kissgml/LLnextgen-0.5.5/Changelog

Version 0.5.5:
    Fixed bugs:
    - The LLscnt_ array for thread-safe parsers was dimensioned incorrectly.
    - Handling of string escapes was incorrect for unknown escapes.

Version 0.5.4:
    Fixed bugs:
    - LLsymb is now properly initialized before the first call to LLread and
      the lexer

Version 0.5.3:
    New features:
    - added possibility to mark a code section with %top, which will cause it
      to be included at the top of the output

Version 0.5.2:
    Fixed bugs:
    - correct check for overflow in addition, which caused incorrect length
      calculations under some compiler optimisations. The user visible result
      was that a "Recursion in default for non-terminal" error was reported
      where there should be none
    - octal escapes between 300 and 377 are now properly parsed, as well as
      two character octal escapes between \30 and \77

Version 0.5.1:
    New features:
    - variables created to hold return values are now initialised the same way
      that LLretval is initialised. This prevents uninitialised use of these
      values and is better aligned with user expectation.
    - warnings are generated for some cases in which it is clear that a
      returned value is not used in subsequent code.

Version 0.5.0:
    New features:
    - a new repetition operator (..?) is implemented. With it, users can
      specify that the last element in a repeating term is optional for the
      last repetition of that term.
    - an option to specify the extensions of the generated files was
      added.

    Fixed bugs:
    - in case of a %while controlled repetition conflict in a ? repetition,
      the sets controlling error recovery would not be adjusted correctly if
      the %while expression evaluated to false
    - when tracing tokens the EOFILE token resulting from %start directives
      would not be printed

Version 0.4.1:
    New features:
    - a new element, the back-reference operator is introduced. By using ...
      elements preceding the current term in its enclosing term are copied in
      its place. Most usefull for comma-separated lists and similar constructs

    Fixed bugs:
    - moved initialisation of LLretval to before user supplied declarations,
      such that assignments to LLretval like: rule<int> { LLretval = 1; }: ...
      will not be undone
    - return value overrides specified on tokens will now trigger an error

Version 0.4.0:
    New features:
    - return values for rules
    - the --depend option has an extra modifier phony, that adds phony targets
      for all the dependencies to avoid make problems when renaming or
      removing dependencies
    - the --help option does not show all options by default anymore. An
      optional argument has been added to --help to select which options to
      show
    - detection of symbols that are likely to clash with generated symbols
    - LLindex and LLsets internal variables are now const, to make sure they
      don't get overwritten and to allow more compiler optimizations

    Fixed bugs:
    - removed a typo, which caused erroneous arguments to --dump-tokens to
      report an erroneous argument to --depend
    - fixed typos in the output --dump-llmessage
    - added volatile keyword to local variables in the generated parser if
      --abort and --reentrant are used because setjmp/longjmp is not
      guaranteed to correctly save local variables otherwise
    - printing of contains sets has been restored for --verbose=3
    - LLnextgen will no longer stop if a 0 byte is found in the input

Version 0.3.3:
    Fixed bugs:
    - using characters in the range 0x80 - 0xff in the input could cause
      segmentation faults

Version 0.3.2:
    New features:
    - an option to facilitate gettext use by adding macros in the symbol table
      definition for all user defined strings, and a call to gettext in
      LLgetSymbol

    Fixed bugs:
    - minor issues with symbol table generation
    - using --generate-lexer-wrapper caused compile errors for generated
      parser due to undeclared symbol

Version 0.3.1:
    Fixed bugs:
    - the configure script has been fixed

Version 0.3.0:
    New features:
    - an option to specify a regular expression for which token definitions are
      automatically generated, plus an option to dump the automatically
      generated token definitions [Only available if the POSIX regular
      expression API is available on your platform]
    - more error messages are now generated in input-line order
    - fewer aborts for "ambiguous input", because several errors are now
      considered non-critical for further analysis
    - new implementation of LLreissue that automatically saves the token to be
      reissued
    - include guards in generated header file
    - defines added to header file to allow for version testing
    - an option to generate thread-safe parsers

    NOTE: the new LLreissue implementation is incompatible with previous
    versions of LLnextgen. If you use LLreissue, and do not use the generated
    lexical analyser wrapper, you will have to adapt your code. See
    documentation for details.

    Fixed bugs:
    - fixed bug in handling of \x escape sequences in include-file names and
      option strings
    - fixed bug that would stop LLnextgen reading the trailing portiong of a
      file that included another but was missing a ';' after the %include
      directive

Version 0.2.5:
    New features:
    - added option to generate dependency information about the generated files
    - added option to dump top-level C-code to allow the C preprocessor to
      generate dependency information for the generated files

    Fixed bugs:
    - added warning about using --generate-lexer-wrapper and --reentrant at the
      same time.
    - removed double warnings about certain combinations of options
    - reduced the size of the LLscnt array by a factor LL_SSETS
    - fixed error about unterminated comments for line comments at the end
      of a file

Version 0.2.4:
    Fixed bugs:
    - finding the base name of file names on MS-Windows was implemented
      incorrectly. It assumed that \ was the only directory separator, while
      / is also allowed.

Version 0.2.3:
    New features:
    - detailed suppression of warnings. NOTE: the option --no-warn-eof has been
      removed in favour of the new warning suppression mechanism

    Fixed bugs:
    - warning suppression wasn't implemented for most warnings. This has now
      been fixed
    - a warning is now issued regarding use of --reentrant if multiple %start
      directives are used in one grammar. The warning can be suppressed with
      --suppress-warnings=multiple-parser

Version 0.2.2:
    Fixed bugs:
    - long options would be recognised even if only a leading substring would
      be specified
    - the alternation conflict example was corrected

Version 0.2.1:
    Fixed bugs:
    - fixed %first code for sets with one or more tokens

Version 0.2:
    New features
    - symbol table generation
    - dumping of lexer wrapper
    - dumping of default LLmessage
    - LLabort
    - %include
    - %options

    Fixed bugs:
    - lexer-wrapper and default LLmessage compile problems
    - prevent potential name space clobbering with token names by moving the
      #include directive
    - defered reading of next token for
      [ 'a' | 'b' ]
      constructions so that a following action can assume the token text to be
      valid (LLgen compatibility fix)
    - labels were generated incorrectly when using a %while to solve
      repetition conflicts in + repetitions, leading to uncompilable code
    - file name handling on MS Windows platform
    - symbol renaming bugs
    - removed spurious error messages about:
      - alternatives never being chosen, while there was only one (be it empty)
        alternative in a rule
      - Terms without symbols
    - fixed %first code for empty sets and sets with only one token
    - fixed add length routine, so it now takes the factor into account. This
      corrects an error in default alternative determination for alternatives
      including a fixed repetition term with a repetition count larger than 1
    - many minor bugs

Version 0.1.1:
    Fixed bugs:
    - Now print warning message at top of Lpars.c and Lpars.h
    - bug fix in token tracking code, which caused erroneous error messages