[bison.git] / NEWS

Bison News
----------

* Changes in version 2.5 (????-??-??):

** Grammar symbol names can now contain non-initial dashes:

  Consistently with directives (such as %error-verbose) and with
  %define variables (e.g. push-pull), grammar symbol names may contain
  dashes in any position except the beginning.  This is a GNU
  extension over POSIX Yacc.  Thus, use of this extension is reported
  by -Wyacc and rejected in Yacc mode (--yacc).

** Named references:

  Historically, Yacc and Bison have supported positional references
  ($n, $$) to allow access to symbol values from inside of semantic
  actions code.

  Starting from this version, Bison can also accept named references.
  When no ambiguity is possible, original symbol names may be used
  as named references:

    if_stmt : 'if' cond_expr 'then' then_stmt ';'
    { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }

  In the more common case, explicit names may be declared:

    stmt[res] : 'if' expr[cond] 'then' stmt[then] 'else' stmt[else] ';'
    { $res = mk_if_stmt($cond, $then, $else); }

  Location information is also accessible using @name syntax.  When
  accessing symbol names containing dots or dashes, explicit bracketing
  ($[sym.1]) must be used.

  These features are experimental in this version.  More user feedback
  will help to stabilize them.

** IELR(1) and canonical LR(1):

  IELR(1) is a minimal LR(1) parser table generation algorithm.  That
  is, given any context-free grammar, IELR(1) generates parser tables
  with the full language recognition power of canonical LR(1) but with
  nearly the same number of parser states as LALR(1).  This reduction
  in parser states is often an order of magnitude.  More importantly,
  because canonical LR(1)'s extra parser states may contain duplicate
  conflicts in the case of non-LR(1) grammars, the number of conflicts
  for IELR(1) is often an order of magnitude less as well.  This can
  significantly reduce the complexity of developing of a grammar.

  Bison can now generate IELR(1) and canonical LR(1) parser tables in
  place of its traditional LALR(1) parser tables, which remain the
  default.  You can specify the type of parser tables in the grammar
  file with these directives:

    %define lr.type lalr
    %define lr.type ielr
    %define lr.type canonical-lr

  The default-reduction optimization in the parser tables can also be
  adjusted using `%define lr.default-reductions'.  For details on both
  of these features, see the new section `Tuning LR' in the Bison
  manual.

  These features are experimental.  More user feedback will help to
  stabilize them.

** LAC (Lookahead Correction) for syntax error handling:

  Canonical LR, IELR, and LALR can suffer from a couple of problems
  upon encountering a syntax error.  First, the parser might perform
  additional parser stack reductions before discovering the syntax
  error.  Such reductions can perform user semantic actions that are
  unexpected because they are based on an invalid token, and they
  cause error recovery to begin in a different syntactic context than
  the one in which the invalid token was encountered.  Second, when
  verbose error messages are enabled (with %error-verbose or the
  obsolete `#define YYERROR_VERBOSE'), the expected token list in the
  syntax error message can both contain invalid tokens and omit valid
  tokens.

  The culprits for the above problems are %nonassoc, default
  reductions in inconsistent states, and parser state merging.  Thus,
  IELR and LALR suffer the most.  Canonical LR can suffer only if
  %nonassoc is used or if default reductions are enabled for
  inconsistent states.

  LAC is a new mechanism within the parsing algorithm that solves
  these problems for canonical LR, IELR, and LALR without sacrificing
  %nonassoc, default reductions, or state merging.  When LAC is in
  use, canonical LR and IELR behave almost exactly the same for both
  syntactically acceptable and syntactically unacceptable input.
  While LALR still does not support the full language-recognition
  power of canonical LR and IELR, LAC at least enables LALR's syntax
  error handling to correctly reflect LALR's language-recognition
  power.

  Currently, LAC is only supported for deterministic parsers in C.
  You can enable LAC with the following directive:

    %define parse.lac full

  See the new section `LAC' in the Bison manual for additional
  details including a few caveats.

  LAC is an experimental feature.  More user feedback will help to
  stabilize it.

** %define improvements:

*** Can now be invoked via the command line:

  Each of these command-line options

    -D NAME[=VALUE]
    --define=NAME[=VALUE]

    -F NAME[=VALUE]
    --force-define=NAME[=VALUE]

  is equivalent to this grammar file declaration

    %define NAME ["VALUE"]

  except that the manner in which Bison processes multiple definitions
  for the same NAME differs.  Most importantly, -F and --force-define
  quietly override %define, but -D and --define do not.  For further
  details, see the section "Bison Options" in the Bison manual.

*** Variables renamed:

  The following %define variables

    api.push_pull
    lr.keep_unreachable_states

  have been renamed to

    api.push-pull
    lr.keep-unreachable-states

  The old names are now deprecated but will be maintained indefinitely
  for backward compatibility.

*** Values no longer need to be quoted in grammar file:

  If a %define value is an identifier, it no longer needs to be placed
  within quotations marks.  For example,

    %define api.push-pull "push"

  can be rewritten as

    %define api.push-pull push

*** Unrecognized variables are now errors not warnings.

*** Multiple invocations for any variable is now an error not a warning.

** Unrecognized %code qualifiers are now errors not warnings.

** Character literals not of length one:

  Previously, Bison quietly converted all character literals to length
  one.  For example, without warning, Bison interpreted the operators in
  the following grammar to be the same token:

    exp: exp '++'
       | exp '+' exp
       ;

  Bison now warns when a character literal is not of length one.  In
  some future release, Bison will start reporting an error instead.

** Destructor calls fixed for lookaheads altered in semantic actions:

  Previously for deterministic parsers in C, if a user semantic action
  altered yychar, the parser in some cases used the old yychar value to
  determine which destructor to call for the lookahead upon a syntax
  error or upon parser return.  This bug has been fixed.

