[BackupPC.git] / ChangeLog

#========================================================================
#
# ChangeLog - change log for BackupPC.
#
# DESCRIPTION
#   Revision history for BackupPC, detailing significant changes between
#   versions, most recent first.
#
# AUTHOR
#   Craig Barratt  <cbarratt@users.sourceforge.net>
#
#========================================================================
#
# Version __VERSION__, released __RELEASEDATE__.
#
# See http://backuppc.sourceforge.net.
#
#========================================================================

#------------------------------------------------------------------------
# Version __VERSION__, __RELEASEDATE__
#------------------------------------------------------------------------

* Support for rsync and rsyncd backup and restore.  Changes to
  BackupPC_dump, BackupPC_restore, and new modules BackupPC::Xfer::Rsync
  and BackupPC::Xfer::RsyncFileIO.

* Added internationalization (i18n) code from Xavier Nicollet,
  with additions from Guillaume Filion.  Voila!  BackupPC_Admin
  now supports English and French, and adding more languages is
  now easy.  New config paramater $Conf{Language} sets the language.

* Added optional user-defined pre/post dump/restore commands, allowing
  things like database shutdown/startup for dumps.

* Changed the way hosts are found.

* Added $Conf{ClientNameAlias}, which allows the name of the physical
  client machine to be set.  This allows several different backup
  "hosts" to all refer to the same physical machine, which is
  convenient if several different types of data need to be backed
  up, or if different parameters are needed for different parts of
  the host.

* Replaced $Conf{PingArgs} with $Conf{PingCmd}, added $Conf{DfCmd},
  $Conf{NmbLookupCmd} allowing all these commands to be fully
  configured.  Also, all commands can also now be fragments of
  perl code.

* Moved all smbclient commands into the config.pl file so the specific
  arguments can be customized.  New config parameters are
  $Conf{SmbClientFullCmd}, $Conf{SmbClientIncrCmd} and
  $Conf{SmbClientRestoreCmd}.

* Added new BackupPC::View module that creates views of backups
  (handling merging etc).  Updated BackupPC_Admin, BackupPC_zipCreate
  and BackupPC_tarCreate to use BackupPC::View.  This removes lots
  of merging and mangling code from the higher-level code.

* Added code from Toby Johnson that allows additional users to be
  specified in the hosts file; these users can also view/start/stop
  and restore backups for that host.  Also added a new config
  setting $Conf{CgiNavBarAdminAllHosts} that allows all hosts to
  be listed in the left nav bar for admins.

* Added $Conf{HardLinkMax} (default 31999) which sets the limit on
  the maximum number of hardlinks per file in the pool.  If a file
  ever gets to this number of links a new pool file is created to
  handle additional links.

* Added $Conf{PerlModuleLoad}, which allows optional additional perl
  modules to be loaded.

* Added $Conf{EMailUserDestDomain} and other EMail config settings to
  allow language-specific default messages to be overridden.

* Added BPC_FTYPE_DELETED to lib/BackupPC/Attrib.pm, allowing deleted
  files to be represented in the attrib file correctly.

* Added support for environment variable BPC_SMB_PASSWD, which is the
  client's smb password.  This overrides the old environment variable
  PASSWD.

* Added taint cleanup for perl5.8 to lib/BackupPC/Lib.pm.

* Changed $tar_unpack_header format in BackupPC_tarExtract to correctly
  handle files with trailing spaces.

* Added catching of SIG_PIPE to BackupPC_dump, and changed catch_signal
  to ignore multiple signals of the same type.

* Added reporting of the largest number of hardlinks in the pool to the
  log file.

* Adding reporting of syntax errors in the per-PC config.pl file.

* Updated BackupPC_sendEmail to handle language-specific email messages.

* Allow client (host) names to contain spaces.  Spaces in host names
  need to be escaped via "\" in the hosts file.  The user of spaces in
  host names is discouraged, but they should work.  One feature that
  doesn't work with host names that contain spaces is the highlighting
  of that name in the log file display in the CGI interface.  There are
  no plans to fix this.

* Renamed $Conf{SmbClientTimeout} to $Conf{ClientTimeout}.

* Fixed all open() calls to use 3 argument form to fix handling of file
  names with trailing whitespace.  Also fixed CGI interface so these
  file names are displayed correctly.

