Translated by Koji Arai with www.DeepL.com/Translator (free version)
This file describes changes of the Autoconf version of the LHa for UNIX.
The compilation procedure is as follows.
gzip -dc lha-1.14i-acXXXXXXXX.tar.gz | tar xvf -
cd lha-114i-acXXXXXXXX
sh ./configure
make
make check
make install
MinGW support is in alpha. It has not been tested much (I only confirmed to success "make check"). To try out the MinGW version under Cygwin, use the:
sh ./configure \
--build=i686-pc-mingw32 \
--host=x86_64-w64-mingw32 \
--with-tmp-file=no
Note: In Windows environments such as Cygwin and MinGW, make check
may fail if you compile in a deep directory.
This is because the full path length limit of Windows is exceeded when checking for long path test.
In this case, run configure in a shallow directory as follows
cd /tmp/build
sh ~/src/lha/configure ....
make
make check
make install
If autoconf/automake is installed and you want to maintain lha source, configure.ac and Makefile.am, follow the steps below. The version of autoconf/automake is assumed to be autoconf 2.5x or later and automake 1.6.x or later.
gzip -dc lha-1.14i-acXXXXXXXX.tar.gz | tar xvf -
cd lha-114i-acXXXXXXXX
aclocal
autoheader
automake -a
autoconf
# Above steps from aclocal to autoconf can replace with `autoreconf -is`
sh ./configure
make
make check
make install
The Autoconf version of the LHa for UNIX has been modified based on the LHa for UNIX ver1.14i http://www2m.biglobe.ne.jp/~dolphin/lha/lha.htm
The main changes are listed below.
In the original LHa for UNIX 1.14i, if you compile without defining the SUPPORT_LH7
macro,
it was not possible to archive with -lh6-
and -lh7-
methods.
This is not a problem, but when the SUPPORT_LH7
is defined, the -lh7-
method is created by default.
So, to make it more flexible for lha users to choose these methods:
SUPPORT_LH7
should always be defined.- default archive method can be specified with the configure option
--with-default-method=[567]
.
The default value of this configure option is -lh5-
. The command line option of lha, -o6
or -o7
options
can create -lh6-
or -lh7-
archives at using time, and the configure option can change the default method.
The original LHa for UNIX 1.14i is careless about encoding of filenames in archives.
Even when the MULTIBYTE_CHAR
macro is defined at compile time, it does not convert encoding between
filenames in the archive and the system filename.
In the autoconf version, the configure option --enable-multibyte-filename
allows the use of Japanese Kanji
filename. it considers encoding of filenames stored in the archive as fixed to Japanese Shift JIS (it is
specs of LHA).
The value of --enable-multibyte-filename
is specifying Japanese Kanji character encoding of the system
filename. These are follows:
--enable-multibyte-filename=sjis
Specifies filename encoding of the system as Shift_JIS.
--enable-multibyte-filename=euc
Specifies filename encoding of the system as EUC-JP.
--enable-multibyte-filename=utf8
Specifies filename encoding of the system as UTF-8.
--enable-multibyte-filename=auto (yes or omit value)
Automatically determine the filename encoding of the system.
Automatic currently means:
Shift_JIS: for Cygwin, MinGW and HP-UX environment
UTF-8: the environment that the iconv is available
EUC-JP: otherwise
--enable-multibyte-filename=no
--disable-multibyte-filename
Disables multibyte support for filenames.
The default value is auto
.
The lha command line option allows you to change the compile-time value. This command line option is specified with the GNU style long option (with two leading dashes).
--system-kanji-code=xxx
Specifies the system kanji code.
--archive-kanji-code=xxx
Specifies the encoding of filename to be stored in the archive.
This is usually fixed to Shift_JIS (`sjis`) and should not be changed.
xxx
is one of sjis
, euc
, utf8
, or cap
. The cap
is a code used by the Samba and others to represent
multibyte filename with a ":" and two hexadecimal character.
For example:
$ touch 漢字
$ lha c foo.lzh 漢字
$ lha l foo.lzh
PERMISSION UID GID SIZE RATIO STAMP NAME
---------- ----------- ------- ------ ------------ --------------------
-rw-r--r-- 1000/1000 0 ****** Mar 23 21:23 漢字
---------- ----------- ------- ------ ------------ --------------------
Total 1 file 0 ****** Mar 23 21:23
$ lha l --system-kanji-code=cap foo.lzh
PERMISSION UID GID SIZE RATIO STAMP NAME
---------- ----------- ------- ------ ------------ --------------------
-rw-r--r-- 1000/1000 0 ****** Mar 23 21:23 :8a:bf:8e:9a
---------- ----------- ------- ------ ------------ --------------------
You can also use UTF-8 outside of Mac OS X by using the iconv library.