** C++ parsers use YYRHSLOC:

  Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
  macro and use it in the default YYLLOC_DEFAULT.  You are encouraged
  to use it.  If, for instance, your location structure has "first"
  and "last" members, instead of

      # define YYLLOC_DEFAULT(Current, Rhs, N)                             \
        do                                                                 \
          if (N)                                                           \
            {                                                              \
              (Current).first = (Rhs)[1].location.first;                   \
              (Current).last  = (Rhs)[N].location.last;                    \
            }                                                              \
          else                                                             \
            {                                                              \
              (Current).first = (Current).last = (Rhs)[0].location.last;   \
            }                                                              \
        while (false)

  use:

      # define YYLLOC_DEFAULT(Current, Rhs, N)                             \
        do                                                                 \
          if (N)                                                           \
            {                                                              \
              (Current).first = YYRHSLOC (Rhs, 1).first;                   \
              (Current).last  = YYRHSLOC (Rhs, N).last;                    \
            }                                                              \
          else                                                             \
            {                                                              \
              (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last;   \
            }                                                              \
        while (false)

** YYLLOC_DEFAULT in C++:

  The default implementation of YYLLOC_DEFAULT used to be issued in
  the header file.  It is now output in the implementation file, after
  the user %code sections so that its #ifndef guard does not try to
  override the user's YYLLOC_DEFAULT if provided.

** YYFAIL now produces warnings and Java parsers no longer implement it:

  YYFAIL has existed for many years as an undocumented feature of
  deterministic parsers in C generated by Bison.  More recently, it was
  a documented feature of Bison's experimental Java parsers.  As
  promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
  semantic action now produces a deprecation warning, and Java parsers
  no longer implement YYFAIL at all.  For further details, including a
  discussion of how to suppress C preprocessor warnings about YYFAIL
  being unused, see the Bison 2.4.2 NEWS entry.

** Temporary hack for adding a semicolon to the user action:

  Previously, Bison appended a semicolon to every user action for
  reductions when the output language defaulted to C (specifically, when
  neither %yacc, %language, %skeleton, or equivalent command-line
  options were specified).  This allowed actions such as

    exp: exp "+" exp { $$ = $1 + $3 };

  instead of

    exp: exp "+" exp { $$ = $1 + $3; };

  As a first step in removing this misfeature, Bison now issues a
  warning when it appends a semicolon.  Moreover, in cases where Bison
  cannot easily determine whether a semicolon is needed (for example, an
  action ending with a cpp directive or a braced compound initializer),
  it no longer appends one.  Thus, the C compiler might now complain
  about a missing semicolon where it did not before.  Future releases of
  Bison will cease to append semicolons entirely.

** Verbose syntax error message fixes:

  When %error-verbose or the obsolete `#define YYERROR_VERBOSE' is
  specified, syntax error messages produced by the generated parser
  include the unexpected token as well as a list of expected tokens.
  The effect of %nonassoc on these verbose messages has been corrected
  in two ways, but a more complete fix requires LAC, described above:

*** When %nonassoc is used, there can exist parser states that accept no
    tokens, and so the parser does not always require a lookahead token
    in order to detect a syntax error.  Because no unexpected token or
    expected tokens can then be reported, the verbose syntax error
    message described above is suppressed, and the parser instead
    reports the simpler message, "syntax error".  Previously, this
    suppression was sometimes erroneously triggered by %nonassoc when a
    lookahead was actually required.  Now verbose messages are
    suppressed only when all previous lookaheads have already been
    shifted or discarded.

*** Previously, the list of expected tokens erroneously included tokens
    that would actually induce a syntax error because conflicts for them
    were resolved with %nonassoc in the current parser state.  Such
    tokens are now properly omitted from the list.

*** Expected token lists are still often wrong due to state merging
    (from LALR or IELR) and default reductions, which can both add
    invalid tokens and subtract valid tokens.  Canonical LR almost
    completely fixes this problem by eliminating state merging and
    default reductions.  However, there is one minor problem left even
    when using canonical LR and even after the fixes above.  That is,
    if the resolution of a conflict with %nonassoc appears in a later
    parser state than the one at which some syntax error is
    discovered, the conflicted token is still erroneously included in
    the expected token list.  Bison's new LAC implementation,
    described above, eliminates this problem and the need for
    canonical LR.  However, LAC is still experimental and is disabled
    by default.

** Java skeleton fixes:

*** A location handling bug has been fixed.

*** The top element of each of the value stack and location stack is now
    cleared when popped so that it can be garbage collected.

* Changes in version 2.4.3 (2010-08-05):

** Bison now obeys -Werror and --warnings=error for warnings about
   grammar rules that are useless in the parser due to conflicts.

** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
   been fixed.

** Failures in the test suite for GCC 4.5 have been fixed.

** Failures in the test suite for some versions of Sun Studio C++ have
   been fixed.

** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
   warnings about undefined %prec identifiers will not be converted to
   errors in Bison 2.5.  They will remain warnings, which should be
   sufficient for POSIX while avoiding backward compatibility issues.

** Minor documentation fixes.

* Changes in version 2.4.2 (2010-03-20):

** Some portability problems that resulted in failures and livelocks
   in the test suite on some versions of at least Solaris, AIX, HP-UX,
   RHEL4, and Tru64 have been addressed.  As a result, fatal Bison
   errors should no longer cause M4 to report a broken pipe on the
   affected platforms.

** `%prec IDENTIFIER' requires IDENTIFIER to be defined separately.

  POSIX specifies that an error be reported for any identifier that does
  not appear on the LHS of a grammar rule and that is not defined by
  %token, %left, %right, or %nonassoc.  Bison 2.3b and later lost this
  error report for the case when an identifier appears only after a
  %prec directive.  It is now restored.  However, for backward
  compatibility with recent Bison releases, it is only a warning for
  now.  In Bison 2.5 and later, it will return to being an error.
  [Between the 2.4.2 and 2.4.3 releases, it was decided that this
  warning will not be converted to an error in Bison 2.5.]

** Detection of GNU M4 1.4.6 or newer during configure is improved.

** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
   YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
   avoided.

