
281 unless( $fh->open( $file, $mode ) ) {
282 $self->_error(qq[Could not create filehandle for '$file': $!]);
286 ### enable bin mode on tar archives
297 my $handle = shift or return;
298 my $opts = shift || {};
300 my $count = $opts->{limit} || 0;
301 my $filter = $opts->{filter};
302 my $extract = $opts->{extract} || 0;
304 ### set a cap on the amount of files to extract ###
306 $limit = 1 if $count > 0;
311 my $real_name; # to set the name of a file when
312 # we're encountering @longlink
316 while( $handle->read( $chunk, HEAD ) ) {
317 ### IO::Zlib doesn't support this yet
318 my $offset = eval { tell $handle } || 'unknown';
321 my $gzip = GZIP_MAGIC_NUM;
322 if( $chunk =~ /$gzip/ ) {
323 $self->_error( qq[Cannot read compressed format in tar-mode] );
328 ### if we can't read in all bytes... ###
329 last if length $chunk != HEAD;
331 ### Apparently this should really be two blocks of 512 zeroes,
332 ### but GNU tar sometimes gets it wrong. See comment in the
333 ### source code (tar.c) to GNU cpio.
334 next if $chunk eq TAR_END;
336 ### according to the posix spec, the last 12 bytes of the header are
337 ### null bytes, to pad it to a 512 byte block. That means if these
338 ### bytes are NOT null bytes, it's a corrrupt header. See:
339 ### www.koders.com/c/fidCE473AD3D9F835D690259D60AD5654591D91D5BA.aspx
341 { my $nulls = join '', "\0" x 12;
342 unless( $nulls eq substr( $chunk, 500, 12 ) ) {
343 $self->_error( qq[Invalid header block at offset $offset] );
348 ### pass the realname, so we can set it 'proper' right away
349 ### some of the heuristics are done on the name, so important
352 { my %extra_args = ();
353 $extra_args{'name'} = $$real_name if defined $real_name;
355 unless( $entry = Archive::Tar::File->new( chunk => $chunk,
358 $self->_error( qq[Couldn't read chunk at offset $offset] );
364 ### http://www.gnu.org/manual/tar/html_node/tar_139.html
365 next if $entry->is_label;
367 if( length $entry->type and ($entry->is_file || $entry->is_longlink) ) {
369 if ( $entry->is_file && !$entry->validate ) {
370 ### sometimes the chunk is rather fux0r3d and a whole 512
371 ### bytes ends up in the ->name area.
372 ### clean it up, if need be
373 my $name = $entry->name;
374 $name = substr($name, 0, 100) if length $name > 100;
377 $self->_error( $name . qq[: checksum error] );
381 my $block = BLOCK_SIZE->( $entry->size );
383 $data = $entry->get_content_by_ref;
385 ### just read everything into memory
386 ### can't do lazy loading since IO::Zlib doesn't support 'seek'
387 ### this is because Compress::Zlib doesn't support it =/
388 ### this reads in the whole data in one read() call.
389 if( $handle->read( $$data, $block ) < $block ) {
390 $self->_error( qq[Read error on tarfile (missing data) '].
391 $entry->full_path ."' at offset $offset" );
395 ### throw away trailing garbage ###
396 substr ($$data, $entry->size) = "" if defined $$data;
398 ### part II of the @LongLink munging -- need to do /after/
399 ### the checksum check.
400 if( $entry->is_longlink ) {
401 ### weird thing in tarfiles -- if the file is actually a
402 ### @LongLink, the data part seems to have a trailing ^@
403 ### (unprintable) char. to display, pipe output through less.
404 ### but that doesn't *always* happen.. so check if the last
405 ### character is a control character, and if so remove it
406 ### at any rate, we better remove that character here, or tests
407 ### like 'eq' and hashlook ups based on names will SO not work
408 ### remove it by calculating the proper size, and then
409 ### tossing out everything that's longer than that size.
411 ### count number of nulls
412 my $nulls = $$data =~ tr/\0/\0/;
414 ### cut data + size by that many bytes
415 $entry->size( $entry->size - $nulls );
416 substr ($$data, $entry->size) = "";
420 ### clean up of the entries.. posix tar /apparently/ has some
421 ### weird 'feature' that allows for filenames > 255 characters
422 ### they'll put a header in with as name '././@LongLink' and the
423 ### contents will be the name of the /next/ file in the archive
424 ### pretty crappy and kludgy if you ask me
426 ### set the name for the next entry if this is a @LongLink;
427 ### this is one ugly hack =/ but needed for direct extraction
428 if( $entry->is_longlink ) {
431 } elsif ( defined $real_name ) {
432 $entry->name( $$real_name );
437 ### skip this entry if we're filtering
438 if ($filter && $entry->name !~ $filter) {
441 ### skip this entry if it's a pax header. This is a special file added
442 ### by, among others, git-generated tarballs. It holds comments and is
443 ### not meant for extracting. See #38932: pax_global_header extracted
444 } elsif ( $entry->name eq PAX_HEADER ) {
448 $self->_extract_file( $entry ) if $extract
449 && !$entry->is_longlink
450 && !$entry->is_unknown
451 && !$entry->is_label;
453 ### Guard against tarfiles with garbage at the end
454 last LOOP if $entry->name eq '';
456 ### push only the name on the rv if we're extracting
457 ### -- for extract_archive
458 push @$tarfile, ($extract ? $entry->name : $entry);
461 $count-- unless $entry->is_longlink || $entry->is_dir;
462 last LOOP unless $count;
471 =head2 $tar->contains_file( $filename )
473 Check if the archive contains a certain file.
474 It will return true if the file is in the archive, false otherwise.
476 Note however, that this function does an exact match using C<eq>
477 on the full path. So it cannot compensate for case-insensitive file-
478 systems or compare 2 paths to see if they would point to the same
487 return unless defined $full;
489 ### don't warn if the entry isn't there.. that's what this function
490 ### is for after all.
492 return 1 if $self->_find_entry($full);
496 =head2 $tar->extract( [@filenames] )
498 Write files whose names are equivalent to any of the names in
499 C<@filenames> to disk, creating subdirectories as necessary. This
500 might not work too well under VMS.
501 Under MacPerl, the file's modification time will be converted to the
502 MacOS zero of time, and appropriate conversions will be done to the
503 path. However, the length of each element of the path is not
504 inspected to see whether it's longer than MacOS currently allows (32
507 If C<extract> is called without a list of file names, the entire
508 contents of the archive are extracted.
