path: root/doc/tarlz.texi

\input texinfo @c -*-texinfo-*-
@c %**start of header
@setfilename tarlz.info
@documentencoding ISO-8859-15
@settitle Tarlz Manual
@finalout
@c %**end of header

@set UPDATED 3 January 2024
@set VERSION 0.25

@dircategory Archiving
@direntry
* Tarlz: (tarlz).               Archiver with multimember lzip compression
@end direntry


@ifnothtml
@titlepage
@title Tarlz
@subtitle Archiver with multimember lzip compression
@subtitle for Tarlz version @value{VERSION}, @value{UPDATED}
@author by Antonio Diaz Diaz

@page
@vskip 0pt plus 1filll
@end titlepage

@contents
@end ifnothtml

@ifnottex
@node Top
@top

This manual is for Tarlz (version @value{VERSION}, @value{UPDATED}).

@menu
* Introduction::              Purpose and features of tarlz
* Invoking tarlz::            Command-line interface
* Portable character set::    POSIX portable filename character set
* File format::               Detailed format of the compressed archive
* Amendments to pax format::  The reasons for the differences with pax
* Program design::            Internal structure of tarlz
* Multi-threaded decoding::   Limitations of parallel tar decoding
* Minimum archive sizes::     Sizes required for full multi-threaded speed
* Examples::                  A small tutorial with examples
* Problems::                  Reporting bugs
* Concept index::             Index of concepts
@end menu

@sp 1
Copyright @copyright{} 2013-2024 Antonio Diaz Diaz.

This manual is free documentation: you have unlimited permission to copy,
distribute, and modify it.
@end ifnottex


@node Introduction
@chapter Introduction
@cindex introduction

@uref{http://www.nongnu.org/lzip/tarlz.html,,Tarlz} is a massively parallel
(multi-threaded) combined implementation of the tar archiver and the
@uref{http://www.nongnu.org/lzip/lzip.html,,lzip} compressor. Tarlz uses the
compression library @uref{http://www.nongnu.org/lzip/lzlib.html,,lzlib}.

Tarlz creates tar archives using a simplified and safer variant of the POSIX
pax format compressed in lzip format, keeping the alignment between tar
members and lzip members. The resulting multimember tar.lz archive is
backward compatible with standard tar tools like GNU tar, which treat it
like any other tar.lz archive. Tarlz can append files to the end of such
compressed archives.

Keeping the alignment between tar members and lzip members has two
advantages. It adds an indexed lzip layer on top of the tar archive, making
it possible to decode the archive safely in parallel. It also minimizes the
amount of data lost in case of corruption. Compressing a tar archive with
plzip may even double the amount of files lost for each lzip member damaged
because it does not keep the members aligned.

Tarlz can create tar archives with five levels of compression granularity:
per file (@option{--no-solid}), per block (@option{--bsolid}, default), per
directory (@option{--dsolid}), appendable solid (@option{--asolid}), and
solid (@option{--solid}). It can also create uncompressed tar archives.

@noindent
Of course, compressing each file (or each directory) individually can't
achieve a compression ratio as high as compressing solidly the whole tar
archive, but it has the following advantages:

@itemize @bullet
@item
The resulting multimember tar.lz archive can be decompressed in
parallel, multiplying the decompression speed.

@item
New members can be appended to the archive (by removing the
end-of-archive member), and unwanted members can be deleted from the
archive. Just like an uncompressed tar archive.

@item
It is a safe POSIX-style backup format. In case of corruption, tarlz
can extract all the undamaged members from the tar.lz archive,
skipping over the damaged members, just like the standard
(uncompressed) tar. Moreover, the option @option{--keep-damaged} can be used
to recover as much data as possible from each damaged member, and
lziprecover can be used to recover some of the damaged members.

@item
A multimember tar.lz archive is usually smaller than the corresponding
solidly compressed tar.gz archive, except when individually
compressing files smaller than about @w{32 KiB}.
@end itemize

Tarlz protects the extended records with a Cyclic Redundancy Check (CRC) in
a way compatible with standard tar tools. @xref{crc32}.

Tarlz does not understand other tar formats like @samp{gnu}, @samp{oldgnu},
@samp{star}, or @samp{v7}. The command
@w{@samp{tarlz -t -f archive.tar.lz > /dev/null}} can be used to check that
the format of the archive is compatible with tarlz.


@node Invoking tarlz
@chapter Invoking tarlz
@cindex invoking
@cindex options
@cindex usage
@cindex version

The format for running tarlz is:

@example
tarlz @var{operation} [@var{options}] [@var{files}]
@end example

@noindent
All operations except @option{--concatenate} and @option{--compress} operate
on whole trees if any @var{file} is a directory. All operations except
@option{--compress} overwrite output files without warning. If no archive is
specified, tarlz tries to read it from standard input or write it to
standard output. Tarlz refuses to read archive data from a terminal or write
archive data to a terminal. Tarlz detects when the archive being created or
enlarged is among the files to be archived, appended, or concatenated, and
skips it.

Tarlz does not use absolute file names nor file names above the current
working directory (perhaps changed by option @option{-C}). On archive creation
or appending tarlz archives the files specified, but removes from member
names any leading and trailing slashes and any file name prefixes containing
a @samp{..} component. On extraction, leading and trailing slashes are also
removed from member names, and archive members containing a @samp{..}
component in the file name are skipped. Tarlz does not follow symbolic links
during extraction; not even symbolic links replacing intermediate
directories.

On extraction and listing, tarlz removes leading @samp{./} strings from
member names in the archive or given in the command line, so that
@w{@samp{tarlz -xf foo ./bar baz}} extracts members @samp{bar} and
@samp{./baz} from archive @samp{foo}.

If several compression levels or @option{--*solid} options are given, the last
setting is used. For example @w{@option{-9 --solid --uncompressed -1}} is
equivalent to @w{@option{-1 --solid}}.

tarlz supports the following operations:

@table @code
@item --help
Print an informative help message describing the options and exit.

@item -V
@itemx --version
Print the version number of tarlz on the standard output and exit.
This version number should be included in all bug reports.

@item -A
@itemx --concatenate
Append one or more archives to the end of an archive. If no archive is
specified with the option @option{-f}, concatenate the input archives to
standard output. All the archives involved must be regular (seekable) files,
and must be either all compressed or all uncompressed. Compressed and
uncompressed archives can't be mixed. Compressed archives must be
multimember lzip files with the two end-of-archive blocks plus any zero
padding contained in the last lzip member of each archive. The intermediate
end-of-archive blocks are removed as each new archive is concatenated. If
the archive is uncompressed, tarlz parses tar headers until it finds the
end-of-archive blocks. Exit with status 0 without modifying the archive if
no @var{files} have been specified.