** %code is now a permanent feature.

  A traditional Yacc prologue directive is written in the form:

    %{CODE%}

  To provide a more flexible alternative, Bison 2.3b introduced the
  %code directive with the following forms for C/C++:

    %code          {CODE}
    %code requires {CODE}
    %code provides {CODE}
    %code top      {CODE}

  These forms are now considered permanent features of Bison.  See the
  %code entries in the section "Bison Declaration Summary" in the Bison
  manual for a summary of their functionality.  See the section
  "Prologue Alternatives" for a detailed discussion including the
  advantages of %code over the traditional Yacc prologue directive.

  Bison's Java feature as a whole including its current usage of %code
  is still considered experimental.

** YYFAIL is deprecated and will eventually be removed.

  YYFAIL has existed for many years as an undocumented feature of
  deterministic parsers in C generated by Bison.  Previously, it was
  documented for Bison's experimental Java parsers.  YYFAIL is no longer
  documented for Java parsers and is formally deprecated in both cases.
  Users are strongly encouraged to migrate to YYERROR, which is
  specified by POSIX.

  Like YYERROR, you can invoke YYFAIL from a semantic action in order to
  induce a syntax error.  The most obvious difference from YYERROR is
  that YYFAIL will automatically invoke yyerror to report the syntax
  error so that you don't have to.  However, there are several other
  subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
  inherent flaws when %error-verbose or `#define YYERROR_VERBOSE' is
  used.  For a more detailed discussion, see:

    http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html

  The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
  deterministic parsers in C will continue to implement it.  However,
  because YYFAIL is already flawed, it seems futile to try to make new
  Bison features compatible with it.  Thus, during parser generation,
  Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
  rule action.  In a later release, YYFAIL will be disabled for
  %error-verbose and `#define YYERROR_VERBOSE'.  Eventually, YYFAIL will
  be removed altogether.

  There exists at least one case where Bison 2.5's YYFAIL warning will
  be a false positive.  Some projects add phony uses of YYFAIL and other
  Bison-defined macros for the sole purpose of suppressing C
  preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
  To avoid Bison's future warning, such YYFAIL uses can be moved to the
  epilogue (that is, after the second `%%') in the Bison input file.  In
  this release (2.4.2), Bison already generates its own code to suppress
  C preprocessor warnings for YYFAIL, so projects can remove their own
  phony uses of YYFAIL if compatibility with Bison releases prior to
  2.4.2 is not necessary.

** Internationalization.

  Fix a regression introduced in Bison 2.4: Under some circumstances,
  message translations were not installed although supported by the
  host system.

* Changes in version 2.4.1 (2008-12-11):

** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
   declarations have been fixed.

** Temporary hack for adding a semicolon to the user action.

  Bison used to prepend a trailing semicolon at the end of the user
  action for reductions.  This allowed actions such as

    exp: exp "+" exp { $$ = $1 + $3 };

  instead of

    exp: exp "+" exp { $$ = $1 + $3; };

  Some grammars still depend on this `feature'.  Bison 2.4.1 restores
  the previous behavior in the case of C output (specifically, when
  neither %language or %skeleton or equivalent command-line options
  are used) to leave more time for grammars depending on the old
  behavior to be adjusted.  Future releases of Bison will disable this
  feature.

** A few minor improvements to the Bison manual.

* Changes in version 2.4 (2008-11-02):

** %language is an experimental feature.

  We first introduced this feature in test release 2.3b as a cleaner
  alternative to %skeleton.  Since then, we have discussed the possibility of
  modifying its effect on Bison's output file names.  Thus, in this release,
  we consider %language to be an experimental feature that will likely evolve
  in future releases.

** Forward compatibility with GNU M4 has been improved.

** Several bugs in the C++ skeleton and the experimental Java skeleton have been
  fixed.

* Changes in version 2.3b (2008-05-27):

** The quotes around NAME that used to be required in the following directive
  are now deprecated:

    %define NAME "VALUE"

** The directive `%pure-parser' is now deprecated in favor of:

    %define api.pure

  which has the same effect except that Bison is more careful to warn about
  unreasonable usage in the latter case.

** Push Parsing

  Bison can now generate an LALR(1) parser in C with a push interface.  That
  is, instead of invoking `yyparse', which pulls tokens from `yylex', you can
  push one token at a time to the parser using `yypush_parse', which will
  return to the caller after processing each token.  By default, the push
  interface is disabled.  Either of the following directives will enable it:

    %define api.push_pull "push" // Just push; does not require yylex.
    %define api.push_pull "both" // Push and pull; requires yylex.

  See the new section `A Push Parser' in the Bison manual for details.

  The current push parsing interface is experimental and may evolve.  More user
  feedback will help to stabilize it.

** The -g and --graph options now output graphs in Graphviz DOT format,
  not VCG format.  Like --graph, -g now also takes an optional FILE argument
  and thus cannot be bundled with other short options.

** Java

  Bison can now generate an LALR(1) parser in Java.  The skeleton is
  `data/lalr1.java'.  Consider using the new %language directive instead of
  %skeleton to select it.

  See the new section `Java Parsers' in the Bison manual for details.

  The current Java interface is experimental and may evolve.  More user
  feedback will help to stabilize it.

** %language

  This new directive specifies the programming language of the generated
  parser, which can be C (the default), C++, or Java.  Besides the skeleton
  that Bison uses, the directive affects the names of the generated files if
  the grammar file's name ends in ".y".

** XML Automaton Report

  Bison can now generate an XML report of the LALR(1) automaton using the new
  `--xml' option.  The current XML schema is experimental and may evolve.  More
  user feedback will help to stabilize it.

** The grammar file may now specify the name of the parser header file using
  %defines.  For example:

    %defines "parser.h"

** When reporting useless rules, useless nonterminals, and unused terminals,
  Bison now employs the terms "useless in grammar" instead of "useless",
  "useless in parser" instead of "never reduced", and "unused in grammar"
  instead of "unused".