510 Returns a list of filenames extracted.
519 # use the speed optimization for all extracted files
520 local($self->{cwd}) = cwd() unless $self->{cwd};
522 ### you requested the extraction of only certian files
524 for my $file ( @args ) {
526 ### it's already an object?
527 if( UNIVERSAL::isa( $file, 'Archive::Tar::File' ) ) {
535 for my $entry ( @{$self->_data} ) {
536 next unless $file eq $entry->full_path;
538 ### we found the file you're looking for
544 return $self->_error(
545 qq[Could not find '$file' in archive] );
550 ### just grab all the file items
552 @files = $self->get_files;
555 ### nothing found? that's an error
556 unless( scalar @files ) {
557 $self->_error( qq[No files found for ] . $self->_file );
562 for my $entry ( @files ) {
563 unless( $self->_extract_file( $entry ) ) {
564 $self->_error(q[Could not extract ']. $entry->full_path .q['] );
572 =head2 $tar->extract_file( $file, [$extract_path] )
574 Write an entry, whose name is equivalent to the file name provided to
575 disk. Optionally takes a second parameter, which is the full native
576 path (including filename) the entry will be written to.
580 $tar->extract_file( 'name/in/archive', 'name/i/want/to/give/it' );
582 $tar->extract_file( $at_file_object, 'name/i/want/to/give/it' );
584 Returns true on success, false on failure.
590 my $file = shift; return unless defined $file;
593 my $entry = $self->_find_entry( $file )
594 or $self->_error( qq[Could not find an entry for '$file'] ), return;
596 return $self->_extract_file( $entry, $alt );
601 my $entry = shift or return;
604 ### you wanted an alternate extraction location ###
605 my $name = defined $alt ? $alt : $entry->full_path;
607 ### splitpath takes a bool at the end to indicate
608 ### that it's splitting a dir
609 my ($vol,$dirs,$file);
610 if ( defined $alt ) { # It's a local-OS path
611 ($vol,$dirs,$file) = File::Spec->splitpath( $alt,
614 ($vol,$dirs,$file) = File::Spec::Unix->splitpath( $name,
619 ### is $name an absolute path? ###
620 if( $vol || File::Spec->file_name_is_absolute( $dirs ) ) {
622 ### absolute names are not allowed to be in tarballs under
623 ### strict mode, so only allow it if a user tells us to do it
624 if( not defined $alt and not $INSECURE_EXTRACT_MODE ) {
626 q[Entry ']. $entry->full_path .q[' is an absolute path. ].
627 q[Not extracting absolute paths under SECURE EXTRACT MODE]
632 ### user asked us to, it's fine.
633 $dir = File::Spec->catpath( $vol, $dirs, "" );
635 ### it's a relative path ###
637 my $cwd = (ref $self and defined $self->{cwd})
641 my @dirs = defined $alt
642 ? File::Spec->splitdir( $dirs ) # It's a local-OS path
643 : File::Spec::Unix->splitdir( $dirs ); # it's UNIX-style, likely
644 # straight from the tarball
646 if( not defined $alt and
647 not $INSECURE_EXTRACT_MODE
650 ### paths that leave the current directory are not allowed under
651 ### strict mode, so only allow it if a user tells us to do this.
652 if( grep { $_ eq '..' } @dirs ) {
655 q[Entry ']. $entry->full_path .q[' is attempting to leave ].
656 q[the current working directory. Not extracting under ].
657 q[SECURE EXTRACT MODE]
662 ### the archive may be asking us to extract into a symlink. This
663 ### is not sane and a possible security issue, as outlined here:
664 ### https://rt.cpan.org/Ticket/Display.html?id=30380
665 ### https://bugzilla.redhat.com/show_bug.cgi?id=295021
666 ### https://issues.rpath.com/browse/RPL-1716
667 my $full_path = $cwd;
668 for my $d ( @dirs ) {
669 $full_path = File::Spec->catdir( $full_path, $d );
671 ### we've already checked this one, and it's safe. Move on.
672 next if ref $self and $self->{_link_cache}->{$full_path};
674 if( -l $full_path ) {
675 my $to = readlink $full_path;
676 my $diag = "symlinked directory ($full_path => $to)";
679 q[Entry ']. $entry->full_path .q[' is attempting to ].
680 qq[extract to a $diag. This is considered a security ].
681 q[vulnerability and not allowed under SECURE EXTRACT ].
687 ### XXX keep a cache if possible, so the stats become cheaper:
688 $self->{_link_cache}->{$full_path} = 1 if ref $self;
692 ### '.' is the directory delimiter on VMS, which has to be escaped
693 ### or changed to '_' on vms. vmsify is used, because older versions
694 ### of vmspath do not handle this properly.
695 ### Must not add a '/' to an empty directory though.
696 map { length() ? VMS::Filespec::vmsify($_.'/') : $_ } @dirs if ON_VMS;
698 my ($cwd_vol,$cwd_dir,$cwd_file)
699 = File::Spec->splitpath( $cwd );
700 my @cwd = File::Spec->splitdir( $cwd_dir );
701 push @cwd, $cwd_file if length $cwd_file;
703 ### We need to pass '' as the last elemant to catpath. Craig Berry
704 ### explains why (msgid <p0624083dc311ae541393@[172.16.52.1]>):
705 ### The root problem is that splitpath on UNIX always returns the
706 ### final path element as a file even if it is a directory, and of
707 ### course there is no way it can know the difference without checking
708 ### against the filesystem, which it is documented as not doing. When
709 ### you turn around and call catpath, on VMS you have to know which bits
710 ### are directory bits and which bits are file bits. In this case we
711 ### know the result should be a directory. I had thought you could omit
712 ### the file argument to catpath in such a case, but apparently on UNIX
714 $dir = File::Spec->catpath(
715 $cwd_vol, File::Spec->catdir( @cwd, @dirs ), ''
718 ### catdir() returns undef if the path is longer than 255 chars on
719 ### older VMS systems.
