3 * Noteworthy changes in release ?.? (????-??-??) [?]
7 *** C++: Support for move semantics (lalr1.cc)
9 The lalr1.cc skeleton now fully supports C++ move semantics, while
10 maintaining compatibility with C++98. You may now store move-only types
11 when using Bison's variants. For instance:
19 %define api.value.type variant
23 %token <int> INT "int";
24 %nterm <std::unique_ptr<int>> int;
25 %nterm <std::vector<std::unique_ptr<int>>> list;
29 | list int { $$ = std::move($1); $$.emplace_back(std::move($2)); }
31 int: "int" { $$ = std::make_unique<int>($1); }
33 *** C++: Implicit move of right-hand side values (lalr1.cc)
35 In modern C++ (C++11 and later), you should always use 'std::move' with
36 the values of the right-hand side symbols ($1, $2, etc.), as they will be
37 popped from the stack anyway. Using 'std::move' is mandatory for
38 move-only types such as unique_ptr, and it provides a significant speedup
39 for large types such as std::string, or std::vector, etc.
41 If '%define api.value.automove' is set, every occurrence '$n' is replaced
42 by 'std::move ($n)'. The second rule in the previous grammar can be
45 list: list int { $$ = $1; $$.emplace_back($2); }
47 With automove enabled, the semantic values are no longer lvalues, so do
48 not use the swap idiom:
50 list: list int { std::swap($$, $1); $$.emplace_back($2); }
52 This idiom is anyway obsolete: it is preferable to move than to swap.
54 A warning is issued when automove is enabled, and a value is used several
57 input.yy:16.31-32: warning: multiple occurrences of $2 with api.value.automove enabled [-Wother]
58 exp: "twice" exp { $$ = $2 + $2; }
61 Enabling api.value.automove does not require support for modern C++. The
62 generated code is valid C++98/03, but will use copies instead of moves.
64 The new examples/variant-11.yy shows these features in action.
66 * Noteworthy changes in release 3.1 (2018-08-27) [stable]
68 ** Backward incompatible changes
70 Compiling Bison now requires a C99 compiler---as announced during the
71 release of Bison 3.0, five years ago. Generated parsers do not require a
74 Support for DJGPP, which have been unmaintained and untested for years, is
75 obsolete. Unless there is activity to revive it, the next release of Bison
80 *** Typed midrule actions
82 Because their type is unknown to Bison, the values of midrule actions are
83 not treated like the others: they don't have %printer and %destructor
84 support. It also prevents C++ (Bison) variants to handle them properly.
86 Typed midrule actions address these issues. Instead of:
88 exp: { $<ival>$ = 1; } { $<ival>$ = 2; } { $$ = $<ival>1 + $<ival>2; }
92 exp: <ival>{ $$ = 1; } <ival>{ $$ = 2; } { $$ = $1 + $2; }
94 *** Reports include the type of the symbols
96 The sections about terminal and nonterminal symbols of the '*.output' file
97 now specify their declared type. For instance, for:
101 the report now shows '<ival>':
103 Terminals, with rules where they appear
107 *** Diagnostics about useless rules
109 In the following grammar, the 'exp' nonterminal is trivially useless. So,
110 of course, its rules are useless too.
114 exp: exp '+' exp | exp '-' exp | '(' exp ')'
116 Previously all the useless rules were reported, including those whose
117 left-hand side is the 'exp' nonterminal:
119 warning: 1 nonterminal useless in grammar [-Wother]
120 warning: 4 rules useless in grammar [-Wother]
121 2.14-16: warning: nonterminal useless in grammar: exp [-Wother]
124 2.14-16: warning: rule useless in grammar [-Wother]
127 3.6-16: warning: rule useless in grammar [-Wother]
128 exp: exp '+' exp | exp '-' exp | '(' exp ')'
130 3.20-30: warning: rule useless in grammar [-Wother]
131 exp: exp '+' exp | exp '-' exp | '(' exp ')'
133 3.34-44: warning: rule useless in grammar [-Wother]
134 exp: exp '+' exp | exp '-' exp | '(' exp ')'
137 Now, rules whose left-hand side symbol is useless are no longer reported
138 as useless. The locations of the errors have also been adjusted to point
139 to the first use of the nonterminal as a left-hand side of a rule:
141 warning: 1 nonterminal useless in grammar [-Wother]
142 warning: 4 rules useless in grammar [-Wother]
143 3.1-3: warning: nonterminal useless in grammar: exp [-Wother]
144 exp: exp '+' exp | exp '-' exp | '(' exp ')'
146 2.14-16: warning: rule useless in grammar [-Wother]
150 *** C++: Generated parsers can be compiled with -fno-exceptions (lalr1.cc)
152 When compiled with exceptions disabled, the generated parsers no longer
153 uses try/catch clauses.
155 Currently only GCC and Clang are supported.
159 *** A demonstration of variants
161 A new example was added (installed in .../share/doc/bison/examples),
162 'variant.yy', which shows how to use (Bison) variants in C++.
164 The other examples were made nicer to read.
166 *** Some features are no longer 'experimental'
168 The following features, mature enough, are no longer flagged as
169 experimental in the documentation: push parsers, default %printer and
170 %destructor (typed: <*> and untyped: <>), %define api.value.type union and
171 variant, Java parsers, XML output, LR family (lr, ielr, lalr), and
172 semantic predicates (%?).
176 *** GLR: Predicates support broken by #line directives
178 Predicates (%?) in GLR such as
181 %? {new_syntax} 'w' id new_args
182 | %?{!new_syntax} 'w' id old_args
184 were issued with #lines in the middle of C code.
186 *** Printer and destructor with broken #line directives
188 The #line directives were not properly escaped when emitting the code for
189 %printer/%destructor, which resulted in compiler errors if there are
190 backslashes or double-quotes in the grammar file name.
192 *** Portability on ICC
194 The Intel compiler claims compatibility with GCC, yet rejects its _Pragma.
195 Generated parsers now work around this.
199 There were several small fixes in the test suite and in the build system,
200 many warnings in bison and in the generated parsers were eliminated. The
201 documentation also received its share of minor improvements.
203 Useless code was removed from C++ parsers, and some of the generated
204 constructors are more 'natural'.
206 * Noteworthy changes in release 3.0.5 (2018-05-27) [stable]
210 *** C++: Fix support of 'syntax_error'
212 One incorrect 'inline' resulted in linking errors about the constructor of
213 the syntax_error exception.
215 *** C++: Fix warnings
217 GCC 7.3 (with -O1 or -O2 but not -O0 or -O3) issued null-dereference
218 warnings about yyformat being possibly null. It also warned about the
219 deprecated implicit definition of copy constructors when there's a
220 user-defined (copy) assignment operator.
222 *** Location of errors
224 In C++ parsers, out-of-bounds errors can happen when a rule with an empty
225 ride-hand side raises a syntax error. The behavior of the default parser
226 (yacc.c) in such a condition was undefined.
228 Now all the parsers match the behavior of glr.c: @$ is used as the
229 location of the error. This handles gracefully rules with and without
232 *** Portability fixes in the test suite
234 On some platforms, some Java and/or C++ tests were failing.
236 * Noteworthy changes in release 3.0.4 (2015-01-23) [stable]
240 *** C++ with Variants (lalr1.cc)
242 Fix a compiler warning when no %destructor use $$.
246 Several portability issues in tests were fixed.
248 * Noteworthy changes in release 3.0.3 (2015-01-15) [stable]
252 *** C++ with Variants (lalr1.cc)
254 Problems with %destructor and '%define parse.assert' have been fixed.
256 *** Named %union support (yacc.c, glr.c)
258 Bison 3.0 introduced a regression on named %union such as
260 %union foo { int ival; };
262 The possibility to use a name was introduced "for Yacc compatibility".
263 It is however not required by POSIX Yacc, and its usefulness is not clear.
265 *** %define api.value.type union with %defines (yacc.c, glr.c)
267 The C parsers were broken when %defines was used together with "%define
268 api.value.type union".
270 *** Redeclarations are reported in proper order
278 bison used to report:
280 /tmp/foo.yy:2.10-11: error: %printer redeclaration for FOO
283 /tmp/foo.yy:3.10-11: previous declaration
287 Now, the "previous" declaration is always the first one.
292 Bison now installs various files in its docdir (which defaults to
293 '/usr/local/share/doc/bison'), including the three fully blown examples
294 extracted from the documentation:
297 Reverse Polish Calculator, a simple introductory example.
299 Multi-function Calc, a calculator with memory and functions and located
302 a calculator in C++ using variant support and token constructors.
304 * Noteworthy changes in release 3.0.2 (2013-12-05) [stable]
308 *** Generated source files when errors are reported
310 When warnings are issued and -Werror is set, bison would still generate
311 the source files (*.c, *.h...). As a consequence, some runs of "make"
312 could fail the first time, but not the second (as the files were generated
315 This is fixed: bison no longer generates this source files, but, of
316 course, still produces the various reports (*.output, *.xml, etc.).
318 *** %empty is used in reports
320 Empty right-hand sides are denoted by '%empty' in all the reports (text,
321 dot, XML and formats derived from it).
323 *** YYERROR and variants
325 When C++ variant support is enabled, an error triggered via YYERROR, but
326 not caught via error recovery, resulted in a double deletion.
328 * Noteworthy changes in release 3.0.1 (2013-11-12) [stable]
332 *** Errors in caret diagnostics
334 On some platforms, some errors could result in endless diagnostics.
336 *** Fixes of the -Werror option
338 Options such as "-Werror -Wno-error=foo" were still turning "foo"
339 diagnostics into errors instead of warnings. This is fixed.
341 Actually, for consistency with GCC, "-Wno-error=foo -Werror" now also
342 leaves "foo" diagnostics as warnings. Similarly, with "-Werror=foo
343 -Wno-error", "foo" diagnostics are now errors.
347 As demonstrated in the documentation, one can now leave spaces between
352 The yacc.1 man page is no longer installed if --disable-yacc was
355 *** Fixes in the test suite
357 Bugs and portability issues.
359 * Noteworthy changes in release 3.0 (2013-07-25) [stable]
361 ** WARNING: Future backward-incompatibilities!
363 Like other GNU packages, Bison will start using some of the C99 features
364 for its own code, especially the definition of variables after statements.
365 The generated C parsers still aim at C90.
367 ** Backward incompatible changes
369 *** Obsolete features
371 Support for YYFAIL is removed (deprecated in Bison 2.4.2): use YYERROR.
373 Support for yystype and yyltype is removed (deprecated in Bison 1.875):
374 use YYSTYPE and YYLTYPE.
376 Support for YYLEX_PARAM and YYPARSE_PARAM is removed (deprecated in Bison
377 1.875): use %lex-param, %parse-param, or %param.
379 Missing semicolons at the end of actions are no longer added (as announced
382 *** Use of YACC='bison -y'
384 TL;DR: With Autoconf <= 2.69, pass -Wno-yacc to (AM_)YFLAGS if you use
387 Traditional Yacc generates 'y.tab.c' whatever the name of the input file.
388 Therefore Makefiles written for Yacc expect 'y.tab.c' (and possibly
389 'y.tab.h' and 'y.outout') to be generated from 'foo.y'.
391 To this end, for ages, AC_PROG_YACC, Autoconf's macro to look for an
392 implementation of Yacc, was using Bison as 'bison -y'. While it does
393 ensure compatible output file names, it also enables warnings for
394 incompatibilities with POSIX Yacc. In other words, 'bison -y' triggers
395 warnings for Bison extensions.
