This is the file README for the gzip distribution, version 1.2.2.
gzip (GNU zip) is a compression utility designed to be a replacement
for 'compress'. Its main advantages over compress are much better
compression and freedom from patented algorithms. The GNU Project
uses it as the standard compression program for its system.
gzip currently uses by default the LZ77 algorithm used in zip 1.9 (the
portable pkzip compatible archiver). The gzip format was however
designed to accommodate several compression algorithms. See below
for a comparison of zip and gzip.
gunzip can currently decompress files created by gzip, compress or
pack. The detection of the input format is automatic. For the
gzip format, gunzip checks a 32 bit CRC. For pack, gunzip checks the
uncompressed length. The 'compress' format was not designed to allow
consistency checks. However gunzip is sometimes able to detect a bad
.Z file because there is some redundancy in the .Z compression format.
If you get an error when uncompressing a .Z file, do not assume that
the .Z file is correct simply because the standard uncompress does not
complain. This generally means that the standard uncompress does not
check its input, and happily generates garbage output.
gzip produces files with a .gz extension. Previous versions of gzip
used the .z extension, which was already used by the 'pack'
Huffman encoder. gunzip is able to decompress .z files (packed
or gzip'ed).
Several planned features are not yet supported (see the file TODO).
See the file NEWS for a summary of changes since 0.5. See the file
INSTALL for installation instructions. Some answers to frequently
asked questions are given in the file INSTALL, please read it. (In
particular, please don't ask me once more for an /etc/magic entry.)
WARNINGS about broken optimizers:
- on the NeXT, "cc -finline-functions" is broken. gzip produces
valid .gz files but they are much too large because the string
matching code misses most matches. Use "cc -O" instead.
- on the Mips R4000, gcc -O (version 2.3.1) generates bad code, use cc
or just gcc -g instead.
- gcc 2.3.3 and 2.4.3 on the SGI Indigo IRIX 4.0.5 also produce bad code.
Use instead: make CC='cc -O2' or gcc without -O.
- On Solaris 2.1 for x86, the January 1993 "OEM" compiler release
generates bad code. This is fixed in the June 1993 "FCS" release.
- on Sparc with SunOS 4.1.1 and the SC1.0 compiler, the optimizer
works up to -O3 but -O4 does not work.
- MSC 5.1 with -Ox and -DDYN_ALLOC generates bad code in inflate.c.
The default is static allocation (no DYN_ALLOC) and -Ox works on inflate.c.
But -Ox does not work on util.c, so you must use -Oait -Gs.
- On dnix 5.3 2.2 cc version 2.37c is buggy. Version 2.38d works.
- On an Alliant running Concentrix, cc (even without optimization) generates
incorrect code. You have to use gcc.
For all machines, Use "make check" to check that gzip was compiled correctly.
Please send all comments and bug reports by electronic mail to:
Jean-loup Gailly <jloup@chorus.fr>
or, if this fails, to bug-gnu-utils@prep.ai.mit.edu.
Bug reports should ideally include:
* The complete output of "gzip -V" (or the contents of revision.h
if you can't get gzip to compile)
* The hardware and operating system (try "uname -a")
* The compiler used to compile (if it is gcc, use "gcc -v")
* A description of the bug behavior
* The input to gzip, that triggered the bug
If you send me patches for machines I don't have access to, please test them
very carefully. gzip is used for backups, it must be extremely reliable.
The package crypt++.el is highly recommended to manipulate gzip'ed
file from emacs. It recognizes automatically encrypted and compressed
files when they are first visited or written. It is available via
anonymous ftp to roebling.poly.edu [128.238.5.31] in /pub/crypt++.el.
The same directory contains also patches to dired, ange-ftp and info.
GNU tar 1.11.2 has a -z option to invoke directly gzip, so you don't have
to patch it.
The znew and gzexe shell scripts provided with gzip benefit from
(but do not require) the cpmod utility to transfer file attributes.
It is available by anonymous ftp on gatekeeper.dec.com in
/.0/usenet/comp.sources.unix/volume11/cpmod.Z.
The sample programs zread.c, sub.c and add.c are provided as examples
of useful complements to gzip. Read the comments inside each source file.
The perl script ztouch is also provided as example (not installed
by default since it relies on perl).
gzip is free software, you can redistribute it and/or modify it under
the terms of the GNU General Public License, a copy of which is
provided under the name COPYING. The latest version of gzip are always
available by ftp in prep.ai.mit.edu:/pub/gnu, or in any of the prep
mirror sites:
- sources in gzip-*.tar (or .shar or .tar.gz)
- Solaris 2 executables in sparc-sun-solaris2/gzip-binaries-*.tar
- MSDOS lha self-extracting exe in gzip-msdos-*.exe. Once extracted,
copy gzip.exe to gunzip.exe and zcat.exe, or use "gzip -d" to decompress.
gzip386.exe runs much faster but only on 386 and above; it needs go32.exe,
available in omnigate.clarkson.edu:/pub/msdos/djgpp/djdev110.zip
A VMS executable is available in ftp.spc.edu:[.macro32.savesets]gzip-1-*.zip
(use [.macro32]unzip.exe to extract). A PRIMOS executable is available
in ftp.lysator.liu.se:/pub/primos/run/gzip.run.
Many thanks to those who provided me with bug reports and feedback.
See the files THANKS and ChangeLog for more details.
Note about zip vs. gzip:
The name 'gzip' was a very unfortunate choice, because zip and gzip
are two really different programs, although the actual compression and
decompression sources were written by the same persons. A different
name should have been used for gzip, but it is too late to change now.
zip is an archiver: it compresses several files into a single archive
file. gzip is a simple compressor: each file is compressed separately.
Both share the same compression and decompression code for the
'deflate' method. unzip can also decompress old zip archives
(implode, shrink and reduce methods). gunzip can also decompress files
created by compress and pack. zip 1.9 and gzip do not support
compression methods other than deflation. (zip 1.0 supports shrink and
implode). Better compression methods may be added in future versions
of gzip. zip will always stick to absolute compatibility with pkzip,
it is thus constrained by PKWare, which is a commercial company. The
gzip header format is deliberately different from that of pkzip to
avoid such a constraint.
On Unix, gzip is mostly useful in combination with tar. GNU tar
1.11.2 has a -z option to invoke gzip automatically. "tar -z"
compresses better than zip, since gzip can then take advantage of
redundancy between distinct files. The drawback is that you must
scan the whole tar.gz file in order to extract a single file near
the end; unzip can directly seek to the end of the zip file. There
is no overhead when you extract the whole archive anyway.
If a member of a .zip archive is damaged, other files can still
be recovered. If a .tar.gz file is damaged, files beyond the failure
point cannot be recovered. (Future versions of gzip will have
error recovery features.)
gzip and gunzip are distributed as a single program. zip and unzip
are, for historical reasons, two separate programs, although the
authors of these two programs work closely together in the info-zip
team. zip and unzip are not associated with the GNU project.
The sources are available by ftp in
oak.oakland.edu:/pub/misc/unix/zip19p1.zip
oak.oakland.edu:/pub/misc/unix/unz50p1.tar-z
