ChangeLog

See below for a list of changes.

----------------------------------------------------------------------
Version 2.0.6, 2018-11-06
- implement Checkpoint LEA transfer format
  ... at least if we guess right that this is the format name. This
  type of format seems to be seen in syslog message. Checkpoint does
  not provide a spec, so everything is guesswork... :-(
  closes https://github.com/rsyslog/liblognorm/issues/309
- made build on AIX
  Thanks to Philippe Duveau for the patch.
- fixes and improvements in bash scripting
  mostly based on shellcheck recommandations (via CodeFactor.com)
- string parser: add "lazy" matching mode
  This introduces paramter "matching.lazy". See doc for details.
- bugfix: suppress invalid param error for field name "-"
  Suppress invalid param error for name for hexnumber, float, number,
  date-rfc3164 and date-rfc5424. It will just check if name is "-" to
  make sure that we only suppress the error message in case we do not
  want to capture something.
  Thanks to Sol Huebner for the patch.
  closes https://github.com/rsyslog/liblognorm/issues/270
- bugfix: cisco-interface-spec did not succeed when at end of line
  Thanks to Sol Huebner for the patch.
  closes https://github.com/rsyslog/liblognorm/issues/229
----------------------------------------------------------------------
Version 2.0.5, 2018-04-26
- bugfix: es_str2cstr leak in string-to v1 parser
  Thanks to Harshvardhan Shrivastava for the patch.
- make "make check" "succeed" on solaris 10
  actually, we just ignore the CI failures so that OpenCSW can build
  new packages. The problems actually exist on that platform, but
  testing has shown they always existed. We currently run out of time
  to really fixing this, plus we never had any bug report on Solaris
  (I assme nobody uses it on Solaris 10). However, that issues is a
  blocker to make new rsyslog versions available on OpenCSW for
  Solaris 10, so we go the dirty way of pretenting there is no
  problem. Note: the issues was orignally not seen, as the failing
  tests have been added later on. So the problem was always there,
  just not visible.
- some mostly cosmetic fixes detected by Coverity Scan
  e. g. memory leak if and only if system was completely out of memory
----------------------------------------------------------------------
Version 2.0.4, 2017-10-04
- added support for native JSON number formats
  supported by parsers: number, float, hex
- added support for creating unix timestamps
  supported by parsers: date-rfc3164, date-rfc5424
- fixed build problems on Solaris
  ... but there still seem to be some code issues, manifested in
  testbench failures. So use with care!
----------------------------------------------------------------------
Version 2.0.3, 2017-03-22
- add ability to load rulebase from a string
  introduces new API:
  int ln_loadSamplesFromString(ln_ctx ctx, const char *string);
  closes https://github.com/rsyslog/liblognorm/issues/239
- bugfix: string parser did not correctly parse word at end of line
- bugfix: literal parser does not always store value if name is specified
  if
  rule=:%{"type":"literal", "text":"a", "name":"var"}%
  is used and matching message is provided, variable var ist not persisted.
  see also http://lists.adiscon.net/pipermail/rsyslog/2016-December/043985.html
----------------------------------------------------------------------
Version 2.0.2, 2016-11-15
- bugfix: no error was emitted on invalid "annotate" line
- "annnotate": permit inline comments
- fix a problem with cross-compilation
  see also: https://github.com/rsyslog/liblognorm/pull/221
  Thanks to Luca Boccassi for the patch
- testbench: add test for "annotate" functionality
- bugfix: abort in literal path compaction when useing "alternative" parser
  When using the "alternative" parser, literals nodes could be created with
  multiple reference count. This is valid. However, literal path compaction
  did not consider this case, and so "merged" these nodes, which lead to
  pdag corruption and quickly to segfault.
  closes https://github.com/rsyslog/liblognorm/issues/220
  closes https://github.com/rsyslog/liblognorm/issues/153
- bugfix: lognormalizer could loop
  This also caused the testbench to fail on some platforms.
  due too incorrect data type
  Thanks to Michael Biebl for this fix.
- fix misleading compiler warning
  Thanks to Michael Biebl for this fix.