* Fixed new 2.0.0 CGI navigation bug that causes the top-level directory
  to have a URL "&share=//boot&dir=" instead of "&share=/boot&dir=/".
  Reported by Pascal Schelcher.  Fixed similar problem reported by
  Doug Lytle.

* Added "PerlTaintCheck On" to the mod_perl section in the docs,
  suggested by Tim Demarest.

#------------------------------------------------------------------------
# Version 1.5.0, 2 Aug 2002
#------------------------------------------------------------------------

* Changed conf/config.pl so that $Conf{TarIncrArgs} uses the --newer
  option instead of --newer-mtime.  Also removed --atime-preserve from
  $Conf{TarClientCmd}.  This makes the default settings work better
  with tripwire.

* Fixed configure.pl so it correctly detects a running BackupPC <= v1.4.0
  so it can correctly warn the user to stop it before upgrading.  Reported
  by David Holland.

* Added missing ";" to entity escape in EscapeHTML in BackupPC_Admin.
  Reported by Guillaume Filion.

* Added LDAP setup to documentation from David Holland.

* Tar.pm now adds a "." to file paths that start with "/", so that all
  tar paths are relative.  From Ludovic Drolez.

#------------------------------------------------------------------------
# Version 1.5.0beta0, 30 Jun 2002
#------------------------------------------------------------------------

* A full set of restore options is now supported, including direct
  restore via smbclient or tar or downloading a zip or tar file.

* Major additions to CGI script to support better directory navigation,
  restore features and mod_perl.  Also, file downloads from the CGI
  interface now correctly preserve the file name and provide the
  correct Content-Type for the most common types of files.  Improved
  directory navigation was contributed by Ryan Kucera.

* New script BackupPC_zipCreate (contributed by Guillaume Filion) is the
  zip analog of BackupPC_tarCreate.  BackupPC_zipCreate can be used to
  create a zip archive of any portion of a backup.

* Substantial additions to BackupPC_tarCreate to support restore,
  including modifying path names, handling hardlinks, fixing
  support of old backups without attributes (pre-v1.4.0).  Plus
  BackupPC_tarCreate is now an offical part of the release.
  (Lack of support for hardlinks was reported by John Stanley.)

* BackupPC_tarExtract now supports hardlinks and fixed pooling of
  attribute files.

* A unix domain socket is now used for communication between the CGI
  interface and BackupPC.  The original TCP socket is optional.  Sockets
  are correctly re-initialized if config.pl is updated with new socket
  settings.

* For improved security messages over the unix or TCP socket are protected
  via an MD5 digest based on a shared secret, a sequence number, a time
  stamp and a unique per-connection number.

* Additions to configure.pl to support install of directory navigation
  images.

* Fixed case where $Conf{BackupFilesOnly} or $Conf{BackupFilesExclude}
  were set to a single string or list (in v1.4.0 only the case of
  hash worked correctly).  Reported by Phillip Bertolus.

* Fixed case of $Conf{BackoutGoodCnt} == 0.  This setting now makes the
  client always subject to blackout, matching the comments in config.pl.
  Also fixed handling of $Conf{BackoutGoodCnt} < 0 in the CGI script
  reported by Pascal Schelcher.

* Fixed byte and file totals for tar backups, reported by several users.

* Fixed --newer-mtime date/timestamp format to make it ISO 8601 compliant,
  suggested by Erminio Baranzini.

* Fixed handling of $Conf{BackupFilesOnly} in BackupPC::Xfer::Tar.pm, as
  well as shell escaping of tar arguments.

* Fixed entity encoding of 8-bit characters in the CGI interface.

* Added optional CGI headers in $Conf{CgiHeaders} that by default
  is set to a no-cache pragma.  Suggested by Benno Zuure.

#------------------------------------------------------------------------
# Version 1.4.0, 16 Mar 2002
#------------------------------------------------------------------------

* BackupPC now supports tar (in addition to smb) for extracting host    
  data.  This is the most convenient option for linux/unix hosts.
  Tar can be configured to run over ssh, rsh or to backup a local
  nfs mount from the host.

* Support for special files, including symbolic links, fifo, character
  and block device files has been added, so that all native linux/unix
  file types can be correctly backed up when using tar transport.
  Special files are all stored as regular files and the type attributes
  are used to remember the original file type.

