Imported Debian patch 2.23.05-1

[hcoop/zz_old/debian/webalizer.git] / CHANGES

--------------------------------------------------------------------
2.23-xx changes from 2.21-xx
--------------------------------------------------------------------

Fixes:
 o Fix sporadic eol problem with some IIS/W3C logs

 o Fix compiler directive syntax error (broke some 64 bit systems)

Changes/Additions:
 o Modest speed improvements in hash table code

--------------------------------------------------------------------
2.21-xx changes from 2.20-xx
--------------------------------------------------------------------

Fixes:
 o Added missing memory deallocation call in DNS lookup code.

 o Minor fixes to configure script

Changes/Additions:
 o Added "YearTotals" config option for main index page totals

 o Rename local stricmp() function to ouricmp() to prevent name
   confilict on systems that happen to provide it already.

--------------------------------------------------------------------
2.20-xx changes from 2.01-xx
--------------------------------------------------------------------

Fixes:
 o Fixed problem with timing totals.

 o Fixed referrer linking to avoid possible xss injection.

 o Fixed month change detection error that caused incorrect report
   dates when logs had a 'gap' longer than a year.

 o Fixed buffer overrun possibility in parsing code and user agent
   mangle logic.

 o Added symbolic link checks for file I/O to prevent possible
   privilege escalation exploits.  Disallows reading from or writing
   to any file that is a symlink.  Thanks to Julien Danjou.

 o Added code to preserve the history and incremental data files in
   the event of a crash before writing to them completely.  Thanks
   to Robert Millan for the idea and initial code.

Changes/Additions:
 o Added native geolocation services, which fully supports both IPv4
   and IPv6 lookups.  Adds the configuration keywords 'GeoDB' and
   'GeoDBDatabase' along with the '-j' and '-J' command line options.

 o Added 'wcmgr', "The Webalizer (DNS) Cache file Manager" to the
   distribution to provide cache file maintenance.  See the supplied
   man page for a description and usage information.

 o Changed history code and main index page to allow for more than
   12 months of reports to be displayed.  Added the config keywords
   'IndexMonths' (-K command line option), 'GraphMonths' (-k command
   line option) and 'YearHeaders' to control how index is displayed.

 o Changed Berkeley DB code to use current 4.x APIs.

 o Added support for bzip2 compressed log files (.bz2) as a compile
   time option (--enable-bz2).  If enabled, bzipped files will be
   decompressed automatically during processing.

 o Added support for W3C formatted logs.  Based on code submitted
   by Klaus Reimer.

 o Added GeoIP support as compile time option (--enable-geoip). Adds
   'GeoIP' and 'GeoIPDatabase' config keywords, '-w' and '-W'
   command line options. (http://www.maxmind.com/)

 o Added IPv6 support.  Based on initial code by Jose Carlos Meneiros
   and modified to support Solaris and other problematic platforms.

 o Added 'CacheIPs' config option to allow saving unresolved addresses
   in the DNS cache.

 o Added 'CacheTTL' config option which allows the DNS cache time to
   live (TTL) value to be specified at run-time.

 o Added 'SearchCaseI' config option to specify if search strings
   should be treated as case insensitive or not.  The default value,
   'yes', causes search strings to be treated as case insensitive.

 o Added 'HTAccess' config option.  Allows writing a default .htaccess
   file to the output directory.

 o Added ability to display flags in the top country table.  Adds the
   config keywords 'CountryFlags' and 'FlagDir', and -z command line
   option.

 o Added 'StripCGI' config option to configure how CGI variables on
   the end of URLs are treated (can now be stripped or left in place).

 o Added 'DefaultIndex' config option to enable/disable the use of
   "index." as a default index name to be stripped from the end of URLs.

 o Added 'TrimSquidURL' config option to allow squid log URLs to be
   reduced in granularity by a user definable amount.  Thanks to code
   submitted by Stuart Gall.

 o Added 'OmitPage' config option (and the '-O' command line switch)
   to prevent specified URLs from being counted as pages even if they
   otherwise would be.  Thanks to code submitted by Adam Morton.

 o Added 'IgnoreState' config option (and the -b command line switch)
   to allow ignoring any existing incremental data file (similar to
   the IgnoreHist/-i option).

 o Changed logic to always generate summary report (index.html),
   even if no records were processed.

 o Added color support to allow changing graph colors.  Based on the
   Webalizer-usecolor code submitted by Benoit Rouits.  Adds 11 new
   config options, see the README file for complete descriptions.