Concatenating archives containing files in common results in two or more tar
members with the same name in the resulting archive, which may produce
nondeterministic behavior during multi-threaded extraction.
@xref{mt-extraction}.

@item -c
@itemx --create
Create a new archive from @var{files}.

@item -d
@itemx --diff
Compare and report differences between archive and file system. For each tar
member in the archive, check that the corresponding file in the file system
exists and is of the same type (regular file, directory, etc). Report on
standard output the differences found in type, mode (permissions), owner and
group IDs, modification time, file size, file contents (of regular files),
target (of symlinks) and device number (of block/character special files).

As tarlz removes leading slashes from member names, the option @option{-C} may
be used in combination with @option{--diff} when absolute file names were used
on archive creation: @w{@samp{tarlz -C / -d}}. Alternatively, tarlz may be
run from the root directory to perform the comparison.

@item --delete
Delete files and directories from an archive in place. It currently can
delete only from uncompressed archives and from archives with files
compressed individually (@option{--no-solid} archives). Note that files of
about @option{--data-size} or larger are compressed individually even if
@option{--bsolid} is used, and can therefore be deleted. Tarlz takes care to
not delete a tar member unless it is possible to do so. For example it won't
try to delete a tar member that is not compressed individually. Even in the
case of finding a corrupt member after having deleted some member(s), tarlz
stops and copies the rest of the file as soon as corruption is found,
leaving it just as corrupt as it was, but not worse.

To delete a directory without deleting the files under it, use
@w{@samp{tarlz --delete -f foo --exclude='dir/*' dir}}. Deleting in place
may be dangerous. A corrupt archive, a power cut, or an I/O error may cause
data loss.

@item -r
@itemx --append
Append files to the end of an archive. The archive must be a regular
(seekable) file either compressed or uncompressed. Compressed members can't
be appended to an uncompressed archive, nor vice versa. If the archive is
compressed, it must be a multimember lzip file with the two end-of-archive
blocks plus any zero padding contained in the last lzip member of the
archive. It is possible to append files to an archive with a different
compression granularity. Appending works as follows; first the
end-of-archive blocks are removed, then the new members are appended, and
finally two new end-of-archive blocks are appended to the archive. If the
archive is uncompressed, tarlz parses and skips tar headers until it finds
the end-of-archive blocks. Exit with status 0 without modifying the archive
if no @var{files} have been specified.

Appending files already present in the archive results in two or more tar
members with the same name, which may produce nondeterministic behavior
during multi-threaded extraction. @xref{mt-extraction}.

@item -t
@itemx --list
List the contents of an archive. If @var{files} are given, list only the
@var{files} given.

@item -x
@itemx --extract
Extract files from an archive. If @var{files} are given, extract only the
@var{files} given. Else extract all the files in the archive. To extract a
directory without extracting the files under it, use
@w{@samp{tarlz -xf foo --exclude='dir/*' dir}}. Tarlz removes files and
empty directories unconditionally before extracting over them. Other than
that, it does not make any special effort to extract a file over an
incompatible type of file. For example, extracting a file over a non-empty
directory usually fails.

@item -z
@itemx --compress
Compress existing POSIX tar archives aligning the lzip members to the tar
members with choice of granularity (@option{--bsolid} by default,
@option{--dsolid} works like @option{--asolid}). Exit with error status 2 if
any input archive is an empty file. The input archives are kept unchanged.
Existing compressed archives are not overwritten. A hyphen @samp{-} used as
the name of an input archive reads from standard input and writes to
standard output (unless the option @option{--output} is used). Tarlz can be
used as compressor for GNU tar by using a command like
@w{@samp{tar -c -Hustar foo | tarlz -z -o foo.tar.lz}}. Tarlz can be used as
compressor for zupdate (zutils) by using a command like
@w{@samp{zupdate --lz="tarlz -z" foo.tar.gz}}. Note that tarlz only works
reliably on archives without global headers, or with global headers whose
content can be ignored.

The compression is reversible, including any garbage present after the
end-of-archive blocks. Tarlz stops parsing after the first end-of-archive
block is found, and then compresses the rest of the archive. Unless solid
compression is requested, the end-of-archive blocks are compressed in a lzip
member separated from the preceding members and from any non-zero garbage
following the end-of-archive blocks. @option{--compress} implies plzip
argument style, not tar style. Each input archive is compressed to a file
with the extension @samp{.lz} added unless the option @option{--output} is
used. When @option{--output} is used, only one input archive can be specified.
@option{-f} can't be used with @option{--compress}.

@item --check-lib
Compare the
@uref{http://www.nongnu.org/lzip/manual/lzlib_manual.html#Library-version,,version of lzlib}
used to compile tarlz with the version actually being used at run time and
exit. Report any differences found. Exit with error status 1 if differences
are found. A mismatch may indicate that lzlib is not correctly installed or
that a different version of lzlib has been installed after compiling tarlz.
Exit with error status 2 if LZ_API_VERSION and LZ_version_string don't
match. @w{@samp{tarlz -v --check-lib}} shows the version of lzlib being used
and the value of LZ_API_VERSION (if defined).
@ifnothtml
@xref{Library version,,,lzlib}.
@end ifnothtml

@end table

tarlz supports the following
@uref{http://www.nongnu.org/arg-parser/manual/arg_parser_manual.html#Argument-syntax,,options}:
@ifnothtml
@xref{Argument syntax,,,arg_parser}.
@end ifnothtml

@table @code
@anchor{--data-size}
@item -B @var{bytes}
@itemx --data-size=@var{bytes}
Set target size of input data blocks for the option @option{--bsolid}.
@xref{--bsolid}. Valid values range from @w{8 KiB} to @w{1 GiB}. Default
value is two times the dictionary size, except for option @option{-0} where it
defaults to @w{1 MiB}. @xref{Minimum archive sizes}.

@item -C @var{dir}
@itemx --directory=@var{dir}
Change to directory @var{dir}. When creating, appending, comparing, or
extracting, the position of each @option{-C} option in the command line is
significant; it changes the current working directory for the following
@var{files} until a new @option{-C} option appears in the command line.
@option{--list} and @option{--delete} ignore any @option{-C} options
specified. @var{dir} is relative to the then current working directory,
perhaps changed by a previous @option{-C} option.

Note that a process can only have one current working directory (CWD).
Therefore multi-threading can't be used to create or decode an archive if a
@option{-C} option appears after a (relative) file name in the command line.
(All file names are made relative when decoding).

