]> git.saurik.com Git - bison.git/blob - NEWS
gnulib: update
[bison.git] / NEWS
1 GNU Bison NEWS
2
3 * Noteworthy changes in release ?.? (????-??-??) [?]
4
5 ** WARNING: Future backward-incompatibilities!
6
7 Bison will stop adding a semicolon at the end of the actions (as announced
8 in the release 2.5):
9
10 foo.y:2.25: warning: a ';' might be needed at the end of action code
11 exp: "number" { $$ = $1 }
12 ^
13 foo.y:2.25: future versions of Bison will not add the ';'
14
15 Like other GNU packages, Bison will start using some of the C99 features
16 for its own code, especially the definition of variables after statements.
17 The generated C parsers still aim at C90.
18
19 ** Backward incompatible changes
20
21 *** Obsolete features
22
23 Support for YYFAIL is removed (deprecated in Bison 2.4.2): use YYERROR.
24
25 Support for yystype and yyltype is removed (deprecated in Bison 1.875):
26 use YYSTYPE and YYLTYPE.
27
28 Support for YYLEX_PARAM and YYPARSE_PARAM is removed (deprecated in Bison
29 1.875): use %lex-param, %parse-param, or %param.
30
31 ** Bug fixes
32
33 *** The epilogue is no longer affected by internal #defines (glr.c)
34
35 The glr.c skeleton uses defines such as #define yylval (yystackp->yyval) in
36 generated code. These weren't properly undefined before the inclusion of
37 the user epilogue, so functions such as the following were butchered by the
38 preprocessor expansion:
39
40 int yylex (YYSTYPE *yylval);
41
42 This is has been fixed: yylval, yynerrs, yychar, and yylloc are now valid
43 identifiers for user-provided variables.
44
45 *** stdio.h is no longer needed when locations are enabled (yacc.c)
46
47 Changes in Bison 2.7 introduced a dependency on FILE and fprintf when
48 locations are enabled. This is fixed.
49
50 ** Diagnostics reported by Bison
51
52 Most of these features were contributed by Théophile Ranquet and Victor
53 Santet.
54
55 *** Carets
56
57 Version 2.7 introduced caret errors, for a prettier output. These are now
58 activated by default. The old format can still be used by invoking Bison
59 with -fno-caret (or -fnone).
60
61 Some error messages that reproduced excerpts of the grammar are now using
62 the caret information only. For instance on:
63
64 %%
65 exp: 'a' | 'a';
66
67 Bison 2.7 reports:
68
69 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
70 in.y:2.12-14: warning: rule useless in parser due to conflicts: exp: 'a' [-Wother]
71
72 Now bison reports:
73
74 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
75 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
76 exp: 'a' | 'a';
77 ^^^
78
79 and "bison -fno-caret" reports:
80
81 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
82 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
83
84 *** Enhancements of the -Werror option
85
86 The -Werror=CATEGORY option is now recognized, and will treat specified
87 warnings as errors. The warnings need not have been explicitly activated
88 using the -W option, this is similar to what GCC 4.7 does.
89
90 For example, given the following command line, Bison will treat both
91 warnings related to POSIX Yacc incompatibilities and S/R conflicts as
92 errors (and only those):
93
94 $ bison -Werror=yacc,error=conflicts-sr input.y
95
96 If no categories are specified, -Werror will make all active warnings into
97 errors. For example, the following line does the same the previous example:
98
99 $ bison -Werror -Wnone -Wyacc -Wconflicts-sr input.y
100
101 (By default -Wconflicts-sr,conflicts-rr,deprecated,other is enabled.)
102
103 Note that the categories in this -Werror option may not be prefixed with
104 "no-". However, -Wno-error[=CATEGORY] is valid.
105
106 Note that -y enables -Werror=yacc. Therefore it is now possible to require
107 Yacc-like behavior (e.g., always generate y.tab.c), but to report
108 incompatibilities as warnings: "-y -Wno-error=yacc".
109
110 *** The display of warnings is now richer
111
112 The option that controls a given warning is now displayed:
113
114 foo.y:4.6: warning: type clash on default action: <foo> != <bar> [-Wother]
115
116 In the case of warnings treated as errors, the prefix is changed from
117 "warning: " to "error: ", and the suffix is displayed, in a manner similar
118 to GCC, as [-Werror=CATEGORY].
119
120 For instance, where the previous version of Bison would report (and exit
121 with failure):
122
123 bison: warnings being treated as errors
124 input.y:1.1: warning: stray ',' treated as white space
125
126 it now reports:
127
128 input.y:1.1: error: stray ',' treated as white space [-Werror=other]
129
130 *** Deprecated constructs
131
132 The new 'deprecated' warning category flags obsolete constructs whose
133 support will be discontinued. It is enabled by default. These warnings
134 used to be reported as 'other' warnings.
135
136 *** Useless semantic types
137
138 Bison now warns about useless (uninhabited) semantic types. Since
139 semantic types are not declared to Bison (they are defined in the opaque
140 %union structure), it is %printer/%destructor directives about useless
141 types that trigger the warning:
142
143 %token <type1> term
144 %type <type2> nterm
145 %printer {} <type1> <type3>
146 %destructor {} <type2> <type4>
147 %%
148 nterm: term { $$ = $1; };
149
150 3.28-34: warning: type <type3> is used, but is not associated to any symbol
151 4.28-34: warning: type <type4> is used, but is not associated to any symbol
152
153 *** Undefined but unused symbols
154
155 Bison used to raise an error for undefined symbols that are not used in
156 the grammar. This is now only a warning.
157
158 %printer {} symbol1
159 %destructor {} symbol2
160 %type <type> symbol3
161 %%
162 exp: "a";
163
164 *** Useless destructors or printers
165
166 Bison now warns about useless destructors or printers. In the following
167 example, the printer for <type1>, and the destructor for <type2> are
168 useless: all symbols of <type1> (token1) already have a printer, and all
169 symbols of type <type2> (token2) already have a destructor.
170
171 %token <type1> token1
172 <type2> token2
173 <type3> token3
174 <type4> token4
175 %printer {} token1 <type1> <type3>
176 %destructor {} token2 <type2> <type4>
177
178 *** Conflicts
179
180 The warnings and error messages about shift/reduce and reduce/reduce
181 conflicts have been normalized. For instance on the following foo.y file:
182
183 %glr-parser
184 %%
185 exp: exp '+' exp | '0' | '0';
186
187 compare the previous version of bison:
188
189 $ bison foo.y
190 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
191 $ bison -Werror foo.y
192 bison: warnings being treated as errors
193 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
194
195 with the new behavior:
196
197 $ bison foo.y
198 foo.y: warning: 1 shift/reduce conflict [-Wconflicts-sr]
199 foo.y: warning: 2 reduce/reduce conflicts [-Wconflicts-rr]
200 $ bison -Werror foo.y
201 foo.y: error: 1 shift/reduce conflict [-Werror=conflicts-sr]
202 foo.y: error: 2 reduce/reduce conflicts [-Werror=conflicts-rr]
203
204 When %expect or %expect-rr is used, such as with bar.y:
205
206 %expect 0
207 %glr-parser
208 %%
209 exp: exp '+' exp | '0' | '0';
210
211 Former behavior:
212
213 $ bison bar.y
214 bar.y: conflicts: 1 shift/reduce, 2 reduce/reduce
215 bar.y: expected 0 shift/reduce conflicts
216 bar.y: expected 0 reduce/reduce conflicts
217
218 New one:
219
220 $ bison bar.y
221 bar.y: error: shift/reduce conflicts: 1 found, 0 expected
222 bar.y: error: reduce/reduce conflicts: 2 found, 0 expected
223
224 ** Additional yylex/yyparse arguments
225
226 The new directive %param declares additional arguments to both yylex and
227 yyparse. The %lex-param, %parse-param, and %param directives support one
228 or more arguments. Instead of
229
230 %lex-param {arg1_type *arg1}
231 %lex-param {arg2_type *arg2}
232 %parse-param {arg1_type *arg1}
233 %parse-param {arg2_type *arg2}
234
235 one may now declare
236
237 %param {arg1_type *arg1} {arg2_type *arg2}
238
239 ** Variable api.token.prefix
240
241 The variable api.token.prefix changes the way tokens are identified in
242 the generated files. This is especially useful to avoid collisions
243 with identifiers in the target language. For instance
244
245 %token FILE for ERROR
246 %define api.token.prefix "TOK_"
247 %%
248 start: FILE for ERROR;
249
250 will generate the definition of the symbols TOK_FILE, TOK_for, and
251 TOK_ERROR in the generated sources. In particular, the scanner must
252 use these prefixed token names, although the grammar itself still
253 uses the short names (as in the sample rule given above).
254
255 ** Variable parse.error
256
257 This variable controls the verbosity of error messages. The use of the
258 %error-verbose directive is deprecated in favor of "%define parse.error
259 verbose".
260
261 ** Renamed %define variables
262
263 The following variables have been renamed for consistency. Backward
264 compatibility is ensured, but upgrading is recommended.
