bison-2.5 released [stable]

"Joel E. Denny" <>
Sat, 14 May 2011 21:29:57 -0400 (EDT)

          From comp.compilers

Related articles
bison-2.5 released [stable] (Joel E. Denny) (2011-05-14)
| List of all articles for this month |

From: "Joel E. Denny" <>
Newsgroups: comp.compilers
Date: Sat, 14 May 2011 21:29:57 -0400 (EDT)
Organization: Compilers Central
Keywords: bison, available
Posted-Date: 14 May 2011 22:36:38 EDT

Bison version 2.5 is now available. Relative to 2.4.3, this version
contains many bug fixes and many exciting new features that need more
user feedback. Some of the most notable features are:

    1. Support for dashes in grammar symbol names.
    2. Named references, such as $sym instead of $1, in semantic actions.
    3. IELR(1) and canonical LR(1) parser tables.
    4. LAC (Lookahead Correction), which improves syntax error handling.
    5. %error-verbose syntax error message fixes, mostly from LAC.

LALR(1) parser tables and non-LAC parsers are still the default. See
NEWS below for further details.

Here are the compressed sources: (2.6MB) (2.0MB)

Here are the GPG detached signatures[*]:

To reduce load on the main server, use a mirror listed at:

[*] You can use either of the above signature files to verify that
the corresponding file (without the .sig suffix) is intact. First,
be sure to download both the .sig file and the corresponding tarball.
Then, run a command like this:

    gpg --verify bison-2.5.tar.gz.sig

If that command fails because you don't have the required public key,
then run this command to import it:

    gpg --keyserver --recv-keys 1D233519

and rerun the `gpg --verify' command.

This release was bootstrapped with the following tools:
    Autoconf 2.65
    Automake 1.11.1
    Flex 2.5.35
    Gettext 0.18.1
    Gnulib v0.0-5111-geff4c03


* Changes in version 2.5 (2011-05-14):

** 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

    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

    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

    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]

        -F 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


    have been renamed to


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

*** Values no longer need to be quoted in the 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)


            # 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)


    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.

*** Parser traces now print the top element of the stack.

** -W/--warnings fixes:

*** Bison now properly recognizes the `no-' versions of categories:

    For example, given the following command line, Bison now enables all
    warnings except warnings for incompatibilities with POSIX Yacc:

        bison -Wall,no-yacc gram.y

*** Bison now treats S/R and R/R conflicts like other warnings:

    Previously, conflict reports were independent of Bison's normal
    warning system. Now, Bison recognizes the warning categories
    `conflicts-sr' and `conflicts-rr'. This change has important
    consequences for the -W and --warnings command-line options. For

        bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
        bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
        bison -Wnone gram.y # no conflicts are reported
        bison -Werror gram.y # any conflict is an error

    However, as before, if the %expect or %expect-rr directive is
    specified, an unexpected number of conflicts is an error, and an
    expected number of conflicts is not reported, so -W and --warning
    then have no effect on the conflict report.

*** The `none' category no longer disables a preceding `error':

    For example, for the following command line, Bison now reports
    errors instead of warnings for incompatibilities with POSIX Yacc:

        bison -Werror,none,yacc gram.y

*** The `none' category now disables all Bison warnings:

    Previously, the `none' category disabled only Bison warnings for
    which there existed a specific -W/--warning category. However,
    given the following command line, Bison is now guaranteed to
    suppress all warnings:

        bison -Wnone gram.y

** Precedence directives can now assign token number 0:

    Since Bison 2.3b, which restored the ability of precedence
    directives to assign token numbers, doing so for token number 0 has
    produced an assertion failure. For example:

        %left END 0

    This bug has been fixed.

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.