FIXME:
 update web page and links etc
---


when repacking archives, don't specify '.'. Instead specify each extracted file or
directory. (Jens Schleusener)

---

FIXME: implement -O fully! also, use if for compression only! or?

---

$ arepack today.log.bz2 today.log.sz		[today.log.bz2 exists, today.log.sz doesn't]
arepack: today.log.sz: no such file and cannot identify format from extension
Use of uninitialized value $fmt2 in string eq at /home/usel/sys/bin/arepack line 354.
today.log.bz2: extracted to `Unpack-4162/today.log'
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1348.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1349.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1350.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1351.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1352.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1353.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1354.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1355.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1356.
Use of uninitialized value $fmt in string eq at /home/usel/sys/bin/arepack line 1357.
arepack: ../today.log.sz: no such file and cannot identify format from extension

---

From Francois Marier: omit_garbage option, do not unpack
  Thumbs.db
  __MACOSX/
  ._filename
  .Trashes
  .DS_Store
See also: http://unmac.rubyforge.org/
          http://github.com/fxn/unmac/blob/HEAD/lib/unmacer.rb


add compression level control arguments: store (0) or 1-9
allow user to add customer parameters to each archiver??

make use of modern tar options, like '--use-compress-program=prog'
cfg_use_tar_compress_option?

if making RPM, then upload SRPM as well!

improve messages printed from arepack. sometimes prints unnecessary messages like "data.tar.gz: extracted to `Unpack-2962/data'"

cat mode is not supported for lzop, because it won't write to standard out
unless we give it the -f option. Because of this we use lzop as if -c was
not supported. This should be fixed in the future.

--no-umask option from Reuben to fix permissions of extracted files. perhaps change owner etc.

aunpack -e is default - or perhaps with option to make it default?

aremove tool
  how it should work:
    unpack archive to a separate directory
    for each unpacked file, check if it exists and is the same in the current directory.
    if yes, remove that file in the current directory, and (optionally) from unpacked directory as well.
    if no, remove that file from the unpacked directory

From Francois Marier <francois@debian.org>: check if pbzip2 is in the path and use it

rewrite atool implement extract_archive that is sensitive or not

add option to assume format with -F in some situations (like tar.gz)?
add option to always assume -e? and add some option for the reverse (-s, --single)
assume -e when passing a single directory as argument?

change back error messages to:
   cannot foo `bar': No such file or directory
  make function to format error message?

have autoconf generate files in debian subdir as well?

in readconfig, add datatype checking to options (bool/str/int/..)

add 'atest': test archive integrity
make sure atool has non-zero rc when acat can't find a file (.arc files and possibly other)

improve error message when some program doesn't exist (now gets unnecessary lines)

  arepack -F.tar.bz2 foo.tar.gz   would only gunzip then bzip2. tar not involved!

tests:
  does gzip support --?
  does tar support -z?
  does tar support -I?
  does tar support --?
  does zip treat *?[]\ chars in filenames specially?

use "--" with gzip and others?

Reasons to rewrite in C:
  - use manual ZIP code so that we can extract archive with '../foo' files too!
    tar fails entirely when the archive contains '../xxx' members.
  - the option --save-outdir in extract-to mode could be implemented
    easier.
  - multiple commands per action more easily (same with perl tho)
  - always identify format using magic values, ignoring extension (fast!)
  - i18n much easier (and faster)

Bugs
  implement list command for single compressed archives? only gzip supports -l
  consider:
      apack dummy.tar.gz /etc/hostname
      mkdir etc
      aunpack dummy.tar.gz
    Archive is now extracted with file as dummy/etc/hostname
    Should it be dummy/hostname or etc-1/hostname instead?

Other formats
  add support for SEA arc.
  add support for szip (hard because its interface is different from gzip)
  make an ACE archive (using winace stuff), test ACE extraction!
  add support for DEB files
     list: 	dpkg-deb -c FILE.deb
     extract:	dpkg-deb -x FILE.deb $outdir/CONTENTS
		dpkg-deb -e FILE.deb $outdir/DEBIAN
  JAR/ZIP:
    Implement 'use_jar_for_zip' - use jar for zip archives (fastjar) - test
    Rename 'use_zip_for_jar' - use zip for jar archives - test
    Enable add command for jar archives!