@item -f @var{archive}
@itemx --file=@var{archive}
Use archive file @var{archive}. A hyphen @samp{-} used as an @var{archive}
argument reads from standard input or writes to standard output.

@item -h
@itemx --dereference
Follow symbolic links during archive creation, appending or comparison.
Archive or compare the files they point to instead of the links themselves.

@item -n @var{n}
@itemx --threads=@var{n}
Set the number of (de)compression threads, overriding the system's default.
Valid values range from 0 to "as many as your system can support". A value
of 0 disables threads entirely. If this option is not used, tarlz tries to
detect the number of processors in the system and use it as default value.
@w{@samp{tarlz --help}} shows the system's default value. See the note about
multi-threading in the option @option{-C} above.

Note that the number of usable threads is limited during compression to
@w{ceil( uncompressed_size / data_size )} (@pxref{Minimum archive sizes}),
and during decompression to the number of lzip members in the tar.lz
archive, which you can find by running @w{@samp{lzip -lv archive.tar.lz}}.

@item -o @var{file}
@itemx --output=@var{file}
Write the compressed output to @var{file}. @w{@option{-o -}} writes the
compressed output to standard output. Currently @option{--output} only works
with @option{--compress}.

@item -p
@itemx --preserve-permissions
On extraction, set file permissions as they appear in the archive. This is
the default behavior when tarlz is run by the superuser. The default for
other users is to subtract the umask of the user running tarlz from the
permissions specified in the archive.

@item -q
@itemx --quiet
Quiet operation. Suppress all messages.

@item -v
@itemx --verbose
Verbosely list files processed. Further -v's (up to 4) increase the
verbosity level.

@item -0 .. -9
Set the compression level for @option{--create}, @option{--append}, and
@option{--compress}. The default compression level is @option{-6}. Like lzip,
tarlz also minimizes the dictionary size of the lzip members it creates,
reducing the amount of memory required for decompression.

@multitable {Level} {Dictionary size} {Match length limit}
@item Level @tab Dictionary size @tab Match length limit
@item -0 @tab 64 KiB @tab  16 bytes
@item -1 @tab  1 MiB @tab   5 bytes
@item -2 @tab  1.5 MiB @tab   6 bytes
@item -3 @tab  2 MiB @tab   8 bytes
@item -4 @tab  3 MiB @tab  12 bytes
@item -5 @tab  4 MiB @tab  20 bytes
@item -6 @tab  8 MiB @tab  36 bytes
@item -7 @tab 16 MiB @tab  68 bytes
@item -8 @tab 24 MiB @tab 132 bytes
@item -9 @tab 32 MiB @tab 273 bytes
@end multitable

@item --uncompressed
With @option{--create}, don't compress the tar archive created. Create an
uncompressed tar archive instead. With @option{--append}, don't compress the
new members appended to the tar archive. Compressed members can't be
appended to an uncompressed archive, nor vice versa. @option{--uncompressed}
can be omitted if it can be deduced from the archive name. (An uncompressed
archive name lacks a @samp{.lz} or @samp{.tlz} extension).

@item --asolid
When creating or appending to a compressed archive, use appendable solid
compression. All the files being added to the archive are compressed into a
single lzip member, but the end-of-archive blocks are compressed into a
separate lzip member. This creates a solidly compressed appendable archive.
Solid archives can't be created nor decoded in parallel.

@anchor{--bsolid}
@item --bsolid
When creating or appending to a compressed archive, use block compression.
Tar members are compressed together in a lzip member until they approximate
a target uncompressed size. The size can't be exact because each solidly
compressed data block must contain an integer number of tar members. Block
compression is the default because it improves compression ratio for
archives with many files smaller than the block size. This option allows
tarlz revert to default behavior if, for example, it is invoked through an
alias like @w{@samp{tar='tarlz --solid'}}. @xref{--data-size}, to set the
target block size.

@item --dsolid
When creating or appending to a compressed archive, compress each file
specified in the command line separately in its own lzip member, and use
solid compression for each directory specified in the command line. The
end-of-archive blocks are compressed into a separate lzip member. This
creates a compressed appendable archive with a separate lzip member for each
file or top-level directory specified.

@item --no-solid
When creating or appending to a compressed archive, compress each file
separately in its own lzip member. The end-of-archive blocks are compressed
into a separate lzip member. This creates a compressed appendable archive
with a lzip member for each file.

@item --solid
When creating or appending to a compressed archive, use solid compression.
The files being added to the archive, along with the end-of-archive blocks,
are compressed into a single lzip member. The resulting archive is not
appendable. No more files can be later appended to the archive. Solid
archives can't be created nor decoded in parallel.

@item --anonymous
Equivalent to @w{@option{--owner=root --group=root}}.

@item --owner=@var{owner}
When creating or appending, use @var{owner} for files added to the archive.
If @var{owner} is not a valid user name, it is decoded as a decimal numeric
user ID.

@item --group=@var{group}
When creating or appending, use @var{group} for files added to the archive.
If @var{group} is not a valid group name, it is decoded as a decimal numeric
group ID.

@item --exclude=@var{pattern}
Exclude files matching a shell pattern like @samp{*.o}. A file is considered
to match if any component of the file name matches. For example, @samp{*.o}
matches @samp{foo.o}, @samp{foo.o/bar} and @samp{foo/bar.o}. If
@var{pattern} contains a @samp{/}, it matches a corresponding @samp{/} in
the file name. For example, @samp{foo/*.o} matches @samp{foo/bar.o}.
Multiple @option{--exclude} options can be specified.

@item --ignore-ids
Make @option{--diff} ignore differences in owner and group IDs. This option is
useful when comparing an @option{--anonymous} archive.

@item --ignore-metadata
Make @option{--diff} ignore any differences in metadata (file permissions,
owner and group IDs, modification time). Compare only file type, file size,
and file content. This option is useful when file permissions have not been
fully restored because uid/gid changed on extraction.

@item --ignore-overflow
Make @option{--diff} ignore differences in mtime caused by overflow on 32-bit
systems with a 32-bit time_t.

@item --keep-damaged
Don't delete partially extracted files. If a decompression error happens
while extracting a file, keep the partial data extracted. Use this option to
recover as much data as possible from each damaged member. It is recommended
to run tarlz in single-threaded mode (@option{--threads=0}) when using this
option.