** Unreachable State Removal

  Previously, Bison sometimes generated parser tables containing unreachable
  states.  A state can become unreachable during conflict resolution if Bison
  disables a shift action leading to it from a predecessor state.  Bison now:

    1. Removes unreachable states.

    2. Does not report any conflicts that appeared in unreachable states.
       WARNING: As a result, you may need to update %expect and %expect-rr
       directives in existing grammar files.

    3. For any rule used only in such states, Bison now reports the rule as
       "useless in parser due to conflicts".

  This feature can be disabled with the following directive:

    %define lr.keep_unreachable_states

  See the %define entry in the `Bison Declaration Summary' in the Bison manual
  for further discussion.

** Lookahead Set Correction in the `.output' Report

  When instructed to generate a `.output' file including lookahead sets
  (using `--report=lookahead', for example), Bison now prints each reduction's
  lookahead set only next to the associated state's one item that (1) is
  associated with the same rule as the reduction and (2) has its dot at the end
  of its RHS.  Previously, Bison also erroneously printed the lookahead set
  next to all of the state's other items associated with the same rule.  This
  bug affected only the `.output' file and not the generated parser source
  code.

** --report-file=FILE is a new option to override the default `.output' file
  name.

** The `=' that used to be required in the following directives is now
  deprecated:

    %file-prefix "parser"
    %name-prefix "c_"
    %output "parser.c"

** An Alternative to `%{...%}' -- `%code QUALIFIER {CODE}'

  Bison 2.3a provided a new set of directives as a more flexible alternative to
  the traditional Yacc prologue blocks.  Those have now been consolidated into
  a single %code directive with an optional qualifier field, which identifies
  the purpose of the code and thus the location(s) where Bison should generate
  it:

    1. `%code          {CODE}' replaces `%after-header  {CODE}'
    2. `%code requires {CODE}' replaces `%start-header  {CODE}'
    3. `%code provides {CODE}' replaces `%end-header    {CODE}'
    4. `%code top      {CODE}' replaces `%before-header {CODE}'

  See the %code entries in section `Bison Declaration Summary' in the Bison
  manual for a summary of the new functionality.  See the new section `Prologue
  Alternatives' for a detailed discussion including the advantages of %code
  over the traditional Yacc prologues.

  The prologue alternatives are experimental.  More user feedback will help to
  determine whether they should become permanent features.

** Revised warning: unset or unused mid-rule values

  Since Bison 2.2, Bison has warned about mid-rule values that are set but not
  used within any of the actions of the parent rule.  For example, Bison warns
  about unused $2 in:

    exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };

  Now, Bison also warns about mid-rule values that are used but not set.  For
  example, Bison warns about unset $$ in the mid-rule action in:

    exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };

  However, Bison now disables both of these warnings by default since they
  sometimes prove to be false alarms in existing grammars employing the Yacc
  constructs $0 or $-N (where N is some positive integer).

  To enable these warnings, specify the option `--warnings=midrule-values' or
  `-W', which is a synonym for `--warnings=all'.

** Default %destructor or %printer with `<*>' or `<>'

  Bison now recognizes two separate kinds of default %destructor's and
  %printer's:

    1. Place `<*>' in a %destructor/%printer symbol list to define a default
       %destructor/%printer for all grammar symbols for which you have formally
       declared semantic type tags.

    2. Place `<>' in a %destructor/%printer symbol list to define a default
       %destructor/%printer for all grammar symbols without declared semantic
       type tags.

  Bison no longer supports the `%symbol-default' notation from Bison 2.3a.
  `<*>' and `<>' combined achieve the same effect with one exception: Bison no
  longer applies any %destructor to a mid-rule value if that mid-rule value is
  not actually ever referenced using either $$ or $n in a semantic action.

  The default %destructor's and %printer's are experimental.  More user
  feedback will help to determine whether they should become permanent
  features.

  See the section `Freeing Discarded Symbols' in the Bison manual for further
  details.

** %left, %right, and %nonassoc can now declare token numbers.  This is required
  by POSIX.  However, see the end of section `Operator Precedence' in the Bison
  manual for a caveat concerning the treatment of literal strings.

** The nonfunctional --no-parser, -n, and %no-parser options have been
  completely removed from Bison.

* Changes in version 2.3a, 2006-09-13:

** Instead of %union, you can define and use your own union type
  YYSTYPE if your grammar contains at least one <type> tag.
  Your YYSTYPE need not be a macro; it can be a typedef.
  This change is for compatibility with other Yacc implementations,
  and is required by POSIX.

** Locations columns and lines start at 1.
  In accordance with the GNU Coding Standards and Emacs.

** You may now declare per-type and default %destructor's and %printer's:

  For example:

    %union { char *string; }
    %token <string> STRING1
    %token <string> STRING2
    %type  <string> string1
    %type  <string> string2
    %union { char character; }
    %token <character> CHR
    %type  <character> chr
    %destructor { free ($$); } %symbol-default
    %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
    %destructor { } <character>

  guarantees that, when the parser discards any user-defined symbol that has a
  semantic type tag other than `<character>', it passes its semantic value to
  `free'.  However, when the parser discards a `STRING1' or a `string1', it
  also prints its line number to `stdout'.  It performs only the second
  `%destructor' in this case, so it invokes `free' only once.

  [Although we failed to mention this here in the 2.3a release, the default
  %destructor's and %printer's were experimental, and they were rewritten in
  future versions.]

** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with `-y',
  `--yacc', or `%yacc'), Bison no longer generates #define statements for
  associating token numbers with token names.  Removing the #define statements
  helps to sanitize the global namespace during preprocessing, but POSIX Yacc
  requires them.  Bison still generates an enum for token names in all cases.

** Handling of traditional Yacc prologue blocks is now more consistent but
  potentially incompatible with previous releases of Bison.

  As before, you declare prologue blocks in your grammar file with the
  `%{ ... %}' syntax.  To generate the pre-prologue, Bison concatenates all
  prologue blocks that you've declared before the first %union.  To generate
  the post-prologue, Bison concatenates all prologue blocks that you've
  declared after the first %union.