397 Autoconf 2.70+ fixes this incompatibility by using YACC='bison -o y.tab.c'
398 (which also generates 'y.tab.h' and 'y.output' when needed).
399 Alternatively, disable Yacc warnings by passing '-Wno-yacc' to your Yacc
400 flags (YFLAGS, or AM_YFLAGS with Automake).
404 *** The epilogue is no longer affected by internal #defines (glr.c)
406 The glr.c skeleton uses defines such as #define yylval (yystackp->yyval) in
407 generated code. These weren't properly undefined before the inclusion of
408 the user epilogue, so functions such as the following were butchered by the
409 preprocessor expansion:
411 int yylex (YYSTYPE *yylval);
413 This is fixed: yylval, yynerrs, yychar, and yylloc are now valid
414 identifiers for user-provided variables.
416 *** stdio.h is no longer needed when locations are enabled (yacc.c)
418 Changes in Bison 2.7 introduced a dependency on FILE and fprintf when
419 locations are enabled. This is fixed.
421 *** Warnings about useless %pure-parser/%define api.pure are restored
423 ** Diagnostics reported by Bison
425 Most of these features were contributed by Théophile Ranquet and Victor
430 Version 2.7 introduced caret errors, for a prettier output. These are now
431 activated by default. The old format can still be used by invoking Bison
432 with -fno-caret (or -fnone).
434 Some error messages that reproduced excerpts of the grammar are now using
435 the caret information only. For instance on:
442 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
443 in.y:2.12-14: warning: rule useless in parser due to conflicts: exp: 'a' [-Wother]
447 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
448 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
452 and "bison -fno-caret" reports:
454 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
455 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
457 *** Enhancements of the -Werror option
459 The -Werror=CATEGORY option is now recognized, and will treat specified
460 warnings as errors. The warnings need not have been explicitly activated
461 using the -W option, this is similar to what GCC 4.7 does.
463 For example, given the following command line, Bison will treat both
464 warnings related to POSIX Yacc incompatibilities and S/R conflicts as
465 errors (and only those):
467 $ bison -Werror=yacc,error=conflicts-sr input.y
469 If no categories are specified, -Werror will make all active warnings into
470 errors. For example, the following line does the same the previous example:
472 $ bison -Werror -Wnone -Wyacc -Wconflicts-sr input.y
474 (By default -Wconflicts-sr,conflicts-rr,deprecated,other is enabled.)
476 Note that the categories in this -Werror option may not be prefixed with
477 "no-". However, -Wno-error[=CATEGORY] is valid.
479 Note that -y enables -Werror=yacc. Therefore it is now possible to require
480 Yacc-like behavior (e.g., always generate y.tab.c), but to report
481 incompatibilities as warnings: "-y -Wno-error=yacc".
483 *** The display of warnings is now richer
485 The option that controls a given warning is now displayed:
487 foo.y:4.6: warning: type clash on default action: <foo> != <bar> [-Wother]
489 In the case of warnings treated as errors, the prefix is changed from
490 "warning: " to "error: ", and the suffix is displayed, in a manner similar
491 to GCC, as [-Werror=CATEGORY].
493 For instance, where the previous version of Bison would report (and exit
496 bison: warnings being treated as errors
497 input.y:1.1: warning: stray ',' treated as white space
501 input.y:1.1: error: stray ',' treated as white space [-Werror=other]
503 *** Deprecated constructs
505 The new 'deprecated' warning category flags obsolete constructs whose
506 support will be discontinued. It is enabled by default. These warnings
507 used to be reported as 'other' warnings.
509 *** Useless semantic types
511 Bison now warns about useless (uninhabited) semantic types. Since
512 semantic types are not declared to Bison (they are defined in the opaque
513 %union structure), it is %printer/%destructor directives about useless
514 types that trigger the warning:
518 %printer {} <type1> <type3>
519 %destructor {} <type2> <type4>
521 nterm: term { $$ = $1; };
523 3.28-34: warning: type <type3> is used, but is not associated to any symbol
524 4.28-34: warning: type <type4> is used, but is not associated to any symbol
526 *** Undefined but unused symbols
528 Bison used to raise an error for undefined symbols that are not used in
529 the grammar. This is now only a warning.
532 %destructor {} symbol2
537 *** Useless destructors or printers
539 Bison now warns about useless destructors or printers. In the following
540 example, the printer for <type1>, and the destructor for <type2> are
541 useless: all symbols of <type1> (token1) already have a printer, and all
542 symbols of type <type2> (token2) already have a destructor.
544 %token <type1> token1
548 %printer {} token1 <type1> <type3>
549 %destructor {} token2 <type2> <type4>
553 The warnings and error messages about shift/reduce and reduce/reduce
554 conflicts have been normalized. For instance on the following foo.y file:
558 exp: exp '+' exp | '0' | '0';
560 compare the previous version of bison:
563 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
564 $ bison -Werror foo.y
565 bison: warnings being treated as errors
566 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
568 with the new behavior:
571 foo.y: warning: 1 shift/reduce conflict [-Wconflicts-sr]
572 foo.y: warning: 2 reduce/reduce conflicts [-Wconflicts-rr]
573 $ bison -Werror foo.y
574 foo.y: error: 1 shift/reduce conflict [-Werror=conflicts-sr]
575 foo.y: error: 2 reduce/reduce conflicts [-Werror=conflicts-rr]
577 When %expect or %expect-rr is used, such as with bar.y:
582 exp: exp '+' exp | '0' | '0';
587 bar.y: conflicts: 1 shift/reduce, 2 reduce/reduce
588 bar.y: expected 0 shift/reduce conflicts
589 bar.y: expected 0 reduce/reduce conflicts
594 bar.y: error: shift/reduce conflicts: 1 found, 0 expected
595 bar.y: error: reduce/reduce conflicts: 2 found, 0 expected
597 ** Incompatibilities with POSIX Yacc
599 The 'yacc' category is no longer part of '-Wall', enable it explicitly
602 ** Additional yylex/yyparse arguments
604 The new directive %param declares additional arguments to both yylex and
605 yyparse. The %lex-param, %parse-param, and %param directives support one
606 or more arguments. Instead of
608 %lex-param {arg1_type *arg1}
609 %lex-param {arg2_type *arg2}
610 %parse-param {arg1_type *arg1}
611 %parse-param {arg2_type *arg2}
615 %param {arg1_type *arg1} {arg2_type *arg2}
617 ** Types of values for %define variables
619 Bison used to make no difference between '%define foo bar' and '%define
620 foo "bar"'. The former is now called a 'keyword value', and the latter a
621 'string value'. A third kind was added: 'code values', such as '%define
624 Keyword variables are used for fixed value sets, e.g.,
628 Code variables are used for value in the target language, e.g.,
630 %define api.value.type {struct semantic_type}
632 String variables are used remaining cases, e.g. file names.
634 ** Variable api.token.prefix
636 The variable api.token.prefix changes the way tokens are identified in
637 the generated files. This is especially useful to avoid collisions
638 with identifiers in the target language. For instance
640 %token FILE for ERROR
641 %define api.token.prefix {TOK_}
643 start: FILE for ERROR;
645 will generate the definition of the symbols TOK_FILE, TOK_for, and
646 TOK_ERROR in the generated sources. In particular, the scanner must
647 use these prefixed token names, although the grammar itself still
648 uses the short names (as in the sample rule given above).
650 ** Variable api.value.type
652 This new %define variable supersedes the #define macro YYSTYPE. The use
653 of YYSTYPE is discouraged. In particular, #defining YYSTYPE *and* either
654 using %union or %defining api.value.type results in undefined behavior.
656 Either define api.value.type, or use "%union":
663 %token <ival> INT "integer"
664 %token <sval> STRING "string"
665 %printer { fprintf (yyo, "%d", $$); } <ival>
666 %destructor { free ($$); } <sval>
669 yylval.ival = 42; return INT;
670 yylval.sval = "42"; return STRING;
672 The %define variable api.value.type supports both keyword and code values.
674 The keyword value 'union' means that the user provides genuine types, not
675 union member names such as "ival" and "sval" above (WARNING: will fail if
676 -y/--yacc/%yacc is enabled).
678 %define api.value.type union
679 %token <int> INT "integer"
680 %token <char *> STRING "string"
681 %printer { fprintf (yyo, "%d", $$); } <int>
682 %destructor { free ($$); } <char *>
685 yylval.INT = 42; return INT;
686 yylval.STRING = "42"; return STRING;
688 The keyword value variant is somewhat equivalent, but for C++ special
689 provision is made to allow classes to be used (more about this below).
691 %define api.value.type variant
692 %token <int> INT "integer"
693 %token <std::string> STRING "string"
695 Code values (in braces) denote user defined types. This is where YYSTYPE
713 %define api.value.type {struct my_value}
714 %token <u.ival> INT "integer"
715 %token <u.sval> STRING "string"
716 %printer { fprintf (yyo, "%d", $$); } <u.ival>
717 %destructor { free ($$); } <u.sval>
720 yylval.u.ival = 42; return INT;
721 yylval.u.sval = "42"; return STRING;
723 ** Variable parse.error
725 This variable controls the verbosity of error messages. The use of the
726 %error-verbose directive is deprecated in favor of "%define parse.error
729 ** Renamed %define variables
731 The following variables have been renamed for consistency. Backward
732 compatibility is ensured, but upgrading is recommended.
734 lr.default-reductions -> lr.default-reduction
735 lr.keep-unreachable-states -> lr.keep-unreachable-state
736 namespace -> api.namespace
737 stype -> api.value.type
739 ** Semantic predicates
741 Contributed by Paul Hilfinger.
743 The new, experimental, semantic-predicate feature allows actions of the
744 form "%?{ BOOLEAN-EXPRESSION }", which cause syntax errors (as for
745 YYERROR) if the expression evaluates to 0, and are evaluated immediately
746 in GLR parsers, rather than being deferred. The result is that they allow
747 the programmer to prune possible parses based on the values of run-time
750 ** The directive %expect-rr is now an error in non GLR mode
752 It used to be an error only if used in non GLR mode, _and_ if there are
753 reduce/reduce conflicts.
755 ** Tokens are numbered in their order of appearance
757 Contributed by Valentin Tolmer.
759 With '%token A B', A had a number less than the one of B. However,
760 precedence declarations used to generate a reversed order. This is now
761 fixed, and introducing tokens with any of %token, %left, %right,
762 %precedence, or %nonassoc yields the same result.
764 When mixing declarations of tokens with a literal character (e.g., 'a') or
765 with an identifier (e.g., B) in a precedence declaration, Bison numbered
766 the literal characters first. For example
770 would lead to the tokens declared in this order: 'c' 'd' A B. Again, the
771 input order is now preserved.
773 These changes were made so that one can remove useless precedence and
774 associativity declarations (i.e., map %nonassoc, %left or %right to
775 %precedence, or to %token) and get exactly the same output.
777 ** Useless precedence and associativity
779 Contributed by Valentin Tolmer.
781 When developing and maintaining a grammar, useless associativity and
782 precedence directives are common. They can be a nuisance: new ambiguities
783 arising are sometimes masked because their conflicts are resolved due to
784 the extra precedence or associativity information. Furthermore, it can
785 hinder the comprehension of a new grammar: one will wonder about the role
786 of a precedence, where in fact it is useless. The following changes aim
787 at detecting and reporting these extra directives.
789 *** Precedence warning category
791 A new category of warning, -Wprecedence, was introduced. It flags the
792 useless precedence and associativity directives.