@item --missing-crc
Exit with error status 2 if the CRC of the extended records is missing. When
this option is used, tarlz detects any corruption in the extended records
(only limited by CRC collisions). But note that a corrupt @samp{GNU.crc32}
keyword, for example @samp{GNU.crc30}, is reported as a missing CRC instead
of as a corrupt record. This misleading @w{@samp{Missing CRC}} message is
the consequence of a flaw in the POSIX pax format; i.e., the lack of a
mandatory check sequence of the extended records. @xref{crc32}.

@item --mtime=@var{date}
When creating or appending, use @var{date} as the modification time for
files added to the archive instead of their actual modification times. The
value of @var{date} may be either @samp{@@} followed by the number of
seconds since (or before) the epoch, or a date in format
@w{@samp{[-]YYYY-MM-DD HH:MM:SS}} or @samp{[-]YYYY-MM-DDTHH:MM:SS}, or the
name of an existing reference file starting with @samp{.} or @samp{/} whose
modification time is used. The time of day @samp{HH:MM:SS} in the date
format is optional and defaults to @samp{00:00:00}. The epoch is
@w{@samp{1970-01-01 00:00:00 UTC}}. Negative seconds or years define a
modification time before the epoch.

@item --out-slots=@var{n}
Number of @w{1 MiB} output packets buffered per worker thread during
multi-threaded creation or appending to compressed archives. Increasing the
number of packets may increase compression speed if the files being archived
are larger than @w{64 MiB} compressed, but requires more memory. Valid
values range from 1 to 1024. The default value is 64.

@item --warn-newer
During archive creation, warn if any file being archived has a modification
time newer than the archive creation time. This option may slow archive
creation somewhat because it makes an extra call to @samp{stat} after
archiving each file, but it guarantees that file contents were not modified
during the creation of the archive. Note that the file must be at least one
second newer than the archive for it to be detected as newer.

@ignore
@item --permissive
Allow some violations of the archive format, like consecutive extended
headers preceding a ustar header, or several records with the same
keyword appearing in the same block of extended records.
@end ignore

@end table

Exit status: 0 for a normal exit, 1 for environmental problems
(file not found, files differ, invalid command-line options, I/O errors,
etc), 2 to indicate a corrupt or invalid input file, 3 for an internal
consistency error (e.g., bug) which caused tarlz to panic.


@node Portable character set
@chapter POSIX portable filename character set
@cindex portable character set

The set of characters from which portable file names are constructed.

@example
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
a b c d e f g h i j k l m n o p q r s t u v w x y z
0 1 2 3 4 5 6 7 8 9 . _ -
@end example

The last three characters are the period, underscore, and hyphen-minus
characters, respectively.

File names are identifiers. Therefore, archiving works better when file
names use only the portable character set without spaces added.


@node File format
@chapter File format
@cindex file format

In the diagram below, a box like this:

@verbatim
+---+
|   | <-- the vertical bars might be missing
+---+
@end verbatim

represents one byte; a box like this:

@verbatim
+==============+
|              |
+==============+
@end verbatim

represents a variable number of bytes or a fixed but large number of
bytes (for example 512).

@sp 1
A tar.lz file consists of one or more lzip members (compressed data sets).
The members simply appear one after another in the file, with no additional
information before, between, or after them.

Each lzip member contains one or more tar members in a simplified POSIX pax
interchange format. The only pax typeflag value supported by tarlz (in
addition to the typeflag values defined by the ustar format) is @samp{x}.
The pax format is an extension on top of the ustar format that removes the
size limitations of the ustar format.

Each tar member contains one file archived, and is represented by the
following sequence:

@itemize @bullet
@item
An optional extended header block followed by one or more blocks that
contain the extended header records as if they were the contents of a file;
i.e., the extended header records are included as the data for this header
block. This header block is of the form described in pax header block, with
a typeflag value of @samp{x}.

@item
A header block in ustar format that describes the file. Any fields defined
in the preceding optional extended header records override the associated
fields in this header block for this file.

@item
Zero or more blocks that contain the contents of the file.
@end itemize

Each tar member must be contiguously stored in a lzip member for the
parallel decoding operations like @option{--list} to work. If any tar member
is split over two or more lzip members, the archive must be decoded
sequentially. @xref{Multi-threaded decoding}.

At the end of the archive file there are two 512-byte blocks filled with
binary zeros, interpreted as an end-of-archive indicator. These EOA blocks
are either compressed in a separate lzip member or compressed along with the
tar members contained in the last lzip member. For a compressed archive to
be recognized by tarlz as appendable, the last lzip member must contain
between 512 and 32256 zeros alone (without any non-zero bytes).

The diagram below shows the correspondence between each tar member (formed
by one or two headers plus optional data) in the tar archive and each
@uref{http://www.nongnu.org/lzip/manual/lzip_manual.html#File-format,,lzip member}
in the resulting multimember tar.lz archive, when per file compression is
used:
@ifnothtml
@xref{File format,,,lzip}.
@end ifnothtml

@verbatim
tar
+========+======+=================+===============+========+======+========+
| header | data | extended header | extended data | header | data |   EOA  |
+========+======+=================+===============+========+======+========+

tar.lz
+===============+=================================================+========+
|     member    |                      member                     | member |
+===============+=================================================+========+
@end verbatim

@ignore
When @option{--permissive} is used, the following violations of the
archive format are allowed:@*
If several extended headers precede an ustar header, only the last
extended header takes effect. The other extended headers are ignored.
Similarly, if several records with the same keyword appear in the same
block of extended records, only the last record for the repeated keyword
takes effect. The other records for the repeated keyword are ignored.@*
A global header inserted between an extended header and an ustar header.@*
An extended header just before the end-of-archive blocks.
@end ignore

@sp 1
@section Pax header block

The pax header block is identical to the ustar header block described below
except that the typeflag has the value @samp{x} (extended). The field
@samp{size} is the size of the extended header data in bytes. Most other
fields in the pax header block are zeroed on archive creation to prevent
trouble if the archive is read by an ustar tool, and are ignored by tarlz on
archive extraction. @xref{flawed-compat}.

Tarlz limits the size of the pax extended header data so that the whole
header set (extended header + extended data + ustar header) can be read and
decoded in a buffer of size INT_MAX.

The pax extended header data consists of one or more records, each of
them constructed as follows:@*
@w{@samp{"%d %s=%s\n", <length>, <keyword>, <value>}}

The fields <length> and <keyword> in the record must be limited to the
portable character set (@pxref{Portable character set}). The field <length>
contains the decimal length of the record in bytes, including the trailing
newline. The field <value> is stored as-is, without conversion to UTF-8 nor
any other transformation. The fields are separated by the ASCII characters
space, equal-sign, and newline.