265
266 lr.default-reductions -> lr.default-reduction
267 lr.keep-unreachable-states -> lr.keep-unreachable-state
268 namespace -> api.namespace
269 stype -> api.value.type
270
271 ** Semantic predicates
272
273 Contributed by Paul Hilfinger.
274
275 The new, experimental, semantic-predicate feature allows actions of the
276 form "%?{ BOOLEAN-EXPRESSION }", which cause syntax errors (as for
277 YYERROR) if the expression evaluates to 0, and are evaluated immediately
278 in GLR parsers, rather than being deferred. The result is that they allow
279 the programmer to prune possible parses based on the values of run-time
280 expressions.
281
282 ** The directive %expect-rr is now an error in non GLR mode
283
284 It used to be an error only if used in non GLR mode, _and_ if there are
285 reduce/reduce conflicts.
286
287 ** Token numbering has changed to preserve the user-defined order
288
289 When declaring %token A B, the numbering for A is inferior to B. Up to now,
290 when declaring associativity at the same time, with %left (or %right,
291 %precedence, %nonassoc), B was inferior to A.
292
293 ** Useless precedence and associativity
294
295 Contributed by Valentin Tolmer.
296
297 When developing and maintaining a grammar, useless associativity and
298 precedence directives are common. They can be a nuisance: new ambiguities
299 arising are sometimes masked because their conflicts are resolved due to
300 the extra precedence or associativity information. Furthermore, it can
301 hinder the comprehension of a new grammar: one will wonder about the role
302 of a precedence, where in fact it is useless. The following changes aim
303 at detecting and reporting these extra directives.
304
305 *** Precedence warning category
306
307 A new category of warning, -Wprecedence, was introduced. It flags the
308 useless precedence and associativity directives.
309
310 *** Useless associativity
311
312 Bison now warns about symbols with a declared associativity that is never
313 used to resolve conflicts. In that case, using %precedence is sufficient;
314 the parsing tables will remain unchanged. Solving these warnings may raise
315 useless precedence warnings, as the symbols no longer have associativity.
316 For example:
317
318 %left '+'
319 %left '*'
320 %%
321 exp:
322 "number"
323 | exp '+' "number"
324 | exp '*' exp
325 ;
326
327 will produce a
328
329 warning: useless associativity for '+', use %precedence [-Wprecedence]
330 %left '+'
331 ^^^
332
333 *** Useless precedence
334
335 Bison now warns about symbols with a declared precedence and no declared
336 associativity (i.e., declared with %precedence), and whose precedence is
337 never used. In that case, the symbol can be safely declared with %token
338 instead, without modifying the parsing tables. For example:
339
340 %precedence '='
341 %%
342 exp: "var" '=' "number";
343
344 will produce a
345
346 warning: useless precedence for '=' [-Wprecedence]
347 %precedence '='
348 ^^^
349
350 *** Useless precedence and associativity
351
352 In case of both useless precedence and associativity, the issue is flagged
353 as follows:
354
355 %nonassoc '='
356 %%
357 exp: "var" '=' "number";
358
359 The warning is:
360
361 warning: useless precedence and associativity for '=' [-Wprecedence]
362 %nonassoc '='
363 ^^^
364
365 ** Empty rules
366
367 With help from Joel E. Denny and Gabriel Rassoul.
368
369 Empty rules (i.e., with an empty right-hand side) can now be explicitly
370 marked by the new %empty directive. Using %empty on a non-empty rule is
371 an error. The new -Wempty-rule warning reports empty rules without
372 %empty. On the following grammar:
373
374 %%
375 s: a b c;
376 a: ;
377 b: %empty;
378 c: 'a' %empty;
379
380 bison reports:
381
382 3.4-5: warning: empty rule without %empty [-Wempty-rule]
383 a: {}
384 ^^
385 5.8-13: error: %empty on non-empty rule
386 c: 'a' %empty {};
387 ^^^^^^
388
389 ** Java skeleton improvements
390
391 Contributed by Paolo Bonzini.
392
393 The constants for token names were moved to the Lexer interface. Also, it
394 is possible to add code to the parser's constructors using "%code init"
395 and "%define init_throws".
396
397 ** C++ skeletons improvements
398
399 *** The parser header is no longer mandatory (lalr1.cc, glr.cc)
400
401 Using %defines is now optional. Without it, the needed support classes
402 are defined in the generated parser, instead of additional files (such as
403 location.hh, position.hh and stack.hh).
404
405 *** Locations are no longer mandatory (lalr1.cc, glr.cc)
406
407 Both lalr1.cc and glr.cc no longer require %location.
408
409 *** syntax_error exception (lalr1.cc)
410
411 The C++ parser features a syntax_error exception, which can be
412 thrown from the scanner or from user rules to raise syntax errors.
413 This facilitates reporting errors caught in sub-functions (e.g.,
414 rejecting too large integral literals from a conversion function
415 used by the scanner, or rejecting invalid combinations from a
416 factory invoked by the user actions).
417
418 *** %define api.value.type variant
419
420 This is based on a submission from Michiel De Wilde. With help
421 from Théophile Ranquet.
422
423 In this mode, complex C++ objects can be used as semantic values. For
424 instance:
425
426 %token <::std::string> TEXT;
427 %token <int> NUMBER;
428 %token SEMICOLON ";"
429 %type <::std::string> item;
430 %type <::std::list<std::string>> list;
431 %%
432 result:
433 list { std::cout << $1 << std::endl; }
434 ;
435
436 list:
437 %empty { /* Generates an empty string list. */ }
438 | list item ";" { std::swap ($$, $1); $$.push_back ($2); }
439 ;
440
441 item:
442 TEXT { std::swap ($$, $1); }
443 | NUMBER { $$ = string_cast ($1); }
444 ;
445
446 *** %define api.token.constructor
447
448 When variants are enabled, Bison can generate functions to build the
449 tokens. This guarantees that the token type (e.g., NUMBER) is consistent
450 with the semantic value (e.g., int):
451
452 parser::symbol_type yylex ()
453 {
454 parser::location_type loc = ...;
455 ...
456 return parser::make_TEXT ("Hello, world!", loc);
457 ...
458 return parser::make_NUMBER (42, loc);
459 ...
460 return parser::make_SEMICOLON (loc);
461 ...
462 }
463
464 * Noteworthy changes in release 2.7 (2012-12-12) [stable]
465
466 ** Bug fixes
467
468 Warnings about uninitialized yylloc in yyparse have been fixed.
469
470 Restored C90 compliance (yet no report was ever made).
471
472 ** Diagnostics are improved
473
474 Contributed by Théophile Ranquet.
475
476 *** Changes in the format of error messages
477
478 This used to be the format of many error reports:
479
480 input.y:2.7-12: %type redeclaration for exp
481 input.y:1.7-12: previous declaration
482
483 It is now:
484
485 input.y:2.7-12: error: %type redeclaration for exp
486 input.y:1.7-12: previous declaration
487
488 *** New format for error reports: carets
489
490 Caret errors have been added to Bison:
491
492 input.y:2.7-12: error: %type redeclaration for exp
493 %type <sval> exp
494 ^^^^^^
495 input.y:1.7-12: previous declaration
496 %type <ival> exp
497 ^^^^^^
498
499 or
500
501 input.y:3.20-23: error: ambiguous reference: '$exp'
502 exp: exp '+' exp { $exp = $1 + $3; };
503 ^^^^
504 input.y:3.1-3: refers to: $exp at $$
505 exp: exp '+' exp { $exp = $1 + $3; };
506 ^^^
507 input.y:3.6-8: refers to: $exp at $1
508 exp: exp '+' exp { $exp = $1 + $3; };
509 ^^^
510 input.y:3.14-16: refers to: $exp at $3
511 exp: exp '+' exp { $exp = $1 + $3; };
512 ^^^
513
514 The default behavior for now is still not to display these unless
515 explicitly asked with -fcaret (or -fall). However, in a later release, it
516 will be made the default behavior (but may still be deactivated with
517 -fno-caret).
518
519 ** New value for %define variable: api.pure full
520
521 The %define variable api.pure requests a pure (reentrant) parser. However,
522 for historical reasons, using it in a location-tracking Yacc parser
523 resulted in a yyerror function that did not take a location as a
524 parameter. With this new value, the user may request a better pure parser,
525 where yyerror does take a location as a parameter (in location-tracking
526 parsers).
527
528 The use of "%define api.pure true" is deprecated in favor of this new
529 "%define api.pure full".
530
531 ** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java)
532
533 The %define variable api.location.type defines the name of the type to use
534 for locations. When defined, Bison no longer generates the position.hh
535 and location.hh files, nor does the parser will include them: the user is
536 then responsible to define her type.
537
538 This can be used in programs with several parsers to factor their location
539 and position files: let one of them generate them, and the others just use
540 them.
541
542 This feature was actually introduced, but not documented, in Bison 2.5,
543 under the name "location_type" (which is maintained for backward
544 compatibility).
545
546 For consistency, lalr1.java's %define variables location_type and
547 position_type are deprecated in favor of api.location.type and
548 api.position.type.
