unix.txt

arj source code

   @PRODUCT_LONG  @{r74}@{d}@{_}


   This  product is an  implementation of ARJ v @COUNTERPARTS  for DOS on UNIX  and
   UNIX-like  systems. It is  assumed that the  user is familiar  with ARJ
   operation on DOS before using this package.

   ***********************************************************************
   ***                                                                 ***
   ***  THIS PRODUCT IS DEVELOPED  SEPARATELY FROM THE MAINSTREAM ARJ  ***
   ***  PACKAGE AND  THEREFORE IS NOT  UPDATED SYNCHRONOUSLY WITH ARJ  ***
   ***  FOR DOS.                                                       ***
   ***                                                                 ***
   ***********************************************************************


   INTRODUCTION

      This file describes the features specific to the UNIX port. It is
      suggested that you read the general README file prior to this one.


   SYSTEM REQUIREMENTS

   Linux:

      *  Kernel version 2.0 or higher.
      *  Intel 80386 or upwardly compatible machine.
      *  GLIBC v 2.1 or higher.

   FreeBSD:

      *  FreeBSD v 3.0 or higher (ELF-based).
      *  Intel 80386 or upwardly compatible machine.

   QNX:

      *  QNX v 6.2 NC/SE/PE or higher.
      *  Intel 80386 or upwardly compatible machine.


   IMPORTANT NOTE

      The  implementation of  ARJ  on UNIX  platforms  requires a major
      change in the  archive  format. To support UNIX-style timestamps,
      we had to introduce a new timestamp format which is  incompatible
      with the previous versions of ARJ.

      If you plan to exchange  with ARJ and ARJ32 users, be sure to set
      up the DOS compatibility mode in /etc/arj.cfg as follows:

      + -2d

      This  will  force  the  archiving  to be  done by  default in the
      compatibility mode (time converted  to local, seconds  rounded to
      even, date  range  from 1980 to  2107); to store  UNIX timestamps
      normally, you will have to override "-2d" from the command line.


   IMPLEMENTATION ISSUES

      *  The usual place for ARJ within UNIX directory  structure is as
         follows:

         /usr/local/bin            for ARJ programs
         /usr/local/lib            for ARJCRYPT
         /usr/local/doc/arj        for accompanying documentation

         Any other locations  are tolerated, provided that the programs
         may  be found  by means  of  execution  environment (the  PATH
         variable).

         It is essential that ARJ is kept both readable and executable.

         In this release, the most  up-to-date information regarding ARJ
         commands and  options is  provided   by "arj -?".   In  certain
         installations, ARJ  manual   pages may be  accessible  as well,
         e.g. "man arj" should display the list of ARJ options.

      *  The search order for ARJ configuration files is as follows:

         arj.cfg:
                1. ~/.arj.cfg
                2. /etc/arj.cfg

         arj.key:
                1. arj.key in the directory of ARJ executable file
                2. ~/.arj.key
                3. /etc/arj.key

         rearj.cfg:
                1. ~/.rearj.cfg
                2. /etc/rearj.cfg

         The configuration files are  used on a  first-come  basis, and
	 not "merged" together.

      *  File creation  date/time is  not restored  but is saved  along
	 with regular timestamp information.

      *  The shortnames  for Windows 95  long filenames are  written as
         w95lname.xxx (that is, in  lower case). The upper case is only
         used for ARJ temporary files.

      *  When operating on  standard streams, the file date/time is not
         restored in  UNIX systems, i.e. "arj p archive myfile>newfile"
         will result in newfile created with the current date/time.

      *  External  garble  modules   are  provided  as  shared  dynamic
	 libraries. The encryption algorithm  remains to  be compatible
	 with ARJ versions on all other platforms.

      *  "UNIX" is  stamped  as  the  host OS whenever  the  archive is
	 updated, unless "-2d" is in effect.

      *  The  archive "L"ist command  has been  remodelled  to  provide
	 better  functionality on  UNIX platforms. The UNIX  attributes
	 are  grouped  in  standard  "-rwxrwxrwx" pattern,  plus  three
	 separate  fields,  G/U/A,  for  "set GID",  "set UID" and  the
	 "append-only  directory" or "sticky bit"  flags, respectively.
	 This is quite divergent from the UNIX listing conventions.