720 unless ( defined $dir ) {
721 $^W && $self->_error( qq[Could not compose a path for '$dirs'\n] );
727 if( -e $dir && !-d _ ) {
728 $^W && $self->_error( qq['$dir' exists, but it's not a directory!\n] );
733 eval { File::Path::mkpath( $dir, 0, 0777 ) };
735 my $fp = $entry->full_path;
736 $self->_error(qq[Could not create directory '$dir' for '$fp': $@]);
740 ### XXX chown here? that might not be the same as in the archive
741 ### as we're only chown'ing to the owner of the file we're extracting
742 ### not to the owner of the directory itself, which may or may not
743 ### be another entry in the archive
744 ### Answer: no, gnu tar doesn't do it either, it'd be the wrong
746 #if( $CHOWN && CAN_CHOWN ) {
747 # chown $entry->uid, $entry->gid, $dir or
748 # $self->_error( qq[Could not set uid/gid on '$dir'] );
752 ### we're done if we just needed to create a dir ###
753 return 1 if $entry->is_dir;
755 my $full = File::Spec->catfile( $dir, $file );
757 if( $entry->is_unknown ) {
758 $self->_error( qq[Unknown file type for file '$full'] );
762 if( length $entry->type && $entry->is_file ) {
763 my $fh = IO::File->new;
764 $fh->open( '>' . $full ) or (
765 $self->_error( qq[Could not open file '$full': $!] ),
771 syswrite $fh, $entry->data or (
772 $self->_error( qq[Could not write data to '$full'] ),
778 $self->_error( qq[Could not close file '$full'] ),
783 $self->_make_special_file( $entry, $full ) or return;
786 ### only update the timestamp if it's not a symlink; that will change the
787 ### timestamp of the original. This addresses bug #33669: Could not update
788 ### timestamp warning on symlinks
790 utime time, $entry->mtime - TIME_OFFSET, $full or
791 $self->_error( qq[Could not update timestamp] );
794 if( $CHOWN && CAN_CHOWN->() ) {
795 chown $entry->uid, $entry->gid, $full or
796 $self->_error( qq[Could not set uid/gid on '$full'] );
799 ### only chmod if we're allowed to, but never chmod symlinks, since they'll
800 ### change the perms on the file they're linking too...
801 if( $CHMOD and not -l $full ) {
802 chmod $entry->mode, $full or
803 $self->_error( qq[Could not chown '$full' to ] . $entry->mode );
809 sub _make_special_file {
811 my $entry = shift or return;
812 my $file = shift; return unless defined $file;
816 if( $entry->is_symlink ) {
819 symlink( $entry->linkname, $file ) or $fail++;
822 $self->_extract_special_file_as_plain_file( $entry, $file )
826 $err = qq[Making symbolic link '$file' to '] .
827 $entry->linkname .q[' failed] if $fail;
829 } elsif ( $entry->is_hardlink ) {
832 link( $entry->linkname, $file ) or $fail++;
835 $self->_extract_special_file_as_plain_file( $entry, $file )
839 $err = qq[Making hard link from '] . $entry->linkname .
840 qq[' to '$file' failed] if $fail;
842 } elsif ( $entry->is_fifo ) {
843 ON_UNIX && !system('mknod', $file, 'p') or
844 $err = qq[Making fifo ']. $entry->name .qq[' failed];
846 } elsif ( $entry->is_blockdev or $entry->is_chardev ) {
847 my $mode = $entry->is_blockdev ? 'b' : 'c';
849 ON_UNIX && !system('mknod', $file, $mode,
850 $entry->devmajor, $entry->devminor) or
851 $err = qq[Making block device ']. $entry->name .qq[' (maj=] .
852 $entry->devmajor . qq[ min=] . $entry->devminor .
855 } elsif ( $entry->is_socket ) {
856 ### the original doesn't do anything special for sockets.... ###
860 return $err ? $self->_error( $err ) : 1;
863 ### don't know how to make symlinks, let's just extract the file as
865 sub _extract_special_file_as_plain_file {
867 my $entry = shift or return;
868 my $file = shift; return unless defined $file;
872 my $orig = $self->_find_entry( $entry->linkname );
875 $err = qq[Could not find file '] . $entry->linkname .
880 ### clone the entry, make it appear as a normal file ###
881 my $clone = $entry->clone;
882 $clone->_downgrade_to_plainfile;
883 $self->_extract_file( $clone, $file ) or last TRY;
888 return $self->_error($err);
891 =head2 $tar->list_files( [\@properties] )
893 Returns a list of the names of all the files in the archive.
895 If C<list_files()> is passed an array reference as its first argument
896 it returns a list of hash references containing the requested
897 properties of each file. The following list of properties is
898 supported: name, size, mtime (last modified date), mode, uid, gid,
899 linkname, uname, gname, devmajor, devminor, prefix.
901 Passing an array reference containing only one element, 'name', is
902 special cased to return a list of names rather than a list of hash
903 references, making it equivalent to calling C<list_files> without
910 my $aref = shift || [ ];
912 unless( $self->_data ) {
913 $self->read() or return;
916 if( @$aref == 0 or ( @$aref == 1 and $aref->[0] eq 'name' ) ) {
917 return map { $_->full_path } @{$self->_data};
921 #for my $obj ( @{$self->_data} ) {
922 # push @rv, { map { $_ => $obj->$_() } @$aref };
926 ### this does the same as the above.. just needs a +{ }
927 ### to make sure perl doesn't confuse it for a block
928 return map { my $o=$_;
929 +{ map { $_ => $o->$_() } @$aref }
938 unless( defined $file ) {
939 $self->_error( qq[No file specified] );
943 ### it's an object already
944 return $file if UNIVERSAL::isa( $file, 'Archive::Tar::File' );
946 for my $entry ( @{$self->_data} ) {
947 my $path = $entry->full_path;
948 return $entry if $path eq $file;
951 $self->_error( qq[No such file in archive: '$file'] );
955 =head2 $tar->get_files( [@filenames] )
957 Returns the C<Archive::Tar::File> objects matching the filenames
958 provided. If no filename list was passed, all C<Archive::Tar::File>
959 objects in the current Tar object are returned.
961 Please refer to the C<Archive::Tar::File> documentation on how to
962 handle these objects.
969 return @{ $self->_data } unless @_;
972 for my $file ( @_ ) {
973 push @list, grep { defined } $self->_find_entry( $file );
979 =head2 $tar->get_content( $file )
981 Return the content of the named file.
987 my $entry = $self->_find_entry( shift ) or return;
992 =head2 $tar->replace_content( $file, $content )
994 Make the string $content be the content for the file named $file.
998 sub replace_content {
1000 my $entry = $self->_find_entry( shift ) or return;
1002 return $entry->replace_content( shift );
1005 =head2 $tar->rename( $file, $new_name )
1007 Rename the file of the in-memory archive to $new_name.
1009 Note that you must specify a Unix path for $new_name, since per tar
1010 standard, all files in the archive must be Unix paths.