- testbench: add test for "annotate" functionality
----------------------------------------------------------------------
Version 2.0.1, 2016-08-01
- fix public headers, which invalidly contained a strndup() definition
  Thanks to Michael Biebl for this fix.
- fix some issues in pkgconfig file
  Thanks to Michael Biebl for this fix.
- enhance build system to natively support systems with older
  autoconf versions and/or missing autoconf-archive. In this case we
  gracefully degrade functionality, but the build still is possible.
  Among others, this enables builds on CentOS 5.
----------------------------------------------------------------------
Version 2.0.0, 2016-07-21
- completely rewritten, much feature-enhanced version
- requires libfastjson instead of json-c
- big improvements to testbench runs, especially on travis
  among others, the static analyzer is now run and testbench throws
  an error if the static analyzer (via clang) is not clean
- lognormalizer tool can now handle lines larger 10k characters
  Thanks to Janmejay Singh for the patch
----------------------------------------------------------------------
Version 1.1.3, 2015-??-?? [no official release]
- make work on Solaris
- check for runaway rules.
  A runaway rule is one that has unmatched percent signs and thus
  is not terminated properly at its end. This also means we no longer
  accept "rule=" at the first column of a continuation line, which is
  no problem (see doc for more information).
- fix: process last line if it misses the terminating LF
  This problem occurs with the very last line of a rulebase (at EOF).
  If it is not properly terminated (LF missing), it is silently ignored.
  Previous versions did obviously process lines in this case. While
  technically this is invalid input, we can't outrule that such rulebases
  exist. For example, they do in the rsyslog testbench, which made
  us aware of the problem (see https://github.com/rsyslog/rsyslog/issues/489 )
  I think the proper way of addressing this is to process such lines without
  termination, as many other tools do as well.
  closes https://github.com/rsyslog/liblognorm/issues/135
----------------------------------------------------------------------
Version 1.1.2, 2015-07-20
- permit newline inside parser definition
- new parser "cisco-interface-spec"
- new parser "json" to process json parts of the message
- new parser "mac48" to process mac layer addresses
- new parser "name-value-list" (currently inofficial, experimental)
- some parsers did incorrectly report success when an error occurred
  this was caused by inconsistencies between various macros. We have
  changed the parser-generation macros to match the semantics of the
  broader CHKN/CHKR macros and also restructured/simplified the 
  parser generation macros.
  closes https://github.com/rsyslog/liblognorm/issues/41
- call "rest" parser only if nothing else matches.
  Versions prior to 1.1.2 did execute "rest" during regular parser
  processing, and thus parser matches have been more or less random.
  With 1.1.2 this is now always the last parser called. This may cause
  problems with existing rulesets, HOWEVER, adding any other rule or
  changing the load order would also have caused problems, so there
  really is no compatibility to preserve.
  see also:
  http://blog.gerhards.net/2015/04/liblognorms-rest-parser-now-more-useful.html
- new API to support error callbacks 
  This permits callers to forward messages in regard to e.g. wrong rule
  bases to their users, which is very useful and actually missing in the
  previous code base. So far, we only have few error messages.
  However, we will review the code and add more. The important part is
  that callers can begin to use the new API and thus will benefit when
  we add more error messages.
- testbench is now enabled by default
- bugfix: misadressing on some constant values
  see also https://github.com/rsyslog/liblognorm/pull/67
  Thanks to github user ontholerian for the patch
- bugfix: add missing function prototypes
  This could potentially lead to problems on some platforms,
  especially those with 64 bit pointers.
----------------------------------------------------------------------
Version 1.1.1, 2015-03-09
- fixed library version numbering
  Thanks to Tomas Heinreich for reporting the problem.
- added new parser syntaxes
  Thanks to Janmejay Singh for implementing most of them.
- bugfix: function ln_parseFieldDescr() returns state value
  due to unitialized variable. This can also lead to invalid
  returning no sample node where one would have to be created.
----------------------------------------------------------------------
Version 1.1.0, 2015-01-08
- added regular expression support
  use this feature with great care, as it thrashes performance
  Thanks to Janmejay Singh for implementing this feature.