549
550 ** Exception safety (lalr1.cc)
551
552 The parse function now catches exceptions, uses the %destructors to
553 release memory (the lookahead symbol and the symbols pushed on the stack)
554 before re-throwing the exception.
555
556 This feature is somewhat experimental. User feedback would be
557 appreciated.
558
559 ** Graph improvements in DOT and XSLT
560
561 Contributed by Théophile Ranquet.
562
563 The graphical presentation of the states is more readable: their shape is
564 now rectangular, the state number is clearly displayed, and the items are
565 numbered and left-justified.
566
567 The reductions are now explicitly represented as transitions to other
568 diamond shaped nodes.
569
570 These changes are present in both --graph output and xml2dot.xsl XSLT
571 processing, with minor (documented) differences.
572
573 ** %language is no longer an experimental feature.
574
575 The introduction of this feature, in 2.4, was four years ago. The
576 --language option and the %language directive are no longer experimental.
577
578 ** Documentation
579
580 The sections about shift/reduce and reduce/reduce conflicts resolution
581 have been fixed and extended.
582
583 Although introduced more than four years ago, XML and Graphviz reports
584 were not properly documented.
585
586 The translation of mid-rule actions is now described.
587
588 * Noteworthy changes in release 2.6.5 (2012-11-07) [stable]
589
590 We consider compiler warnings about Bison generated parsers to be bugs.
591 Rather than working around them in your own project, please consider
592 reporting them to us.
593
594 ** Bug fixes
595
596 Warnings about uninitialized yylval and/or yylloc for push parsers with a
597 pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to
598 3.2.
599
600 Other issues in the test suite have been addressed.
601
602 Null characters are correctly displayed in error messages.
603
604 When possible, yylloc is correctly initialized before calling yylex. It
605 is no longer necessary to initialize it in the %initial-action.
606
607 * Noteworthy changes in release 2.6.4 (2012-10-23) [stable]
608
609 Bison 2.6.3's --version was incorrect. This release fixes this issue.
610
611 * Noteworthy changes in release 2.6.3 (2012-10-22) [stable]
612
613 ** Bug fixes
614
615 Bugs and portability issues in the test suite have been fixed.
616
617 Some errors in translations have been addressed, and --help now directs
618 users to the appropriate place to report them.
619
620 Stray Info files shipped by accident are removed.
621
622 Incorrect definitions of YY_, issued by yacc.c when no parser header is
623 generated, are removed.
624
625 All the generated headers are self-contained.
626
627 ** Header guards (yacc.c, glr.c, glr.cc)
628
629 In order to avoid collisions, the header guards are now
630 YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>.
631 For instance the header generated from
632
633 %define api.prefix "calc"
634 %defines "lib/parse.h"
635
636 will use YY_CALC_LIB_PARSE_H_INCLUDED as guard.
637
638 ** Fix compiler warnings in the generated parser (yacc.c, glr.c)
639
640 The compilation of pure parsers (%define api.pure) can trigger GCC
641 warnings such as:
642
643 input.c: In function 'yyparse':
644 input.c:1503:12: warning: 'yylval' may be used uninitialized in this
645 function [-Wmaybe-uninitialized]
646 *++yyvsp = yylval;
647 ^
648
649 This is now fixed; pragmas to avoid these warnings are no longer needed.
650
651 Warnings from clang ("equality comparison with extraneous parentheses" and
652 "function declared 'noreturn' should not return") have also been
653 addressed.
654
655 * Noteworthy changes in release 2.6.2 (2012-08-03) [stable]
656
657 ** Bug fixes
658
659 Buffer overruns, complaints from Flex, and portability issues in the test
660 suite have been fixed.
661
662 ** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc)
663
664 Trailing end-of-lines in %parse-param or %lex-param would result in
665 invalid C++. This is fixed.
666
667 ** Spurious spaces and end-of-lines
668
669 The generated files no longer end (nor start) with empty lines.
670
671 * Noteworthy changes in release 2.6.1 (2012-07-30) [stable]
672
673 Bison no longer executes user-specified M4 code when processing a grammar.
674
675 ** Future Changes
676
677 In addition to the removal of the features announced in Bison 2.6, the
678 next major release will remove the "Temporary hack for adding a semicolon
679 to the user action", as announced in the release 2.5. Instead of:
680
681 exp: exp "+" exp { $$ = $1 + $3 };
682
683 write:
684
685 exp: exp "+" exp { $$ = $1 + $3; };
686
687 ** Bug fixes
688
689 *** Type names are now properly escaped.
690
691 *** glr.cc: set_debug_level and debug_level work as expected.
692
693 *** Stray @ or $ in actions
694
695 While Bison used to warn about stray $ or @ in action rules, it did not
696 for other actions such as printers, destructors, or initial actions. It
697 now does.
698
699 ** Type names in actions
700
701 For consistency with rule actions, it is now possible to qualify $$ by a
702 type-name in destructors, printers, and initial actions. For instance:
703
704 %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>;
705
706 will display two values for each typed and untyped symbol (provided
707 that YYSTYPE has both "ival" and "fval" fields).
708
709 * Noteworthy changes in release 2.6 (2012-07-19) [stable]
710
711 ** Future changes
712
713 The next major release of Bison will drop support for the following
714 deprecated features. Please report disagreements to bug-bison@gnu.org.
715
716 *** K&R C parsers
717
718 Support for generating parsers in K&R C will be removed. Parsers
719 generated for C support ISO C90, and are tested with ISO C99 and ISO C11
720 compilers.
721
722 *** Features deprecated since Bison 1.875
723
724 The definitions of yystype and yyltype will be removed; use YYSTYPE and
725 YYLTYPE.
726
727 YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and
728 %lex-param, will no longer be supported.
729
730 Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use
731 %error-verbose.
732
733 *** The generated header will be included (yacc.c)
734
735 Instead of duplicating the content of the generated header (definition of
736 YYSTYPE, yyparse declaration etc.), the generated parser will include it,
737 as is already the case for GLR or C++ parsers. This change is deferred
738 because existing versions of ylwrap (e.g., Automake 1.12.1) do not support
739 it.
740
741 ** Generated Parser Headers
742
743 *** Guards (yacc.c, glr.c, glr.cc)
744
745 The generated headers are now guarded, as is already the case for C++
746 parsers (lalr1.cc). For instance, with --defines=foo.h:
747
748 #ifndef YY_FOO_H
749 # define YY_FOO_H
750 ...
751 #endif /* !YY_FOO_H */
752
753 *** New declarations (yacc.c, glr.c)
754
755 The generated header now declares yydebug and yyparse. Both honor
756 --name-prefix=bar_, and yield
757
758 int bar_parse (void);
759
760 rather than
761
762 #define yyparse bar_parse
763 int yyparse (void);
764
765 in order to facilitate the inclusion of several parser headers inside a
766 single compilation unit.
767
768 *** Exported symbols in C++
769
770 The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the
771 header, are removed, as they prevent the possibility of including several
772 generated headers from a single compilation unit.
773
774 *** YYLSP_NEEDED
775
776 For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no
777 longer defined.
778
779 ** New %define variable: api.prefix
780
781 Now that the generated headers are more complete and properly protected
782 against multiple inclusions, constant names, such as YYSTYPE are a
783 problem. While yyparse and others are properly renamed by %name-prefix,
784 YYSTYPE, YYDEBUG and others have never been affected by it. Because it
785 would introduce backward compatibility issues in projects not expecting
786 YYSTYPE to be renamed, instead of changing the behavior of %name-prefix,
787 it is deprecated in favor of a new %define variable: api.prefix.
788
789 The following examples compares both:
790
791 %name-prefix "bar_" | %define api.prefix "bar_"
792 %token <ival> FOO %token <ival> FOO
793 %union { int ival; } %union { int ival; }
794 %% %%
795 exp: 'a'; exp: 'a';
796
797 bison generates:
798
799 #ifndef BAR_FOO_H #ifndef BAR_FOO_H
800 # define BAR_FOO_H # define BAR_FOO_H
801
802 /* Enabling traces. */ /* Enabling traces. */
803 # ifndef YYDEBUG | # ifndef BAR_DEBUG
804 > # if defined YYDEBUG
805 > # if YYDEBUG
806 > # define BAR_DEBUG 1
807 > # else
808 > # define BAR_DEBUG 0
809 > # endif
810 > # else
811 # define YYDEBUG 0 | # define BAR_DEBUG 0
812 > # endif
813 # endif | # endif
814
815 # if YYDEBUG | # if BAR_DEBUG
816 extern int bar_debug; extern int bar_debug;
817 # endif # endif
818
819 /* Tokens. */ /* Tokens. */
820 # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE
821 # define YYTOKENTYPE | # define BAR_TOKENTYPE
822 enum yytokentype { | enum bar_tokentype {
823 FOO = 258 FOO = 258
824 }; };
825 # endif # endif
826
827 #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \
828 && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED
829 typedef union YYSTYPE | typedef union BAR_STYPE
830 { {
831 int ival; int ival;
832 } YYSTYPE; | } BAR_STYPE;
833 # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1
834 #endif #endif
835
836 extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval;
837
838 int bar_parse (void); int bar_parse (void);
839
840 #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */
841
842 * Noteworthy changes in release 2.5.1 (2012-06-05) [stable]
843
844 ** Future changes:
845
846 The next major release will drop support for generating parsers in K&R C.
847
848 ** yacc.c: YYBACKUP works as expected.