1012 Returns true on success and false on failure.
1018 my $file = shift; return unless defined $file;
1019 my $new = shift; return unless defined $new;
1021 my $entry = $self->_find_entry( $file ) or return;
1023 return $entry->rename( $new );
1026 =head2 $tar->remove (@filenamelist)
1028 Removes any entries with names matching any of the given filenames
1029 from the in-memory archive. Returns a list of C<Archive::Tar::File>
1030 objects that remain.
1038 my %seen = map { $_->full_path => $_ } @{$self->_data};
1039 delete $seen{ $_ } for @list;
1041 $self->_data( [values %seen] );
1043 return values %seen;
1048 C<clear> clears the current in-memory archive. This effectively gives
1049 you a 'blank' object, ready to be filled again. Note that C<clear>
1050 only has effect on the object, not the underlying tarfile.
1055 my $self = shift or return;
1064 =head2 $tar->write ( [$file, $compressed, $prefix] )
1066 Write the in-memory archive to disk. The first argument can either
1067 be the name of a file or a reference to an already open filehandle (a
1070 The second argument is used to indicate compression. You can either
1071 compress using C<gzip> or C<bzip2>. If you pass a digit, it's assumed
1072 to be the C<gzip> compression level (between 1 and 9), but the use of
1073 constants is prefered:
1075 # write a gzip compressed file
1076 $tar->write( 'out.tgz', COMPRESSION_GZIP );
1078 # write a bzip compressed file
1079 $tar->write( 'out.tbz', COMPRESSION_BZIP );
1081 Note that when you pass in a filehandle, the compression argument
1082 is ignored, as all files are printed verbatim to your filehandle.
1083 If you wish to enable compression with filehandles, use an
1084 C<IO::Zlib> or C<IO::Compress::Bzip2> filehandle instead.
1086 The third argument is an optional prefix. All files will be tucked
1087 away in the directory you specify as prefix. So if you have files
1088 'a' and 'b' in your archive, and you specify 'foo' as prefix, they
1089 will be written to the archive as 'foo/a' and 'foo/b'.
1091 If no arguments are given, C<write> returns the entire formatted
1092 archive as a string, which could be useful if you'd like to stuff the
1093 archive into a socket or a pipe to gzip or something.
1100 my $file = shift; $file = '' unless defined $file;
1101 my $gzip = shift || 0;
1102 my $ext_prefix = shift; $ext_prefix = '' unless defined $ext_prefix;
1105 ### only need a handle if we have a file to print to ###
1106 my $handle = length($file)
1107 ? ( $self->_get_handle($file, $gzip, WRITE_ONLY->($gzip) )
1109 : $HAS_PERLIO ? do { open my $h, '>', \$dummy; $h }
1110 : $HAS_IO_STRING ? IO::String->new
1111 : __PACKAGE__->no_string_support();
1115 for my $entry ( @{$self->_data} ) {
1116 ### entries to be written to the tarfile ###
1119 ### only now will we change the object to reflect the current state
1120 ### of the name and prefix fields -- this needs to be limited to
1122 my $clone = $entry->clone;
1125 ### so, if you don't want use to use the prefix, we'll stuff
1126 ### everything in the name field instead
1127 if( $DO_NOT_USE_PREFIX ) {
1129 ### you might have an extended prefix, if so, set it in the clone
1130 ### XXX is ::Unix right?
1131 $clone->name( length $ext_prefix
1132 ? File::Spec::Unix->catdir( $ext_prefix,
1134 : $clone->full_path );
1135 $clone->prefix( '' );
1137 ### otherwise, we'll have to set it properly -- prefix part in the
1138 ### prefix and name part in the name field.
1141 ### split them here, not before!
1142 my ($prefix,$name) = $clone->_prefix_and_file( $clone->full_path );
1144 ### you might have an extended prefix, if so, set it in the clone
1145 ### XXX is ::Unix right?
1146 $prefix = File::Spec::Unix->catdir( $ext_prefix, $prefix )
1147 if length $ext_prefix;
1149 $clone->prefix( $prefix );
1150 $clone->name( $name );
1153 ### names are too long, and will get truncated if we don't add a
1154 ### '@LongLink' file...
1155 my $make_longlink = ( length($clone->name) > NAME_LENGTH or
1156 length($clone->prefix) > PREFIX_LENGTH
1159 ### perhaps we need to make a longlink file?
1160 if( $make_longlink ) {
1161 my $longlink = Archive::Tar::File->new(
1162 data => LONGLINK_NAME,
1164 { type => LONGLINK }
1167 unless( $longlink ) {
1168 $self->_error( qq[Could not create 'LongLink' entry for ] .
1169 qq[oversize file '] . $clone->full_path ."'" );
1173 push @write_me, $longlink;
1176 push @write_me, $clone;
1178 ### write the one, optionally 2 a::t::file objects to the handle
1179 for my $clone (@write_me) {
1181 ### if the file is a symlink, there are 2 options:
1182 ### either we leave the symlink intact, but then we don't write any
1183 ### data OR we follow the symlink, which means we actually make a
1184 ### copy. if we do the latter, we have to change the TYPE of the
1186 my $link_ok = $clone->is_symlink && $Archive::Tar::FOLLOW_SYMLINK;
1187 my $data_ok = !$clone->is_symlink && $clone->has_content;
1189 ### downgrade to a 'normal' file if it's a symlink we're going to
1190 ### treat as a regular file
1191 $clone->_downgrade_to_plainfile if $link_ok;
1193 ### get the header for this block
1194 my $header = $self->_format_tar_entry( $clone );
1196 $self->_error(q[Could not format header for: ] .
1197 $clone->full_path );
1201 unless( print $handle $header ) {
1202 $self->_error(q[Could not write header for: ] .
1207 if( $link_ok or $data_ok ) {
1208 unless( print $handle $clone->data ) {
1209 $self->_error(q[Could not write data for: ] .