 o Added language 'lang=' specification in generated HTML files.
   
 o Added 'LinkReferrer' config option to allow/disallow links in the
   top referrers table.

 o Added 'PagePrefix' config option to allow URL prefix matches to
   be counted as pages, regardless of file extension or type.  Thanks
   to code submitted by Remco Van de Meent.

 o Enabled large file support (LFS) to support logs greater than 2Gb
   in size on systems that support LFS.  Also increased the size of
   most internal counters to handle larger sites.

 o Minor changes to generated HTML output

 o Updated language files country codes for current IANA TLDs

 o Changed the meaning of the -v command line switch.  It now
   causes verbose information to be displayed at run-time
   (Informational and Debug messages).

 o Changed Group* config options to allow a quoted string for
   the match string.  This allows spaces to be embedded in the
   string.

 o Changed log record parsing logic to allow spaces in URLs.

 o Made configuration keywords, boolean configuration values
   (yes/no), and log file types case insensitive.  Also fixed
   defaults for invalid values to reflect documented defaults.

 o Changed configure script to use --sysconfdir to specify the
   location of the default webalizer.conf configuration file.
   Also added support for DESTDIR during install to aid binary
   package builds.

--------------------------------------------------------------------
2.01-xx changes from 1.30-04
--------------------------------------------------------------------

Fixes:
 o Fix posible obscure buffer overflow bug in DNS resolver code

 o Added additional extended character fixes

 o Let code accept partial content response codes along with 200's

 o Added code to catch blank hostnames (yes, they have been found!)
   Will convert them into 'Unknown'

 o Security fix for cross-site scripting vulnerability found by
   Flavio Veloso (www.magnux.com).

 o Fixed a TOTAL_RC off by one error, which would prevent the last
   response code from being saved when using incremental mode.

 o Fixed possible segfault condition in MangleAgent code on
   some malformed user agent names.

 o Fixed DNS to prevent hangs on blank and malformed hostnames.

 o Fixed problem calculating visits.  Changed timestamps to use
   seconds since epoch (1/1/1970) which results in more accurate
   analysis.  Also changed normal out of sequence code to handle
   up to 1 hour of 'slop' in the timestamps.  This changed the
   semantics of the VisitTimeout and -m configuration options, as
   the values are now specified in number of seconds.

 o Fixed hostname lowercase problem (wasn't) when using DNS lookups.

 o Fixed problem with incremental datafile which could cause a read
   error under certain circumstances (removes control characters).
   Also changed code to now abort on a read error.

 o Fixed problem with hash table node creation where objects that
   were exactly the maximum length would wind up leaving a garbage
   byte at the end of the memory space allocated.  This was causing
   some very infrequent and widely different problems.

 o Fixed problem where country graph could be produced incorrectly
   if using a non-english language and the country name overlapped
   the pie chart.

 o Found and fixed a problem with a possible 32-bit wrap around
   problem using incremental mode on large sites.  The problem
   would cause the KBytes data on large groups to become inaccuate.

Changes/Additions:
 o Modified configure to allow specification of the default config
   directory.  If not given, will use /etc (/etc/webalizer.conf).

 o Added DailyGraph and DailyStats configuration options to enable
   or disable the Daily usage graph and stats table from output.

 o Improved visit calculation logic to reduce 'false' counts generated
   by external image referrals.

 o Added reverse DNS lookup capability.  This adds the command
   line switchs -D and -N, and configuration keywords "DNSCache"
   and "DNSChildren".  See the DNS.README for additional info.
   Based in part on code submitted by Henning P. Schmiedehausen
   (hps@tanstaafl.de).

 o Added ability to dump Sites, URLs, Referrers, User Agents,
   Usernames and Search Strings to tab delimited files, suitable
   for import into most database and spreadsheet programs.  The
   location of this file may be specified using the "DumpPath"
   configuration keyword, allowing the data to be kept someplace
   outside the web servers document tree.  The configuration
   keywords "DumpSites", "DumpURLs", "DumpReferrers", "DumpAgents",
   "DumpUsers" and "DumpSearchStr"  have been added to control the
   file dumps.  Column headers can be included in the file with
   the "DumpHeader" keyword.  Dump filename extensions may be
   specified using the "DumpExtension" keyword (default is .tab).