Other features
  make a script to detect system settings and configure atool appropriately
  Make it possible to send additional arguments to RAR etc. engine configurable via command line?
    especially adiff diff -ruN (note the N) option

--------------------------------------------------------------------------------

libarchive:
  OpenArchive(fd)->archive
  CloseArchive(archive)
  GetType(archive)

  BeginJob(archive)->job
  PerformJob(job)
  EndJob(job)
  
  GetError(job)
  ErrorManger(job,callback)
  ProgressManager(job,callback)

  CreateArchive(file)->archive

  FileDescriptor
    close()
    read(data,amount)
    write(data,amount)
    seek(pos,whence)
  
  FileIterator(job,fileselector,filecallback)
  
  DeleteFile(job,fh)
  ExtractFile(job,fh,outfd)
  ReplaceFile(job,fh,infile)
  
  AddFile(job,filename,extfile)

---

verbosity levels:
  0 default, quiet
  1 verbose, print filenames
  2 very verbose, print long list format

extract: what to when local file exists, only when archive contains one file
(and not in interactive mode)
  * (m)ove to subdirectory
  * (r)ename
  * move files to subdir, rename directory
  * for interactive mode only: (o)verwrite

how to extract single compressed file
  * delete original file afterwards
  * keep original file
  * don't extract at all (refuse)		

add:
  * apack: create new archive (refuse if existing exists)
  * apack -a: add to existing archives (create if necessary)

---

MAJOR options:

  aunpack	-X	-x [DIR]	extract to dir
  apack		-A	-a		add files to archive (or create)
  apack		-C	-c		create only
  als		-L	-l		list files in archive
  acat		-P	-c/-p		print file (to standard out) in archive

  adiff		-D	-d		print difference between two archives
  afind		-F	-f		find files in archive (emulate find)		[?]

  arepack	-R	-r		extract and pack archives



als ARCHIVE1..
  -l		-l (long) like ls
  -f		-f (classify) like ls
  -m[TYPE]	from all archives, list only files matching (remaining args are matches)

prev: als -e *.tar		now:  als *.tar
prev: als foo.tar f*		now:  als foo.tar -m f*

aunpack ARCHIVE1..
  -X DIR	where to extract to (otherwise be smart)
  -m[TYPE]	from all archives, extract only files matching (remaining args are matches)

apack ARCHIVE FILE1..
  -a		add files if archive already exists (without -a, refuse to overwrite existing)
  -e		all arguments are files (or dirs) to compress individually
  -F FORMAT	force file format

  -f		allow overwriting of archives (clobber). NOTE: for safety, this implies


--------------------------------------------------------------------------------

New rules format:

# This document shows how a future rules file format for atool could look. It
# was written to outline what was necessary for such a format. It is quite
# possible this format will never become a reality due to its complexity.

# use ; to separate commands!

# rules and configuration files are the same

noinclude		# don't include default rules/configuration file!
			# (otherwise /etc/atool.rules will be included first)
			# useful if you want to make your own rules completely

set path_zip "foobar"	# example of variable set in a configuration file
			# (can only be done after option declared in format)

# the extract command defined here also handles extract-in (only outdir changes)
# the create command refuses to create an existing archive
# the add command will turn into a create command if the archive doesn't exist
# the print command outputs a file in the archive to standard out

# variable namespace:
#   read-only variables - $outdir, $archive, $args
#   format-local variables - declared with 'variable'
#   options - declared with option

# FIXME: what about about special exit codes (zip)? CHECK atool

format {
  name { zip jar }		# a name for users to use in -F option (force format)
  extension { .zip .jar }	# possible file extensions (ignored?)
  magic	    { PK\003\004 }	# magic bytes (kinda regex?)
  option string path_zip "zip"
  option string path_unzip "unzip"
  command extract    { $path_unzip "-d" $outdir $archive $args }
  command add,create { $path_zip "-r" $archive $args }
  command print	     { $path_unzip "-p" $archive $args }
  command list 	     { $path_unzip "-l" $archive $args }
  command common  {
    [ $verbosity >  1 ] "-v"		# this format means: add "-v" only if $verbosity > 1
    [ $verbosity <  0 ] "-qq"		# (the first string after [] applies, or block if { } )
    [ $verbosity == 0 ] "-q"
  }
  multiarchive true	# specifies that this archive may contain multiple files
  needs_cwd false	# specifies that this archive doesn't need to cwd to outdir on extract
  			# (technically we could autodetect this by seeing of $outdir is referenced
			# in the extract command)
}