849
850 ** glr.c improvements:
851
852 *** Location support is eliminated when not requested:
853
854 GLR parsers used to include location-related code even when locations were
855 not requested, and therefore not even usable.
856
857 *** __attribute__ is preserved:
858
859 __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e.,
860 when -std is passed to GCC).
861
862 ** lalr1.java: several fixes:
863
864 The Java parser no longer throws ArrayIndexOutOfBoundsException if the
865 first token leads to a syntax error. Some minor clean ups.
866
867 ** Changes for C++:
868
869 *** C++11 compatibility:
870
871 C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L
872 or higher.
873
874 *** Header guards
875
876 The header files such as "parser.hh", "location.hh", etc. used a constant
877 name for preprocessor guards, for instance:
878
879 #ifndef BISON_LOCATION_HH
880 # define BISON_LOCATION_HH
881 ...
882 #endif // !BISON_LOCATION_HH
883
884 The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower
885 case characters are converted to upper case, and series of
886 non-alphanumerical characters are converted to an underscore.
887
888 With "bison -o lang++/parser.cc", "location.hh" would now include:
889
890 #ifndef YY_LANG_LOCATION_HH
891 # define YY_LANG_LOCATION_HH
892 ...
893 #endif // !YY_LANG_LOCATION_HH
894
895 *** C++ locations:
896
897 The position and location constructors (and their initialize methods)
898 accept new arguments for line and column. Several issues in the
899 documentation were fixed.
900
901 ** liby is no longer asking for "rpl_fprintf" on some platforms.
902
903 ** Changes in the manual:
904
905 *** %printer is documented
906
907 The "%printer" directive, supported since at least Bison 1.50, is finally
908 documented. The "mfcalc" example is extended to demonstrate it.
909
910 For consistency with the C skeletons, the C++ parsers now also support
911 "yyoutput" (as an alias to "debug_stream ()").
912
913 *** Several improvements have been made:
914
915 The layout for grammar excerpts was changed to a more compact scheme.
916 Named references are motivated. The description of the automaton
917 description file (*.output) is updated to the current format. Incorrect
918 index entries were fixed. Some other errors were fixed.
919
920 ** Building bison:
921
922 *** Conflicting prototypes with recent/modified Flex.
923
924 Fixed build problems with the current, unreleased, version of Flex, and
925 some modified versions of 2.5.35, which have modified function prototypes.
926
927 *** Warnings during the build procedure have been eliminated.
928
929 *** Several portability problems in the test suite have been fixed:
930
931 This includes warnings with some compilers, unexpected behavior of tools
932 such as diff, warning messages from the test suite itself, etc.
933
934 *** The install-pdf target works properly:
935
936 Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer
937 halts in the middle of its course.
938
939 * Changes in version 2.5 (2011-05-14):
940
941 ** Grammar symbol names can now contain non-initial dashes:
942
943 Consistently with directives (such as %error-verbose) and with
944 %define variables (e.g. push-pull), grammar symbol names may contain
945 dashes in any position except the beginning. This is a GNU
946 extension over POSIX Yacc. Thus, use of this extension is reported
947 by -Wyacc and rejected in Yacc mode (--yacc).
948
949 ** Named references:
950
951 Historically, Yacc and Bison have supported positional references
952 ($n, $$) to allow access to symbol values from inside of semantic
953 actions code.
954
955 Starting from this version, Bison can also accept named references.
956 When no ambiguity is possible, original symbol names may be used
957 as named references:
958
959 if_stmt : "if" cond_expr "then" then_stmt ';'
960 { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }
961
962 In the more common case, explicit names may be declared:
963
964 stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';'
965 { $res = mk_if_stmt($cond, $then, $else); }
966
967 Location information is also accessible using @name syntax. When
968 accessing symbol names containing dots or dashes, explicit bracketing
969 ($[sym.1]) must be used.
970
971 These features are experimental in this version. More user feedback
972 will help to stabilize them.
973 Contributed by Alex Rozenman.
974
975 ** IELR(1) and canonical LR(1):
976
977 IELR(1) is a minimal LR(1) parser table generation algorithm. That
978 is, given any context-free grammar, IELR(1) generates parser tables
979 with the full language-recognition power of canonical LR(1) but with
980 nearly the same number of parser states as LALR(1). This reduction
981 in parser states is often an order of magnitude. More importantly,
982 because canonical LR(1)'s extra parser states may contain duplicate
983 conflicts in the case of non-LR(1) grammars, the number of conflicts
984 for IELR(1) is often an order of magnitude less as well. This can
985 significantly reduce the complexity of developing of a grammar.
986
987 Bison can now generate IELR(1) and canonical LR(1) parser tables in
988 place of its traditional LALR(1) parser tables, which remain the
989 default. You can specify the type of parser tables in the grammar
990 file with these directives:
991
992 %define lr.type lalr
993 %define lr.type ielr
994 %define lr.type canonical-lr
995
996 The default-reduction optimization in the parser tables can also be
997 adjusted using "%define lr.default-reductions". For details on both
998 of these features, see the new section "Tuning LR" in the Bison
999 manual.
1000
1001 These features are experimental. More user feedback will help to
1002 stabilize them.
1003
1004 ** LAC (Lookahead Correction) for syntax error handling
1005
1006 Contributed by Joel E. Denny.
1007
1008 Canonical LR, IELR, and LALR can suffer from a couple of problems
1009 upon encountering a syntax error. First, the parser might perform
1010 additional parser stack reductions before discovering the syntax
1011 error. Such reductions can perform user semantic actions that are
1012 unexpected because they are based on an invalid token, and they
1013 cause error recovery to begin in a different syntactic context than
1014 the one in which the invalid token was encountered. Second, when
1015 verbose error messages are enabled (with %error-verbose or the
1016 obsolete "#define YYERROR_VERBOSE"), the expected token list in the
1017 syntax error message can both contain invalid tokens and omit valid
1018 tokens.
1019
1020 The culprits for the above problems are %nonassoc, default
1021 reductions in inconsistent states, and parser state merging. Thus,
1022 IELR and LALR suffer the most. Canonical LR can suffer only if
1023 %nonassoc is used or if default reductions are enabled for
1024 inconsistent states.
1025
1026 LAC is a new mechanism within the parsing algorithm that solves
1027 these problems for canonical LR, IELR, and LALR without sacrificing
1028 %nonassoc, default reductions, or state merging. When LAC is in
1029 use, canonical LR and IELR behave almost exactly the same for both
1030 syntactically acceptable and syntactically unacceptable input.
1031 While LALR still does not support the full language-recognition
1032 power of canonical LR and IELR, LAC at least enables LALR's syntax
1033 error handling to correctly reflect LALR's language-recognition
1034 power.
1035
1036 Currently, LAC is only supported for deterministic parsers in C.
1037 You can enable LAC with the following directive:
1038
1039 %define parse.lac full
1040
1041 See the new section "LAC" in the Bison manual for additional
1042 details including a few caveats.
1043
1044 LAC is an experimental feature. More user feedback will help to
1045 stabilize it.
1046
1047 ** %define improvements:
1048
1049 *** Can now be invoked via the command line:
1050
1051 Each of these command-line options
1052
1053 -D NAME[=VALUE]
1054 --define=NAME[=VALUE]
1055
1056 -F NAME[=VALUE]
1057 --force-define=NAME[=VALUE]
1058
1059 is equivalent to this grammar file declaration
1060
1061 %define NAME ["VALUE"]
1062
1063 except that the manner in which Bison processes multiple definitions
1064 for the same NAME differs. Most importantly, -F and --force-define
1065 quietly override %define, but -D and --define do not. For further
1066 details, see the section "Bison Options" in the Bison manual.
1067
1068 *** Variables renamed:
1069
1070 The following %define variables
1071
1072 api.push_pull
1073 lr.keep_unreachable_states
1074
1075 have been renamed to
1076
1077 api.push-pull
1078 lr.keep-unreachable-states
1079
1080 The old names are now deprecated but will be maintained indefinitely
1081 for backward compatibility.
1082
1083 *** Values no longer need to be quoted in the grammar file:
1084
1085 If a %define value is an identifier, it no longer needs to be placed
1086 within quotations marks. For example,
1087
1088 %define api.push-pull "push"
1089
1090 can be rewritten as
1091
1092 %define api.push-pull push
1093
1094 *** Unrecognized variables are now errors not warnings.
1095
1096 *** Multiple invocations for any variable is now an error not a warning.
1097
1098 ** Unrecognized %code qualifiers are now errors not warnings.
1099
1100 ** Character literals not of length one:
1101
1102 Previously, Bison quietly converted all character literals to length
1103 one. For example, without warning, Bison interpreted the operators in
1104 the following grammar to be the same token:
1105
1106 exp: exp '++'
1107 | exp '+' exp
1108 ;
1109
1110 Bison now warns when a character literal is not of length one. In
1111 some future release, Bison will start reporting an error instead.