 o Added username analysis, based on usernames found in the log,
   and only available if username information is present in the
   log (ie: http authentication or wu-ftpd xferlog).  The keywords
   'GroupUser', 'HideUser', 'IgnoreUser', 'IncludeUser', 'AllUsers',
   and 'TopUsers' have been added to the configuration file code.
   This change also modified the format of the incremental data file.

 o Added the ability to display ALL sites, URLs, Referrers,
   User Agents and Search Strings on a seperate HTML page from
   the normal statistics page.  This adds the configuration
   keywords 'AllSites', 'AllURLs', 'AllReferrers', 'AllAgents'
   and 'AllSearchStr', which can have either a "yes" or "no"
   value (default is "no").  Will add a "View All..." link to
   the bottom of the appropriate "Top" table if enabled.

 o Added support for squid proxy logs, thanks to code submitted
   by Steinar H. Gunderson (sgunderson@bigfoot.com).  To use
   squid logs, specify a LogType of 'squid' in the configuration
   file.  This also changed the behaviour of the '-F' command
   line switch, which now requires a second argument of either
   'clf', 'ftp' or 'squid'.

 o Completely modified the way the various TOP tables are handled
   and sorted, which now allows extremely large top tables without
   any performance degredation.  Previously, tables greater than
   a few hundred elements produced a noticable perfomance penalty
   during processing.

 o Added the ability to group domains automatically and to hide
   individual host names from the report, using the 'GroupDomains'
   and 'HideAllSites' configuration keywords (-g and -X command
   line options).  Domain Grouping is configurable as to the level
   of grouping (second level domain, third, etc...).  HideAllSites
   forces only grouped site records to be displayed if any.  Based
   on ideas/code by Michael Klemme (mklemme@gmx.de).  This changes
   the behaviour of the '-g' switch, which previously was used to
   force the use of GMT time for reports.

 o Added user configurable search engine specification, used for
   search string analysis.  This adds the 'SearchEngine' keyword
   in configuration files.  Based on idea/code by Alexey Kizilov.

 o Changed code to use the latest version of GD which supports PNG
   images instead of GIF images.  Also included changes in configure
   script to ensure the presence of the libpng and libz libraries.

 o Added ability to override log file to STDIN by use of '-' on
   the command line.

 o Added gzipped logfile support.  The program will automatically
   detect logfiles with a '.gz' extension and uncompress on the
   fly.  Uses gz file support of zlib, since it's required for
   our gd/png stuff anyway.  Please note that using gzipped logs
   will incur a small performance penality.

 o Minor changes to search string code to increase accuracy.  This
   also removes a previous condition that would occasionally cause
   search strings to incorrectly be counted twice or to be counted
   as different search strings when only differing by a space.

 o Minor changes to URL parse code to allow additional characters.
   Also changed unescape code to properly handle extended chars.

 o Major changes to hash table node format for reduced memory usage.
   Instead of fixed size strings, the new format will dynamically
   allocate string memory and use pointers to existing table data
   under certain circumstances.  The memory savings is significant
   and will be greatly noticed with large sites.  Because of these
   changes, the formatting of the incremental data file had to be
   changed, therefore it is incompatible with previous versions.

 o Major code reorganization and cleanup.  This was to facilitate
   future developent and make things more managable.

 o Usual documentation updates for new features/functions.

--------------------------------------------------------------------
1.30-xx changes from 1.22-06
--------------------------------------------------------------------

Fixes:
 o Fixed minor bug that would allow incorrect site totals for the
   first day of the month under certain conditions.

Changes/Additions:
 o Added Top Entry and Exit Page tables.  Added configuration file
   keywords TopEntry (-e command line) and TopExit (-E command line)
   to specify the number of entries to display for each table.  The
   default for both is 10.  See README for additional information.
   
 o Added 'Group' labels.  Allows display of a specified label for
   grouped entries (in 'Top' tables).  Based on patch submitted
   by Oliver Graf (ograf@rhein-zeitung.de).  See sample.conf for
   examples.

 o Added 'Visits' totals.  The length of time that constitutes a
   'visit' can be set using the VisitTimeout configuration keyword
   (-m command line option).  The value must be given in HHMMSS
   format, you can omit leading zeros. Default is 30 minutes (3000).

 o Added 'Pages' totals, based on user specified extensions.  Changes
   made to generated graphs as well.  Configuration keyword PageType
   (and command line -P switch) allows specification of extensions
   to use (defaults to 'htm*' and 'cgi').  Also called "pageviews".