1214 ### pad the end of the clone if required ###
1215 print $handle TAR_PAD->( $clone->size ) if $clone->size % BLOCK
1218 } ### done writing these entries
1221 ### write the end markers ###
1222 print $handle TAR_END x 2 or
1223 return $self->_error( qq[Could not write tar end markers] );
1225 ### did you want it written to a file, or returned as a string? ###
1226 my $rv = length($file) ? 1
1227 : $HAS_PERLIO ? $dummy
1228 : do { seek $handle, 0, 0; local $/; <$handle> };
1230 ### make sure to close the handle;
1236 sub _format_tar_entry {
1238 my $entry = shift or return;
1239 my $ext_prefix = shift; $ext_prefix = '' unless defined $ext_prefix;
1240 my $no_prefix = shift || 0;
1242 my $file = $entry->name;
1243 my $prefix = $entry->prefix; $prefix = '' unless defined $prefix;
1245 ### remove the prefix from the file name
1246 ### not sure if this is still neeeded --kane
1247 ### no it's not -- Archive::Tar::File->_new_from_file will take care of
1248 ### this for us. Even worse, this would break if we tried to add a file
1250 #if( length $prefix ) {
1251 # $file =~ s/^$match//;
1254 $prefix = File::Spec::Unix->catdir($ext_prefix, $prefix)
1255 if length $ext_prefix;
1257 ### not sure why this is... ###
1258 my $l = PREFIX_LENGTH; # is ambiguous otherwise...
1259 substr ($prefix, 0, -$l) = "" if length $prefix >= PREFIX_LENGTH;
1261 my $f1 = "%06o"; my $f2 = "%11o";
1263 ### this might be optimizable with a 'changed' flag in the file objects ###
1268 (map { sprintf( $f1, $entry->$_() ) } qw[mode uid gid]),
1269 (map { sprintf( $f2, $entry->$_() ) } qw[size mtime]),
1271 "", # checksum field - space padded a bit down
1273 (map { $entry->$_() } qw[type linkname magic]),
1275 $entry->version || TAR_VERSION,
1277 (map { $entry->$_() } qw[uname gname]),
1278 (map { sprintf( $f1, $entry->$_() ) } qw[devmajor devminor]),
1280 ($no_prefix ? '' : $prefix)
1283 ### add the checksum ###
1284 substr($tar,148,7) = sprintf("%6o\0", unpack("%16C*",$tar));
1289 =head2 $tar->add_files( @filenamelist )
1291 Takes a list of filenames and adds them to the in-memory archive.
1293 The path to the file is automatically converted to a Unix like
1294 equivalent for use in the archive, and, if on MacOS, the file's
1295 modification time is converted from the MacOS epoch to the Unix epoch.
1296 So tar archives created on MacOS with B<Archive::Tar> can be read
1297 both with I<tar> on Unix and applications like I<suntar> or
1298 I<Stuffit Expander> on MacOS.
1300 Be aware that the file's type/creator and resource fork will be lost,
1301 which is usually what you want in cross-platform archives.
1303 Instead of a filename, you can also pass it an existing C<Archive::Tar::File>
1304 object from, for example, another archive. The object will be clone, and
1305 effectively be a copy of the original, not an alias.
1307 Returns a list of C<Archive::Tar::File> objects that were just added.
1313 my @files = @_ or return;
1316 for my $file ( @files ) {
1318 ### you passed an Archive::Tar::File object
1319 ### clone it so we don't accidentally have a reference to
1320 ### an object from another archive
1321 if( UNIVERSAL::isa( $file,'Archive::Tar::File' ) ) {
1322 push @rv, $file->clone;
1326 unless( -e $file || -l $file ) {
1327 $self->_error( qq[No such file: '$file'] );
1331 my $obj = Archive::Tar::File->new( file => $file );
1333 $self->_error( qq[Unable to add file: '$file'] );
1340 push @{$self->{_data}}, @rv;
1345 =head2 $tar->add_data ( $filename, $data, [$opthashref] )
1347 Takes a filename, a scalar full of data and optionally a reference to
1348 a hash with specific options.
1350 Will add a file to the in-memory archive, with name C<$filename> and
1351 content C<$data>. Specific properties can be set using C<$opthashref>.
1352 The following list of properties is supported: name, size, mtime
1353 (last modified date), mode, uid, gid, linkname, uname, gname,
1354 devmajor, devminor, prefix, type. (On MacOS, the file's path and
1355 modification times are converted to Unix equivalents.)
1357 Valid values for the file type are the following constants defined in
1358 Archive::Tar::Constants:
1370 Hard and symbolic ("soft") links; linkname should specify target.
1376 Character and block devices. devmajor and devminor should specify the major
1377 and minor device numbers.
1393 Returns the C<Archive::Tar::File> object that was just added, or
1394 C<undef> on failure.
1400 my ($file, $data, $opt) = @_;
1402 my $obj = Archive::Tar::File->new( data => $file, $data, $opt );
1404 $self->_error( qq[Unable to add file: '$file'] );
1408 push @{$self->{_data}}, $obj;
1413 =head2 $tar->error( [$BOOL] )
1415 Returns the current errorstring (usually, the last error reported).
1416 If a true value was specified, it will give the C<Carp::longmess>
1417 equivalent of the error, in effect giving you a stacktrace.
1419 For backwards compatibility, this error is also available as
1420 C<$Archive::Tar::error> although it is much recommended you use the
1421 method call instead.
1431 my $msg = $error = shift;
1432 $longmess = Carp::longmess($error);
1434 ### set Archive::Tar::WARN to 0 to disable printing
1437 carp $DEBUG ? $longmess : $msg;
1445 return shift() ? $longmess : $error;
1449 =head2 $tar->setcwd( $cwd );
1451 C<Archive::Tar> needs to know the current directory, and it will run
1452 C<Cwd::cwd()> I<every> time it extracts a I<relative> entry from the
1453 tarfile and saves it in the file system. (As of version 1.30, however,
1454 C<Archive::Tar> will use the speed optimization described below
1455 automatically, so it's only relevant if you're using C<extract_file()>).
1457 Since C<Archive::Tar> doesn't change the current directory internally
1458 while it is extracting the items in a tarball, all calls to C<Cwd::cwd()>
1459 can be avoided if we can guarantee that the current directory doesn't
1460 get changed externally.
1462 To use this performance boost, set the current directory via
1465 $tar->setcwd( cwd() );
1467 once before calling a function like C<extract_file> and
1468 C<Archive::Tar> will use the current directory setting from then on
1469 and won't call C<Cwd::cwd()> internally.