* All unix file attributes are now saved (and pooled when possible).
  This includes user and group ownership, permissions, and modification
  time.  Smbclient also does a reasonable job of emulating unix 
  permissions (such as mtime), and these attributes get saved too.

* The new default is to not fill incremental dumps.  configure.pl
  automatically sets $Conf{IncrFill} to 0.  The default was 1
  (incrementals were filled with hardlinks).  Since the CGI
  script does filling at browsing time, there is no need to
  fill incremental dumps.

* Backup file names are now stored in "mangled" form. Each node of a
  path is preceded by "f", and special characters (\n, \r, % and /) are
  URI-encoded as "%xx", where xx is the ascii character's hex value. So
  c:/craig/example.txt is now stored as fc/fcraig/fexample.txt. This
  was done mainly so meta-data could be stored alongside the backup
  files without name collisions.  In particular, the attributes for the
  files in a directory are stored in a file called "attrib", and
  mangling avoids file name collisions (I discarded the idea of having
  a duplicate directory tree for every backup just to store the
  attributes). Other meta-data (eg: rsync checksums) could be stored in
  file names preceded by, eg, "c".  There are two other benefits to
  mangling: the share name might contain "/" (eg: "/home/craig" for tar
  transport), and I wanted that represented as a single level in the
  storage tree.  Secondly, as files are written to NewFileList for later
  processing by BackupPC_link, embedded newlines in the file's path
  will cause problems which are avoided by mangling.

  The CGI script undoes the mangling, so it is invisibe to the user.
  Of course, old (unmangled) backups are still supported by the CGI
  interface.

* Various changes to the CGI interface, BackupPC_Admin:

  + Added button that allows users to manually start a full dump in
    addition to the existing incremental dump.

  + Added display of file attributes when browsing backups.

  + Added an optional holdoff time specified by the user when canceling
    a backup. BackupPC will not attempt any new backups for at least the
    specified time. This holdoff time can be changed whether or not a
    backup is running.

  + Added supports for file mangling, and correct merging of unfilled
    backups from mangled or unmangled (and compressed or uncompressed)
    fulls when browsing or restoring.

  + Only displays a "Start Incr Backup" button if there are already some
    backups.

  + For DHCP hosts, when a user tries to manually start a backup, add
    a check for the netbios name of both the host the request came
    from (REMOTE_ADDR) and the last known DHCP address for that host
    to see if either address matches the host. If not, an error
    message is display. The previous behavior was that only requests
    from the client itself succeeded, and requests from other machines
    quietly failed.

* Changed the version numbering to X.Y.Z, instead of X.0Y. This release
  is 1.4.0. The first digit is for major new releases, the middle digit
  is for significant feature releases and improvements, and the last
  digit is for bug fixes. You should think of the old 1.00, 1.01, 1.02
  and 1.03 as 1.0.0, ..., 1.3.0.

* BackupPC and the CGI script BackupPC_Admin now check that the effective
  user id is correct to avoid accidentally launching BackupPC as the
  wrong user or detecting CGI configuration problems.  This behavior
  can be turned off using the $Conf{BackupPCUserVerify} option.

* In numerous places changed "Smb" to "Xfer" (eg: log file names) to
  support generic names for both smb and tar transport methods.  The
  CGI script checks for old names for backward compatibility.

* Major changed to Backup_dump to support new tar transport.  All transport
  specific code moved into BackupPC::Xfer::Smb and BackupPC::Xfer::Tar
  objects.

* Added workaround for a bug in Samba's smbclient for files between 2GB
  and 4GB.  The file size in the tar header is incorrect.  This allows
  files up to 4GB to work with smbclient, rather than 2GB.  To support
  files larger than 2GB you must make sure perl is compiled with the
  uselargefiles option (use "perl -V | egrep largefiles" to check) and
  the pool directory must be on a file system that supports large files.

* Moved the pool writing code into a module BackupPC::PoolWrite.  This
  allows the clever file pool checking (digest, uncompressing, comparing
  etc with minimum disk IO) to be used easily in multiple places (eg: it
  is now used for writing attribute files so they can be pooled).

* Changed MD5 to Digest::MD5 to avoid use of the depreceated MD5 module.