format {
  name		{ lha lzh }
  extension	{ .lha .lzh }
  magic		{ "..-lh[01234567 d]-" "..-lz[45s]-" }	# what about >20 byte x ; 2 string -lh6-
  option string path_lha "lha"
  command common { $path_lha }
  command extract { "x" $archive $args $outdirslash }	# outdirslash = outdir that ends in slash
  command add,create { "a" $archive $args }
  command print { "p" $archive $args }
  command list {
    "a"
    [ $verbosity >= 3 ] "v"
    [ $verbosity =  2 ] "l"
    [ $verbosity <= 1 ] "lq"
  }
  multiarchive true
  needs_cwd false
}

format {
  name		{ tar[+.]gz(ip)? }
  extension	{ .tgz .tar.gz }
  option string path_tar "tar"				# this one assume tar supports -z
  variable string opts
  command common { $path_tar [ $verbosity >= 1 ] [ $opts = "v" ] }	# a conditional
  command print { xO ++ $opts ++ zf }			# ++ means concatenate
  command extract { x ++ $opts ++ zf -C $outdir }	# -C works on Sun too
  command list { t ++ $opts ++ zf }
  command create { c ++ $opts ++ zf }			# tar can append (-r), but not tar+gz
  command common { $archive $args }
  multiarchive true
  needs_cwd false
}

format {
  disabled						# do not use this format at all
  name		{ tar[+.]gz(ip)? }
  extension	{ .tgz .tar.gz }
  option string path_tar "tar"				# this one does NOT assume tar supports -z
  option string path_gzip "gzip"
  variable string opts
  command common { [ $verbosity >= 1 ] [ $opts = "v" ] }
  command print { $path_gzip -cd $archive | $path_tar xO ++ $opts $args }
  command extract { $path_gzip -cd $archive | $path_tar x ++ $opts $args }
  command list { $path_gzip -cd $archive | $path_tar t ++ $opts $args }
  command create { $path_tar c ++ $opts $args | $path_gzip > $archive }
  multiarchive true
  needs_cwd false
  create_only true					# FIXME: maybe not necessary? (see create/add above)
}


--------

TODO.v2:

better extraction semantics, cleaner code

 * always use change to directory for extraction, never specify dir on cmd command-line?
   use ../for instead of absolute dir
   merges extract-to and extract

 * core support for multiple commands per mode (think debian)

 * better control of user arguments to external command (@args)
 
 * better solution than multiarchivecmd - cleaned up

 * more general commands:
     test

 * more general options
     compression level

 * stripext only based on format! (e.g. unpack gz then strip .gz, not .tar.gz)

want aunpack -e to be default!
remove dir: failure to remove a single file is fatal. chmod error isn't.
chmod errors should be ignored altogheter?

always chdir on extract?

unified handling of single & non-single (wrt extract etc)

acat  displays *all* files or just first?
apack creating a single archive with many files = ERROR
extract-logic with -S:
  tar xvfz ...
  content-move DIR   or  atool --handle-extracted DIR

does unlink_dir handle . and .. (i.e. cur and up)?
write safety check for those

use File::Spec abs2rel



---

COMMANDS

apack	- create a new archive
aunpack - extract files from an archive
als	- list archive like ls
afind	- list archive like find
acat	- extract files from an archive to standard out
aless	- extract files from an archive and pipe into a pager
amore	- extract files from an archive and pipe into a pager
adiff	- compare archives using diff
agrep	- search files for a regular expression using grep
aegrep	- search files for a regular expression using egrep
afgrep	- search files for a regular expression using fgrep
arepack	- extract archives of one type and create archives of another
acmp	- compare archives using cmp
aadd	- add files to an existing archive (may require unpacking)
atest	- test archive integrity (may unpack)
amcfs	- midnight commander virtual file system (extfs)

OPTIONS

-F,--format	major format
--format-in	format of files being read
--format-out	format of files being created
-f,--files	hereafter comes files (not archives)
-o,--options	hereafter comes options for archiver programs
-X,--extract-to
-E,--explain
-S,--simulate
-q,--quiet	display no regular output
-v,--verbose	display a little more output
-z,--compression-level	compression level for gzip, bzip2, "1" to "9"

ERRORS

display at max two errors: one detailed ($!), one general ("command failed")

LOCALIZATION

FORMATS

add SEA arc
add szip