1471 To switch back to the default behaviour, use
1473 $tar->setcwd( undef );
1475 and C<Archive::Tar> will call C<Cwd::cwd()> internally again.
1477 If you're using C<Archive::Tar>'s C<exract()> method, C<setcwd()> will
1486 $self->{cwd} = $cwd;
1489 =head1 Class Methods
1491 =head2 Archive::Tar->create_archive($file, $compressed, @filelist)
1493 Creates a tar file from the list of files provided. The first
1494 argument can either be the name of the tar file to create or a
1495 reference to an open file handle (e.g. a GLOB reference).
1497 The second argument is used to indicate compression. You can either
1498 compress using C<gzip> or C<bzip2>. If you pass a digit, it's assumed
1499 to be the C<gzip> compression level (between 1 and 9), but the use of
1500 constants is prefered:
1502 # write a gzip compressed file
1503 Archive::Tar->create_archive( 'out.tgz', COMPRESSION_GZIP, @filelist );
1505 # write a bzip compressed file
1506 Archive::Tar->create_archive( 'out.tbz', COMPRESSION_BZIP, @filelist );
1508 Note that when you pass in a filehandle, the compression argument
1509 is ignored, as all files are printed verbatim to your filehandle.
1510 If you wish to enable compression with filehandles, use an
1511 C<IO::Zlib> or C<IO::Compress::Bzip2> filehandle instead.
1513 The remaining arguments list the files to be included in the tar file.
1514 These files must all exist. Any files which don't exist or can't be
1515 read are silently ignored.
1517 If the archive creation fails for any reason, C<create_archive> will
1518 return false. Please use the C<error> method to find the cause of the
1521 Note that this method does not write C<on the fly> as it were; it
1522 still reads all the files into memory before writing out the archive.
1523 Consult the FAQ below if this is a problem.
1527 sub create_archive {
1530 my $file = shift; return unless defined $file;
1531 my $gzip = shift || 0;
1535 return $class->_error( qq[Cowardly refusing to create empty archive!] );
1538 my $tar = $class->new;
1539 $tar->add_files( @files );
1540 return $tar->write( $file, $gzip );
1543 =head2 Archive::Tar->iter( $filename, [ $compressed, {opt => $val} ] )
1545 Returns an iterator function that reads the tar file without loading
1546 it all in memory. Each time the function is called it will return the
1547 next file in the tarball. The files are returned as
1548 C<Archive::Tar::File> objects. The iterator function returns the
1549 empty list once it has exhausted the the files contained.
1551 The second argument can be a hash reference with options, which are
1552 identical to the arguments passed to C<read()>.
1556 my $next = Archive::Tar->iter( "example.tar.gz", 1, {filter => qr/\.pm$/} );
1558 while( my $f = $next->() ) {
1559 print $f->name, "\n";
1561 $f->extract or warn "Extraction failed";
1571 my $filename = shift or return;
1572 my $compressed = shift or 0;
1573 my $opts = shift || {};
1575 ### get a handle to read from.
1576 my $handle = $class->_get_handle(
1584 return shift(@data) if @data; # more than one file returned?
1585 return unless $handle; # handle exhausted?
1587 ### read data, should only return file
1588 @data = @{ $class->_read_tar($handle, { %$opts, limit => 1 }) };
1590 ### return one piece of data
1591 return shift(@data) if @data;
1593 ### data is exhausted, free the filehandle
1599 =head2 Archive::Tar->list_archive($file, $compressed, [\@properties])
1601 Returns a list of the names of all the files in the archive. The
1602 first argument can either be the name of the tar file to list or a
1603 reference to an open file handle (e.g. a GLOB reference).
1605 If C<list_archive()> is passed an array reference as its third
1606 argument it returns a list of hash references containing the requested
1607 properties of each file. The following list of properties is
1608 supported: full_path, name, size, mtime (last modified date), mode,
1609 uid, gid, linkname, uname, gname, devmajor, devminor, prefix.
1611 See C<Archive::Tar::File> for details about supported properties.
1613 Passing an array reference containing only one element, 'name', is
1614 special cased to return a list of names rather than a list of hash
1621 my $file = shift; return unless defined $file;
1622 my $gzip = shift || 0;
1624 my $tar = $class->new($file, $gzip);
1627 return $tar->list_files( @_ );
1630 =head2 Archive::Tar->extract_archive($file, $compressed)
1632 Extracts the contents of the tar file. The first argument can either
1633 be the name of the tar file to create or a reference to an open file
1634 handle (e.g. a GLOB reference). All relative paths in the tar file will
1635 be created underneath the current working directory.
1637 C<extract_archive> will return a list of files it extracted.
1638 If the archive extraction fails for any reason, C<extract_archive>
1639 will return false. Please use the C<error> method to find the cause
1644 sub extract_archive {
1646 my $file = shift; return unless defined $file;
1647 my $gzip = shift || 0;
1649 my $tar = $class->new( ) or return;
1651 return $tar->read( $file, $gzip, { extract => 1 } );
1654 =head2 $bool = Archive::Tar->has_io_string
1656 Returns true if we currently have C<IO::String> support loaded.
1658 Either C<IO::String> or C<perlio> support is needed to support writing
1659 stringified archives. Currently, C<perlio> is the preferred method, if
1662 See the C<GLOBAL VARIABLES> section to see how to change this preference.
1666 sub has_io_string { return $HAS_IO_STRING; }
1668 =head2 $bool = Archive::Tar->has_perlio
1670 Returns true if we currently have C<perlio> support loaded.
1672 This requires C<perl-5.8> or higher, compiled with C<perlio>
1674 Either C<IO::String> or C<perlio> support is needed to support writing
1675 stringified archives. Currently, C<perlio> is the preferred method, if
1678 See the C<GLOBAL VARIABLES> section to see how to change this preference.
1682 sub has_perlio { return $HAS_PERLIO; }
1684 =head2 $bool = Archive::Tar->has_zlib_support
1686 Returns true if C<Archive::Tar> can extract C<zlib> compressed archives
1690 sub has_zlib_support { return ZLIB }
1692 =head2 $bool = Archive::Tar->has_bzip2_support
1694 Returns true if C<Archive::Tar> can extract C<bzip2> compressed archives
1698 sub has_bzip2_support { return BZIP }
1700 =head2 Archive::Tar->can_handle_compressed_files
1702 A simple checking routine, which will return true if C<Archive::Tar>
1703 is able to uncompress compressed archives on the fly with C<IO::Zlib>
1704 and C<IO::Compress::Bzip2> or false if not both are installed.