These are the <keyword> values currently supported by tarlz:

@table @code
@item atime
The signed decimal representation of the access time of the following file
in seconds since (or before) the epoch, obtained from the function
@samp{stat}. The atime record is created only for files with a modification
time outside of the ustar range. @xref{ustar-mtime}.

@item gid
The unsigned decimal representation of the group ID of the group that owns
the following file. The gid record is created only for files with a group ID
greater than 2_097_151 @w{(octal 7_777_777)}. @xref{ustar-uid-gid}.

@item linkpath
The file name of a link being created to another file, of any type,
previously archived. This record overrides the field @samp{linkname} in the
following ustar header block. The following ustar header block determines
the type of link created. If typeflag of the following header block is 1, a
hard link is created. If typeflag is 2, a symbolic link is created and the
linkpath value is used as the contents of the symbolic link. The linkpath
record is created only for links with a link name that does not fit in the
space provided by the ustar header.

@item mtime
The signed decimal representation of the modification time of the following
file in seconds since (or before) the epoch, obtained from the function
@samp{stat}. This record overrides the field @samp{mtime} in the following
ustar header block. The mtime record is created only for files with a
modification time outside of the ustar range. @xref{ustar-mtime}.

@item path
The file name of the following file. This record overrides the fields
@samp{name} and @samp{prefix} in the following ustar header block. The path
record is created for files with a name that does not fit in the space
provided by the ustar header, but is also created for files that require any
other extended record so that the fields @samp{name} and @samp{prefix} in
the following ustar header block can be zeroed.

@item size
The size of the file in bytes, expressed as a decimal number using digits
from the ISO/IEC 646:1991 (ASCII) standard. This record overrides the field
@samp{size} in the following ustar header block. The size record is created
only for files with a size value greater than 8_589_934_591
@w{(octal 77_777_777_777)}; that is, @w{8 GiB} (2^33 bytes) or larger.

@item uid
The unsigned decimal representation of the user ID of the file owner of the
following file. The uid record is created only for files with a user ID
greater than 2_097_151 @w{(octal 7_777_777)}. @xref{ustar-uid-gid}.

@anchor{key_crc32}
@item GNU.crc32
CRC32-C (Castagnoli) of the extended header data excluding the 8 bytes
representing the CRC <value> itself. The <value> is represented as 8
hexadecimal digits in big endian order,
@w{@samp{22 GNU.crc32=00000000\n}}. The keyword of the CRC record is
protected by the CRC to guarantee that corruption is always detected when
using @option{--missing-crc} (except in case of CRC collision). A CRC was
chosen because a checksum is too weak for a potentially large list of
variable sized records. A checksum can't detect simple errors like the
swapping of two bytes.

@end table

At verbosity level 1 or higher tarlz prints a diagnostic for each unknown
extended header keyword found in an archive, once per keyword.

@sp 1
@section Ustar header block

The ustar header block has a length of 512 bytes and is structured as
shown in the following table. All lengths and offsets are in decimal.

@multitable {Field Name} {Offset} {Length (in bytes)}
@item Field Name @tab Offset @tab Length (in bytes)
@item name     @tab   0 @tab 100
@item mode     @tab 100 @tab   8
@item uid      @tab 108 @tab   8
@item gid      @tab 116 @tab   8
@item size     @tab 124 @tab  12
@item mtime    @tab 136 @tab  12
@item chksum   @tab 148 @tab   8
@item typeflag @tab 156 @tab   1
@item linkname @tab 157 @tab 100
@item magic    @tab 257 @tab   6
@item version  @tab 263 @tab   2
@item uname    @tab 265 @tab  32
@item gname    @tab 297 @tab  32
@item devmajor @tab 329 @tab   8
@item devminor @tab 337 @tab   8
@item prefix   @tab 345 @tab 155
@end multitable

All characters in the header block are coded using the ISO/IEC 646:1991
(ASCII) standard, except in fields storing names for files, users, and
groups. For maximum portability between implementations, names should only
contain characters from the portable character set (@pxref{Portable
character set}), but if an implementation supports the use of characters
outside of @samp{/} and the portable character set in names for files,
users, and groups, tarlz will use the byte values in these names unmodified.

The fields @samp{name}, @samp{linkname}, and @samp{prefix} are
null-terminated character strings except when all characters in the array
contain non-null characters including the last character.

The fields @samp{name} and @samp{prefix} produce the file name. A new file
name is formed, if prefix is not an empty string (its first character is not
null), by concatenating prefix (up to the first null character), a slash
character, and name; otherwise, name is used alone. In either case, name is
terminated at the first null character. If prefix begins with a null
character, it is ignored. In this manner, file names of at most 256
characters can be supported. If a file name does not fit in the space
provided, an extended record is used to store the file name.

The field @samp{linkname} does not use the prefix to produce a file name. If
the link name does not fit in the 100 characters provided, an extended
record is used to store the link name.

The field @samp{mode} provides 12 access permission bits. The following
table shows the symbolic name of each bit and its octal value:

@multitable {Bit Name} {Value} {Bit Name} {Value} {Bit Name} {Value}
@headitem Bit Name @tab Value @tab Bit Name @tab Value @tab Bit Name @tab Value
@item S_ISUID @tab 04000 @tab S_ISGID @tab 02000 @tab S_ISVTX @tab 01000
@item S_IRUSR @tab 00400 @tab S_IWUSR @tab 00200 @tab S_IXUSR @tab 00100
@item S_IRGRP @tab 00040 @tab S_IWGRP @tab 00020 @tab S_IXGRP @tab 00010
@item S_IROTH @tab 00004 @tab S_IWOTH @tab 00002 @tab S_IXOTH @tab 00001
@end multitable

@anchor{ustar-uid-gid}
The fields @samp{uid} and @samp{gid} are the user and group IDs of the owner
and group of the file, respectively. If the file uid or gid are greater than
2_097_151 @w{(octal 7_777_777)}, an extended record is used to store the uid
or gid.

The field @samp{size} contains the octal representation of the size of the
file in bytes. If the field @samp{typeflag} specifies a file of type '0'
(regular file) or '7' (high performance regular file), the number of logical
records following the header is @w{(size / 512)} rounded to the next
integer. For all other values of typeflag, tarlz either sets the size field
to 0 or ignores it, and does not store or expect any logical records
following the header. If the file size is larger than 8_589_934_591 bytes
@w{(octal 77_777_777_777)}, an extended record is used to store the file size.