1112
1113 ** Destructor calls fixed for lookaheads altered in semantic actions:
1114
1115 Previously for deterministic parsers in C, if a user semantic action
1116 altered yychar, the parser in some cases used the old yychar value to
1117 determine which destructor to call for the lookahead upon a syntax
1118 error or upon parser return. This bug has been fixed.
1119
1120 ** C++ parsers use YYRHSLOC:
1121
1122 Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
1123 macro and use it in the default YYLLOC_DEFAULT. You are encouraged
1124 to use it. If, for instance, your location structure has "first"
1125 and "last" members, instead of
1126
1127 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1128 do \
1129 if (N) \
1130 { \
1131 (Current).first = (Rhs)[1].location.first; \
1132 (Current).last = (Rhs)[N].location.last; \
1133 } \
1134 else \
1135 { \
1136 (Current).first = (Current).last = (Rhs)[0].location.last; \
1137 } \
1138 while (false)
1139
1140 use:
1141
1142 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1143 do \
1144 if (N) \
1145 { \
1146 (Current).first = YYRHSLOC (Rhs, 1).first; \
1147 (Current).last = YYRHSLOC (Rhs, N).last; \
1148 } \
1149 else \
1150 { \
1151 (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \
1152 } \
1153 while (false)
1154
1155 ** YYLLOC_DEFAULT in C++:
1156
1157 The default implementation of YYLLOC_DEFAULT used to be issued in
1158 the header file. It is now output in the implementation file, after
1159 the user %code sections so that its #ifndef guard does not try to
1160 override the user's YYLLOC_DEFAULT if provided.
1161
1162 ** YYFAIL now produces warnings and Java parsers no longer implement it:
1163
1164 YYFAIL has existed for many years as an undocumented feature of
1165 deterministic parsers in C generated by Bison. More recently, it was
1166 a documented feature of Bison's experimental Java parsers. As
1167 promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
1168 semantic action now produces a deprecation warning, and Java parsers
1169 no longer implement YYFAIL at all. For further details, including a
1170 discussion of how to suppress C preprocessor warnings about YYFAIL
1171 being unused, see the Bison 2.4.2 NEWS entry.
1172
1173 ** Temporary hack for adding a semicolon to the user action:
1174
1175 Previously, Bison appended a semicolon to every user action for
1176 reductions when the output language defaulted to C (specifically, when
1177 neither %yacc, %language, %skeleton, or equivalent command-line
1178 options were specified). This allowed actions such as
1179
1180 exp: exp "+" exp { $$ = $1 + $3 };
1181
1182 instead of
1183
1184 exp: exp "+" exp { $$ = $1 + $3; };
1185
1186 As a first step in removing this misfeature, Bison now issues a
1187 warning when it appends a semicolon. Moreover, in cases where Bison
1188 cannot easily determine whether a semicolon is needed (for example, an
1189 action ending with a cpp directive or a braced compound initializer),
1190 it no longer appends one. Thus, the C compiler might now complain
1191 about a missing semicolon where it did not before. Future releases of
1192 Bison will cease to append semicolons entirely.
1193
1194 ** Verbose syntax error message fixes:
1195
1196 When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is
1197 specified, syntax error messages produced by the generated parser
1198 include the unexpected token as well as a list of expected tokens.
1199 The effect of %nonassoc on these verbose messages has been corrected
1200 in two ways, but a more complete fix requires LAC, described above:
1201
1202 *** When %nonassoc is used, there can exist parser states that accept no
1203 tokens, and so the parser does not always require a lookahead token
1204 in order to detect a syntax error. Because no unexpected token or
1205 expected tokens can then be reported, the verbose syntax error
1206 message described above is suppressed, and the parser instead
1207 reports the simpler message, "syntax error". Previously, this
1208 suppression was sometimes erroneously triggered by %nonassoc when a
1209 lookahead was actually required. Now verbose messages are
1210 suppressed only when all previous lookaheads have already been
1211 shifted or discarded.
1212
1213 *** Previously, the list of expected tokens erroneously included tokens
1214 that would actually induce a syntax error because conflicts for them
1215 were resolved with %nonassoc in the current parser state. Such
1216 tokens are now properly omitted from the list.
1217
1218 *** Expected token lists are still often wrong due to state merging
1219 (from LALR or IELR) and default reductions, which can both add
1220 invalid tokens and subtract valid tokens. Canonical LR almost
1221 completely fixes this problem by eliminating state merging and
1222 default reductions. However, there is one minor problem left even
1223 when using canonical LR and even after the fixes above. That is,
1224 if the resolution of a conflict with %nonassoc appears in a later
1225 parser state than the one at which some syntax error is
1226 discovered, the conflicted token is still erroneously included in
1227 the expected token list. Bison's new LAC implementation,
1228 described above, eliminates this problem and the need for
1229 canonical LR. However, LAC is still experimental and is disabled
1230 by default.
1231
1232 ** Java skeleton fixes:
1233
1234 *** A location handling bug has been fixed.
1235
1236 *** The top element of each of the value stack and location stack is now
1237 cleared when popped so that it can be garbage collected.
1238
1239 *** Parser traces now print the top element of the stack.
1240
1241 ** -W/--warnings fixes:
1242
1243 *** Bison now properly recognizes the "no-" versions of categories:
1244
1245 For example, given the following command line, Bison now enables all
1246 warnings except warnings for incompatibilities with POSIX Yacc:
1247
1248 bison -Wall,no-yacc gram.y
1249
1250 *** Bison now treats S/R and R/R conflicts like other warnings:
1251
1252 Previously, conflict reports were independent of Bison's normal
1253 warning system. Now, Bison recognizes the warning categories
1254 "conflicts-sr" and "conflicts-rr". This change has important
1255 consequences for the -W and --warnings command-line options. For
1256 example:
1257
1258 bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
1259 bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
1260 bison -Wnone gram.y # no conflicts are reported
1261 bison -Werror gram.y # any conflict is an error
1262
1263 However, as before, if the %expect or %expect-rr directive is
1264 specified, an unexpected number of conflicts is an error, and an
1265 expected number of conflicts is not reported, so -W and --warning
1266 then have no effect on the conflict report.
1267
1268 *** The "none" category no longer disables a preceding "error":
1269
1270 For example, for the following command line, Bison now reports
1271 errors instead of warnings for incompatibilities with POSIX Yacc:
1272
1273 bison -Werror,none,yacc gram.y
1274
1275 *** The "none" category now disables all Bison warnings:
1276
1277 Previously, the "none" category disabled only Bison warnings for
1278 which there existed a specific -W/--warning category. However,
1279 given the following command line, Bison is now guaranteed to
1280 suppress all warnings:
1281
1282 bison -Wnone gram.y
1283
1284 ** Precedence directives can now assign token number 0:
1285
1286 Since Bison 2.3b, which restored the ability of precedence
1287 directives to assign token numbers, doing so for token number 0 has
1288 produced an assertion failure. For example:
1289
1290 %left END 0
1291
1292 This bug has been fixed.
1293
1294 * Changes in version 2.4.3 (2010-08-05):
1295
1296 ** Bison now obeys -Werror and --warnings=error for warnings about
1297 grammar rules that are useless in the parser due to conflicts.
1298
1299 ** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
1300 been fixed.
1301
1302 ** Failures in the test suite for GCC 4.5 have been fixed.
1303
1304 ** Failures in the test suite for some versions of Sun Studio C++ have
1305 been fixed.
1306
1307 ** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
1308 warnings about undefined %prec identifiers will not be converted to
1309 errors in Bison 2.5. They will remain warnings, which should be
1310 sufficient for POSIX while avoiding backward compatibility issues.
1311
1312 ** Minor documentation fixes.
1313
1314 * Changes in version 2.4.2 (2010-03-20):
1315
1316 ** Some portability problems that resulted in failures and livelocks
1317 in the test suite on some versions of at least Solaris, AIX, HP-UX,
1318 RHEL4, and Tru64 have been addressed. As a result, fatal Bison
1319 errors should no longer cause M4 to report a broken pipe on the
1320 affected platforms.
1321
1322 ** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately.
1323
1324 POSIX specifies that an error be reported for any identifier that does
1325 not appear on the LHS of a grammar rule and that is not defined by
1326 %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this
1327 error report for the case when an identifier appears only after a
1328 %prec directive. It is now restored. However, for backward
1329 compatibility with recent Bison releases, it is only a warning for
1330 now. In Bison 2.5 and later, it will return to being an error.
1331 [Between the 2.4.2 and 2.4.3 releases, it was decided that this
1332 warning will not be converted to an error in Bison 2.5.]
1333
1334 ** Detection of GNU M4 1.4.6 or newer during configure is improved.
1335
1336 ** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
1337 YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
1338 avoided.
1339
1340 ** %code is now a permanent feature.