         The "append-only directory" bit may be stated for normal files
         on some  systems. This is not the  POSIX-defined behavior, and
         should be treated accordingly with the platform specifics.

         The CRC32 field has been removed  from list  command output on
	 UNIX. The file  type has  been relocated  from grouped list at
	 the rightmost  column to  the  letter prepending  "rwxrwxrwx".
	 Additional  file  type  ID, "u" is  introduced, meaning  "UNIX
	 special file".

         The  "V"  and  "X"  flags ("continues   on   next volume"  and
	 "continued  from  previous volume") have  been  grouped in "S"
	 flag ("section"). It has the following meanings:

         (blank) - file is kept on a single volume
               < - continued from previous volume
               > - continues on the next volume
               * - intermediate section (other parts reside both on the
		   previous and next volumes)

      *  The low 12 bits  are stored  for the  file modes, allowing for
	 POSIX-defined  attributes. When  converted  from  DOS mode and
	 vice-versa, only  the  "read-only" flag  is  meaningful, which
	 corresponds to the write permission for the current user under
	 UNIX.

      *  The  concept  of "UNIX special  files" is  introduced. In  the
         technical implementation, it incorporates all structures which
	 are  unavailable  under  DOS  into a  single  file type, whose
	 extended header  bears the  UNIX-specific information. So  are
	 the  links (both symbolic and  hard links), pipes  and devices
	 (character or block).

         The archiving of special files is  enabled either by "-hbu" or
         "-a1".  Host-specific  special  files, such  as  devices,  are
         checked to  suit the  current platform  type definitions  when
         archive extraction is done.

         If the special files are not enabled, then all of them are not
         included in archive, except for symbolic links, whose contents
         are added under the same name from the resolved link.

         "-hbn" forbids any handling of UNIX special files. This can be
         used in  cases  when  the system  files (devices, pipes, etc.)
	 arrive from a different system or platform.

      *  Hard  links are detected, unless  explicitly disabled by user.
         This is  done  on  per-session  basis  (so, for  links  to  be
	 resolved, the  archiving of  their pertinent  files  should be
	 done in a single ARJ operation). The first file is used as the
	 reference  point  for subsequent  linkages, so it  must  exist
	 before all files linked to it are unpacked.

         The "-2h" option disables the hard  link handling  both on the
         compression  and  extraction  stages. In this  case, the  hard
	 links will not be  resolved  during  archiving,  resulting  in
	 duplicate  content. On extraction of  existing hard links with
	 "-2h", the new files  will not be  linked  to their  reference
	 entries.

         There is also a  "-2h1" option  to convert  the hard  links to
         symbolic links  during extraction. All  links  are directed to
         their  reference file.  This feature is  advantageous when the
         target file system does not support hard links (e.g. HPFS). 

         No other  tracking is  done for  hard links. When  a reference
	 entry is  deleted  from  the  archive, its links  have  to  be
	 located and  deleted manually. No "reference count" is kept in
	 archives!

         On unarchiving, if a hard link  breaks into  cross-device link
         due to different mount points, a warning message is generated.
	 No attempt to establish a symbolic link is made.

      *  The  links  and  special files  are honored  in  hollow  mode,
         contrary to  the extended attributes. Password protection does
	 not apply to them, so, to unpack such entry, the user does not
	 need to have the password, as with any zero-sized file.

      *  The symbolic links, being a sort  of special files, are stored
	 with the target file attributes, and restored with the current
	 system time  and default  attribute mask. To save  and restore
	 their  specific  attributes, the "-2s" switch has to  be  used
	 both  during  archiving  and   extraction.  When  used  during
	 extraction, "-2s" implies that you  have write  permissions to
	 the  place  where  the links do  point to, since  it  involves
	 indirect modification of the target file attributes.

      *  File ownership is preserved and restored when the "-2o" option
	 is given. On  archiving,  there  are  two  options: "-2o"  and
	 "-2o1".