@anchor{ustar-mtime}
The field @samp{mtime} contains the octal representation of the modification
time of the file at the time it was archived, obtained from the function
@samp{stat}. If the modification time is negative or larger than
8_589_934_591 @w{(octal 77_777_777_777)} seconds since the epoch, an extended
record is used to store the modification time. The ustar range of mtime goes
from @w{@samp{1970-01-01 00:00:00 UTC}} to @w{@samp{2242-03-16 12:56:31 UTC}}.

The field @samp{chksum} contains the octal representation of the value of
the simple sum of all bytes in the header logical record. Each byte in the
header is treated as an unsigned value. When calculating the checksum, the
chksum field is treated as if it were all space characters.

The field @samp{typeflag} contains a single character specifying the type of
file archived:

@table @code
@item '0'
Regular file.

@item '1'
Hard link to another file, of any type, previously archived. Hard links must
not contain file data.

@item '2'
Symbolic link.

@item '3', '4'
Character special file and block special file respectively. In this case the
fields @samp{devmajor} and @samp{devminor} contain information defining the
device in unspecified format.

@item '5'
Directory.

@item '6'
FIFO special file.

@item '7'
Reserved to represent a file to which an implementation has associated some
high-performance attribute (contiguous file). Tarlz treats this type of file
as a regular file (type 0).

@end table

The field @samp{magic} contains the ASCII null-terminated string "ustar".
The field @samp{version} contains the characters "00" (0x30,0x30). The
fields @samp{uname} and @samp{gname} are null-terminated character strings
except when all characters in the array contain non-null characters
including the last character. Each numeric field contains a leading space-
or zero-filled, optionally null-terminated octal number using digits from
the ISO/IEC 646:1991 (ASCII) standard. Tarlz is able to decode numeric
fields 1 byte longer than standard ustar by not requiring a terminating null
character.


@node Amendments to pax format
@chapter The reasons for the differences with pax
@cindex Amendments to pax format

Tarlz creates safe archives that allow the reliable detection of invalid or
corrupt metadata during decoding even when the integrity checking of lzip
can't be used because the lzip members are only decompressed partially, as
it happens in parallel @option{--diff}, @option{--list}, and @option{--extract}.
In order to achieve this goal and avoid some other flaws in the pax format,
tarlz makes some changes to the variant of the pax format that it uses. This
chapter describes these changes and the concrete reasons to implement them.

@sp 1
@anchor{crc32}
@section Add a CRC of the extended records

The POSIX pax format has a serious flaw. The metadata stored in pax extended
records are not protected by any kind of check sequence. Corruption in a
long file name may cause the extraction of the file in the wrong place
without warning. Corruption in a large file size may cause the truncation of
the file or the appending of garbage to the file, both followed by a
spurious warning about a corrupt header far from the place of the undetected
corruption.

Metadata like file name and file size must be always protected in an archive
format because of the adverse effects of undetected corruption in them,
potentially much worse that undetected corruption in the data. Even more so
in the case of pax because the amount of metadata it stores is potentially
large, making undetected corruption and archiver misbehavior more probable.

Headers and metadata must be protected separately from data because the
integrity checking of lzip may not be able to detect the corruption before
the metadata have been used, for example, to create a new file in the wrong
place.

Because of the above, tarlz protects the extended records with a Cyclic
Redundancy Check (CRC) in a way compatible with standard tar tools.
@xref{key_crc32}.

@sp 1
@anchor{flawed-compat}
@section Remove flawed backward compatibility

In order to allow the extraction of pax archives by a tar utility conforming
to the POSIX-2:1993 standard, POSIX.1-2008 recommends selecting extended
header field values that allow such tar to create a regular file containing
the extended header records as data. This approach is broken because if the
extended header is needed because of a long file name, the fields
@samp{name} and @samp{prefix} are unable to contain the full file name.
(Some tar implementations store the truncated name in the field @samp{name}
alone, truncating the name to only 100 bytes instead of 256). Therefore the
files corresponding to both the extended header and the overridden ustar
header are extracted using truncated file names, perhaps overwriting
existing files or directories. It may be a security risk to extract a file
with a truncated file name.

To avoid this problem, tarlz writes extended headers with all fields zeroed
except @samp{size} (which contains the size of the extended records),
@samp{chksum}, @samp{typeflag}, @samp{magic}, and @samp{version}. In
particular, tarlz sets the fields @samp{name} and @samp{prefix} to zero.
This prevents old tar programs from extracting the extended records as a
file in the wrong place. Tarlz also sets to zero those fields of the ustar
header overridden by extended records. Finally, tarlz skips members with
zeroed @samp{name} and @samp{prefix} when decoding, except when listing.
This is needed to detect certain format violations during parallel
extraction.

If an extended header is required for any reason (for example a file size of
@w{8 GiB} or larger, or a link name longer than 100 bytes), tarlz also moves
the file name to the extended records to prevent an ustar tool from trying
to extract the file or link. This also makes easier during parallel decoding
the detection of a tar member split between two lzip members at the boundary
between the extended header and the ustar header.

@sp 1
@section As simple as possible (but not simpler)

The tarlz format is mainly ustar. Extended pax headers are used only when
needed because the length of a file name or link name, or the size or other
attribute of a file exceed the limits of the ustar format. Adding @w{1 KiB}
of extended header and records to each member just to save subsecond
timestamps seems wasteful for a backup format. Moreover, minimizing the
overhead may help recovering the archive with lziprecover in case of
corruption.

Global pax headers are tolerated, but not supported; they are parsed and
ignored. Some operations may not behave as expected if the archive contains
global headers.

@sp 1
@section Improve reproducibility

Pax includes by default the process ID of the pax process in the ustar name
of the extended headers, making the archive not reproducible. Tarlz stores
the true name of the file just once, either in the ustar header or in the
extended records, making it easier to produce reproducible archives.

Pax allows an extended record to have length x-1 or x if x is a power of
ten; @samp{99<97_bytes>} or @samp{100<97_bytes>}. Tarlz minimizes the length
of the record and always produces a length of x-1 in these cases.

@sp 1
@section No data in hard links

Tarlz does not allow data in hard link members. The data (if any) must be in
the member determining the type of the file (which can't be a link). If all
the names of a file are stored as hard links, the type of the file is lost.
Not allowing data in hard links also prevents invalid actions like
extracting file data for a hard link to a symbolic link or to a directory.

@sp 1
@section Avoid misconversions to/from UTF-8

There is no portable way to tell what charset a text string is coded into.
Therefore, tarlz stores all fields representing text strings unmodified,
without conversion to UTF-8 nor any other transformation. This prevents
accidental double UTF-8 conversions. If the need arises this behavior will
be adjusted with a command-line option in the future.