* Shortened default $Conf{MyPath} so that perl's taint mode is more likely
  to be happy.  The old $Conf{MyPath} contained /usr/local/bin, which
  on one user's machine was world writable and perl -T correctly
  complained about it.

* Fixed ping command options in Lib.pm so that it works on OpenBSD.
  Thanks to Kyle Amon for sending the fix.  Decided to move the
  ping options from Lib.pm into config.pl (as $Conf{PingArgs}) and
  now configure.pl tries to come up with a sensible default based on
  the OS.

* Fixed argument checking in BackupPC_tarExtract to allow '$' in the
  share name (eg: C$).  Thanks to Jules Agee for this fix.  Also
  changed the default config.pl so that single quotes are used
  everywhere so that people don't get tripped up putting '$' inside
  double-quoted strings.

#------------------------------------------------------------------------
# Version 1.03, 9 Dec 2001
#------------------------------------------------------------------------

* BackupPC now has full support for compression.  There are now two
  pool areas, the original pool for uncompressed files, and cpool for
  compressed files. The compression is done by Compress::Zlib.
  Compression reduces the pool disk usage by around 40%, although your
  mileage may vary.  Compression is optional and can also be specified on
  a per-PC basis (although this will cost more pool storage since many
  backup files will have to be stored in both compressed and
  uncompressed forms.

* A new script, BackupPC_compressPool, can be run to compress the entire
  pool.  This is used once to migrate all the pool data from uncompressed
  to compressed on existing installations.  Read the documentation
  (Installing BackupPC/Compressing an existing pool) before running
  BackupPC_compressPool!

  Alternatively, compression can simply be turned on and all new backups
  will be compressed. Both old (uncompressed) and new (compressed)
  backups can be browsed and viewed. Eventually, the old backups will
  expire and all the pool data will be compressed.  However, until the
  old backups expire, this approach could require 60% or more additional
  pool storage space to store both uncompressed and compressed versions
  of the backup files.

* Significant improvements to the cgi interface, BackupPC_Admin:

    - much better layout navigation
    - handles compressed backup files and compressed log files
    - handles unfilled incremental dumps
    - better backup directory browsing navigation
    - reports compression statistics
    - $Conf{CgiDateFormatMMDD} allows you to set date format (MM/DD or DD/MM)
    - Additional customization with $Conf{CgiHeaderFontType},
      $Conf{CgiHeaderFontSize}, $Conf{CgiNavBarBgColor}, and
      $Conf{CgiHeaderBgColor}.

* Eliminated BackupPC_queueAll.  BackupPC directly reads the hosts
  file and queues the PCs itself.  Like config.pl, BackupPC will
  re-read the hosts file on each wakeup if its modification time
  changes, or upon a SIGHUP.  This also makes for better behavior
  when adding a host: if you add hosts, simply send a SIGHUP to
  BackupPC or wait for the next wakeup.

* BackupPC_dump now compresses the SmbLOG file if compression is enabled.

* BackupPC_dump keeps track of compressed file sizes so that compression
  statistics can be reported by the cgi interface.

* Aging of old log files now handles compressed log files (.z extension).

* Added configuration option $Conf{IncrFill} to specify whether
  incremental dumps should be filled in.  Old behavior was that
  filling was on.  Now it's optional.  See config.pl for more
  details.

* BackupPC_nightly now cleans and generates statistics for both
  the uncompressed pool and compressed pool (cpool).

* Added new utility script BackupPC_zcat that can be used to
  uncompresses BackupPC files.

* configure.pl offers various options related to compression,
  depending upon whether this is a new install or upgrade,
  and whether or not Compress::Zlib is installed.

* configure.pl now makes a backup copy of config.pl before
  config.pl is updated.

* added three new fields to the backups file to handle optional
  filling and compression stats.

* Added -e option to BackupPC_dump.  BackupPC now invokes BackupPC_dump -e
  on each dhcp host once each night to verify that very old backups are
  expired.  This ensures that very old backups are expired even if
  the dhcp host has not been on the network for a long time.

* fixed bug in BackupPC::FileZIO.pm that required Compress::Zlib,
  even if compression was off.  Thanks to Steve Holmes for reporting
  this.

* fixed bug that caused a BackupPC queue to get blocked when a backup
  cancel attempt was made during the BackupPC_link phase.