794 *** Useless associativity
796 Bison now warns about symbols with a declared associativity that is never
797 used to resolve conflicts. In that case, using %precedence is sufficient;
798 the parsing tables will remain unchanged. Solving these warnings may raise
799 useless precedence warnings, as the symbols no longer have associativity.
813 warning: useless associativity for '+', use %precedence [-Wprecedence]
817 *** Useless precedence
819 Bison now warns about symbols with a declared precedence and no declared
820 associativity (i.e., declared with %precedence), and whose precedence is
821 never used. In that case, the symbol can be safely declared with %token
822 instead, without modifying the parsing tables. For example:
826 exp: "var" '=' "number";
830 warning: useless precedence for '=' [-Wprecedence]
834 *** Useless precedence and associativity
836 In case of both useless precedence and associativity, the issue is flagged
841 exp: "var" '=' "number";
845 warning: useless precedence and associativity for '=' [-Wprecedence]
851 With help from Joel E. Denny and Gabriel Rassoul.
853 Empty rules (i.e., with an empty right-hand side) can now be explicitly
854 marked by the new %empty directive. Using %empty on a non-empty rule is
855 an error. The new -Wempty-rule warning reports empty rules without
856 %empty. On the following grammar:
866 3.4-5: warning: empty rule without %empty [-Wempty-rule]
869 5.8-13: error: %empty on non-empty rule
873 ** Java skeleton improvements
875 The constants for token names were moved to the Lexer interface. Also, it
876 is possible to add code to the parser's constructors using "%code init"
877 and "%define init_throws".
878 Contributed by Paolo Bonzini.
880 The Java skeleton now supports push parsing.
881 Contributed by Dennis Heimbigner.
883 ** C++ skeletons improvements
885 *** The parser header is no longer mandatory (lalr1.cc, glr.cc)
887 Using %defines is now optional. Without it, the needed support classes
888 are defined in the generated parser, instead of additional files (such as
889 location.hh, position.hh and stack.hh).
891 *** Locations are no longer mandatory (lalr1.cc, glr.cc)
893 Both lalr1.cc and glr.cc no longer require %location.
895 *** syntax_error exception (lalr1.cc)
897 The C++ parser features a syntax_error exception, which can be
898 thrown from the scanner or from user rules to raise syntax errors.
899 This facilitates reporting errors caught in sub-functions (e.g.,
900 rejecting too large integral literals from a conversion function
901 used by the scanner, or rejecting invalid combinations from a
902 factory invoked by the user actions).
904 *** %define api.value.type variant
906 This is based on a submission from Michiel De Wilde. With help
907 from Théophile Ranquet.
909 In this mode, complex C++ objects can be used as semantic values. For
912 %token <::std::string> TEXT;
915 %type <::std::string> item;
916 %type <::std::list<std::string>> list;
919 list { std::cout << $1 << std::endl; }
923 %empty { /* Generates an empty string list. */ }
924 | list item ";" { std::swap ($$, $1); $$.push_back ($2); }
928 TEXT { std::swap ($$, $1); }
929 | NUMBER { $$ = string_cast ($1); }
932 *** %define api.token.constructor
934 When variants are enabled, Bison can generate functions to build the
935 tokens. This guarantees that the token type (e.g., NUMBER) is consistent
936 with the semantic value (e.g., int):
938 parser::symbol_type yylex ()
940 parser::location_type loc = ...;
942 return parser::make_TEXT ("Hello, world!", loc);
944 return parser::make_NUMBER (42, loc);
946 return parser::make_SEMICOLON (loc);
952 There are operator- and operator-= for 'location'. Negative line/column
953 increments can no longer underflow the resulting value.
955 * Noteworthy changes in release 2.7.1 (2013-04-15) [stable]
959 *** Fix compiler attribute portability (yacc.c)
961 With locations enabled, __attribute__ was used unprotected.
963 *** Fix some compiler warnings (lalr1.cc)
965 * Noteworthy changes in release 2.7 (2012-12-12) [stable]
969 Warnings about uninitialized yylloc in yyparse have been fixed.
971 Restored C90 compliance (yet no report was ever made).
973 ** Diagnostics are improved
975 Contributed by Théophile Ranquet.
977 *** Changes in the format of error messages
979 This used to be the format of many error reports:
981 input.y:2.7-12: %type redeclaration for exp
982 input.y:1.7-12: previous declaration
986 input.y:2.7-12: error: %type redeclaration for exp
987 input.y:1.7-12: previous declaration
989 *** New format for error reports: carets
991 Caret errors have been added to Bison:
993 input.y:2.7-12: error: %type redeclaration for exp
996 input.y:1.7-12: previous declaration
1002 input.y:3.20-23: error: ambiguous reference: '$exp'
1003 exp: exp '+' exp { $exp = $1 + $3; };
1005 input.y:3.1-3: refers to: $exp at $$
1006 exp: exp '+' exp { $exp = $1 + $3; };
1008 input.y:3.6-8: refers to: $exp at $1
1009 exp: exp '+' exp { $exp = $1 + $3; };
1011 input.y:3.14-16: refers to: $exp at $3
1012 exp: exp '+' exp { $exp = $1 + $3; };
1015 The default behavior for now is still not to display these unless
1016 explicitly asked with -fcaret (or -fall). However, in a later release, it
1017 will be made the default behavior (but may still be deactivated with
1020 ** New value for %define variable: api.pure full
1022 The %define variable api.pure requests a pure (reentrant) parser. However,
1023 for historical reasons, using it in a location-tracking Yacc parser
1024 resulted in a yyerror function that did not take a location as a
1025 parameter. With this new value, the user may request a better pure parser,
1026 where yyerror does take a location as a parameter (in location-tracking
1029 The use of "%define api.pure true" is deprecated in favor of this new
1030 "%define api.pure full".
1032 ** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java)
1034 The %define variable api.location.type defines the name of the type to use
1035 for locations. When defined, Bison no longer generates the position.hh
1036 and location.hh files, nor does the parser will include them: the user is
1037 then responsible to define her type.
1039 This can be used in programs with several parsers to factor their location
1040 and position files: let one of them generate them, and the others just use
1043 This feature was actually introduced, but not documented, in Bison 2.5,
1044 under the name "location_type" (which is maintained for backward
1047 For consistency, lalr1.java's %define variables location_type and
1048 position_type are deprecated in favor of api.location.type and
1051 ** Exception safety (lalr1.cc)
1053 The parse function now catches exceptions, uses the %destructors to
1054 release memory (the lookahead symbol and the symbols pushed on the stack)
1055 before re-throwing the exception.
1057 This feature is somewhat experimental. User feedback would be
1060 ** Graph improvements in DOT and XSLT
1062 Contributed by Théophile Ranquet.
1064 The graphical presentation of the states is more readable: their shape is
1065 now rectangular, the state number is clearly displayed, and the items are
1066 numbered and left-justified.
1068 The reductions are now explicitly represented as transitions to other
1069 diamond shaped nodes.
1071 These changes are present in both --graph output and xml2dot.xsl XSLT
1072 processing, with minor (documented) differences.
1074 ** %language is no longer an experimental feature.
1076 The introduction of this feature, in 2.4, was four years ago. The
1077 --language option and the %language directive are no longer experimental.
1081 The sections about shift/reduce and reduce/reduce conflicts resolution
1082 have been fixed and extended.
1084 Although introduced more than four years ago, XML and Graphviz reports
1085 were not properly documented.
1087 The translation of midrule actions is now described.
1089 * Noteworthy changes in release 2.6.5 (2012-11-07) [stable]
1091 We consider compiler warnings about Bison generated parsers to be bugs.
1092 Rather than working around them in your own project, please consider
1093 reporting them to us.
1097 Warnings about uninitialized yylval and/or yylloc for push parsers with a
1098 pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to
1101 Other issues in the test suite have been addressed.
1103 Null characters are correctly displayed in error messages.
1105 When possible, yylloc is correctly initialized before calling yylex. It
1106 is no longer necessary to initialize it in the %initial-action.
1108 * Noteworthy changes in release 2.6.4 (2012-10-23) [stable]
1110 Bison 2.6.3's --version was incorrect. This release fixes this issue.
1112 * Noteworthy changes in release 2.6.3 (2012-10-22) [stable]
1116 Bugs and portability issues in the test suite have been fixed.
1118 Some errors in translations have been addressed, and --help now directs
1119 users to the appropriate place to report them.
1121 Stray Info files shipped by accident are removed.
1123 Incorrect definitions of YY_, issued by yacc.c when no parser header is
1124 generated, are removed.
1126 All the generated headers are self-contained.
1128 ** Header guards (yacc.c, glr.c, glr.cc)
1130 In order to avoid collisions, the header guards are now
1131 YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>.
1132 For instance the header generated from
1134 %define api.prefix "calc"
1135 %defines "lib/parse.h"
1137 will use YY_CALC_LIB_PARSE_H_INCLUDED as guard.
1139 ** Fix compiler warnings in the generated parser (yacc.c, glr.c)
1141 The compilation of pure parsers (%define api.pure) can trigger GCC
1144 input.c: In function 'yyparse':
1145 input.c:1503:12: warning: 'yylval' may be used uninitialized in this
1146 function [-Wmaybe-uninitialized]
1150 This is now fixed; pragmas to avoid these warnings are no longer needed.
1152 Warnings from clang ("equality comparison with extraneous parentheses" and
1153 "function declared 'noreturn' should not return") have also been
1156 * Noteworthy changes in release 2.6.2 (2012-08-03) [stable]
1160 Buffer overruns, complaints from Flex, and portability issues in the test
1161 suite have been fixed.
1163 ** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc)
1165 Trailing end-of-lines in %parse-param or %lex-param would result in
1166 invalid C++. This is fixed.
1168 ** Spurious spaces and end-of-lines
1170 The generated files no longer end (nor start) with empty lines.
1172 * Noteworthy changes in release 2.6.1 (2012-07-30) [stable]
1174 Bison no longer executes user-specified M4 code when processing a grammar.
1178 In addition to the removal of the features announced in Bison 2.6, the
1179 next major release will remove the "Temporary hack for adding a semicolon
1180 to the user action", as announced in the release 2.5. Instead of:
1182 exp: exp "+" exp { $$ = $1 + $3 };
1186 exp: exp "+" exp { $$ = $1 + $3; };
1190 *** Type names are now properly escaped.
1192 *** glr.cc: set_debug_level and debug_level work as expected.
1194 *** Stray @ or $ in actions
1196 While Bison used to warn about stray $ or @ in action rules, it did not
1197 for other actions such as printers, destructors, or initial actions. It
1200 ** Type names in actions
1202 For consistency with rule actions, it is now possible to qualify $$ by a
1203 type-name in destructors, printers, and initial actions. For instance:
1205 %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>;
1207 will display two values for each typed and untyped symbol (provided
1208 that YYSTYPE has both "ival" and "fval" fields).
1210 * Noteworthy changes in release 2.6 (2012-07-19) [stable]
1214 The next major release of Bison will drop support for the following
1215 deprecated features. Please report disagreements to bug-bison@gnu.org.
1219 Support for generating parsers in K&R C will be removed. Parsers
1220 generated for C support ISO C90, and are tested with ISO C99 and ISO C11
1223 *** Features deprecated since Bison 1.875
1225 The definitions of yystype and yyltype will be removed; use YYSTYPE and
1228 YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and
1229 %lex-param, will no longer be supported.