- fix build problem when --enable-debug was set
  closes: https://github.com/rsyslog/liblognorm/issues/5
----------------------------------------------------------------------
Version 1.0.1, 2014-04-11
- improved doc (via RST/Sphinx)
- bugfix: unparsed fields were copied incorrectly from non-terminated
  string. Thanks to Josh Blum for the fix.
- bugfix: mandatory tag did not work in lognormalizer
----------------------------------------------------------------------
Version 1.0.0, 2013-11-28
- WARNING: this version has incompatible interface and older programs 
  will not compile with it.
  For details see http://www.liblognorm.com/news/on-liblognorm-1-0-0/
- libestr is not used any more in interface functions. Traditional
  C strings are used instead. Internally, libestr is still used, but
  scheduled for removal.
- libee is not used any more. JSON-C is used for object handling
  instead. Parsers and formatters are now part of liblognorm.
- added new field type "rest", which simply sinks all up to end of
  the string.
- added support for glueing two fields together, without literal 
  between them. It allows for constructs like:
  %volume:number%%unit:word%
  which matches string "1000Kbps"
- Fix incorrect merging of trees with empty literal at end
  Thanks to Pavel Levshin for the patch
- this version has survived many bugfixes
----------------------------------------------------------------------
================================================================================
The versions below is liblognorm0, which has a different API
================================================================================
----------------------------------------------------------------------
Version 0.3.7, 2013-07-17
- added support to load single samples
  Thanks to John Hopper for the patch
----------------------------------------------------------------------
Version 0.3.6, 2013-03-22
- bugfix: unitialized variable could lead to rulebase load error
----------------------------------------------------------------------
Version 0.3.5 (rgerhards), 2012-09-18
- renamed "normalizer" tool to "lognormalizer" to solve name clashes
  Thanks to the Fedora folks for pointing this out.
----------------------------------------------------------------------
Version 0.3.4 (rgerhards), 2012-04-16
- bugfix: normalizer tool had a memory leak
  Thanks to Brian Know for alerting me and helping to debug
----------------------------------------------------------------------
Version 0.3.3 (rgerhards), 2012-02-06
- required header file was not installed, resulting in compile error
  closes: http://bugzilla.adiscon.com/show_bug.cgi?id=307
  Thanks to Andreis Vinogradovs for alerting us on this bug.
----------------------------------------------------------------------
Version 0.3.2 (rgerhards), 2011-11-21
- added rfc5424 parser (requires libee >= 0.3.2)
- added "-" to serve as name for filler fields. Value is extracted,
  but no field is written
- special handling for iptables log via %iptables% parser added 
  (currently experimental pending practical verification)
- normalizer tool on its way to a full-blow stand-alone tool
- support for annotations added, for the time being see
  http://blog.gerhards.net/2011/11/log-annotation-with-liblognorm.html
----------------------------------------------------------------------
Version 0.3.1 (rgerhards), 2011-04-18
- added -t option to normalizer so that only messages with a 
  specified tag will be output
- bugfix: abort if a tag was assigned to a message without any 
  fields parsed out (uncommon scenario)
- bugfix: mem leak on parse tree destruct -- associated tags were
  not deleted
- bugfix: potential abort in normalizer due to misadressing in debug
  message generation
----------------------------------------------------------------------
Version 0.3.0 (rgerhards), 2011-04-06
- support for message classification via tags added
- bugfix: partial messages were invalidly matched
  closes: http://bugzilla.adiscon.com/show_bug.cgi?id=247
----------------------------------------------------------------------
Version 0.2.0 (rgerhards), 2011-04-01
- added -E option to normalizer tool, permits to specify data for
  encoders
- support for new libee parsers:
  * Time12hr
  * Time24hr
  * ISODate
  * QuotedString
- support for csv encoding added
- added -p option to normalizer tool (output only correctly parsed
  entries)
- bugfix: segfault if a parse tree prefix had exactly buffer size,
  in which case it was invalidly assumed that an external buffer had
  been allocated
----------------------------------------------------------------------
Version 0.1.0 (rgerhards), 2010-12-09
Initial public release.