 o Added Search String analysis.  Keyword 'TopSearch' defines how
   many of the top search strings to display.  Default is 20. Can
   be disabled by using zero (0).

 o Added native support for ftp logs (xferlog ala wu-ftpd).  Added
   'LogType' configuration file keyword (-F command line option)
   to specify log type.  Values can be either 'web' or 'ftp', with
   the default of 'web'.

 o Changed graphs to handle pages and visits totals. Also  added
   color coded legends, which can be disabled using the GraphLegend 
   configuration keyword (-L command line option).  Default is to
   display them.

 o Added background lines to graphs.  Default is 2 lines, and can
   be set to any number using the GraphLines configuration keyword
   (-l command line option).  Can use anywhere from none (0) to
   twenty lines.  They will be drawn in all but the country graph.

 o Added CountryGraph configuration file keyword (-Y command line
   option) to enable/disable display of country usage pie chart.

 o Added FoldSeqErr keyword (-f command line option).  Normally,
   the program will ignore log records that are out of sequence
   (chronological order).  This option lets them be folded into
   the analysis anyway, as if the were the same date/time as the
   last good record.  Apache users can safely ignore :)

 o Added additonal 'Top' tables for SITES and URLs, sorted by
   KBytes instead of hits.  Two new configuration file keywords,
   TopKSites and TopKURLs, can be used to specify the number of
   entries for each (zero to disable).  Default for both is 10.

 o Added additional calculations for max/avg files, pages, visits
   and KBytes in monthly statistics.

 o Updated generated HTML code to fully comply with the HTML 4.0
   Transitional spec.  DOCTYPE header reflects this change as well.

 o Changed code to use 4 digit years in filenames.  Purely for the
   Y2K phobes who couldn't deal with only two digits (even though
   it was _purely_ for humans, the program couldn't care less).
   Unfortunately, this means that you will have to rename previous
   month files to the new format.  Not a big deal if you plan on
   re-running all your logs to take advantage of the new features.

 o Major changes to both history file and incremental file formats
   to handle additional totals (pages/visits data).  As a result,
   this version is INCOMPATABLE with previous versions.  See the
   file README.FIRST for important information on upgrading.

 o Language files and documentation updated for new functions.

--------------------------------------------------------------------
1.22-xx changes from 1.20-11
--------------------------------------------------------------------

Fixes:
 o Fixed bug in country total generation.  Caused country table
   to show bogus entries if logs contain hostnames that were not
   fully qualified (ie: don't have the domain name/TLD portion).

 o Changed/fixed incremental data I/O routines to better detect and
   handle error conditions.  This involved some minor incremental
   data file format changes as well.  Fixes problem large sites were
   having where random tables were getting munged.

 o Fixed record parse code to better detect and strip query portion
   from URLs and Referrer strings.

 o Fixed segfault condition when more than MAX_CTRY entries were
   specified for the "Top Countries" table.

Changes/Additions:
 o Added code to detect negative byte transfer sizes in logs (another
   netscape server kludge :)  Could cause KByte xfer sizes to become
   corrupt.

 o Several small changes (mostly ifdef/endif's) to make code compile
   clean 'out-of-the-box' across more platforms (ala SunOS).  Also
   added a GNU autoconf 'configure' script which helps a bit as well.

 o Added Include* keywords.  Allows forcing the inclusion of specified
   log records.  Takes precedence over counterpart Ignore* keywords.

 o Added HTMLPre, HTMLBody, HTMLEnd and HTMLExtension keywords, and
   changed behaviour of HTMLHead keyword.  Previous versions need
   only change the 'HTMLHead' keword in existing files to 'HTMLBody'
   to upgrade.  Thanks to Colin Viebrock <cmv@privateworld.com> for
   the idea and code examples.

 o Changed mangle agent code to support Opera and other browsers.
   Also updated response codes to IETF HTTP/1.1 Rev 6 draft.
   Thanks to Yves Lafon <ylafon@w3.org> for this these.

 o Added HistoryName and IncrementalName keywords, which allow the
   specification of the history and incremental data filenames.

 o Added UseHTTPS keyword, which allows using 'https://' instead
   of 'http://' for links to URLS in the 'Top URLs' table.  Also
   added check for URLs that already have the protocol specified
   (such as on virtual web and proxy servers), and to use unmodified
   if found (will only force to lowercase for matching).

 o Added code to ignore out-of-sequence log records.

 o Added code to force hostnames to lowercase (was causing country skew).

 o Disabled display of blank (zero hit) days at start of daily stat table.

 o Added records per second calculation to timing totals.

 o ALT= tags now use translated strings instead of forcing english.

 o Updated documentation for new functions/features.