1231 Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use
1234 *** The generated header will be included (yacc.c)
1236 Instead of duplicating the content of the generated header (definition of
1237 YYSTYPE, yyparse declaration etc.), the generated parser will include it,
1238 as is already the case for GLR or C++ parsers. This change is deferred
1239 because existing versions of ylwrap (e.g., Automake 1.12.1) do not support
1242 ** Generated Parser Headers
1244 *** Guards (yacc.c, glr.c, glr.cc)
1246 The generated headers are now guarded, as is already the case for C++
1247 parsers (lalr1.cc). For instance, with --defines=foo.h:
1252 #endif /* !YY_FOO_H */
1254 *** New declarations (yacc.c, glr.c)
1256 The generated header now declares yydebug and yyparse. Both honor
1257 --name-prefix=bar_, and yield
1259 int bar_parse (void);
1263 #define yyparse bar_parse
1266 in order to facilitate the inclusion of several parser headers inside a
1267 single compilation unit.
1269 *** Exported symbols in C++
1271 The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the
1272 header, are removed, as they prevent the possibility of including several
1273 generated headers from a single compilation unit.
1277 For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no
1280 ** New %define variable: api.prefix
1282 Now that the generated headers are more complete and properly protected
1283 against multiple inclusions, constant names, such as YYSTYPE are a
1284 problem. While yyparse and others are properly renamed by %name-prefix,
1285 YYSTYPE, YYDEBUG and others have never been affected by it. Because it
1286 would introduce backward compatibility issues in projects not expecting
1287 YYSTYPE to be renamed, instead of changing the behavior of %name-prefix,
1288 it is deprecated in favor of a new %define variable: api.prefix.
1290 The following examples compares both:
1292 %name-prefix "bar_" | %define api.prefix "bar_"
1293 %token <ival> FOO %token <ival> FOO
1294 %union { int ival; } %union { int ival; }
1300 #ifndef BAR_FOO_H #ifndef BAR_FOO_H
1301 # define BAR_FOO_H # define BAR_FOO_H
1303 /* Enabling traces. */ /* Enabling traces. */
1304 # ifndef YYDEBUG | # ifndef BAR_DEBUG
1305 > # if defined YYDEBUG
1307 > # define BAR_DEBUG 1
1309 > # define BAR_DEBUG 0
1312 # define YYDEBUG 0 | # define BAR_DEBUG 0
1316 # if YYDEBUG | # if BAR_DEBUG
1317 extern int bar_debug; extern int bar_debug;
1320 /* Tokens. */ /* Tokens. */
1321 # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE
1322 # define YYTOKENTYPE | # define BAR_TOKENTYPE
1323 enum yytokentype { | enum bar_tokentype {
1328 #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \
1329 && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED
1330 typedef union YYSTYPE | typedef union BAR_STYPE
1333 } YYSTYPE; | } BAR_STYPE;
1334 # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1
1337 extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval;
1339 int bar_parse (void); int bar_parse (void);
1341 #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */
1343 * Noteworthy changes in release 2.5.1 (2012-06-05) [stable]
1347 The next major release will drop support for generating parsers in K&R C.
1349 ** yacc.c: YYBACKUP works as expected.
1351 ** glr.c improvements:
1353 *** Location support is eliminated when not requested:
1355 GLR parsers used to include location-related code even when locations were
1356 not requested, and therefore not even usable.
1358 *** __attribute__ is preserved:
1360 __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e.,
1361 when -std is passed to GCC).
1363 ** lalr1.java: several fixes:
1365 The Java parser no longer throws ArrayIndexOutOfBoundsException if the
1366 first token leads to a syntax error. Some minor clean ups.
1370 *** C++11 compatibility:
1372 C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L
1377 The header files such as "parser.hh", "location.hh", etc. used a constant
1378 name for preprocessor guards, for instance:
1380 #ifndef BISON_LOCATION_HH
1381 # define BISON_LOCATION_HH
1383 #endif // !BISON_LOCATION_HH
1385 The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower
1386 case characters are converted to upper case, and series of
1387 non-alphanumerical characters are converted to an underscore.
1389 With "bison -o lang++/parser.cc", "location.hh" would now include:
1391 #ifndef YY_LANG_LOCATION_HH
1392 # define YY_LANG_LOCATION_HH
1394 #endif // !YY_LANG_LOCATION_HH
1398 The position and location constructors (and their initialize methods)
1399 accept new arguments for line and column. Several issues in the
1400 documentation were fixed.
1402 ** liby is no longer asking for "rpl_fprintf" on some platforms.
1404 ** Changes in the manual:
1406 *** %printer is documented
1408 The "%printer" directive, supported since at least Bison 1.50, is finally
1409 documented. The "mfcalc" example is extended to demonstrate it.
1411 For consistency with the C skeletons, the C++ parsers now also support
1412 "yyoutput" (as an alias to "debug_stream ()").
1414 *** Several improvements have been made:
1416 The layout for grammar excerpts was changed to a more compact scheme.
1417 Named references are motivated. The description of the automaton
1418 description file (*.output) is updated to the current format. Incorrect
1419 index entries were fixed. Some other errors were fixed.
1423 *** Conflicting prototypes with recent/modified Flex.
1425 Fixed build problems with the current, unreleased, version of Flex, and
1426 some modified versions of 2.5.35, which have modified function prototypes.
1428 *** Warnings during the build procedure have been eliminated.
1430 *** Several portability problems in the test suite have been fixed:
1432 This includes warnings with some compilers, unexpected behavior of tools
1433 such as diff, warning messages from the test suite itself, etc.
1435 *** The install-pdf target works properly:
1437 Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer
1438 halts in the middle of its course.
1440 * Changes in version 2.5 (2011-05-14):
1442 ** Grammar symbol names can now contain non-initial dashes:
1444 Consistently with directives (such as %error-verbose) and with
1445 %define variables (e.g. push-pull), grammar symbol names may contain
1446 dashes in any position except the beginning. This is a GNU
1447 extension over POSIX Yacc. Thus, use of this extension is reported
1448 by -Wyacc and rejected in Yacc mode (--yacc).
1450 ** Named references:
1452 Historically, Yacc and Bison have supported positional references
1453 ($n, $$) to allow access to symbol values from inside of semantic
1456 Starting from this version, Bison can also accept named references.
1457 When no ambiguity is possible, original symbol names may be used
1458 as named references:
1460 if_stmt : "if" cond_expr "then" then_stmt ';'
1461 { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }
1463 In the more common case, explicit names may be declared:
1465 stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';'
1466 { $res = mk_if_stmt($cond, $then, $else); }
1468 Location information is also accessible using @name syntax. When
1469 accessing symbol names containing dots or dashes, explicit bracketing
1470 ($[sym.1]) must be used.
1472 These features are experimental in this version. More user feedback
1473 will help to stabilize them.
1474 Contributed by Alex Rozenman.
1476 ** IELR(1) and canonical LR(1):
1478 IELR(1) is a minimal LR(1) parser table generation algorithm. That
1479 is, given any context-free grammar, IELR(1) generates parser tables
1480 with the full language-recognition power of canonical LR(1) but with
1481 nearly the same number of parser states as LALR(1). This reduction
1482 in parser states is often an order of magnitude. More importantly,
1483 because canonical LR(1)'s extra parser states may contain duplicate
1484 conflicts in the case of non-LR(1) grammars, the number of conflicts
1485 for IELR(1) is often an order of magnitude less as well. This can
1486 significantly reduce the complexity of developing of a grammar.
1488 Bison can now generate IELR(1) and canonical LR(1) parser tables in
1489 place of its traditional LALR(1) parser tables, which remain the
1490 default. You can specify the type of parser tables in the grammar
1491 file with these directives:
1493 %define lr.type lalr
1494 %define lr.type ielr
1495 %define lr.type canonical-lr
1497 The default-reduction optimization in the parser tables can also be
1498 adjusted using "%define lr.default-reductions". For details on both
1499 of these features, see the new section "Tuning LR" in the Bison
1502 These features are experimental. More user feedback will help to
1505 ** LAC (Lookahead Correction) for syntax error handling
1507 Contributed by Joel E. Denny.
1509 Canonical LR, IELR, and LALR can suffer from a couple of problems
1510 upon encountering a syntax error. First, the parser might perform
1511 additional parser stack reductions before discovering the syntax
1512 error. Such reductions can perform user semantic actions that are
1513 unexpected because they are based on an invalid token, and they
1514 cause error recovery to begin in a different syntactic context than
1515 the one in which the invalid token was encountered. Second, when
1516 verbose error messages are enabled (with %error-verbose or the
1517 obsolete "#define YYERROR_VERBOSE"), the expected token list in the
1518 syntax error message can both contain invalid tokens and omit valid
1521 The culprits for the above problems are %nonassoc, default
1522 reductions in inconsistent states, and parser state merging. Thus,
1523 IELR and LALR suffer the most. Canonical LR can suffer only if
1524 %nonassoc is used or if default reductions are enabled for
1525 inconsistent states.
1527 LAC is a new mechanism within the parsing algorithm that solves
1528 these problems for canonical LR, IELR, and LALR without sacrificing
1529 %nonassoc, default reductions, or state merging. When LAC is in
1530 use, canonical LR and IELR behave almost exactly the same for both
1531 syntactically acceptable and syntactically unacceptable input.
1532 While LALR still does not support the full language-recognition
1533 power of canonical LR and IELR, LAC at least enables LALR's syntax
1534 error handling to correctly reflect LALR's language-recognition
1537 Currently, LAC is only supported for deterministic parsers in C.
1538 You can enable LAC with the following directive:
1540 %define parse.lac full
1542 See the new section "LAC" in the Bison manual for additional
1543 details including a few caveats.
1545 LAC is an experimental feature. More user feedback will help to
1548 ** %define improvements:
1550 *** Can now be invoked via the command line:
1552 Each of these command-line options
1555 --define=NAME[=VALUE]
1558 --force-define=NAME[=VALUE]
1560 is equivalent to this grammar file declaration
1562 %define NAME ["VALUE"]
1564 except that the manner in which Bison processes multiple definitions
1565 for the same NAME differs. Most importantly, -F and --force-define
1566 quietly override %define, but -D and --define do not. For further
1567 details, see the section "Bison Options" in the Bison manual.
1569 *** Variables renamed:
1571 The following %define variables
1574 lr.keep_unreachable_states
1576 have been renamed to
1579 lr.keep-unreachable-states
1581 The old names are now deprecated but will be maintained indefinitely
1582 for backward compatibility.
1584 *** Values no longer need to be quoted in the grammar file:
1586 If a %define value is an identifier, it no longer needs to be placed
1587 within quotations marks. For example,
1589 %define api.push-pull "push"
1593 %define api.push-pull push
1595 *** Unrecognized variables are now errors not warnings.
1597 *** Multiple invocations for any variable is now an error not a warning.
1599 ** Unrecognized %code qualifiers are now errors not warnings.
1601 ** Character literals not of length one:
1603 Previously, Bison quietly converted all character literals to length
1604 one. For example, without warning, Bison interpreted the operators in
1605 the following grammar to be the same token:
1611 Bison now warns when a character literal is not of length one. In
1612 some future release, Bison will start reporting an error instead.
1614 ** Destructor calls fixed for lookaheads altered in semantic actions:
1616 Previously for deterministic parsers in C, if a user semantic action
1617 altered yychar, the parser in some cases used the old yychar value to
1618 determine which destructor to call for the lookahead upon a syntax
1619 error or upon parser return. This bug has been fixed.