  Previous releases of Bison inserted the pre-prologue into both the header
  file and the code file in all cases except for LALR(1) parsers in C.  In the
  latter case, Bison inserted it only into the code file.  For parsers in C++,
  the point of insertion was before any token definitions (which associate
  token numbers with names).  For parsers in C, the point of insertion was
  after the token definitions.

  Now, Bison never inserts the pre-prologue into the header file.  In the code
  file, it always inserts it before the token definitions.

** Bison now provides a more flexible alternative to the traditional Yacc
  prologue blocks: %before-header, %start-header, %end-header, and
  %after-header.

  For example, the following declaration order in the grammar file reflects the
  order in which Bison will output these code blocks.  However, you are free to
  declare these code blocks in your grammar file in whatever order is most
  convenient for you:

    %before-header {
      /* Bison treats this block like a pre-prologue block: it inserts it into
       * the code file before the contents of the header file.  It does *not*
       * insert it into the header file.  This is a good place to put
       * #include's that you want at the top of your code file.  A common
       * example is `#include "system.h"'.  */
    }
    %start-header {
      /* Bison inserts this block into both the header file and the code file.
       * In both files, the point of insertion is before any Bison-generated
       * token, semantic type, location type, and class definitions.  This is a
       * good place to define %union dependencies, for example.  */
    }
    %union {
      /* Unlike the traditional Yacc prologue blocks, the output order for the
       * new %*-header blocks is not affected by their declaration position
       * relative to any %union in the grammar file.  */
    }
    %end-header {
      /* Bison inserts this block into both the header file and the code file.
       * In both files, the point of insertion is after the Bison-generated
       * definitions.  This is a good place to declare or define public
       * functions or data structures that depend on the Bison-generated
       * definitions.  */
    }
    %after-header {
      /* Bison treats this block like a post-prologue block: it inserts it into
       * the code file after the contents of the header file.  It does *not*
       * insert it into the header file.  This is a good place to declare or
       * define internal functions or data structures that depend on the
       * Bison-generated definitions.  */
    }

  If you have multiple occurrences of any one of the above declarations, Bison
  will concatenate the contents in declaration order.

  [Although we failed to mention this here in the 2.3a release, the prologue
  alternatives were experimental, and they were rewritten in future versions.]

** The option `--report=look-ahead' has been changed to `--report=lookahead'.
  The old spelling still works, but is not documented and may be removed
  in a future release.

* Changes in version 2.3, 2006-06-05:

** GLR grammars should now use `YYRECOVERING ()' instead of `YYRECOVERING',
  for compatibility with LALR(1) grammars.

** It is now documented that any definition of YYSTYPE or YYLTYPE should
  be to a type name that does not contain parentheses or brackets.

* Changes in version 2.2, 2006-05-19:

** The distribution terms for all Bison-generated parsers now permit
  using the parsers in nonfree programs.  Previously, this permission
  was granted only for Bison-generated LALR(1) parsers in C.

** %name-prefix changes the namespace name in C++ outputs.

** The C++ parsers export their token_type.

** Bison now allows multiple %union declarations, and concatenates
  their contents together.

** New warning: unused values
  Right-hand side symbols whose values are not used are reported,
  if the symbols have destructors.  For instance:

     exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
	| exp "+" exp
	;

  will trigger a warning about $$ and $5 in the first rule, and $3 in
  the second ($1 is copied to $$ by the default rule).  This example
  most likely contains three errors, and could be rewritten as:

     exp: exp "?" exp ":" exp
	    { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
	| exp "+" exp
	    { $$ = $1 ? $1 : $3; if ($1) free ($3); }
	;

  However, if the original actions were really intended, memory leaks
  and all, the warnings can be suppressed by letting Bison believe the
  values are used, e.g.:

     exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
	| exp "+" exp         { $$ = $1; (void) $3; }
	;

  If there are mid-rule actions, the warning is issued if no action
  uses it.  The following triggers no warning: $1 and $3 are used.

     exp: exp { push ($1); } '+' exp { push ($3); sum (); };

  The warning is intended to help catching lost values and memory leaks.
  If a value is ignored, its associated memory typically is not reclaimed.

** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
  Destructors are now called when user code invokes YYABORT, YYACCEPT,
  and YYERROR, for all objects on the stack, other than objects
  corresponding to the right-hand side of the current rule.

** %expect, %expect-rr
  Incorrect numbers of expected conflicts are now actual errors,
  instead of warnings.

** GLR, YACC parsers.
  The %parse-params are available in the destructors (and the
  experimental printers) as per the documentation.

** Bison now warns if it finds a stray `$' or `@' in an action.

** %require "VERSION"
  This specifies that the grammar file depends on features implemented
  in Bison version VERSION or higher.

** lalr1.cc: The token and value types are now class members.
  The tokens were defined as free form enums and cpp macros.  YYSTYPE
  was defined as a free form union.  They are now class members:
  tokens are enumerations of the `yy::parser::token' struct, and the
  semantic values have the `yy::parser::semantic_type' type.

  If you do not want or can update to this scheme, the directive
  `%define "global_tokens_and_yystype" "1"' triggers the global
  definition of tokens and YYSTYPE.  This change is suitable both
  for previous releases of Bison, and this one.

  If you wish to update, then make sure older version of Bison will
  fail using `%require "2.2"'.

** DJGPP support added.
\f
* Changes in version 2.1, 2005-09-16:

** The C++ lalr1.cc skeleton supports %lex-param.

** Bison-generated parsers now support the translation of diagnostics like
  "syntax error" into languages other than English.  The default
  language is still English.  For details, please see the new
  Internationalization section of the Bison manual.  Software
  distributors should also see the new PACKAGING file.  Thanks to
  Bruno Haible for this new feature.

** Wording in the Bison-generated parsers has been changed slightly to
  simplify translation.  In particular, the message "memory exhausted"
  has replaced "parser stack overflow", as the old message was not
  always accurate for modern Bison-generated parsers.

** Destructors are now called when the parser aborts, for all symbols left
  behind on the stack.  Also, the start symbol is now destroyed after a
  successful parse.  In both cases, the behavior was formerly inconsistent.