1341
1342 A traditional Yacc prologue directive is written in the form:
1343
1344 %{CODE%}
1345
1346 To provide a more flexible alternative, Bison 2.3b introduced the
1347 %code directive with the following forms for C/C++:
1348
1349 %code {CODE}
1350 %code requires {CODE}
1351 %code provides {CODE}
1352 %code top {CODE}
1353
1354 These forms are now considered permanent features of Bison. See the
1355 %code entries in the section "Bison Declaration Summary" in the Bison
1356 manual for a summary of their functionality. See the section
1357 "Prologue Alternatives" for a detailed discussion including the
1358 advantages of %code over the traditional Yacc prologue directive.
1359
1360 Bison's Java feature as a whole including its current usage of %code
1361 is still considered experimental.
1362
1363 ** YYFAIL is deprecated and will eventually be removed.
1364
1365 YYFAIL has existed for many years as an undocumented feature of
1366 deterministic parsers in C generated by Bison. Previously, it was
1367 documented for Bison's experimental Java parsers. YYFAIL is no longer
1368 documented for Java parsers and is formally deprecated in both cases.
1369 Users are strongly encouraged to migrate to YYERROR, which is
1370 specified by POSIX.
1371
1372 Like YYERROR, you can invoke YYFAIL from a semantic action in order to
1373 induce a syntax error. The most obvious difference from YYERROR is
1374 that YYFAIL will automatically invoke yyerror to report the syntax
1375 error so that you don't have to. However, there are several other
1376 subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
1377 inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is
1378 used. For a more detailed discussion, see:
1379
1380 http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
1381
1382 The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
1383 deterministic parsers in C will continue to implement it. However,
1384 because YYFAIL is already flawed, it seems futile to try to make new
1385 Bison features compatible with it. Thus, during parser generation,
1386 Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
1387 rule action. In a later release, YYFAIL will be disabled for
1388 %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will
1389 be removed altogether.
1390
1391 There exists at least one case where Bison 2.5's YYFAIL warning will
1392 be a false positive. Some projects add phony uses of YYFAIL and other
1393 Bison-defined macros for the sole purpose of suppressing C
1394 preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
1395 To avoid Bison's future warning, such YYFAIL uses can be moved to the
1396 epilogue (that is, after the second "%%") in the Bison input file. In
1397 this release (2.4.2), Bison already generates its own code to suppress
1398 C preprocessor warnings for YYFAIL, so projects can remove their own
1399 phony uses of YYFAIL if compatibility with Bison releases prior to
1400 2.4.2 is not necessary.
1401
1402 ** Internationalization.
1403
1404 Fix a regression introduced in Bison 2.4: Under some circumstances,
1405 message translations were not installed although supported by the
1406 host system.
1407
1408 * Changes in version 2.4.1 (2008-12-11):
1409
1410 ** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
1411 declarations have been fixed.
1412
1413 ** Temporary hack for adding a semicolon to the user action.
1414
1415 Bison used to prepend a trailing semicolon at the end of the user
1416 action for reductions. This allowed actions such as
1417
1418 exp: exp "+" exp { $$ = $1 + $3 };
1419
1420 instead of
1421
1422 exp: exp "+" exp { $$ = $1 + $3; };
1423
1424 Some grammars still depend on this "feature". Bison 2.4.1 restores
1425 the previous behavior in the case of C output (specifically, when
1426 neither %language or %skeleton or equivalent command-line options
1427 are used) to leave more time for grammars depending on the old
1428 behavior to be adjusted. Future releases of Bison will disable this
1429 feature.
1430
1431 ** A few minor improvements to the Bison manual.
1432
1433 * Changes in version 2.4 (2008-11-02):
1434
1435 ** %language is an experimental feature.
1436
1437 We first introduced this feature in test release 2.3b as a cleaner
1438 alternative to %skeleton. Since then, we have discussed the possibility of
1439 modifying its effect on Bison's output file names. Thus, in this release,
1440 we consider %language to be an experimental feature that will likely evolve
1441 in future releases.
1442
1443 ** Forward compatibility with GNU M4 has been improved.
1444
1445 ** Several bugs in the C++ skeleton and the experimental Java skeleton have been
1446 fixed.
1447
1448 * Changes in version 2.3b (2008-05-27):
1449
1450 ** The quotes around NAME that used to be required in the following directive
1451 are now deprecated:
1452
1453 %define NAME "VALUE"
1454
1455 ** The directive "%pure-parser" is now deprecated in favor of:
1456
1457 %define api.pure
1458
1459 which has the same effect except that Bison is more careful to warn about
1460 unreasonable usage in the latter case.
1461
1462 ** Push Parsing
1463
1464 Bison can now generate an LALR(1) parser in C with a push interface. That
1465 is, instead of invoking "yyparse", which pulls tokens from "yylex", you can
1466 push one token at a time to the parser using "yypush_parse", which will
1467 return to the caller after processing each token. By default, the push
1468 interface is disabled. Either of the following directives will enable it:
1469
1470 %define api.push_pull "push" // Just push; does not require yylex.
1471 %define api.push_pull "both" // Push and pull; requires yylex.
1472
1473 See the new section "A Push Parser" in the Bison manual for details.
1474
1475 The current push parsing interface is experimental and may evolve. More user
1476 feedback will help to stabilize it.
1477
1478 ** The -g and --graph options now output graphs in Graphviz DOT format,
1479 not VCG format. Like --graph, -g now also takes an optional FILE argument
1480 and thus cannot be bundled with other short options.
1481
1482 ** Java
1483
1484 Bison can now generate an LALR(1) parser in Java. The skeleton is
1485 "data/lalr1.java". Consider using the new %language directive instead of
1486 %skeleton to select it.
1487
1488 See the new section "Java Parsers" in the Bison manual for details.
1489
1490 The current Java interface is experimental and may evolve. More user
1491 feedback will help to stabilize it.
1492 Contributed by Paolo Bonzini.
1493
1494 ** %language
1495
1496 This new directive specifies the programming language of the generated
1497 parser, which can be C (the default), C++, or Java. Besides the skeleton
1498 that Bison uses, the directive affects the names of the generated files if
1499 the grammar file's name ends in ".y".
1500
1501 ** XML Automaton Report
1502
1503 Bison can now generate an XML report of the LALR(1) automaton using the new
1504 "--xml" option. The current XML schema is experimental and may evolve. More
1505 user feedback will help to stabilize it.
1506 Contributed by Wojciech Polak.
1507
1508 ** The grammar file may now specify the name of the parser header file using
1509 %defines. For example:
1510
1511 %defines "parser.h"
1512
1513 ** When reporting useless rules, useless nonterminals, and unused terminals,
1514 Bison now employs the terms "useless in grammar" instead of "useless",
1515 "useless in parser" instead of "never reduced", and "unused in grammar"
1516 instead of "unused".
1517
1518 ** Unreachable State Removal
1519
1520 Previously, Bison sometimes generated parser tables containing unreachable
1521 states. A state can become unreachable during conflict resolution if Bison
1522 disables a shift action leading to it from a predecessor state. Bison now:
1523
1524 1. Removes unreachable states.
1525
1526 2. Does not report any conflicts that appeared in unreachable states.
1527 WARNING: As a result, you may need to update %expect and %expect-rr
1528 directives in existing grammar files.
1529
1530 3. For any rule used only in such states, Bison now reports the rule as
1531 "useless in parser due to conflicts".
1532
1533 This feature can be disabled with the following directive:
1534
1535 %define lr.keep_unreachable_states
1536
1537 See the %define entry in the "Bison Declaration Summary" in the Bison manual
1538 for further discussion.
1539
1540 ** Lookahead Set Correction in the ".output" Report
1541
1542 When instructed to generate a ".output" file including lookahead sets
1543 (using "--report=lookahead", for example), Bison now prints each reduction's
1544 lookahead set only next to the associated state's one item that (1) is
1545 associated with the same rule as the reduction and (2) has its dot at the end
1546 of its RHS. Previously, Bison also erroneously printed the lookahead set
1547 next to all of the state's other items associated with the same rule. This
1548 bug affected only the ".output" file and not the generated parser source
1549 code.
1550
1551 ** --report-file=FILE is a new option to override the default ".output" file
1552 name.
1553
1554 ** The "=" that used to be required in the following directives is now
1555 deprecated:
1556
1557 %file-prefix "parser"
1558 %name-prefix "c_"
1559 %output "parser.c"
1560
1561 ** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}"
1562
1563 Bison 2.3a provided a new set of directives as a more flexible alternative to
1564 the traditional Yacc prologue blocks. Those have now been consolidated into
1565 a single %code directive with an optional qualifier field, which identifies
1566 the purpose of the code and thus the location(s) where Bison should generate
1567 it:
1568
1569 1. "%code {CODE}" replaces "%after-header {CODE}"
1570 2. "%code requires {CODE}" replaces "%start-header {CODE}"
1571 3. "%code provides {CODE}" replaces "%end-header {CODE}"
1572 4. "%code top {CODE}" replaces "%before-header {CODE}"
1573
1574 See the %code entries in section "Bison Declaration Summary" in the Bison
1575 manual for a summary of the new functionality. See the new section "Prologue
1576 Alternatives" for a detailed discussion including the advantages of %code
1577 over the traditional Yacc prologues.