1621 ** C++ parsers use YYRHSLOC:
1623 Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
1624 macro and use it in the default YYLLOC_DEFAULT. You are encouraged
1625 to use it. If, for instance, your location structure has "first"
1626 and "last" members, instead of
1628 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1632 (Current).first = (Rhs)[1].location.first; \
1633 (Current).last = (Rhs)[N].location.last; \
1637 (Current).first = (Current).last = (Rhs)[0].location.last; \
1643 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1647 (Current).first = YYRHSLOC (Rhs, 1).first; \
1648 (Current).last = YYRHSLOC (Rhs, N).last; \
1652 (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \
1656 ** YYLLOC_DEFAULT in C++:
1658 The default implementation of YYLLOC_DEFAULT used to be issued in
1659 the header file. It is now output in the implementation file, after
1660 the user %code sections so that its #ifndef guard does not try to
1661 override the user's YYLLOC_DEFAULT if provided.
1663 ** YYFAIL now produces warnings and Java parsers no longer implement it:
1665 YYFAIL has existed for many years as an undocumented feature of
1666 deterministic parsers in C generated by Bison. More recently, it was
1667 a documented feature of Bison's experimental Java parsers. As
1668 promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
1669 semantic action now produces a deprecation warning, and Java parsers
1670 no longer implement YYFAIL at all. For further details, including a
1671 discussion of how to suppress C preprocessor warnings about YYFAIL
1672 being unused, see the Bison 2.4.2 NEWS entry.
1674 ** Temporary hack for adding a semicolon to the user action:
1676 Previously, Bison appended a semicolon to every user action for
1677 reductions when the output language defaulted to C (specifically, when
1678 neither %yacc, %language, %skeleton, or equivalent command-line
1679 options were specified). This allowed actions such as
1681 exp: exp "+" exp { $$ = $1 + $3 };
1685 exp: exp "+" exp { $$ = $1 + $3; };
1687 As a first step in removing this misfeature, Bison now issues a
1688 warning when it appends a semicolon. Moreover, in cases where Bison
1689 cannot easily determine whether a semicolon is needed (for example, an
1690 action ending with a cpp directive or a braced compound initializer),
1691 it no longer appends one. Thus, the C compiler might now complain
1692 about a missing semicolon where it did not before. Future releases of
1693 Bison will cease to append semicolons entirely.
1695 ** Verbose syntax error message fixes:
1697 When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is
1698 specified, syntax error messages produced by the generated parser
1699 include the unexpected token as well as a list of expected tokens.
1700 The effect of %nonassoc on these verbose messages has been corrected
1701 in two ways, but a more complete fix requires LAC, described above:
1703 *** When %nonassoc is used, there can exist parser states that accept no
1704 tokens, and so the parser does not always require a lookahead token
1705 in order to detect a syntax error. Because no unexpected token or
1706 expected tokens can then be reported, the verbose syntax error
1707 message described above is suppressed, and the parser instead
1708 reports the simpler message, "syntax error". Previously, this
1709 suppression was sometimes erroneously triggered by %nonassoc when a
1710 lookahead was actually required. Now verbose messages are
1711 suppressed only when all previous lookaheads have already been
1712 shifted or discarded.
1714 *** Previously, the list of expected tokens erroneously included tokens
1715 that would actually induce a syntax error because conflicts for them
1716 were resolved with %nonassoc in the current parser state. Such
1717 tokens are now properly omitted from the list.
1719 *** Expected token lists are still often wrong due to state merging
1720 (from LALR or IELR) and default reductions, which can both add
1721 invalid tokens and subtract valid tokens. Canonical LR almost
1722 completely fixes this problem by eliminating state merging and
1723 default reductions. However, there is one minor problem left even
1724 when using canonical LR and even after the fixes above. That is,
1725 if the resolution of a conflict with %nonassoc appears in a later
1726 parser state than the one at which some syntax error is
1727 discovered, the conflicted token is still erroneously included in
1728 the expected token list. Bison's new LAC implementation,
1729 described above, eliminates this problem and the need for
1730 canonical LR. However, LAC is still experimental and is disabled
1733 ** Java skeleton fixes:
1735 *** A location handling bug has been fixed.
1737 *** The top element of each of the value stack and location stack is now
1738 cleared when popped so that it can be garbage collected.
1740 *** Parser traces now print the top element of the stack.
1742 ** -W/--warnings fixes:
1744 *** Bison now properly recognizes the "no-" versions of categories:
1746 For example, given the following command line, Bison now enables all
1747 warnings except warnings for incompatibilities with POSIX Yacc:
1749 bison -Wall,no-yacc gram.y
1751 *** Bison now treats S/R and R/R conflicts like other warnings:
1753 Previously, conflict reports were independent of Bison's normal
1754 warning system. Now, Bison recognizes the warning categories
1755 "conflicts-sr" and "conflicts-rr". This change has important
1756 consequences for the -W and --warnings command-line options. For
1759 bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
1760 bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
1761 bison -Wnone gram.y # no conflicts are reported
1762 bison -Werror gram.y # any conflict is an error
1764 However, as before, if the %expect or %expect-rr directive is
1765 specified, an unexpected number of conflicts is an error, and an
1766 expected number of conflicts is not reported, so -W and --warning
1767 then have no effect on the conflict report.
1769 *** The "none" category no longer disables a preceding "error":
1771 For example, for the following command line, Bison now reports
1772 errors instead of warnings for incompatibilities with POSIX Yacc:
1774 bison -Werror,none,yacc gram.y
1776 *** The "none" category now disables all Bison warnings:
1778 Previously, the "none" category disabled only Bison warnings for
1779 which there existed a specific -W/--warning category. However,
1780 given the following command line, Bison is now guaranteed to
1781 suppress all warnings:
1785 ** Precedence directives can now assign token number 0:
1787 Since Bison 2.3b, which restored the ability of precedence
1788 directives to assign token numbers, doing so for token number 0 has
1789 produced an assertion failure. For example:
1793 This bug has been fixed.
1795 * Changes in version 2.4.3 (2010-08-05):
1797 ** Bison now obeys -Werror and --warnings=error for warnings about
1798 grammar rules that are useless in the parser due to conflicts.
1800 ** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
1803 ** Failures in the test suite for GCC 4.5 have been fixed.
1805 ** Failures in the test suite for some versions of Sun Studio C++ have
1808 ** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
1809 warnings about undefined %prec identifiers will not be converted to
1810 errors in Bison 2.5. They will remain warnings, which should be
1811 sufficient for POSIX while avoiding backward compatibility issues.
1813 ** Minor documentation fixes.
1815 * Changes in version 2.4.2 (2010-03-20):
1817 ** Some portability problems that resulted in failures and livelocks
1818 in the test suite on some versions of at least Solaris, AIX, HP-UX,
1819 RHEL4, and Tru64 have been addressed. As a result, fatal Bison
1820 errors should no longer cause M4 to report a broken pipe on the
1823 ** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately.
1825 POSIX specifies that an error be reported for any identifier that does
1826 not appear on the LHS of a grammar rule and that is not defined by
1827 %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this
1828 error report for the case when an identifier appears only after a
1829 %prec directive. It is now restored. However, for backward
1830 compatibility with recent Bison releases, it is only a warning for
1831 now. In Bison 2.5 and later, it will return to being an error.
1832 [Between the 2.4.2 and 2.4.3 releases, it was decided that this
1833 warning will not be converted to an error in Bison 2.5.]
1835 ** Detection of GNU M4 1.4.6 or newer during configure is improved.
1837 ** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
1838 YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
1841 ** %code is now a permanent feature.
1843 A traditional Yacc prologue directive is written in the form:
1847 To provide a more flexible alternative, Bison 2.3b introduced the
1848 %code directive with the following forms for C/C++:
1851 %code requires {CODE}
1852 %code provides {CODE}
1855 These forms are now considered permanent features of Bison. See the
1856 %code entries in the section "Bison Declaration Summary" in the Bison
1857 manual for a summary of their functionality. See the section
1858 "Prologue Alternatives" for a detailed discussion including the
1859 advantages of %code over the traditional Yacc prologue directive.
1861 Bison's Java feature as a whole including its current usage of %code
1862 is still considered experimental.
1864 ** YYFAIL is deprecated and will eventually be removed.
1866 YYFAIL has existed for many years as an undocumented feature of
1867 deterministic parsers in C generated by Bison. Previously, it was
1868 documented for Bison's experimental Java parsers. YYFAIL is no longer
1869 documented for Java parsers and is formally deprecated in both cases.
1870 Users are strongly encouraged to migrate to YYERROR, which is
1873 Like YYERROR, you can invoke YYFAIL from a semantic action in order to
1874 induce a syntax error. The most obvious difference from YYERROR is
1875 that YYFAIL will automatically invoke yyerror to report the syntax
1876 error so that you don't have to. However, there are several other
1877 subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
1878 inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is
1879 used. For a more detailed discussion, see:
1881 http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
1883 The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
1884 deterministic parsers in C will continue to implement it. However,
1885 because YYFAIL is already flawed, it seems futile to try to make new
1886 Bison features compatible with it. Thus, during parser generation,
1887 Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
1888 rule action. In a later release, YYFAIL will be disabled for
1889 %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will
1890 be removed altogether.
1892 There exists at least one case where Bison 2.5's YYFAIL warning will
1893 be a false positive. Some projects add phony uses of YYFAIL and other
1894 Bison-defined macros for the sole purpose of suppressing C
1895 preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
1896 To avoid Bison's future warning, such YYFAIL uses can be moved to the
1897 epilogue (that is, after the second "%%") in the Bison input file. In
1898 this release (2.4.2), Bison already generates its own code to suppress
1899 C preprocessor warnings for YYFAIL, so projects can remove their own
1900 phony uses of YYFAIL if compatibility with Bison releases prior to
1901 2.4.2 is not necessary.
1903 ** Internationalization.
1905 Fix a regression introduced in Bison 2.4: Under some circumstances,
1906 message translations were not installed although supported by the
1909 * Changes in version 2.4.1 (2008-12-11):
1911 ** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
1912 declarations have been fixed.
1914 ** Temporary hack for adding a semicolon to the user action.
1916 Bison used to prepend a trailing semicolon at the end of the user
1917 action for reductions. This allowed actions such as
1919 exp: exp "+" exp { $$ = $1 + $3 };
1923 exp: exp "+" exp { $$ = $1 + $3; };
1925 Some grammars still depend on this "feature". Bison 2.4.1 restores
1926 the previous behavior in the case of C output (specifically, when
1927 neither %language or %skeleton or equivalent command-line options
1928 are used) to leave more time for grammars depending on the old
1929 behavior to be adjusted. Future releases of Bison will disable this
1932 ** A few minor improvements to the Bison manual.
1934 * Changes in version 2.4 (2008-11-02):
1936 ** %language is an experimental feature.
1938 We first introduced this feature in test release 2.3b as a cleaner
1939 alternative to %skeleton. Since then, we have discussed the possibility of
1940 modifying its effect on Bison's output file names. Thus, in this release,
1941 we consider %language to be an experimental feature that will likely evolve
1944 ** Forward compatibility with GNU M4 has been improved.
1946 ** Several bugs in the C++ skeleton and the experimental Java skeleton have been
1949 * Changes in version 2.3b (2008-05-27):
1951 ** The quotes around NAME that used to be required in the following directive
1954 %define NAME "VALUE"
1956 ** The directive "%pure-parser" is now deprecated in favor of:
1960 which has the same effect except that Bison is more careful to warn about
1961 unreasonable usage in the latter case.