1706 You can use this as a shortcut to determine whether C<Archive::Tar>
1707 will do what you think before passing compressed archives to its
1712 sub can_handle_compressed_files { return ZLIB && BZIP ? 1 : 0 }
1714 sub no_string_support {
1715 croak("You have to install IO::String to support writing archives to strings");
1722 =head1 GLOBAL VARIABLES
1724 =head2 $Archive::Tar::FOLLOW_SYMLINK
1726 Set this variable to C<1> to make C<Archive::Tar> effectively make a
1727 copy of the file when extracting. Default is C<0>, which
1728 means the symlink stays intact. Of course, you will have to pack the
1729 file linked to as well.
1731 This option is checked when you write out the tarfile using C<write>
1732 or C<create_archive>.
1734 This works just like C</bin/tar>'s C<-h> option.
1736 =head2 $Archive::Tar::CHOWN
1738 By default, C<Archive::Tar> will try to C<chown> your files if it is
1739 able to. In some cases, this may not be desired. In that case, set
1740 this variable to C<0> to disable C<chown>-ing, even if it were
1743 The default is C<1>.
1745 =head2 $Archive::Tar::CHMOD
1747 By default, C<Archive::Tar> will try to C<chmod> your files to
1748 whatever mode was specified for the particular file in the archive.
1749 In some cases, this may not be desired. In that case, set this
1750 variable to C<0> to disable C<chmod>-ing.
1752 The default is C<1>.
1754 =head2 $Archive::Tar::DO_NOT_USE_PREFIX
1756 By default, C<Archive::Tar> will try to put paths that are over
1757 100 characters in the C<prefix> field of your tar header, as
1758 defined per POSIX-standard. However, some (older) tar programs
1759 do not implement this spec. To retain compatibility with these older
1760 or non-POSIX compliant versions, you can set the C<$DO_NOT_USE_PREFIX>
1761 variable to a true value, and C<Archive::Tar> will use an alternate
1762 way of dealing with paths over 100 characters by using the
1763 C<GNU Extended Header> feature.
1765 Note that clients who do not support the C<GNU Extended Header>
1766 feature will not be able to read these archives. Such clients include
1767 tars on C<Solaris>, C<Irix> and C<AIX>.
1769 The default is C<0>.
1771 =head2 $Archive::Tar::DEBUG
1773 Set this variable to C<1> to always get the C<Carp::longmess> output
1774 of the warnings, instead of the regular C<carp>. This is the same
1775 message you would get by doing:
1781 =head2 $Archive::Tar::WARN
1783 Set this variable to C<0> if you do not want any warnings printed.
1784 Personally I recommend against doing this, but people asked for the
1785 option. Also, be advised that this is of course not threadsafe.
1789 =head2 $Archive::Tar::error
1791 Holds the last reported error. Kept for historical reasons, but its
1792 use is very much discouraged. Use the C<error()> method instead:
1794 warn $tar->error unless $tar->extract;
1796 =head2 $Archive::Tar::INSECURE_EXTRACT_MODE
1798 This variable indicates whether C<Archive::Tar> should allow
1799 files to be extracted outside their current working directory.
1801 Allowing this could have security implications, as a malicious
1802 tar archive could alter or replace any file the extracting user
1803 has permissions to. Therefor, the default is to not allow
1804 insecure extractions.
1806 If you trust the archive, or have other reasons to allow the
1807 archive to write files outside your current working directory,
1808 set this variable to C<true>.
1810 Note that this is a backwards incompatible change from version
1813 =head2 $Archive::Tar::HAS_PERLIO
1815 This variable holds a boolean indicating if we currently have
1816 C<perlio> support loaded. This will be enabled for any perl
1817 greater than C<5.8> compiled with C<perlio>.
1819 If you feel strongly about disabling it, set this variable to
1820 C<false>. Note that you will then need C<IO::String> installed
1821 to support writing stringified archives.
1823 Don't change this variable unless you B<really> know what you're
1826 =head2 $Archive::Tar::HAS_IO_STRING
1828 This variable holds a boolean indicating if we currently have
1829 C<IO::String> support loaded. This will be enabled for any perl
1830 that has a loadable C<IO::String> module.
1832 If you feel strongly about disabling it, set this variable to
1833 C<false>. Note that you will then need C<perlio> support from
1834 your perl to be able to write stringified archives.
1836 Don't change this variable unless you B<really> know what you're
1843 =item What's the minimum perl version required to run Archive::Tar?
1845 You will need perl version 5.005_03 or newer.
1847 =item Isn't Archive::Tar slow?
1849 Yes it is. It's pure perl, so it's a lot slower then your C</bin/tar>
1850 However, it's very portable. If speed is an issue, consider using
1851 C</bin/tar> instead.
1853 =item Isn't Archive::Tar heavier on memory than /bin/tar?
1855 Yes it is, see previous answer. Since C<Compress::Zlib> and therefore
1856 C<IO::Zlib> doesn't support C<seek> on their filehandles, there is little
1857 choice but to read the archive into memory.
1858 This is ok if you want to do in-memory manipulation of the archive.
1860 If you just want to extract, use the C<extract_archive> class method
1861 instead. It will optimize and write to disk immediately.
1863 Another option is to use the C<iter> class method to iterate over
1864 the files in the tarball without reading them all in memory at once.
1866 =item Can you lazy-load data instead?
1868 In some cases, yes. You can use the C<iter> class method to iterate
1869 over the files in the tarball without reading them all in memory at once.
1871 =item How much memory will an X kb tar file need?
1873 Probably more than X kb, since it will all be read into memory. If
1874 this is a problem, and you don't need to do in memory manipulation
1875 of the archive, consider using the C<iter> class method, or C</bin/tar>
1878 =item What do you do with unsupported filetypes in an archive?
1880 C<Unix> has a few filetypes that aren't supported on other platforms,
1881 like C<Win32>. If we encounter a C<hardlink> or C<symlink> we'll just
1882 try to make a copy of the original file, rather than throwing an error.
1884 This does require you to read the entire archive in to memory first,
1885 since otherwise we wouldn't know what data to fill the copy with.
1886 (This means that you cannot use the class methods, including C<iter>
1887 on archives that have incompatible filetypes and still expect things
1890 For other filetypes, like C<chardevs> and C<blockdevs> we'll warn that
1891 the extraction of this particular item didn't work.