If you don't want to use iconv, you can do follows:
sh ./configure --enable-multibyte-filename=utf8 --disable-iconv
In particular, on Mac OS X, when the iconv disabled, it will use the Core Foundation library to support UTF-8.
The converter between UTF-8 and Shift_JIS for Mac OS X was created by Hiroto Sakai in 2002/6. Thank you very much.
The original LHa for UNIX 1.14i uses standard input as follows:
echo foo.txt | lha x foo.lzh
find bar -name '*.[ch]' | lha c bar.lzh
This will extract only foo.txt from foo.lzh, or store files under the bar directory in bar.lzh.
In other words, This is the features to specify compressed/extracted files from the standard input, but it's not really useful for me.
(The compression example is as useful as cpio
, but I think it would be better to create a new option for
this behavior).
I think it is only intended for use from tty, but it will probably cause unintended behavior when running lha from daemon. For this reason, I have taken the liberty of removing this feature.
It is a meaningful feature in the Windows environment where the following things are not possible:
lha x bar.lzh `echo foo.txt`
Unfortunately, isatty() doesn't work well with MinGW, so the feature to use the standard input doesn't work on Windows.
If you want to restore the original feature, change #if 0
to #if 1
at line about 631 in the lharc.c
.
It is now possible to create extended headers (0x52, 0x53) for user and group names for Unix. (default is off). See the header.doc.md file for details.
This is enabled when you build with the configure option --enable-user-name-header
.
Whenever extracting and listing, if this information is in the header, it is used in preference to the ID.
In the original LHa for UNIX 1.14i, when a file is added to the archive, or when a file is deleted
from the archive, the original archive file is saved with extension .bak
.
I found this behavior annoying, so I decided not to create a backup file in the autoconf version.
If you don't trust this patch, please build with the configure option --enable-backup-archive
(which you should definitely do :p). It will behave the same as the original.
The header.c has been rewritten. In addition to the above changes, the following bugs have been fixed since the LHa for UNIX 1.14i.
Could not correctly read archives with total header size
(first 2 bytes of archive header) greater than 256.
And, it created an invalid archive where the total header size was exactly 256.
The LHA header specification does not allow the total header size to be a multiple of 256, since a header
leading 0 is considered to be the end of the archive (e.g. 256 is represented by "0x00 0x01" in
little-endian, so it is leading 0).
CRC check of header is now performed when reading.
Since no extended headers were used for filenames, if you write a filename (excluding directory name) that exceeds 230 bytes to an archive, it created an invalid archive that exceeds the archive header size limit.
Invalid archives were being created for long pathnames (including directory name) that exceeded the size limit of the archive header. (actually, this limit still only applies to the filename length, since the original does not write any directory information in the level 0 header). The autoconf version will output a warning message for pathnames that exceed the limit, and will truncate the end of the pathname. (The level 0 header should not be used).
An empty -lhd- header (with no directory name information) was created.
$ mkdir foo
$ lha c0 foo.lzh foo
$ lha v foo.lzh
PERMSSN UID GID PACKED SIZE RATIO METHOD CRC STAMP NAME
---------- ----------- ------- ------- ------ ---------- ------------ -------------
drwxrwxr-x 1000/1000 0 0 ****** -lhd- 0000 Jul 29 00:18
---------- ----------- ------- ------- ------ ---------- ------------ -------------
Total 1 file 0 0 ****** Jul 29 00:18
There was a theory that the -lhd-
method could not be used with the level 0 header.
http://kuwa.xps.jp/x68k/KGARC/ARC/LHAHED15.ZIP
However, Haruyasu Yoshizaki's original LHA for DOS/Windows (ver 2.55, 2.67) creates -lhd-
method in level 0
header.
This opinion was acknowledged. (http://kuwa.xps.jp/diary/2002-10.html#27_3) The above document have been fixed.
http://kuwa.xps.jp/x68k/KGARC/ARC/LHAHED21.ZIP
The -g
option is said in the man page to suppress the creation of UNIX-specific information in the
archive when creating an archive, but in fact it suppressed directory information as well.