** When generating verbose diagnostics, Bison-generated parsers no longer
  quote the literal strings associated with tokens.  For example, for
  a syntax error associated with '%token NUM "number"' they might
  print 'syntax error, unexpected number' instead of 'syntax error,
  unexpected "number"'.
\f
* Changes in version 2.0, 2004-12-25:

** Possibly-incompatible changes

  - Bison-generated parsers no longer default to using the alloca function
    (when available) to extend the parser stack, due to widespread
    problems in unchecked stack-overflow detection.  You can "#define
    YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
    the manual to determine safe values for YYMAXDEPTH in that case.

  - Error token location.
    During error recovery, the location of the syntax error is updated
    to cover the whole sequence covered by the error token: it includes
    the shifted symbols thrown away during the first part of the error
    recovery, and the lookahead rejected during the second part.

  - Semicolon changes:
    . Stray semicolons are no longer allowed at the start of a grammar.
    . Semicolons are now required after in-grammar declarations.

  - Unescaped newlines are no longer allowed in character constants or
    string literals.  They were never portable, and GCC 3.4.0 has
    dropped support for them.  Better diagnostics are now generated if
    forget a closing quote.

  - NUL bytes are no longer allowed in Bison string literals, unfortunately.

** New features

  - GLR grammars now support locations.

  - New directive: %initial-action.
    This directive allows the user to run arbitrary code (including
    initializing @$) from yyparse before parsing starts.

  - A new directive "%expect-rr N" specifies the expected number of
    reduce/reduce conflicts in GLR parsers.

  - %token numbers can now be hexadecimal integers, e.g., `%token FOO 0x12d'.
    This is a GNU extension.

  - The option `--report=lookahead' was changed to `--report=look-ahead'.
    [However, this was changed back after 2.3.]

  - Experimental %destructor support has been added to lalr1.cc.

  - New configure option --disable-yacc, to disable installation of the
    yacc command and -ly library introduced in 1.875 for POSIX conformance.

** Bug fixes

  - For now, %expect-count violations are now just warnings, not errors.
    This is for compatibility with Bison 1.75 and earlier (when there are
    reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
    are too many or too few shift/reduce conflicts).  However, in future
    versions of Bison we plan to improve the %expect machinery so that
    these violations will become errors again.

  - Within Bison itself, numbers (e.g., goto numbers) are no longer
    arbitrarily limited to 16-bit counts.

  - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
\f
* Changes in version 1.875, 2003-01-01:

** The documentation license has been upgraded to version 1.2
  of the GNU Free Documentation License.

** syntax error processing

  - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
    locations too.  This fixes bugs in error-location computation.

  - %destructor
    It is now possible to reclaim the memory associated to symbols
    discarded during error recovery.  This feature is still experimental.

  - %error-verbose
    This new directive is preferred over YYERROR_VERBOSE.

  - #defining yyerror to steal internal variables is discouraged.
    It is not guaranteed to work forever.

** POSIX conformance

  - Semicolons are once again optional at the end of grammar rules.
    This reverts to the behavior of Bison 1.33 and earlier, and improves
    compatibility with Yacc.

  - `parse error' -> `syntax error'
    Bison now uniformly uses the term `syntax error'; formerly, the code
    and manual sometimes used the term `parse error' instead.  POSIX
    requires `syntax error' in diagnostics, and it was thought better to
    be consistent.

  - The documentation now emphasizes that yylex and yyerror must be
    declared before use.  C99 requires this.

  - Bison now parses C99 lexical constructs like UCNs and
    backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.

  - File names are properly escaped in C output.  E.g., foo\bar.y is
    output as "foo\\bar.y".

  - Yacc command and library now available
    The Bison distribution now installs a `yacc' command, as POSIX requires.
    Also, Bison now installs a small library liby.a containing
    implementations of Yacc-compatible yyerror and main functions.
    This library is normally not useful, but POSIX requires it.

  - Type clashes now generate warnings, not errors.

  - If the user does not define YYSTYPE as a macro, Bison now declares it
    using typedef instead of defining it as a macro.
    For consistency, YYLTYPE is also declared instead of defined.

** Other compatibility issues

  - %union directives can now have a tag before the `{', e.g., the
    directive `%union foo {...}' now generates the C code
    `typedef union foo { ... } YYSTYPE;'; this is for Yacc compatibility.
    The default union tag is `YYSTYPE', for compatibility with Solaris 9 Yacc.
    For consistency, YYLTYPE's struct tag is now `YYLTYPE' not `yyltype'.
    This is for compatibility with both Yacc and Bison 1.35.

  - `;' is output before the terminating `}' of an action, for
    compatibility with Bison 1.35.

  - Bison now uses a Yacc-style format for conflict reports, e.g.,
    `conflicts: 2 shift/reduce, 1 reduce/reduce'.

  - `yystype' and `yyltype' are now obsolescent macros instead of being
    typedefs or tags; they are no longer documented and are planned to be
    withdrawn in a future release.

** GLR parser notes

  - GLR and inline
    Users of Bison have to decide how they handle the portability of the
    C keyword `inline'.

  - `parsing stack overflow...' -> `parser stack overflow'
    GLR parsers now report `parser stack overflow' as per the Bison manual.

** Bison now warns if it detects conflicting outputs to the same file,
  e.g., it generates a warning for `bison -d -o foo.h foo.y' since
  that command outputs both code and header to foo.h.

** #line in output files
  - --no-line works properly.

** Bison can no longer be built by a K&R C compiler; it requires C89 or
  later to be built.  This change originally took place a few versions
  ago, but nobody noticed until we recently asked someone to try
  building Bison with a K&R C compiler.
\f
* Changes in version 1.75, 2002-10-14:

** Bison should now work on 64-bit hosts.

** Indonesian translation thanks to Tedi Heriyanto.

** GLR parsers
  Fix spurious parse errors.

** Pure parsers
  Some people redefine yyerror to steal yyparse' private variables.
  Reenable this trick until an official feature replaces it.

