# NEWS   [plain text]

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

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"'.

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"'.

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
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'. The old spelling still works, but is not documented and will be removed. - 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. 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. 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. 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 and later] explicitly associate look-ahead 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. 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. 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. 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. Changes in version 1.32, 2002-01-23: * Fix Yacc output file names * Portability fixes * Italian, Dutch translations 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'. 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. 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.

Changes in version 1.28, 1999-07-06:

* Should compile better now with K&R compilers.

* Fixed a problem with escaping the double quote character.

* There is now a FAQ.

Changes in version 1.27:

* The make rule which prevented bison.simple from being created on
some systems has been fixed.

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().

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.

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.

Changes in version 1.22:

Changes in version 1.20:

Output file does not redefine const for C++.

Local Variables:
mode: outline
End:

-----

Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
2004, 2005, 2006 Free Software Foundation, Inc.

This file is part of Bison, the GNU Compiler Compiler.

Bison is free software; you can redistribute it and/or modify