1578
1579 The prologue alternatives are experimental. More user feedback will help to
1580 determine whether they should become permanent features.
1581
1582 ** Revised warning: unset or unused mid-rule values
1583
1584 Since Bison 2.2, Bison has warned about mid-rule values that are set but not
1585 used within any of the actions of the parent rule. For example, Bison warns
1586 about unused $2 in:
1587
1588 exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };
1589
1590 Now, Bison also warns about mid-rule values that are used but not set. For
1591 example, Bison warns about unset $$ in the mid-rule action in:
1592
1593 exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };
1594
1595 However, Bison now disables both of these warnings by default since they
1596 sometimes prove to be false alarms in existing grammars employing the Yacc
1597 constructs $0 or $-N (where N is some positive integer).
1598
1599 To enable these warnings, specify the option "--warnings=midrule-values" or
1600 "-W", which is a synonym for "--warnings=all".
1601
1602 ** Default %destructor or %printer with "<*>" or "<>"
1603
1604 Bison now recognizes two separate kinds of default %destructor's and
1605 %printer's:
1606
1607 1. Place "<*>" in a %destructor/%printer symbol list to define a default
1608 %destructor/%printer for all grammar symbols for which you have formally
1609 declared semantic type tags.
1610
1611 2. Place "<>" in a %destructor/%printer symbol list to define a default
1612 %destructor/%printer for all grammar symbols without declared semantic
1613 type tags.
1614
1615 Bison no longer supports the "%symbol-default" notation from Bison 2.3a.
1616 "<*>" and "<>" combined achieve the same effect with one exception: Bison no
1617 longer applies any %destructor to a mid-rule value if that mid-rule value is
1618 not actually ever referenced using either $$ or $n in a semantic action.
1619
1620 The default %destructor's and %printer's are experimental. More user
1621 feedback will help to determine whether they should become permanent
1622 features.
1623
1624 See the section "Freeing Discarded Symbols" in the Bison manual for further
1625 details.
1626
1627 ** %left, %right, and %nonassoc can now declare token numbers. This is required
1628 by POSIX. However, see the end of section "Operator Precedence" in the Bison
1629 manual for a caveat concerning the treatment of literal strings.
1630
1631 ** The nonfunctional --no-parser, -n, and %no-parser options have been
1632 completely removed from Bison.
1633
1634 * Changes in version 2.3a, 2006-09-13:
1635
1636 ** Instead of %union, you can define and use your own union type
1637 YYSTYPE if your grammar contains at least one <type> tag.
1638 Your YYSTYPE need not be a macro; it can be a typedef.
1639 This change is for compatibility with other Yacc implementations,
1640 and is required by POSIX.
1641
1642 ** Locations columns and lines start at 1.
1643 In accordance with the GNU Coding Standards and Emacs.
1644
1645 ** You may now declare per-type and default %destructor's and %printer's:
1646
1647 For example:
1648
1649 %union { char *string; }
1650 %token <string> STRING1
1651 %token <string> STRING2
1652 %type <string> string1
1653 %type <string> string2
1654 %union { char character; }
1655 %token <character> CHR
1656 %type <character> chr
1657 %destructor { free ($$); } %symbol-default
1658 %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
1659 %destructor { } <character>
1660
1661 guarantees that, when the parser discards any user-defined symbol that has a
1662 semantic type tag other than "<character>", it passes its semantic value to
1663 "free". However, when the parser discards a "STRING1" or a "string1", it
1664 also prints its line number to "stdout". It performs only the second
1665 "%destructor" in this case, so it invokes "free" only once.
1666
1667 [Although we failed to mention this here in the 2.3a release, the default
1668 %destructor's and %printer's were experimental, and they were rewritten in
1669 future versions.]
1670
1671 ** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y",
1672 "--yacc", or "%yacc"), Bison no longer generates #define statements for
1673 associating token numbers with token names. Removing the #define statements
1674 helps to sanitize the global namespace during preprocessing, but POSIX Yacc
1675 requires them. Bison still generates an enum for token names in all cases.
1676
1677 ** Handling of traditional Yacc prologue blocks is now more consistent but
1678 potentially incompatible with previous releases of Bison.
1679
1680 As before, you declare prologue blocks in your grammar file with the
1681 "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all
1682 prologue blocks that you've declared before the first %union. To generate
1683 the post-prologue, Bison concatenates all prologue blocks that you've
1684 declared after the first %union.
1685
1686 Previous releases of Bison inserted the pre-prologue into both the header
1687 file and the code file in all cases except for LALR(1) parsers in C. In the
1688 latter case, Bison inserted it only into the code file. For parsers in C++,
1689 the point of insertion was before any token definitions (which associate
1690 token numbers with names). For parsers in C, the point of insertion was
1691 after the token definitions.
1692
1693 Now, Bison never inserts the pre-prologue into the header file. In the code
1694 file, it always inserts it before the token definitions.
1695
1696 ** Bison now provides a more flexible alternative to the traditional Yacc
1697 prologue blocks: %before-header, %start-header, %end-header, and
1698 %after-header.
1699
1700 For example, the following declaration order in the grammar file reflects the
1701 order in which Bison will output these code blocks. However, you are free to
1702 declare these code blocks in your grammar file in whatever order is most
1703 convenient for you:
1704
1705 %before-header {
1706 /* Bison treats this block like a pre-prologue block: it inserts it into
1707 * the code file before the contents of the header file. It does *not*
1708 * insert it into the header file. This is a good place to put
1709 * #include's that you want at the top of your code file. A common
1710 * example is '#include "system.h"'. */
1711 }
1712 %start-header {
1713 /* Bison inserts this block into both the header file and the code file.
1714 * In both files, the point of insertion is before any Bison-generated
1715 * token, semantic type, location type, and class definitions. This is a
1716 * good place to define %union dependencies, for example. */
1717 }
1718 %union {
1719 /* Unlike the traditional Yacc prologue blocks, the output order for the
1720 * new %*-header blocks is not affected by their declaration position
1721 * relative to any %union in the grammar file. */
1722 }
1723 %end-header {
1724 /* Bison inserts this block into both the header file and the code file.
1725 * In both files, the point of insertion is after the Bison-generated
1726 * definitions. This is a good place to declare or define public
1727 * functions or data structures that depend on the Bison-generated
1728 * definitions. */
1729 }
1730 %after-header {
1731 /* Bison treats this block like a post-prologue block: it inserts it into
1732 * the code file after the contents of the header file. It does *not*
1733 * insert it into the header file. This is a good place to declare or
1734 * define internal functions or data structures that depend on the
1735 * Bison-generated definitions. */
1736 }
1737
1738 If you have multiple occurrences of any one of the above declarations, Bison
1739 will concatenate the contents in declaration order.
1740
1741 [Although we failed to mention this here in the 2.3a release, the prologue
1742 alternatives were experimental, and they were rewritten in future versions.]
1743
1744 ** The option "--report=look-ahead" has been changed to "--report=lookahead".
1745 The old spelling still works, but is not documented and may be removed
1746 in a future release.
1747
1748 * Changes in version 2.3, 2006-06-05:
1749
1750 ** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING",
1751 for compatibility with LALR(1) grammars.
1752
1753 ** It is now documented that any definition of YYSTYPE or YYLTYPE should
1754 be to a type name that does not contain parentheses or brackets.
1755
1756 * Changes in version 2.2, 2006-05-19:
1757
1758 ** The distribution terms for all Bison-generated parsers now permit
1759 using the parsers in nonfree programs. Previously, this permission
1760 was granted only for Bison-generated LALR(1) parsers in C.
1761
1762 ** %name-prefix changes the namespace name in C++ outputs.
1763
1764 ** The C++ parsers export their token_type.
1765
1766 ** Bison now allows multiple %union declarations, and concatenates
1767 their contents together.
1768
1769 ** New warning: unused values
1770 Right-hand side symbols whose values are not used are reported,
1771 if the symbols have destructors. For instance:
1772
1773 exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
1774 | exp "+" exp
1775 ;
1776
1777 will trigger a warning about $$ and $5 in the first rule, and $3 in
1778 the second ($1 is copied to $$ by the default rule). This example
1779 most likely contains three errors, and could be rewritten as:
1780
1781 exp: exp "?" exp ":" exp
1782 { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
1783 | exp "+" exp
1784 { $$ = $1 ? $1 : $3; if ($1) free ($3); }
1785 ;
1786
1787 However, if the original actions were really intended, memory leaks
1788 and all, the warnings can be suppressed by letting Bison believe the
1789 values are used, e.g.:
1790
1791 exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
1792 | exp "+" exp { $$ = $1; (void) $3; }
1793 ;
1794
1795 If there are mid-rule actions, the warning is issued if no action
1796 uses it. The following triggers no warning: $1 and $3 are used.