	 "-2o" resolves the user  name to  the character  form, i.e. no
	 direct UID or GID is  stated. Be aware that, when  the file is
         unarchived on  a setup  where a  user  with the  same name and
         different  UID/GID exists, the file  gets the  UID/GID of that
	 particular user. This may lead to problems when unarchiving to
         partitions of one system from another system, but simplify the
	 administrator's job of moving files across  several hosts with
	 a common user base.

	 "-2o1" stores the numeric UID/GID to the archive. It is mostly
	 intended for backup/restore operations.

         "-2o2" is similar to "-2o" but will resolve and save the group
         name to which the user belongs. It is convenient when a single
         user may belong to several GIDs.

      *  It  is  possible to restrict  the  archiving to  a  particular
	 filesystem, or exclude  some  volumes  from the set  of  files
	 being archived.  The "-2b" parameter  enables filtering of the
	 files by their relevant block devices.

         A series  of  one or   more  "-2b+<filename>" parameters  will
	 explicitly limit the    search to a  set  of  volumes. "+"  is
	 optional, so "-2b<filename>" is  completely equivalent to  the
	 above.  "-2b-<filename>", on the   contrary, will exclude  the
	 volumes from search.
 
         It  is most common to specify  the mount point in "<filename>"
	 but any other filename or directory  will be accepted as well.
	 "-2b."  restricts archiving to the current volume. Examples:

	 arj a test / -r -2b/             (store the whole root volume)
	 arj a main / -r -2b-/usr
                     (store all volumes except the one containing /usr)

         There   is no sense   if  both "-2b+"  and  "-2b-" options are
         combined in a single command  line.  Also, it should be  noted
         that giving a physical device  name, such as, "-2b/dev/wd0s3",
         is incorrect and would mean the device containing "/dev/wd0s3"
         entry (usually the root device or devfs) but not /dev/wd0s3.

      *  The "-&" option  has  no effect  under  UNIX. In  fact, it  is
         turned on permanently.

      *  With  the  default  archiving  order, it's  assured  that  the
	 ownership and permission data for a directory is restored only
	 when there are no files  belonging to  this directory ahead in
	 the archive. Besides, the directory  records as such  are only
	 stored with the "-a", "-a1" or "-hbd" parameter.

         If,  however, any  new  files  are appended  to  the  existing
	 archive, or the archiving list has been specified to include a
	 particular  directory  in  two  or  more  locations, then  the
	 directory   information   will   be   finalized   before   the
	 corresponding data files.

      *  ARJSFXV is capable of  handling UNIX special files, hard links
	 and  ownership  data. Other SFX  modules do  not process  this
	 data. ARJSFXV is expected to be readable  on the system  where
	 it is run, besides, it must  be referred  to with  a qualified
	 filename (e.g. "./mysfx"), and not  searched for  through  the
	 path. There is no  SFX extension, unless  explicitly stated by
	 the user. There are native SFX stubs, about third the  size of
	 ARJSFXJR (as of Linux ARJ v 3.10.06). Only the first volume of
	 each SFX set is made executable.

      *  The ANSI comments are displayed as raw text. No terminal setup
         occurs for them.

      *  "-2p<n>" can be used to set the  priority. The priority levels
	 range from  1 (lowest) to  41 (highest), with  default  at 21.
	 Priority  levels  above  21  can   only  be  assigned  by  the
	 super user.

      *  ARJSFXJR  does  not  handle  the  DOS  compatibility  formats.
	 ARJSFXJR archives can only be created with "-2d" set off.

      *  To take advantage of REARJ, a specific set of native archivers
	 has to be  provided (RAR and  InfoZIP are known  to work). The 
	 DOS counterparts can not be utilized. Besides, a new rearj.cfg
	 file  has  to  be created (see /usr/doc/arj/rearj.cfg.example,
	 and  consult  the original  REARJ  documentation  for  further
	 reference). In UNIX version  of REARJ, suffix length limit has
         been raised to 32 characters.


   KNOWN PROBLEMS

      *  Poor  performance  (long  startup  delays)  when  the  shell's
	 wildcard  expansion  takes  effect  for  a  long list  of UNIX
	 special files.


      End of document