#------------------------------------------------------------------------
# Version 1.02, 28 Oct 2001.
#------------------------------------------------------------------------

* Added new script BackupPC_tarExtract to extract the smbclient tar
  archive. This reduces disk writes by perhaps 90-95% and disk reads by
  50%. Previously, tar was used to extract and write everything to disk.
  Then BackupPC_dump would read enough of each file to compute the MD5
  digest, and then compare the full file with candidate pool files.  So
  for each 1MB file that matches a single file in the pool, there would
  be 1MB of disk writes and 2MB of disk reads (to compare two 1MB files).

  BackupPC_tarExtract instead extracts the archive using a 1MB memory
  buffer.  This allows the MD5 digest to be computed without touching the
  disk.  Next, any potential pool file compares are done by comparing the
  pool file against the incoming tar data in memory, which only requires
  the pool file to be read.  So for each 1MB file that matches a single
  file in the pool, there are now no disk writes, and only 1MB of reads.
  BackupPC_tarExtract handles arbitrary size files and repeated
  potential pool matches.  If the incoming file doesn't match the pool
  then it is written to disk (once the pool is mature this happens maybe
  5-10% of the time).

* Substantial changes to BackupPC_dump:

    + BackupPC_tarExtract is now used in place of tar.

    + BackupPC_dump now reads the output from both smbclient and
      BackupPC_tarExtract and merges them into SmbLOG.

    + Named pipes are no longer used to connect smbclient to tar
      (now BackupPC_tarExtract).  Regular pipes are used instead.
      This avoids the need to system mknod or mkfifo.

    + Locked files on the client that can't be read by smbclient
      previously were filled with 0x0 bytes by smbclient, meaning
      tar extracted a useless file filled with 0x0 bytes.  Now,
      BackupPC_dump watches the output of smbclient and removes
      any files that smbclient couldn't read.  This avoids storing
      useless files.  It tries to replace such files with a hard link
      to a previous dump.  These actions appear in the log file.

* added new module lib/BackupPC/FileZIO.pm. This handles pool file
  I/O and is used by BackupPC_tarExtract.  BackupPC::FileIO supports
  reading and writing compressed and regular files and provides all the
  hooks for compression support in BackupPC (should be supported in next
  version). BackupPC::FileIO also does efficient writing of files that
  contain leading 0x0 bytes (by seeking past the 0x0 bytes). This is
  helpful when smbclient reads a locked file, and it fills the tar
  output with a file of the correct size but all 0x0.  Such files will be
  later removed by BackupPC_dump.  But in the meantime, BackupPC::FileIO
  writes such files efficiently (as sparse files), meaning just a few
  blocks of disk space will be needed even if the file is large.

* alive/dead counting for blackout now works correctly for DHCP hosts.

* BackupPC resets activeJob on startup, to fix bug when BackupPC was
  killed and restarted with backups running.

* added extra non blocking select() in BackupPC to make sure the socket
  reads don't block.

* BackupPC avoids queuing multiple BackupPC_queueAll's on the CmdQueue.

* Updated BackupPC_sendEmail to correctly parse the locked file
  error from 2.2.1a smbclient, so that missing Outlook file emails
  can be correctly sent.

* Changed HostInfoRead() in lib/BackupPC/Lib.pm to lowercase the
  hostname read from the hosts file.

* BackupPC_Admin provides general summary when the host name is empty.

* configure.pl (and BackupPC) now requires perl 5.6.0 or later.

* configure.pl complains if BackupPC is already running, reminding you
  to stop it before upgrading.

* updated documentation, and fixed auto-insertion of config.pl into
  BackupPC.pod (previously the last config parameter was left out of
  BackupPC.pod).

#------------------------------------------------------------------------
# Version 1.01, 30 Sep 2001
#------------------------------------------------------------------------

* Documentation cleanup in README, doc/BackupPC.pod, conf/config.pl.

* BackupPC_sendMail now reads the optional per-PC config file, allowing
  email configuration parameters to be set on a per-PC basis.

* Removed the unused 4096-length MD5 digest code in lib/BackupPC/Lib.pm.

#------------------------------------------------------------------------
# Version 1.00, 21 Sep 2001
#------------------------------------------------------------------------

* Initial release of BackupPC on sourceforge.net.