--------------------------------------------------------------------
1.20-xx changes from 1.12-10
--------------------------------------------------------------------

Fixes:
 o Modified record parse routine to not touch stuff between quotes
   (").  Was causing problems parsing some malformed request fields.

 o Fixed memory leak in MangleAgent code, and relocated to elimitate
   un-necessary processing (causing segfault on some machines).

Changes/Additions:
 o Changed transfer totals on host/url structures to support large
   groupings (such as *.gif) on heavly hit servers.  Hopefully, this
   should cure the 32bit overflow problem large sites were having.

 o Changed daily transfer totals to support transfers greater than
   roughly 4.2 gigabytes a day.

 o Added some missing HTML tags and altered the way totals are
   calculated on the 'Top' tables (to correct for grouped records).

 o Added incremental run capability (-p command line option or
   "Incremental" configuration file keyword).

--------------------------------------------------------------------
1.1x-xx changes from 1.00-05
--------------------------------------------------------------------

Fixes:
 o Re-wrote the Group* logic, fixing a bug that allowed hiding of
   objects when they shouldn't be.

 o Fixed broken IgnoreReferrer code.

 o Modified config parse code to handle extended characters.

 o Misc. minor bug fixes/changes.  Added a missing fclose.

 o Cleaned up generated HTML.

 o Fixed duplicate warnings on large referrer fields.

 o Fixed country table bug adding grouped records to totals.

Changes/Additions:
 o Added GroupSite, GroupReferrer and GroupAgent keywords to round
   out the Group* configuration options.

 o Added GroupShading and GroupHighlight keywords to allow selective
   highlight and shading on grouped rows in table.

 o Removed the '-L' command line option.  Groupings can now only
   be specified from a configuration file.  Language files changed
   to reflect change.

 o Added '-V' command line option (identical to '-v') for version.

 o Added additional language support.  Language files will be marked
   /* New for 1.1 */ where changes have been made.

 o Various rewrites to streamline the code, accomidate the new
   group options and make things easier down the road when I implement
   incremental (partial log) processing.

 o Usual README and CHANGES documentation updates.

--------------------------------------------------------------------
1.00-xx changes from 0.99-06
--------------------------------------------------------------------

Fixes:
 o Modify record parser so that spaces in usernames (auth field)
   don't cause record to be skipped (w/'Bad Record' message).

 o Included various error conditions that were being ignored in
   the timing statistics ('bad records' value) totals.

Changes/Additions:
 o Added GMTTime (-g) option to force display of timestamps in
   GMT (UTC) time instead of local timezone.

 o Added GroupURL (-L) option for grouping of URLs as if they
   were a single object.  See README for details.

 o Language support in the form of a language specific header
   file containing all strings used by The Webalizer.  English
   file is used by default unless changed.  Support for other
   languages will be distributed as I receive them.

--------------------------------------------------------------------
0.99-xx changes from 0.98-16
--------------------------------------------------------------------

0.99 is mostly a bug-fix release, with a few added extra goodies.

Fixes:
 o Fixed monthly total transfer size (silent) overflow problem.

 o Fixed the numerous fprintf format errors.  Only seemed to wreak havok
   on non-intel machines though.

 o Fixed core dump condition on certain machines when using stdin for
   input.

 o Fixed floating point code that caused divide by zero errors on some
   platforms (most noticably on SCO OpenServer).

 o Netscape server kludges:  Added code to deal with Netscape log header
   record gracefully.  Also added workaround for timestamp error where
   Netscape sometimes makes a day have 0-24 hours instead of 0-23.  The
   Webalizer will now treat anything greater than 23 as 0.

 o Resized some fixed field sizes to gain memory usage improvements.

Changes/Additions:
 o Ignore* config keywords added.  This allows you to completely ignore
   certain log records based on site name, URL, user agent or referrer.
   * Use will cause inaccurate statistics results.  See documentation.

 o ReallyQuiet config keyword (-Q command line option) added.  Causes
   The Webalizer to supress _all_ messages.  Useful for cron jobs.

 o Removed the "Sites" total at the bottom of the summary by month.
   The total for sites is a useless number and produces a misleadingly
   high value which detracts from the accuracy of the other totals.

 o Updated README and CHANGES