@node Program design
@chapter Internal structure of tarlz
@cindex program design

The parts of tarlz related to sequential processing of the archive are more
or less similar to any other tar and won't be described here. The interesting
parts described here are those related to Multi-threaded processing.

The structure of the part of tarlz performing Multi-threaded archive
creation is somewhat similar to that of
@uref{http://www.nongnu.org/lzip/plzip.html#Program-design,,plzip} with the
added complication of the solidity levels.
@ifnothtml
@xref{Program design,,,plzip}.
@end ifnothtml
A grouper thread and several worker threads are created, acting the main
thread as muxer (multiplexer) thread. A "packet courier" takes care of data
transfers among threads and limits the maximum number of data blocks
(packets) being processed simultaneously.

The grouper traverses the directory tree, groups together the metadata of
the files to be archived in each lzip member, and distributes them to the
workers. The workers compress the metadata received from the grouper along
with the file data read from the file system. The muxer collects processed
packets from the workers, and writes them to the archive.

@verbatim
.--------.
|    data|---> to each worker below
|        |                    .------------.
| file   |                ,-->| worker   0 |--,
| system |                |   `------------'  |
|        |    .---------. |   .------------.  |   .-------.   .---------.
|metadata|--->| grouper |-+-->| worker   1 |--+-->| muxer |-->| archive |
`--------'    `---------' |   `------------'  |   `-------'   `---------'
                          |        ...        |
                          |   .------------.  |
                          `-->| worker N-1 |--'
                              `------------'
@end verbatim

Decoding an archive is somewhat similar to how plzip decompresses a regular
file to standard output, with the differences that it is not the data but
only messages what is written to stdout/stderr, and that each worker may
access files in the file system either to read them (diff) or write them
(extract). As in plzip, each worker reads members directly from the archive.

@verbatim
.--------.
| file   |<---> data to/from each worker below
| system |
`--------'      .------------.
            ,-->| worker   0 |--,
            |   `------------'  |
.---------. |   .------------.  |   .-------.   .--------.
| archive |-+-->| worker   1 |--+-->| muxer |-->| stdout |
`---------' |   `------------'  |   `-------'   | stderr |
            |        ...        |               `--------'
            |   .------------.  |
            `-->| worker N-1 |--'
                `------------'
@end verbatim

As misaligned tar.lz archives can't be decoded in parallel, and the
misalignment can't be detected until after decoding has started, a
"mastership request" mechanism has been designed that allows the decoding to
continue instead of signalling an error.

During parallel decoding, if a worker finds a misalignment, it requests
mastership to decode the rest of the archive. When mastership is requested,
an error_member_id is set, and all subsequently received packets with
member_id > error_member_id are rejected. All workers requesting mastership
are blocked at the request_mastership call until mastership is granted.
Mastership is granted to the delivering worker when its queue is empty to
make sure that all preceding packets have been processed. When mastership is
granted, all packets are deleted and all subsequently received packets not
coming from the master are rejected.

If a worker can't continue decoding for any cause (for example lack of
memory or finding a split tar member at the beginning of a lzip member), it
requests mastership to print an error and terminate the program. Only if
some other worker requests mastership in a previous lzip member can this
error be avoided.


@node Multi-threaded decoding
@chapter Limitations of parallel tar decoding
@cindex parallel tar decoding

Safely decoding an arbitrary tar archive in parallel is only possible if one
decodes the headers sequentially first. For example, if a tar archive
containing another tar archive is decoded starting from some position other
than the beginning, there is no way to know if the first header found there
belongs to the outer tar archive or to the inner tar archive. Tar is a
format inherently serial; it was designed for tapes.

The pax format is even more serial than the ustar format. Two headers need
to be decoded sequentially for each file. The extended header may even need
parsing to reveal something as basic as file size. If a thread decodes the
ustar header skipping the preceding extended header, it may extract a file
of incorrect size at the wrong place. Moreover, a pax archive with global
headers can't be decoded in parallel because each thread can't know about
the global headers decoded by other threads.

In the case of compressed tar archives, the start of each compressed block
determines one point through which the tar archive can be decoded in
parallel. Therefore, in tar.lz archives the decoding operations can't be
parallelized if the tar members are not aligned with the lzip members. Tar
archives compressed with plzip can't be decoded in parallel because tar and
plzip do not have a way to align both sets of members. Certainly one can
decompress one such archive with a multi-threaded tool like plzip, but the
increase in speed is not as large as it could be because plzip must
serialize the decompressed data and pass them to tar, which decodes them
sequentially, one tar member at a time.

On the other hand, if the tar.lz archive is created with a tool like tarlz,
which can guarantee the alignment between tar members and lzip members
because it controls both archiving and compression, then the lzip format
becomes an indexed layer on top of the tar archive which makes possible
decoding it safely in parallel.

Tarlz is able to automatically decode aligned and unaligned multimember
tar.lz archives, keeping backwards compatibility. If tarlz finds a member
misalignment during multi-threaded decoding, it switches to single-threaded
mode and continues decoding the archive.

If the files in the archive are large, multi-threaded @option{--list} on a
regular (seekable) tar.lz archive can be hundreds of times faster than
sequential @option{--list} because, in addition to using several processors,
it only needs to decompress part of each lzip member. See the following
example listing the Silesia corpus on a dual core machine:

@example
tarlz -9 --no-solid -cf silesia.tar.lz silesia
time lzip -cd silesia.tar.lz | tar -tf -            (5.032s)
time plzip -cd silesia.tar.lz | tar -tf -           (3.256s)
time tarlz -tf silesia.tar.lz                       (0.020s)
@end example

On the other hand, multi-threaded @option{--list} won't detect corruption in
the tar member data because it only decodes the part of each lzip member
corresponding to the tar member header. This is another reason why the tar
headers must provide their own integrity checking.

@sp 1
@anchor{mt-extraction}
@section Limitations of multi-threaded extraction

Multi-threaded extraction may produce different output than single-threaded
extraction in some cases:

During multi-threaded extraction, several independent threads are
simultaneously reading the archive and creating files in the file system.
The archive is not read sequentially. As a consequence, any error or
weirdness in the archive (like a corrupt member or an end-of-archive block
in the middle of the archive) won't be usually detected until part of the
archive beyond that point has been processed.

If the archive contains two or more tar members with the same name,
single-threaded extraction extracts the members in the order they appear in
the archive and leaves in the file system the last version of the file. But
multi-threaded extraction may extract the members in any order and leave in
the file system any version of the file nondeterministically. It is
unspecified which of the tar members is extracted.