$ mkdir foo
$ touch foo/bar
$ lha cg1 foo.lzh foo
$ lha foo.lzh
PERMSSN UID GID SIZE RATIO STAMP NAME
---------- ----------- ------- ------ ------------ --------------------
[generic] 0 ****** Jul 29 00:02 bar
---------- ----------- ------- ------ ------------ --------------------
Total 1 file 0 ****** Jul 29 00:02
In the autoconf version, the above becomes foo/bar.
(The original may have done this on purpose, but I saw no reason to do so.)
The -g
option suppresses the creation of -lhd-
, it is same behavior as original one.
In addition, if you specify the -g
option and the header level option (-0
, -1
and so on ) at
the same time, you need to specify the g option first, as shown above(lha cg1
).
This is because the -g
option also has the meaning of the -0
option.
If you specify the -g
option afterwards, such as lha c1g
, a level 0 header will be created.
(this original behavior is a bit confusing)
The level 3 header specification seems to exist, but it is not yet fixed as a specification, so I added support extracting only. Additional extension headers with level 3 header are not supported. (I couldn't find any header that should be supported).
If you want to support largefile, you might want to support this level 3 header.
I added the ability to dump headers. This is completely for debugging purposes.
lha vvv foo.lzh
This will output a dump along with the list of archive contents.
The default header level when creating an archive is now 2. (The original LHa for UNIX 1.14i defaulted to level 1)
The extended header Windows timestamp (0x41) now read (in level 1 header only). It has 64-bit creation date, last modified date, and last accessed date.
For level 2 header and later, the extended header is ignored because the basic part of header contains the 32-bit Unix epoch time. It has not been confirmed whether there is an LHA archiver that outputs Windows timestamp extension headers for level 1 header archives. I don't think this is a very useful fix, but since I made it, I'll leave it. :-)
Added the -x
option to specify the pattern to be excluded from the files to be compressed.
lha c -x '*.o' -x='*.a' -x'*.c' src.lzh src
The source of command line parser has been reconstruct. The usage of the autoconf version is as follows:
usage: lha [-]<commands>[<options>] [-<options> ...] archive_file [file...]
On Cygwin, when extracting archives that do not have permission information, such as MS-DOS types, now extracts them with 0777 & ~umask. This is so that .exe and .dll can have the execute attribute.
It can handle large files (over 2G) if your system supports it. (configure script will provide you with the appropriate compiler options).
However, to support large files on HP-UX 11.0, you need to use the:
CC="cc -Ae +DA2.0W"
It seems that you need to instruct them to do so.
./configure --with-tmp-file=no CC="cc -Ae +DA2.0W" \
ac_cv_have_mktime=yes \
ac_cv_func_mktime=yes
--with-tmp-file=no
will create intermediate files in the same directory as the output destination.
This is in case the temporary directory does not support over 2G.
(This is not related to largefiles support, but ac_cv_*=yes forces HP-UX to use mktime()
because
it fails to detect mktime()
for some reason.)
If you want to "disable" large files support, you can use the:
./configure --disable-largefile
In regular LHA, only files less than 4G can be archived due to the specification of the level 0, 1, and 2 headers (there are only 4 bytes of space to store the file size).
However, UNLHA32.DLL, for example, has an extension header (0x42) that allows it to handle files over 4G (64-bit file size). The autoconf version currently supports files over 4G by referring to this extension header only during extraction. (The reason it does not support creation is that I was not sure if it was correct to simply follow UNLHA32.DLL.)
Files stored in archives created by turning on the "MacBinary" checkbox in MacLHA are MacBinary encoded.
If the -b
option is specified when extracting, MacBinary decoding will be performed after extraction
and only the data fork will be extracted (the resource fork will be ignored).
When extracting a normal archive with the -b
option, the normal extraction process will be performed.
To use this feature, you need the applefile library. The applefile library can be obtained from:
http://sourceforge.net/projects/applefile/
I believe that the src/header.c no longer contains the code that was in the original lha 1.14i. The src/header.c is my source code. However, it is true that the code in lha 1.14i was helpful to me, and I have left the names of the developers in src/header.c as a sign of respect.
This is intention that the redistribution terms in man/lha.man do not apply to src/header.c. I am trying to make the code more free and easy to use as a component of LHa for UNIX. The license to apply to src/header.c is still undecided (so please consider that the redistribution terms of LHa for UNIX apply for now).
License candidates (notes):
Currently, sources of the LHa for UNIX except src/vsnprintf.c, src/fnmatch.[ch], src/getopt_long.[ch] apply the terms contained in man/lha.man. (Note that the terms described in each source also apply to LHa for UNIX made using vsnprintf.c and fnmatch.c.)