1893 =item I'm using WinZip, or some other non-POSIX client, and files are not being extracted properly!
1895 By default, C<Archive::Tar> is in a completely POSIX-compatible
1896 mode, which uses the POSIX-specification of C<tar> to store files.
1897 For paths greather than 100 characters, this is done using the
1898 C<POSIX header prefix>. Non-POSIX-compatible clients may not support
1899 this part of the specification, and may only support the C<GNU Extended
1900 Header> functionality. To facilitate those clients, you can set the
1901 C<$Archive::Tar::DO_NOT_USE_PREFIX> variable to C<true>. See the
1902 C<GLOBAL VARIABLES> section for details on this variable.
1904 Note that GNU tar earlier than version 1.14 does not cope well with
1905 the C<POSIX header prefix>. If you use such a version, consider setting
1906 the C<$Archive::Tar::DO_NOT_USE_PREFIX> variable to C<true>.
1908 =item How do I extract only files that have property X from an archive?
1910 Sometimes, you might not wish to extract a complete archive, just
1911 the files that are relevant to you, based on some criteria.
1913 You can do this by filtering a list of C<Archive::Tar::File> objects
1914 based on your criteria. For example, to extract only files that have
1915 the string C<foo> in their title, you would use:
1918 grep { $_->full_path =~ /foo/ } $tar->get_files
1921 This way, you can filter on any attribute of the files in the archive.
1922 Consult the C<Archive::Tar::File> documentation on how to use these
1925 =item How do I access .tar.Z files?
1927 The C<Archive::Tar> module can optionally use C<Compress::Zlib> (via
1928 the C<IO::Zlib> module) to access tar files that have been compressed
1929 with C<gzip>. Unfortunately tar files compressed with the Unix C<compress>
1930 utility cannot be read by C<Compress::Zlib> and so cannot be directly
1931 accesses by C<Archive::Tar>.
1933 If the C<uncompress> or C<gunzip> programs are available, you can use
1934 one of these workarounds to read C<.tar.Z> files from C<Archive::Tar>
1936 Firstly with C<uncompress>
1940 open F, "uncompress -c $filename |";
1941 my $tar = Archive::Tar->new(*F);
1944 and this with C<gunzip>
1948 open F, "gunzip -c $filename |";
1949 my $tar = Archive::Tar->new(*F);
1952 Similarly, if the C<compress> program is available, you can use this to
1953 write a C<.tar.Z> file
1958 my $fh = new IO::File "| compress -c >$filename";
1959 my $tar = Archive::Tar->new();
1964 =item How do I handle Unicode strings?
1966 C<Archive::Tar> uses byte semantics for any files it reads from or writes
1967 to disk. This is not a problem if you only deal with files and never
1968 look at their content or work solely with byte strings. But if you use
1969 Unicode strings with character semantics, some additional steps need
1972 For example, if you add a Unicode string like
1975 $tar->add_data('file.txt', "Euro: \x{20AC}");
1977 then there will be a problem later when the tarfile gets written out
1978 to disk via C<$tar->write()>:
1980 Wide character in print at .../Archive/Tar.pm line 1014.
1982 The data was added as a Unicode string and when writing it out to disk,
1983 the C<:utf8> line discipline wasn't set by C<Archive::Tar>, so Perl
1984 tried to convert the string to ISO-8859 and failed. The written file
1985 now contains garbage.
1987 For this reason, Unicode strings need to be converted to UTF-8-encoded
1988 bytestrings before they are handed off to C<add_data()>:
1991 my $data = "Accented character: \x{20AC}";
1992 $data = encode('utf8', $data);
1994 $tar->add_data('file.txt', $data);
1996 A opposite problem occurs if you extract a UTF8-encoded file from a
1997 tarball. Using C<get_content()> on the C<Archive::Tar::File> object
1998 will return its content as a bytestring, not as a Unicode string.
2000 If you want it to be a Unicode string (because you want character
2001 semantics with operations like regular expression matching), you need
2002 to decode the UTF8-encoded content and have Perl convert it into
2006 my $data = $tar->get_content();
2008 # Make it a Unicode string
2009 $data = decode('utf8', $data);
2011 There is no easy way to provide this functionality in C<Archive::Tar>,
2012 because a tarball can contain many files, and each of which could be
2013 encoded in a different way.
2021 =item Check if passed in handles are open for read/write
2023 Currently I don't know of any portable pure perl way to do this.
2024 Suggestions welcome.
2026 =item Allow archives to be passed in as string
2028 Currently, we only allow opened filehandles or filenames, but
2029 not strings. The internals would need some reworking to facilitate
2030 stringified archives.
2032 =item Facilitate processing an opened filehandle of a compressed archive
2034 Currently, we only support this if the filehandle is an IO::Zlib object.
2035 Environments, like apache, will present you with an opened filehandle
2036 to an uploaded file, which might be a compressed archive.
2044 =item The GNU tar specification
2046 C<http://www.gnu.org/software/tar/manual/tar.html>
2048 =item The PAX format specication
2050 The specifcation which tar derives from; C< http://www.opengroup.org/onlinepubs/007904975/utilities/pax.html>
2052 =item A comparison of GNU and POSIX tar standards; C<http://www.delorie.com/gnu/docs/tar/tar_114.html>
2054 =item GNU tar intends to switch to POSIX compatibility
2056 GNU Tar authors have expressed their intention to become completely
2057 POSIX-compatible; C<http://www.gnu.org/software/tar/manual/html_node/Formats.html>
2059 =item A Comparison between various tar implementations
2061 Lists known issues and incompatibilities; C<http://gd.tuwien.ac.at/utils/archivers/star/README.otherbugs>
2067 This module by Jos Boumans E<lt>kane@cpan.orgE<gt>.
2069 Please reports bugs to E<lt>bug-archive-tar@rt.cpan.orgE<gt>.
2071 =head1 ACKNOWLEDGEMENTS
2073 Thanks to Sean Burke, Chris Nandor, Chip Salzenberg, Tim Heaney, Gisle Aas
2074 and especially Andrew Savige for their help and suggestions.
2078 This module is copyright (c) 2002 - 2008 Jos Boumans
2079 E<lt>kane@cpan.orgE<gt>. All rights reserved.
2081 This library is free software; you may redistribute and/or modify
2082 it under the same terms as Perl itself.