If the same file is extracted through several paths (different member names
resolve to the same file in the file system), the result is undefined.
(Probably the resulting file will be mangled).

Extraction of a hard link may fail if it is extracted before the file it
links to.


@node Minimum archive sizes
@chapter Minimum archive sizes required for multi-threaded block compression
@cindex minimum archive sizes

When creating or appending to a compressed archive using multi-threaded
block compression, tarlz puts tar members together in blocks and compresses
as many blocks simultaneously as worker threads are chosen, creating a
multimember compressed archive.

For this to work as expected (and roughly multiply the compression speed by
the number of available processors), the uncompressed archive must be at
least as large as the number of worker threads times the block size
(@pxref{--data-size}). Else some processors do not get any data to compress,
and compression is proportionally slower. The maximum speed increase
achievable on a given archive is limited by the ratio
@w{(uncompressed_size / data_size)}. For example, a tarball the size of gcc
or linux scales up to 10 or 14 processors at level -9.

The following table shows the minimum uncompressed archive size needed for
full use of N processors at a given compression level, using the default
data size for each level:

@multitable {Processors} {512 MiB} {512 MiB} {512 MiB} {512 MiB} {512 MiB} {512 MiB}
@headitem Processors @tab 2 @tab 4 @tab 8 @tab 16 @tab 64 @tab 256
@item Level
@item -0 @tab   2 MiB @tab   4 MiB @tab   8 MiB @tab  16 MiB @tab  64 MiB @tab 256 MiB
@item -1 @tab   4 MiB @tab   8 MiB @tab  16 MiB @tab  32 MiB @tab 128 MiB @tab 512 MiB
@item -2 @tab   6 MiB @tab  12 MiB @tab  24 MiB @tab  48 MiB @tab 192 MiB @tab 768 MiB
@item -3 @tab   8 MiB @tab  16 MiB @tab  32 MiB @tab  64 MiB @tab 256 MiB @tab   1 GiB
@item -4 @tab  12 MiB @tab  24 MiB @tab  48 MiB @tab  96 MiB @tab 384 MiB @tab 1.5 GiB
@item -5 @tab  16 MiB @tab  32 MiB @tab  64 MiB @tab 128 MiB @tab 512 MiB @tab   2 GiB
@item -6 @tab  32 MiB @tab  64 MiB @tab 128 MiB @tab 256 MiB @tab   1 GiB @tab   4 GiB
@item -7 @tab  64 MiB @tab 128 MiB @tab 256 MiB @tab 512 MiB @tab   2 GiB @tab   8 GiB
@item -8 @tab  96 MiB @tab 192 MiB @tab 384 MiB @tab 768 MiB @tab   3 GiB @tab  12 GiB
@item -9 @tab 128 MiB @tab 256 MiB @tab 512 MiB @tab   1 GiB @tab   4 GiB @tab  16 GiB
@end multitable


@node Examples
@chapter A small tutorial with examples
@cindex examples

@noindent
Example 1: Create a multimember compressed archive @samp{archive.tar.lz}
containing files @samp{a}, @samp{b} and @samp{c}.

@example
tarlz -cf archive.tar.lz a b c
@end example

@sp 1
@noindent
Example 2: Append files @samp{d} and @samp{e} to the multimember compressed
archive @samp{archive.tar.lz}.

@example
tarlz -rf archive.tar.lz d e
@end example

@sp 1
@noindent
Example 3: Create a solidly compressed appendable archive
@samp{archive.tar.lz} containing files @samp{a}, @samp{b} and @samp{c}.
Then append files @samp{d} and @samp{e} to the archive.

@example
tarlz --asolid -cf archive.tar.lz a b c
tarlz --asolid -rf archive.tar.lz d e
@end example

@sp 1
@noindent
Example 4: Create a compressed appendable archive containing directories
@samp{dir1}, @samp{dir2} and @samp{dir3} with a separate lzip member per
directory. Then append files @samp{a}, @samp{b}, @samp{c}, @samp{d} and
@samp{e} to the archive, all of them contained in a single lzip member.
The resulting archive @samp{archive.tar.lz} contains 5 lzip members
(including the end-of-archive member).

@example
tarlz --dsolid -cf archive.tar.lz dir1 dir2 dir3
tarlz --asolid -rf archive.tar.lz a b c d e
@end example

@sp 1
@noindent
Example 5: Create a solidly compressed archive @samp{archive.tar.lz}
containing files @samp{a}, @samp{b} and @samp{c}. Note that no more
files can be later appended to the archive.

@example
tarlz --solid -cf archive.tar.lz a b c
@end example

@sp 1
@noindent
Example 6: Extract all files from archive @samp{archive.tar.lz}.

@example
tarlz -xf archive.tar.lz
@end example

@sp 1
@noindent
Example 7: Extract files @samp{a} and @samp{c}, and the whole tree under
directory @samp{dir1} from archive @samp{archive.tar.lz}.

@example
tarlz -xf archive.tar.lz a c dir1
@end example

@sp 1
@noindent
Example 8: Copy the contents of directory @samp{sourcedir} to the directory
@samp{destdir}.

@example
tarlz -C sourcedir --uncompressed -cf - . | tarlz -C destdir -xf -
@end example

@sp 1
@noindent
Example 9: Compress the existing POSIX archive @samp{archive.tar} and write
the output to @samp{archive.tar.lz}. Compress each member individually for
maximum availability. (If one member in the compressed archive gets damaged,
the other members can still be extracted).

@example
tarlz -z --no-solid archive.tar
@end example

@sp 1
@noindent
Example 10: Compress the archive @samp{archive.tar} and write the output to
@samp{foo.tar.lz}.

@example
tarlz -z -o foo.tar.lz archive.tar
@end example

@sp 1
@noindent
Example 11: Concatenate and compress two archives @samp{archive1.tar} and
@samp{archive2.tar}, and write the output to @samp{foo.tar.lz}.

@example
tarlz -A archive1.tar archive2.tar | tarlz -z -o foo.tar.lz
@end example


@node Problems
@chapter Reporting bugs
@cindex bugs
@cindex getting help

There are probably bugs in tarlz. There are certainly errors and
omissions in this manual. If you report them, they will get fixed. If
you don't, no one will ever know about them and they will remain unfixed
for all eternity, if not longer.

If you find a bug in tarlz, please send electronic mail to
@email{lzip-bug@@nongnu.org}. Include the version number, which you can
find by running @w{@samp{tarlz --version}} and
@w{@samp{tarlz -v --check-lib}}.


@node Concept index
@unnumbered Concept index

@printindex cp

@bye