** Type Clashes
  In agreement with POSIX and with other Yaccs, leaving a default
  action is valid when $$ is untyped, and $1 typed:

	untyped: ... typed;

  but the converse remains an error:

	typed: ... untyped;

** Values of mid-rule actions
  The following code:

	foo: { ... } { $$ = $1; } ...

  was incorrectly rejected: $1 is defined in the second mid-rule
  action, and is equal to the $$ of the first mid-rule action.
\f
* Changes in version 1.50, 2002-10-04:

** GLR parsing
  The declaration
     %glr-parser
  causes Bison to produce a Generalized LR (GLR) parser, capable of handling
  almost any context-free grammar, ambiguous or not.  The new declarations
  %dprec and %merge on grammar rules allow parse-time resolution of
  ambiguities.  Contributed by Paul Hilfinger.

  Unfortunately Bison 1.50 does not work properly on 64-bit hosts
  like the Alpha, so please stick to 32-bit hosts for now.

** Output Directory
  When not in Yacc compatibility mode, when the output file was not
  specified, running `bison foo/bar.y' created `foo/bar.c'.  It
  now creates `bar.c'.

** Undefined token
  The undefined token was systematically mapped to 2 which prevented
  the use of 2 by the user.  This is no longer the case.

** Unknown token numbers
  If yylex returned an out of range value, yyparse could die.  This is
  no longer the case.

** Error token
  According to POSIX, the error token must be 256.
  Bison extends this requirement by making it a preference: *if* the
  user specified that one of her tokens is numbered 256, then error
  will be mapped onto another number.

** Verbose error messages
  They no longer report `..., expecting error or...' for states where
  error recovery is possible.

** End token
  Defaults to `$end' instead of `$'.

** Error recovery now conforms to documentation and to POSIX
  When a Bison-generated parser encounters a syntax error, it now pops
  the stack until it finds a state that allows shifting the error
  token.  Formerly, it popped the stack until it found a state that
  allowed some non-error action other than a default reduction on the
  error token.  The new behavior has long been the documented behavior,
  and has long been required by POSIX.  For more details, please see
  Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
  <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.

** Traces
  Popped tokens and nonterminals are now reported.

** Larger grammars
  Larger grammars are now supported (larger token numbers, larger grammar
  size (= sum of the LHS and RHS lengths), larger LALR tables).
  Formerly, many of these numbers ran afoul of 16-bit limits;
  now these limits are 32 bits on most hosts.

** Explicit initial rule
  Bison used to play hacks with the initial rule, which the user does
  not write.  It is now explicit, and visible in the reports and
  graphs as rule 0.

** Useless rules
  Before, Bison reported the useless rules, but, although not used,
  included them in the parsers.  They are now actually removed.

** Useless rules, useless nonterminals
  They are now reported, as a warning, with their locations.

** Rules never reduced
  Rules that can never be reduced because of conflicts are now
  reported.

** Incorrect `Token not used'
  On a grammar such as

           %token useless useful
           %%
           exp: '0' %prec useful;

  where a token was used to set the precedence of the last rule,
  bison reported both `useful' and `useless' as useless tokens.

** Revert the C++ namespace changes introduced in 1.31
  as they caused too many portability hassles.

** Default locations
  By an accident of design, the default computation of @$ was
  performed after another default computation was performed: @$ = @1.
  The latter is now removed: YYLLOC_DEFAULT is fully responsible of
  the computation of @$.

** Token end-of-file
  The token end of file may be specified by the user, in which case,
  the user symbol is used in the reports, the graphs, and the verbose
  error messages instead of `$end', which remains being the default.
  For instance
     %token MYEOF 0
  or
     %token MYEOF 0 "end of file"

** Semantic parser
  This old option, which has been broken for ages, is removed.

** New translations
  Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
  Croatian, thanks to Denis Lackovic.

** Incorrect token definitions
  When given `%token 'a' "A"', Bison used to output `#define 'a' 65'.

** Token definitions as enums
  Tokens are output both as the traditional #define's, and, provided
  the compiler supports ANSI C or is a C++ compiler, as enums.
  This lets debuggers display names instead of integers.

** Reports
  In addition to --verbose, bison supports --report=THINGS, which
  produces additional information:
  - itemset
    complete the core item sets with their closure
  - lookahead [changed to `look-ahead' in 1.875e through 2.3, but changed back]
    explicitly associate lookahead tokens to items
  - solved
    describe shift/reduce conflicts solving.
    Bison used to systematically output this information on top of
    the report.  Solved conflicts are now attached to their states.

** Type clashes
  Previous versions don't complain when there is a type clash on
  the default action if the rule has a mid-rule action, such as in:

      %type <foo> bar
      %%
      bar: '0' {} '0';

  This is fixed.

** GNU M4 is now required when using Bison.
\f
* Changes in version 1.35, 2002-03-25:

** C Skeleton
  Some projects use Bison's C parser with C++ compilers, and define
  YYSTYPE as a class.  The recent adjustment of C parsers for data
  alignment and 64 bit architectures made this impossible.

  Because for the time being no real solution for C++ parser
  generation exists, kludges were implemented in the parser to
  maintain this use.  In the future, when Bison has C++ parsers, this
  kludge will be disabled.

  This kludge also addresses some C++ problems when the stack was
  extended.
\f
* Changes in version 1.34, 2002-03-12:

** File name clashes are detected
  $ bison foo.y -d -o foo.x
  fatal error: header and parser would both be named `foo.x'

** A missing `;' at the end of a rule triggers a warning
  In accordance with POSIX, and in agreement with other
  Yacc implementations, Bison will mandate this semicolon in the near
  future.  This eases the implementation of a Bison parser of Bison
  grammars by making this grammar LALR(1) instead of LR(2).  To
  facilitate the transition, this release introduces a warning.

