CHANGELOG

0.8.0 (2013-12-13)
------------------

Major feature:

* Fixes and explicit testing of C89, C90, Visual C++

Minor features:

* Internal implementation of pyparsing (currently disabled)

Various minor fixes.


0.7.2 (2011-07-10)
------------------

Minor features:

* Allow more than one child in a sequenceof.
* Add const correctness to compiled C decoders.
* Add the start of an Microsoft Word specification.

Minor bugfixes:

* Add more detailed copyright and license information.
* Correctly escape the '&' character when printing xml.
* Fix build warnings when compiling with gcc 4.6.


0.7.1 (2011-04-06)
------------------

Bugfixes:

* Fix linking against the variable integer files from c++.
* Fix copyright & license information from the PRESENSE changes.


0.7.0 (2011-02-16)
------------------

Major feature:

* Encoding data structures to binary is now supported. It includes encoding
  from python instances, xml, and C structures.
* When encoding, the values of hidden fields is correctly deduced from
  visible uses of that field. For example, if a 'count' field is hidden but
  used in a sequenceof, it will look at the length of the sequenceof when
  encoding the field.

Features:

* Specifications can now be split into multiple files. Simply pass all of the
  specification files to the decode / encode / compile programs.
* Allow the user to specify the main protocol entry, overriding the default.

Bugfixes:

* Too many to mention!

Backwards incompatible changes:

* The 'bdecode' command now takes the file to decode though the '-f'
  option, not as the second argument.
* The 'bcompile' command now takes the output directory through the '-d'
  option, not as the second argument.


0.6.2 (2010-02-02)
------------------

Bugfixes:

* Correctly decode sequence values in the instance decoder.
* Correctly resolve references found in conditional & constraint expressions.
* Fix exception when we reference an unknown entry in an expression.
* Report missing reference errors from the entry that defined the reference.
* Improve the error messages from data errors.
* Don't throw an exception when choosing and not enough data available.
* Decode sequence entries with both a min and a max.
* Correctly handle conditional references.
* Report errors from the entry opening line & column number, not the closing location.
* Data objects now supports non-seekable files such as stdin.


0.6.1 (2009-12-05)
------------------

Features:

* C representation for choice entries whose children don't contain data is now
  an enum. Previously these entries generated a union without members, which is
  invalid C/C++ (and caused memory errors with g++).

Bugfixes:

* Fix the 'bcompile' script (was crashing).
* Various C compilation fixes (tests now use gcc instead of g++).
* Fix the range of 64 bit values.


0.6.0 (2009-11-30)
------------------

Features:

* Decode floating point, signed integer, and 64 bit values.
* Range analysis on integer types to determine suitable C-types.
* Conditional entries.
* Constraints on sequence values.

Bugfixes:

* Whitespace in xml output
* Fix referenencing length & value of the same entry

Thanks to Fernando Martín for the example patch adding signed integer and
floating point support.


0.5.3 (2009-04-01)
------------------

Bugfixes:

 * Fix issue that could cause the wrong (lower priority) option in a choice be
   decoded.
 * Fix compilation of some sequence integer entries to integers instead of
   structs.


0.5.2 (2009-02-26)
------------------

Bugfixes:

 * Correctly hide common entries that have been referenced with a hidden name.
 * Fix out of data error reporting invalid ranges.
 * Fix compilation errors when  entries named after keywords (such as 'int').
 * Allow references to be used as common entries.


0.5.1 (2009-02-18)
------------------

Bugfixes:

 * Report the correct entry when there is a missing referenced entry.
 * Fix Mako compile exception when running from an egg (ignore .pyc files).


0.5.0 (2009-02-14)
------------------

Features:

 * Compilation speedups (the jpeg spec now only takes 5 seconds to compile
   instead of 25).
 * Python decoding speedups (decoding a small sample pdf document now takes 10
   seconds instead of 300).
 * Compiled choices are now represented by an 'enum' to identify the decoded
   option, and a 'union' for the decoded value. This requires considerably
   less memory allocations, and is easier for client applications to use.
 * Expressions can now directly reference choice entries to reference the
   decoded option.
 * New 'tips' section in the documentation. Includes the ':' hidden entry
   hint.

Bugfixes:
 * Fix entries whose values are passed both in and out of an entry due to 
   expressions.
 * Fix little endian fields with an expected value in compiled specifications.


0.4.1 (2008-11-30)
------------------

Features:

 * Allow the project to be compiled to an egg (including templates).
 * Allow the generated C to be used from C++ programs (eg: extern "c").
 * Install 'bdecode' and 'bcompile' as part of the egg install.

Bugfixes:

 * Generated C language type names are consistent across runs.


0.4.0 (2008-10-08)
------------------

Features:

 * Allow common entries to be given another name.
 * Improve readability of generated C code.
 * Don't include unnecessary data structures in C headers.
 * Change to LGPL license.
 * Documentation updates.

Bugfixes:

 * Various fixes in C decoders.
 * Fix mp3 specification.


0.3.2 (2008-07-06)
------------------

Features:

 * The c decoder memory is now freed
 * Decoders are validated with valgrind
 * A new 'release' script to simplify making new releases

Bugfixes:

 * Handle sequence values in the C decoder
 * Print sequence values in the c decoder output
 * Xml output no longer includes whitespace around decoded values
 * Allow entry names to use the '-' character
 * Improve speed when choosing between embedded choices
 * Bit buffers now allocate their own memory (instead of reusing the input)


0.3.1 (20 April 2008)
---------------------

Docs:

 * Add external documentation (and tutorial), and update internal docs

Format specifications:

 * Add a png specification


0.3.0 (3 April 2008)
--------------------

Core:

 * Detect invalid integer encodings in a specification
 * Nosetests can now find all unit tests
 * Expressions are resolved after the spec as loaded (allows resolving into
   common items not yet specified).
 * Referenced items can no longer use 'end-sequenceof'

Format specifications:

 * Add a basic asf specification (wma/wmv)
 * Add boolean type (pdf)
 * Limit possible comment characters (pdf)
 * Add test data files for pdf, mp3, jpeg, and vfat


0.2.0 (17 September 2007)
-------------------------

Core:

 * Add a 'file' tool that attempts to identify the format of a binary file.
 * Support efficient decoding of files (the whole file is no longer read into
   memory).

Format specifications:

 * More complete pdf protocl specification.
 * Fixes to 'mp3' and 'vfat' specifications.


0.1.1 (8 August 2007)
---------------------

Initial standalone release of bdec library. Allows complex specifications
to be written that can be automatically decoded. The included example
specifications include;

 * pdf documents
 * mp3 and jpg media files
 * fat12/fat16/fat32 filesystems

Encoding is supported only for very basic specifications (no use of
expressions).