1965 Bison can now generate an LALR(1) parser in C with a push interface. That
1966 is, instead of invoking "yyparse", which pulls tokens from "yylex", you can
1967 push one token at a time to the parser using "yypush_parse", which will
1968 return to the caller after processing each token. By default, the push
1969 interface is disabled. Either of the following directives will enable it:
1971 %define api.push_pull "push" // Just push; does not require yylex.
1972 %define api.push_pull "both" // Push and pull; requires yylex.
1974 See the new section "A Push Parser" in the Bison manual for details.
1976 The current push parsing interface is experimental and may evolve. More user
1977 feedback will help to stabilize it.
1979 ** The -g and --graph options now output graphs in Graphviz DOT format,
1980 not VCG format. Like --graph, -g now also takes an optional FILE argument
1981 and thus cannot be bundled with other short options.
1985 Bison can now generate an LALR(1) parser in Java. The skeleton is
1986 "data/lalr1.java". Consider using the new %language directive instead of
1987 %skeleton to select it.
1989 See the new section "Java Parsers" in the Bison manual for details.
1991 The current Java interface is experimental and may evolve. More user
1992 feedback will help to stabilize it.
1993 Contributed by Paolo Bonzini.
1997 This new directive specifies the programming language of the generated
1998 parser, which can be C (the default), C++, or Java. Besides the skeleton
1999 that Bison uses, the directive affects the names of the generated files if
2000 the grammar file's name ends in ".y".
2002 ** XML Automaton Report
2004 Bison can now generate an XML report of the LALR(1) automaton using the new
2005 "--xml" option. The current XML schema is experimental and may evolve. More
2006 user feedback will help to stabilize it.
2007 Contributed by Wojciech Polak.
2009 ** The grammar file may now specify the name of the parser header file using
2010 %defines. For example:
2014 ** When reporting useless rules, useless nonterminals, and unused terminals,
2015 Bison now employs the terms "useless in grammar" instead of "useless",
2016 "useless in parser" instead of "never reduced", and "unused in grammar"
2017 instead of "unused".
2019 ** Unreachable State Removal
2021 Previously, Bison sometimes generated parser tables containing unreachable
2022 states. A state can become unreachable during conflict resolution if Bison
2023 disables a shift action leading to it from a predecessor state. Bison now:
2025 1. Removes unreachable states.
2027 2. Does not report any conflicts that appeared in unreachable states.
2028 WARNING: As a result, you may need to update %expect and %expect-rr
2029 directives in existing grammar files.
2031 3. For any rule used only in such states, Bison now reports the rule as
2032 "useless in parser due to conflicts".
2034 This feature can be disabled with the following directive:
2036 %define lr.keep_unreachable_states
2038 See the %define entry in the "Bison Declaration Summary" in the Bison manual
2039 for further discussion.
2041 ** Lookahead Set Correction in the ".output" Report
2043 When instructed to generate a ".output" file including lookahead sets
2044 (using "--report=lookahead", for example), Bison now prints each reduction's
2045 lookahead set only next to the associated state's one item that (1) is
2046 associated with the same rule as the reduction and (2) has its dot at the end
2047 of its RHS. Previously, Bison also erroneously printed the lookahead set
2048 next to all of the state's other items associated with the same rule. This
2049 bug affected only the ".output" file and not the generated parser source
2052 ** --report-file=FILE is a new option to override the default ".output" file
2055 ** The "=" that used to be required in the following directives is now
2058 %file-prefix "parser"
2062 ** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}"
2064 Bison 2.3a provided a new set of directives as a more flexible alternative to
2065 the traditional Yacc prologue blocks. Those have now been consolidated into
2066 a single %code directive with an optional qualifier field, which identifies
2067 the purpose of the code and thus the location(s) where Bison should generate
2070 1. "%code {CODE}" replaces "%after-header {CODE}"
2071 2. "%code requires {CODE}" replaces "%start-header {CODE}"
2072 3. "%code provides {CODE}" replaces "%end-header {CODE}"
2073 4. "%code top {CODE}" replaces "%before-header {CODE}"
2075 See the %code entries in section "Bison Declaration Summary" in the Bison
2076 manual for a summary of the new functionality. See the new section "Prologue
2077 Alternatives" for a detailed discussion including the advantages of %code
2078 over the traditional Yacc prologues.
2080 The prologue alternatives are experimental. More user feedback will help to
2081 determine whether they should become permanent features.
2083 ** Revised warning: unset or unused midrule values
2085 Since Bison 2.2, Bison has warned about midrule values that are set but not
2086 used within any of the actions of the parent rule. For example, Bison warns
2089 exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };
2091 Now, Bison also warns about midrule values that are used but not set. For
2092 example, Bison warns about unset $$ in the midrule action in:
2094 exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };
2096 However, Bison now disables both of these warnings by default since they
2097 sometimes prove to be false alarms in existing grammars employing the Yacc
2098 constructs $0 or $-N (where N is some positive integer).
2100 To enable these warnings, specify the option "--warnings=midrule-values" or
2101 "-W", which is a synonym for "--warnings=all".
2103 ** Default %destructor or %printer with "<*>" or "<>"
2105 Bison now recognizes two separate kinds of default %destructor's and
2108 1. Place "<*>" in a %destructor/%printer symbol list to define a default
2109 %destructor/%printer for all grammar symbols for which you have formally
2110 declared semantic type tags.
2112 2. Place "<>" in a %destructor/%printer symbol list to define a default
2113 %destructor/%printer for all grammar symbols without declared semantic
2116 Bison no longer supports the "%symbol-default" notation from Bison 2.3a.
2117 "<*>" and "<>" combined achieve the same effect with one exception: Bison no
2118 longer applies any %destructor to a midrule value if that midrule value is
2119 not actually ever referenced using either $$ or $n in a semantic action.
2121 The default %destructor's and %printer's are experimental. More user
2122 feedback will help to determine whether they should become permanent
2125 See the section "Freeing Discarded Symbols" in the Bison manual for further
2128 ** %left, %right, and %nonassoc can now declare token numbers. This is required
2129 by POSIX. However, see the end of section "Operator Precedence" in the Bison
2130 manual for a caveat concerning the treatment of literal strings.
2132 ** The nonfunctional --no-parser, -n, and %no-parser options have been
2133 completely removed from Bison.
2135 * Changes in version 2.3a, 2006-09-13:
2137 ** Instead of %union, you can define and use your own union type
2138 YYSTYPE if your grammar contains at least one <type> tag.
2139 Your YYSTYPE need not be a macro; it can be a typedef.
2140 This change is for compatibility with other Yacc implementations,
2141 and is required by POSIX.
2143 ** Locations columns and lines start at 1.
2144 In accordance with the GNU Coding Standards and Emacs.
2146 ** You may now declare per-type and default %destructor's and %printer's:
2150 %union { char *string; }
2151 %token <string> STRING1
2152 %token <string> STRING2
2153 %type <string> string1
2154 %type <string> string2
2155 %union { char character; }
2156 %token <character> CHR
2157 %type <character> chr
2158 %destructor { free ($$); } %symbol-default
2159 %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
2160 %destructor { } <character>
2162 guarantees that, when the parser discards any user-defined symbol that has a
2163 semantic type tag other than "<character>", it passes its semantic value to
2164 "free". However, when the parser discards a "STRING1" or a "string1", it
2165 also prints its line number to "stdout". It performs only the second
2166 "%destructor" in this case, so it invokes "free" only once.
2168 [Although we failed to mention this here in the 2.3a release, the default
2169 %destructor's and %printer's were experimental, and they were rewritten in
2172 ** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y",
2173 "--yacc", or "%yacc"), Bison no longer generates #define statements for
2174 associating token numbers with token names. Removing the #define statements
2175 helps to sanitize the global namespace during preprocessing, but POSIX Yacc
2176 requires them. Bison still generates an enum for token names in all cases.
2178 ** Handling of traditional Yacc prologue blocks is now more consistent but
2179 potentially incompatible with previous releases of Bison.
2181 As before, you declare prologue blocks in your grammar file with the
2182 "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all
2183 prologue blocks that you've declared before the first %union. To generate
2184 the post-prologue, Bison concatenates all prologue blocks that you've
2185 declared after the first %union.
2187 Previous releases of Bison inserted the pre-prologue into both the header
2188 file and the code file in all cases except for LALR(1) parsers in C. In the
2189 latter case, Bison inserted it only into the code file. For parsers in C++,
2190 the point of insertion was before any token definitions (which associate
2191 token numbers with names). For parsers in C, the point of insertion was
2192 after the token definitions.
2194 Now, Bison never inserts the pre-prologue into the header file. In the code
2195 file, it always inserts it before the token definitions.
2197 ** Bison now provides a more flexible alternative to the traditional Yacc
2198 prologue blocks: %before-header, %start-header, %end-header, and
2201 For example, the following declaration order in the grammar file reflects the
2202 order in which Bison will output these code blocks. However, you are free to
2203 declare these code blocks in your grammar file in whatever order is most
2207 /* Bison treats this block like a pre-prologue block: it inserts it into
2208 * the code file before the contents of the header file. It does *not*
2209 * insert it into the header file. This is a good place to put
2210 * #include's that you want at the top of your code file. A common
2211 * example is '#include "system.h"'. */
2214 /* Bison inserts this block into both the header file and the code file.
2215 * In both files, the point of insertion is before any Bison-generated
2216 * token, semantic type, location type, and class definitions. This is a
2217 * good place to define %union dependencies, for example. */
2220 /* Unlike the traditional Yacc prologue blocks, the output order for the
2221 * new %*-header blocks is not affected by their declaration position
2222 * relative to any %union in the grammar file. */
2225 /* Bison inserts this block into both the header file and the code file.
2226 * In both files, the point of insertion is after the Bison-generated
2227 * definitions. This is a good place to declare or define public
2228 * functions or data structures that depend on the Bison-generated
2232 /* Bison treats this block like a post-prologue block: it inserts it into
2233 * the code file after the contents of the header file. It does *not*
2234 * insert it into the header file. This is a good place to declare or
2235 * define internal functions or data structures that depend on the
2236 * Bison-generated definitions. */
2239 If you have multiple occurrences of any one of the above declarations, Bison
2240 will concatenate the contents in declaration order.
2242 [Although we failed to mention this here in the 2.3a release, the prologue
2243 alternatives were experimental, and they were rewritten in future versions.]
2245 ** The option "--report=look-ahead" has been changed to "--report=lookahead".
2246 The old spelling still works, but is not documented and may be removed
2247 in a future release.
2249 * Changes in version 2.3, 2006-06-05:
2251 ** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING",
2252 for compatibility with LALR(1) grammars.
2254 ** It is now documented that any definition of YYSTYPE or YYLTYPE should
2255 be to a type name that does not contain parentheses or brackets.
2257 * Changes in version 2.2, 2006-05-19:
2259 ** The distribution terms for all Bison-generated parsers now permit
2260 using the parsers in nonfree programs. Previously, this permission
2261 was granted only for Bison-generated LALR(1) parsers in C.
2263 ** %name-prefix changes the namespace name in C++ outputs.
2265 ** The C++ parsers export their token_type.
2267 ** Bison now allows multiple %union declarations, and concatenates
2268 their contents together.