** Revert the C++ namespace changes introduced in 1.31, as they caused too
  many portability hassles.

** DJGPP support added.

** Fix test suite portability problems.
\f
* Changes in version 1.33, 2002-02-07:

** Fix C++ issues
  Groff could not be compiled for the definition of size_t was lacking
  under some conditions.

** Catch invalid @n
  As is done with $n.
\f
* Changes in version 1.32, 2002-01-23:

** Fix Yacc output file names

** Portability fixes

** Italian, Dutch translations
\f
* Changes in version 1.31, 2002-01-14:

** Many Bug Fixes

** GNU Gettext and %expect
  GNU Gettext asserts 10 s/r conflicts, but there are 7.  Now that
  Bison dies on incorrect %expectations, we fear there will be
  too many bug reports for Gettext, so _for the time being_, %expect
  does not trigger an error when the input file is named `plural.y'.

** Use of alloca in parsers
  If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
  malloc exclusively.  Since 1.29, but was not NEWS'ed.

  alloca is used only when compiled with GCC, to avoid portability
  problems as on AIX.

** yyparse now returns 2 if memory is exhausted; formerly it dumped core.

** When the generated parser lacks debugging code, YYDEBUG is now 0
  (as POSIX requires) instead of being undefined.

** User Actions
  Bison has always permitted actions such as { $$ = $1 }: it adds the
  ending semicolon.  Now if in Yacc compatibility mode, the semicolon
  is no longer output: one has to write { $$ = $1; }.

** Better C++ compliance
  The output parsers try to respect C++ namespaces.
  [This turned out to be a failed experiment, and it was reverted later.]

** Reduced Grammars
  Fixed bugs when reporting useless nonterminals.

** 64 bit hosts
  The parsers work properly on 64 bit hosts.

** Error messages
  Some calls to strerror resulted in scrambled or missing error messages.

** %expect
  When the number of shift/reduce conflicts is correct, don't issue
  any warning.

** The verbose report includes the rule line numbers.

** Rule line numbers are fixed in traces.

** Swedish translation

** Parse errors
  Verbose parse error messages from the parsers are better looking.
  Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
     Now: parse error: unexpected '/', expecting "number" or '-' or '('

** Fixed parser memory leaks.
  When the generated parser was using malloc to extend its stacks, the
  previous allocations were not freed.

** Fixed verbose output file.
  Some newlines were missing.
  Some conflicts in state descriptions were missing.

** Fixed conflict report.
  Option -v was needed to get the result.

** %expect
  Was not used.
  Mismatches are errors, not warnings.

** Fixed incorrect processing of some invalid input.

** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.

** Fixed some typos in the documentation.

** %token MY_EOF 0 is supported.
  Before, MY_EOF was silently renumbered as 257.

** doc/refcard.tex is updated.

** %output, %file-prefix, %name-prefix.
  New.

** --output
  New, aliasing `--output-file'.
\f
* Changes in version 1.30, 2001-10-26:

** `--defines' and `--graph' have now an optional argument which is the
  output file name. `-d' and `-g' do not change; they do not take any
  argument.

** `%source_extension' and `%header_extension' are removed, failed
  experiment.

** Portability fixes.
\f
* Changes in version 1.29, 2001-09-07:

** The output file does not define const, as this caused problems when used
  with common autoconfiguration schemes.  If you still use ancient compilers
  that lack const, compile with the equivalent of the C compiler option
  `-Dconst='.  autoconf's AC_C_CONST macro provides one way to do this.

** Added `-g' and `--graph'.

** The Bison manual is now distributed under the terms of the GNU FDL.

** The input and the output files has automatically a similar extension.

** Russian translation added.

** NLS support updated; should hopefully be less troublesome.

** Added the old Bison reference card.

** Added `--locations' and `%locations'.

** Added `-S' and `--skeleton'.

** `%raw', `-r', `--raw' is disabled.

** Special characters are escaped when output.  This solves the problems
  of the #line lines with path names including backslashes.

** New directives.
  `%yacc', `%fixed_output_files', `%defines', `%no_parser', `%verbose',
  `%debug', `%source_extension' and `%header_extension'.

** @$
  Automatic location tracking.
\f
* Changes in version 1.28, 1999-07-06:

** Should compile better now with K&R compilers.

** Added NLS.

** Fixed a problem with escaping the double quote character.

** There is now a FAQ.
\f
* Changes in version 1.27:

** The make rule which prevented bison.simple from being created on
  some systems has been fixed.
\f
* Changes in version 1.26:

** Bison now uses automake.

** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.

** Token numbers now start at 257 as previously documented, not 258.

** Bison honors the TMPDIR environment variable.

** A couple of buffer overruns have been fixed.

** Problems when closing files should now be reported.

** Generated parsers should now work even on operating systems which do
  not provide alloca().
\f
* Changes in version 1.25, 1995-10-16:

** Errors in the input grammar are not fatal; Bison keeps reading
the grammar file, and reports all the errors found in it.

** Tokens can now be specified as multiple-character strings: for
example, you could use "<=" for a token which looks like <=, instead
of chosing a name like LESSEQ.

** The %token_table declaration says to write a table of tokens (names
and numbers) into the parser file.  The yylex function can use this
table to recognize multiple-character string tokens, or for other
purposes.

** The %no_lines declaration says not to generate any #line preprocessor
directives in the parser file.

** The %raw declaration says to use internal Bison token numbers, not
Yacc-compatible token numbers, when token names are defined as macros.

** The --no-parser option produces the parser tables without including
the parser engine; a project can now use its own parser engine.
The actions go into a separate file called NAME.act, in the form of
a switch statement body.
\f
* Changes in version 1.23:

The user can define YYPARSE_PARAM as the name of an argument to be
passed into yyparse.  The argument should have type void *.  It should
actually point to an object.  Grammar actions can access the variable
by casting it to the proper pointer type.

Line numbers in output file corrected.
\f
* Changes in version 1.22:

--help option added.
\f
* Changes in version 1.20:

Output file does not redefine const for C++.

Local Variables:
mode: outline
End:

-----

Copyright (C) 1995-2011 Free Software Foundation, Inc.

This file is part of Bison, the GNU Parser Generator.

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program.  If not, see <http://www.gnu.org/licenses/>.