2270 ** New warning: unused values
2271 Right-hand side symbols whose values are not used are reported,
2272 if the symbols have destructors. For instance:
2274 exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
2278 will trigger a warning about $$ and $5 in the first rule, and $3 in
2279 the second ($1 is copied to $$ by the default rule). This example
2280 most likely contains three errors, and could be rewritten as:
2282 exp: exp "?" exp ":" exp
2283 { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
2285 { $$ = $1 ? $1 : $3; if ($1) free ($3); }
2288 However, if the original actions were really intended, memory leaks
2289 and all, the warnings can be suppressed by letting Bison believe the
2290 values are used, e.g.:
2292 exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
2293 | exp "+" exp { $$ = $1; (void) $3; }
2296 If there are midrule actions, the warning is issued if no action
2297 uses it. The following triggers no warning: $1 and $3 are used.
2299 exp: exp { push ($1); } '+' exp { push ($3); sum (); };
2301 The warning is intended to help catching lost values and memory leaks.
2302 If a value is ignored, its associated memory typically is not reclaimed.
2304 ** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
2305 Destructors are now called when user code invokes YYABORT, YYACCEPT,
2306 and YYERROR, for all objects on the stack, other than objects
2307 corresponding to the right-hand side of the current rule.
2309 ** %expect, %expect-rr
2310 Incorrect numbers of expected conflicts are now actual errors,
2311 instead of warnings.
2313 ** GLR, YACC parsers.
2314 The %parse-params are available in the destructors (and the
2315 experimental printers) as per the documentation.
2317 ** Bison now warns if it finds a stray "$" or "@" in an action.
2319 ** %require "VERSION"
2320 This specifies that the grammar file depends on features implemented
2321 in Bison version VERSION or higher.
2323 ** lalr1.cc: The token and value types are now class members.
2324 The tokens were defined as free form enums and cpp macros. YYSTYPE
2325 was defined as a free form union. They are now class members:
2326 tokens are enumerations of the "yy::parser::token" struct, and the
2327 semantic values have the "yy::parser::semantic_type" type.
2329 If you do not want or can update to this scheme, the directive
2330 '%define "global_tokens_and_yystype" "1"' triggers the global
2331 definition of tokens and YYSTYPE. This change is suitable both
2332 for previous releases of Bison, and this one.
2334 If you wish to update, then make sure older version of Bison will
2335 fail using '%require "2.2"'.
2337 ** DJGPP support added.
2339 * Changes in version 2.1, 2005-09-16:
2341 ** The C++ lalr1.cc skeleton supports %lex-param.
2343 ** Bison-generated parsers now support the translation of diagnostics like
2344 "syntax error" into languages other than English. The default
2345 language is still English. For details, please see the new
2346 Internationalization section of the Bison manual. Software
2347 distributors should also see the new PACKAGING file. Thanks to
2348 Bruno Haible for this new feature.
2350 ** Wording in the Bison-generated parsers has been changed slightly to
2351 simplify translation. In particular, the message "memory exhausted"
2352 has replaced "parser stack overflow", as the old message was not
2353 always accurate for modern Bison-generated parsers.
2355 ** Destructors are now called when the parser aborts, for all symbols left
2356 behind on the stack. Also, the start symbol is now destroyed after a
2357 successful parse. In both cases, the behavior was formerly inconsistent.
2359 ** When generating verbose diagnostics, Bison-generated parsers no longer
2360 quote the literal strings associated with tokens. For example, for
2361 a syntax error associated with '%token NUM "number"' they might
2362 print 'syntax error, unexpected number' instead of 'syntax error,
2363 unexpected "number"'.
2365 * Changes in version 2.0, 2004-12-25:
2367 ** Possibly-incompatible changes
2369 - Bison-generated parsers no longer default to using the alloca function
2370 (when available) to extend the parser stack, due to widespread
2371 problems in unchecked stack-overflow detection. You can "#define
2372 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
2373 the manual to determine safe values for YYMAXDEPTH in that case.
2375 - Error token location.
2376 During error recovery, the location of the syntax error is updated
2377 to cover the whole sequence covered by the error token: it includes
2378 the shifted symbols thrown away during the first part of the error
2379 recovery, and the lookahead rejected during the second part.
2381 - Semicolon changes:
2382 . Stray semicolons are no longer allowed at the start of a grammar.
2383 . Semicolons are now required after in-grammar declarations.
2385 - Unescaped newlines are no longer allowed in character constants or
2386 string literals. They were never portable, and GCC 3.4.0 has
2387 dropped support for them. Better diagnostics are now generated if
2388 forget a closing quote.
2390 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
2394 - GLR grammars now support locations.
2396 - New directive: %initial-action.
2397 This directive allows the user to run arbitrary code (including
2398 initializing @$) from yyparse before parsing starts.
2400 - A new directive "%expect-rr N" specifies the expected number of
2401 reduce/reduce conflicts in GLR parsers.
2403 - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d".
2404 This is a GNU extension.
2406 - The option "--report=lookahead" was changed to "--report=look-ahead".
2407 [However, this was changed back after 2.3.]
2409 - Experimental %destructor support has been added to lalr1.cc.
2411 - New configure option --disable-yacc, to disable installation of the
2412 yacc command and -ly library introduced in 1.875 for POSIX conformance.
2416 - For now, %expect-count violations are now just warnings, not errors.
2417 This is for compatibility with Bison 1.75 and earlier (when there are
2418 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
2419 are too many or too few shift/reduce conflicts). However, in future
2420 versions of Bison we plan to improve the %expect machinery so that
2421 these violations will become errors again.
2423 - Within Bison itself, numbers (e.g., goto numbers) are no longer
2424 arbitrarily limited to 16-bit counts.
2426 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
2428 * Changes in version 1.875, 2003-01-01:
2430 ** The documentation license has been upgraded to version 1.2
2431 of the GNU Free Documentation License.
2433 ** syntax error processing
2435 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
2436 locations too. This fixes bugs in error-location computation.
2439 It is now possible to reclaim the memory associated to symbols
2440 discarded during error recovery. This feature is still experimental.
2443 This new directive is preferred over YYERROR_VERBOSE.
2445 - #defining yyerror to steal internal variables is discouraged.
2446 It is not guaranteed to work forever.
2448 ** POSIX conformance
2450 - Semicolons are once again optional at the end of grammar rules.
2451 This reverts to the behavior of Bison 1.33 and earlier, and improves
2452 compatibility with Yacc.
2454 - "parse error" -> "syntax error"
2455 Bison now uniformly uses the term "syntax error"; formerly, the code
2456 and manual sometimes used the term "parse error" instead. POSIX
2457 requires "syntax error" in diagnostics, and it was thought better to
2460 - The documentation now emphasizes that yylex and yyerror must be
2461 declared before use. C99 requires this.
2463 - Bison now parses C99 lexical constructs like UCNs and
2464 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
2466 - File names are properly escaped in C output. E.g., foo\bar.y is
2467 output as "foo\\bar.y".
2469 - Yacc command and library now available
2470 The Bison distribution now installs a "yacc" command, as POSIX requires.
2471 Also, Bison now installs a small library liby.a containing
2472 implementations of Yacc-compatible yyerror and main functions.
2473 This library is normally not useful, but POSIX requires it.
2475 - Type clashes now generate warnings, not errors.
2477 - If the user does not define YYSTYPE as a macro, Bison now declares it
2478 using typedef instead of defining it as a macro.
2479 For consistency, YYLTYPE is also declared instead of defined.
2481 ** Other compatibility issues
2483 - %union directives can now have a tag before the "{", e.g., the
2484 directive "%union foo {...}" now generates the C code
2485 "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility.
2486 The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc.
2487 For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype".
2488 This is for compatibility with both Yacc and Bison 1.35.
2490 - ";" is output before the terminating "}" of an action, for
2491 compatibility with Bison 1.35.
2493 - Bison now uses a Yacc-style format for conflict reports, e.g.,
2494 "conflicts: 2 shift/reduce, 1 reduce/reduce".
2496 - "yystype" and "yyltype" are now obsolescent macros instead of being
2497 typedefs or tags; they are no longer documented and are planned to be
2498 withdrawn in a future release.
2503 Users of Bison have to decide how they handle the portability of the
2506 - "parsing stack overflow..." -> "parser stack overflow"
2507 GLR parsers now report "parser stack overflow" as per the Bison manual.
2509 ** %parse-param and %lex-param
2510 The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass
2511 additional context to yyparse and yylex. They suffer from several
2514 - a single argument only can be added,
2515 - their types are weak (void *),
2516 - this context is not passed to ancillary functions such as yyerror,
2517 - only yacc.c parsers support them.
2519 The new %parse-param/%lex-param directives provide a more precise control.
2522 %parse-param {int *nastiness}
2523 %lex-param {int *nastiness}
2524 %parse-param {int *randomness}
2526 results in the following signatures:
2528 int yylex (int *nastiness);
2529 int yyparse (int *nastiness, int *randomness);
2531 or, if both %pure-parser and %locations are used:
2533 int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness);
2534 int yyparse (int *nastiness, int *randomness);
2536 ** Bison now warns if it detects conflicting outputs to the same file,
2537 e.g., it generates a warning for "bison -d -o foo.h foo.y" since
2538 that command outputs both code and header to foo.h.
2540 ** #line in output files
2541 - --no-line works properly.
2543 ** Bison can no longer be built by a K&R C compiler; it requires C89 or
2544 later to be built. This change originally took place a few versions
2545 ago, but nobody noticed until we recently asked someone to try
2546 building Bison with a K&R C compiler.
2548 * Changes in version 1.75, 2002-10-14:
2550 ** Bison should now work on 64-bit hosts.
2552 ** Indonesian translation thanks to Tedi Heriyanto.
2555 Fix spurious parse errors.
2558 Some people redefine yyerror to steal yyparse' private variables.
2559 Reenable this trick until an official feature replaces it.
2562 In agreement with POSIX and with other Yaccs, leaving a default
2563 action is valid when $$ is untyped, and $1 typed:
2567 but the converse remains an error:
2571 ** Values of midrule actions
2574 foo: { ... } { $$ = $1; } ...
2576 was incorrectly rejected: $1 is defined in the second midrule
2577 action, and is equal to the $$ of the first midrule action.
2579 * Changes in version 1.50, 2002-10-04:
2584 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
2585 almost any context-free grammar, ambiguous or not. The new declarations
2586 %dprec and %merge on grammar rules allow parse-time resolution of
2587 ambiguities. Contributed by Paul Hilfinger.
2589 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
2590 like the Alpha, so please stick to 32-bit hosts for now.
2593 When not in Yacc compatibility mode, when the output file was not
2594 specified, running "bison foo/bar.y" created "foo/bar.c". It
2595 now creates "bar.c".
2598 The undefined token was systematically mapped to 2 which prevented
2599 the use of 2 by the user. This is no longer the case.
2601 ** Unknown token numbers
2602 If yylex returned an out of range value, yyparse could die. This is
2606 According to POSIX, the error token must be 256.
2607 Bison extends this requirement by making it a preference: *if* the
2608 user specified that one of her tokens is numbered 256, then error
2609 will be mapped onto another number.
2611 ** Verbose error messages
2612 They no longer report "..., expecting error or..." for states where
2613 error recovery is possible.
2616 Defaults to "$end" instead of "$".
2618 ** Error recovery now conforms to documentation and to POSIX
2619 When a Bison-generated parser encounters a syntax error, it now pops
2620 the stack until it finds a state that allows shifting the error
2621 token. Formerly, it popped the stack until it found a state that
2622 allowed some non-error action other than a default reduction on the
2623 error token. The new behavior has long been the documented behavior,
2624 and has long been required by POSIX. For more details, please see
2625 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
2626 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
2629 Popped tokens and nonterminals are now reported.
2632 Larger grammars are now supported (larger token numbers, larger grammar
2633 size (= sum of the LHS and RHS lengths), larger LALR tables).
2634 Formerly, many of these numbers ran afoul of 16-bit limits;
2635 now these limits are 32 bits on most hosts.
2637 ** Explicit initial rule
2638 Bison used to play hacks with the initial rule, which the user does
2639 not write. It is now explicit, and visible in the reports and
2643 Before, Bison reported the useless rules, but, although not used,
2644 included them in the parsers. They are now actually removed.
2646 ** Useless rules, useless nonterminals
2647 They are now reported, as a warning, with their locations.
2649 ** Rules never reduced
2650 Rules that can never be reduced because of conflicts are now
2653 ** Incorrect "Token not used"
2654 On a grammar such as
2656 %token useless useful
2658 exp: '0' %prec useful;
2660 where a token was used to set the precedence of the last rule,
2661 bison reported both "useful" and "useless" as useless tokens.
2663 ** Revert the C++ namespace changes introduced in 1.31
2664 as they caused too many portability hassles.
2666 ** Default locations
2667 By an accident of design, the default computation of @$ was
2668 performed after another default computation was performed: @$ = @1.
2669 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
2670 the computation of @$.
2672 ** Token end-of-file
2673 The token end of file may be specified by the user, in which case,
2674 the user symbol is used in the reports, the graphs, and the verbose
2675 error messages instead of "$end", which remains being the default.
2679 %token MYEOF 0 "end of file"
2682 This old option, which has been broken for ages, is removed.
2685 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
2686 Croatian, thanks to Denis Lackovic.
2688 ** Incorrect token definitions
2691 bison used to output
2694 ** Token definitions as enums
2695 Tokens are output both as the traditional #define's, and, provided
2696 the compiler supports ANSI C or is a C++ compiler, as enums.
2697 This lets debuggers display names instead of integers.
2700 In addition to --verbose, bison supports --report=THINGS, which
2701 produces additional information:
2703 complete the core item sets with their closure
2704 - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back]
2705 explicitly associate lookahead tokens to items
2707 describe shift/reduce conflicts solving.
2708 Bison used to systematically output this information on top of
2709 the report. Solved conflicts are now attached to their states.
2712 Previous versions don't complain when there is a type clash on
2713 the default action if the rule has a midrule action, such as in:
2721 ** GNU M4 is now required when using Bison.
2723 * Changes in version 1.35, 2002-03-25:
2726 Some projects use Bison's C parser with C++ compilers, and define
2727 YYSTYPE as a class. The recent adjustment of C parsers for data
2728 alignment and 64 bit architectures made this impossible.
2730 Because for the time being no real solution for C++ parser
2731 generation exists, kludges were implemented in the parser to
2732 maintain this use. In the future, when Bison has C++ parsers, this
2733 kludge will be disabled.
2735 This kludge also addresses some C++ problems when the stack was
2738 * Changes in version 1.34, 2002-03-12:
2740 ** File name clashes are detected
2741 $ bison foo.y -d -o foo.x
2742 fatal error: header and parser would both be named "foo.x"
2744 ** A missing ";" at the end of a rule triggers a warning
2745 In accordance with POSIX, and in agreement with other
2746 Yacc implementations, Bison will mandate this semicolon in the near
2747 future. This eases the implementation of a Bison parser of Bison
2748 grammars by making this grammar LALR(1) instead of LR(2). To
2749 facilitate the transition, this release introduces a warning.
2751 ** Revert the C++ namespace changes introduced in 1.31, as they caused too
2752 many portability hassles.
2754 ** DJGPP support added.
2756 ** Fix test suite portability problems.
2758 * Changes in version 1.33, 2002-02-07:
2761 Groff could not be compiled for the definition of size_t was lacking
2762 under some conditions.
2767 * Changes in version 1.32, 2002-01-23:
2769 ** Fix Yacc output file names
2771 ** Portability fixes
2773 ** Italian, Dutch translations
2775 * Changes in version 1.31, 2002-01-14:
2779 ** GNU Gettext and %expect
2780 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
2781 Bison dies on incorrect %expectations, we fear there will be
2782 too many bug reports for Gettext, so _for the time being_, %expect
2783 does not trigger an error when the input file is named "plural.y".
2785 ** Use of alloca in parsers
2786 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
2787 malloc exclusively. Since 1.29, but was not NEWS'ed.
2789 alloca is used only when compiled with GCC, to avoid portability
2792 ** yyparse now returns 2 if memory is exhausted; formerly it dumped core.
2794 ** When the generated parser lacks debugging code, YYDEBUG is now 0
2795 (as POSIX requires) instead of being undefined.
2798 Bison has always permitted actions such as { $$ = $1 }: it adds the
2799 ending semicolon. Now if in Yacc compatibility mode, the semicolon
2800 is no longer output: one has to write { $$ = $1; }.
2802 ** Better C++ compliance
2803 The output parsers try to respect C++ namespaces.
2804 [This turned out to be a failed experiment, and it was reverted later.]
2807 Fixed bugs when reporting useless nonterminals.
2810 The parsers work properly on 64 bit hosts.
2813 Some calls to strerror resulted in scrambled or missing error messages.
2816 When the number of shift/reduce conflicts is correct, don't issue
2819 ** The verbose report includes the rule line numbers.
2821 ** Rule line numbers are fixed in traces.
2823 ** Swedish translation
2826 Verbose parse error messages from the parsers are better looking.
2827 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
2828 Now: parse error: unexpected '/', expecting "number" or '-' or '('
2830 ** Fixed parser memory leaks.
2831 When the generated parser was using malloc to extend its stacks, the
2832 previous allocations were not freed.
2834 ** Fixed verbose output file.
2835 Some newlines were missing.
2836 Some conflicts in state descriptions were missing.
2838 ** Fixed conflict report.
2839 Option -v was needed to get the result.
2843 Mismatches are errors, not warnings.
2845 ** Fixed incorrect processing of some invalid input.
2847 ** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
2849 ** Fixed some typos in the documentation.
2851 ** %token MY_EOF 0 is supported.
2852 Before, MY_EOF was silently renumbered as 257.
2854 ** doc/refcard.tex is updated.
2856 ** %output, %file-prefix, %name-prefix.
2860 New, aliasing "--output-file".
2862 * Changes in version 1.30, 2001-10-26:
2864 ** "--defines" and "--graph" have now an optional argument which is the
2865 output file name. "-d" and "-g" do not change; they do not take any
2868 ** "%source_extension" and "%header_extension" are removed, failed
2871 ** Portability fixes.
2873 * Changes in version 1.29, 2001-09-07:
2875 ** The output file does not define const, as this caused problems when used
2876 with common autoconfiguration schemes. If you still use ancient compilers
2877 that lack const, compile with the equivalent of the C compiler option
2878 "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this.
2880 ** Added "-g" and "--graph".
2882 ** The Bison manual is now distributed under the terms of the GNU FDL.
2884 ** The input and the output files has automatically a similar extension.
2886 ** Russian translation added.
2888 ** NLS support updated; should hopefully be less troublesome.
2890 ** Added the old Bison reference card.
2892 ** Added "--locations" and "%locations".
2894 ** Added "-S" and "--skeleton".
2896 ** "%raw", "-r", "--raw" is disabled.
2898 ** Special characters are escaped when output. This solves the problems
2899 of the #line lines with path names including backslashes.
2902 "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose",
2903 "%debug", "%source_extension" and "%header_extension".
2906 Automatic location tracking.
2908 * Changes in version 1.28, 1999-07-06:
2910 ** Should compile better now with K&R compilers.
2914 ** Fixed a problem with escaping the double quote character.
2916 ** There is now a FAQ.
2918 * Changes in version 1.27:
2920 ** The make rule which prevented bison.simple from being created on
2921 some systems has been fixed.
2923 * Changes in version 1.26:
2925 ** Bison now uses Automake.
2927 ** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
2929 ** Token numbers now start at 257 as previously documented, not 258.
2931 ** Bison honors the TMPDIR environment variable.
2933 ** A couple of buffer overruns have been fixed.
2935 ** Problems when closing files should now be reported.
2937 ** Generated parsers should now work even on operating systems which do
2938 not provide alloca().
2940 * Changes in version 1.25, 1995-10-16:
2942 ** Errors in the input grammar are not fatal; Bison keeps reading
2943 the grammar file, and reports all the errors found in it.
2945 ** Tokens can now be specified as multiple-character strings: for
2946 example, you could use "<=" for a token which looks like <=, instead
2947 of choosing a name like LESSEQ.
2949 ** The %token_table declaration says to write a table of tokens (names
2950 and numbers) into the parser file. The yylex function can use this
2951 table to recognize multiple-character string tokens, or for other
2954 ** The %no_lines declaration says not to generate any #line preprocessor
2955 directives in the parser file.
2957 ** The %raw declaration says to use internal Bison token numbers, not
2958 Yacc-compatible token numbers, when token names are defined as macros.
2960 ** The --no-parser option produces the parser tables without including
2961 the parser engine; a project can now use its own parser engine.
2962 The actions go into a separate file called NAME.act, in the form of
2963 a switch statement body.
2965 * Changes in version 1.23:
2967 The user can define YYPARSE_PARAM as the name of an argument to be
2968 passed into yyparse. The argument should have type void *. It should
2969 actually point to an object. Grammar actions can access the variable
2970 by casting it to the proper pointer type.
2972 Line numbers in output file corrected.
2974 * Changes in version 1.22:
2976 --help option added.
2978 * Changes in version 1.20:
2980 Output file does not redefine const for C++.
2984 Copyright (C) 1995-2015, 2018 Free Software Foundation, Inc.
2986 This file is part of Bison, the GNU Parser Generator.
2988 This program is free software: you can redistribute it and/or modify
2989 it under the terms of the GNU General Public License as published by
2990 the Free Software Foundation, either version 3 of the License, or
2991 (at your option) any later version.
2993 This program is distributed in the hope that it will be useful,
2994 but WITHOUT ANY WARRANTY; without even the implied warranty of
2995 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2996 GNU General Public License for more details.
2998 You should have received a copy of the GNU General Public License
2999 along with this program. If not, see <http://www.gnu.org/licenses/>.
3001 LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr
3002 LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr
3003 LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar
3004 LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr
3005 LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks
3006 LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG
3007 LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush
3008 LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly
3009 LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's
3010 LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de
3011 LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto
3012 LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs
3013 LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF
3014 LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY
3015 LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL
3016 LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh
3017 LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf
3018 LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp Wother nterm arg init
3019 LocalWords: TOK calc yyo fval Wconflicts parsers yystackp yyval yynerrs
3020 LocalWords: Théophile Ranquet Santet fno fnone stype associativity Tolmer
3021 LocalWords: Wprecedence Rassoul Wempty Paolo Bonzini parser's Michiel loc
3022 LocalWords: redeclaration sval fcaret reentrant XSLT xsl Wmaybe yyvsp Tedi
3023 LocalWords: pragmas noreturn untyped Rozenman unexpanded Wojciech Polak
3024 LocalWords: Alexandre MERCHANTABILITY yytype emplace ptr automove lvalues
3025 LocalWords: nonterminal yy args Pragma dereference yyformat rhs docdir
3026 LocalWords: Redeclarations rpcalc Autoconf YFLAGS Makefiles outout PROG
3027 LocalWords: Heimbigner