1797
1798 exp: exp { push ($1); } '+' exp { push ($3); sum (); };
1799
1800 The warning is intended to help catching lost values and memory leaks.
1801 If a value is ignored, its associated memory typically is not reclaimed.
1802
1803 ** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
1804 Destructors are now called when user code invokes YYABORT, YYACCEPT,
1805 and YYERROR, for all objects on the stack, other than objects
1806 corresponding to the right-hand side of the current rule.
1807
1808 ** %expect, %expect-rr
1809 Incorrect numbers of expected conflicts are now actual errors,
1810 instead of warnings.
1811
1812 ** GLR, YACC parsers.
1813 The %parse-params are available in the destructors (and the
1814 experimental printers) as per the documentation.
1815
1816 ** Bison now warns if it finds a stray "$" or "@" in an action.
1817
1818 ** %require "VERSION"
1819 This specifies that the grammar file depends on features implemented
1820 in Bison version VERSION or higher.
1821
1822 ** lalr1.cc: The token and value types are now class members.
1823 The tokens were defined as free form enums and cpp macros. YYSTYPE
1824 was defined as a free form union. They are now class members:
1825 tokens are enumerations of the "yy::parser::token" struct, and the
1826 semantic values have the "yy::parser::semantic_type" type.
1827
1828 If you do not want or can update to this scheme, the directive
1829 '%define "global_tokens_and_yystype" "1"' triggers the global
1830 definition of tokens and YYSTYPE. This change is suitable both
1831 for previous releases of Bison, and this one.
1832
1833 If you wish to update, then make sure older version of Bison will
1834 fail using '%require "2.2"'.
1835
1836 ** DJGPP support added.
1837 \f
1838 * Changes in version 2.1, 2005-09-16:
1839
1840 ** The C++ lalr1.cc skeleton supports %lex-param.
1841
1842 ** Bison-generated parsers now support the translation of diagnostics like
1843 "syntax error" into languages other than English. The default
1844 language is still English. For details, please see the new
1845 Internationalization section of the Bison manual. Software
1846 distributors should also see the new PACKAGING file. Thanks to
1847 Bruno Haible for this new feature.
1848
1849 ** Wording in the Bison-generated parsers has been changed slightly to
1850 simplify translation. In particular, the message "memory exhausted"
1851 has replaced "parser stack overflow", as the old message was not
1852 always accurate for modern Bison-generated parsers.
1853
1854 ** Destructors are now called when the parser aborts, for all symbols left
1855 behind on the stack. Also, the start symbol is now destroyed after a
1856 successful parse. In both cases, the behavior was formerly inconsistent.
1857
1858 ** When generating verbose diagnostics, Bison-generated parsers no longer
1859 quote the literal strings associated with tokens. For example, for
1860 a syntax error associated with '%token NUM "number"' they might
1861 print 'syntax error, unexpected number' instead of 'syntax error,
1862 unexpected "number"'.
1863 \f
1864 * Changes in version 2.0, 2004-12-25:
1865
1866 ** Possibly-incompatible changes
1867
1868 - Bison-generated parsers no longer default to using the alloca function
1869 (when available) to extend the parser stack, due to widespread
1870 problems in unchecked stack-overflow detection. You can "#define
1871 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
1872 the manual to determine safe values for YYMAXDEPTH in that case.
1873
1874 - Error token location.
1875 During error recovery, the location of the syntax error is updated
1876 to cover the whole sequence covered by the error token: it includes
1877 the shifted symbols thrown away during the first part of the error
1878 recovery, and the lookahead rejected during the second part.
1879
1880 - Semicolon changes:
1881 . Stray semicolons are no longer allowed at the start of a grammar.
1882 . Semicolons are now required after in-grammar declarations.
1883
1884 - Unescaped newlines are no longer allowed in character constants or
1885 string literals. They were never portable, and GCC 3.4.0 has
1886 dropped support for them. Better diagnostics are now generated if
1887 forget a closing quote.
1888
1889 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
1890
1891 ** New features
1892
1893 - GLR grammars now support locations.
1894
1895 - New directive: %initial-action.
1896 This directive allows the user to run arbitrary code (including
1897 initializing @$) from yyparse before parsing starts.
1898
1899 - A new directive "%expect-rr N" specifies the expected number of
1900 reduce/reduce conflicts in GLR parsers.
1901
1902 - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d".
1903 This is a GNU extension.
1904
1905 - The option "--report=lookahead" was changed to "--report=look-ahead".
1906 [However, this was changed back after 2.3.]
1907
1908 - Experimental %destructor support has been added to lalr1.cc.
1909
1910 - New configure option --disable-yacc, to disable installation of the
1911 yacc command and -ly library introduced in 1.875 for POSIX conformance.
1912
1913 ** Bug fixes
1914
1915 - For now, %expect-count violations are now just warnings, not errors.
1916 This is for compatibility with Bison 1.75 and earlier (when there are
1917 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
1918 are too many or too few shift/reduce conflicts). However, in future
1919 versions of Bison we plan to improve the %expect machinery so that
1920 these violations will become errors again.
1921
1922 - Within Bison itself, numbers (e.g., goto numbers) are no longer
1923 arbitrarily limited to 16-bit counts.
1924
1925 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
1926 \f
1927 * Changes in version 1.875, 2003-01-01:
1928
1929 ** The documentation license has been upgraded to version 1.2
1930 of the GNU Free Documentation License.
1931
1932 ** syntax error processing
1933
1934 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
1935 locations too. This fixes bugs in error-location computation.
1936
1937 - %destructor
1938 It is now possible to reclaim the memory associated to symbols
1939 discarded during error recovery. This feature is still experimental.
1940
1941 - %error-verbose
1942 This new directive is preferred over YYERROR_VERBOSE.
1943
1944 - #defining yyerror to steal internal variables is discouraged.
1945 It is not guaranteed to work forever.
1946
1947 ** POSIX conformance
1948
1949 - Semicolons are once again optional at the end of grammar rules.
1950 This reverts to the behavior of Bison 1.33 and earlier, and improves
1951 compatibility with Yacc.
1952
1953 - "parse error" -> "syntax error"
1954 Bison now uniformly uses the term "syntax error"; formerly, the code
1955 and manual sometimes used the term "parse error" instead. POSIX
1956 requires "syntax error" in diagnostics, and it was thought better to
1957 be consistent.
1958
1959 - The documentation now emphasizes that yylex and yyerror must be
1960 declared before use. C99 requires this.
1961
1962 - Bison now parses C99 lexical constructs like UCNs and
1963 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
1964
1965 - File names are properly escaped in C output. E.g., foo\bar.y is
1966 output as "foo\\bar.y".
1967
1968 - Yacc command and library now available
1969 The Bison distribution now installs a "yacc" command, as POSIX requires.
1970 Also, Bison now installs a small library liby.a containing
1971 implementations of Yacc-compatible yyerror and main functions.
1972 This library is normally not useful, but POSIX requires it.
1973
1974 - Type clashes now generate warnings, not errors.
1975
1976 - If the user does not define YYSTYPE as a macro, Bison now declares it
1977 using typedef instead of defining it as a macro.
1978 For consistency, YYLTYPE is also declared instead of defined.
1979
1980 ** Other compatibility issues
1981
1982 - %union directives can now have a tag before the "{", e.g., the
1983 directive "%union foo {...}" now generates the C code
1984 "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility.
1985 The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc.
1986 For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype".
1987 This is for compatibility with both Yacc and Bison 1.35.
1988
1989 - ";" is output before the terminating "}" of an action, for
1990 compatibility with Bison 1.35.
1991
1992 - Bison now uses a Yacc-style format for conflict reports, e.g.,
1993 "conflicts: 2 shift/reduce, 1 reduce/reduce".
1994
1995 - "yystype" and "yyltype" are now obsolescent macros instead of being
1996 typedefs or tags; they are no longer documented and are planned to be
1997 withdrawn in a future release.
1998
1999 ** GLR parser notes
2000
2001 - GLR and inline
2002 Users of Bison have to decide how they handle the portability of the
2003 C keyword "inline".
2004
2005 - "parsing stack overflow..." -> "parser stack overflow"
2006 GLR parsers now report "parser stack overflow" as per the Bison manual.
2007
2008 ** %parse-param and %lex-param
2009 The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass
2010 additional context to yyparse and yylex. They suffer from several
2011 shortcomings:
2012
2013 - a single argument only can be added,
2014 - their types are weak (void *),
2015 - this context is not passed to ancillary functions such as yyerror,
2016 - only yacc.c parsers support them.
2017
2018 The new %parse-param/%lex-param directives provide a more precise control.
2019 For instance:
2020
2021 %parse-param {int *nastiness}
2022 %lex-param {int *nastiness}
2023 %parse-param {int *randomness}
2024
2025 results in the following signatures:
2026
2027 int yylex (int *nastiness);
2028 int yyparse (int *nastiness, int *randomness);
2029
2030 or, if both %pure-parser and %locations are used:
2031
2032 int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness);
2033 int yyparse (int *nastiness, int *randomness);
2034
2035 ** Bison now warns if it detects conflicting outputs to the same file,
2036 e.g., it generates a warning for "bison -d -o foo.h foo.y" since
2037 that command outputs both code and header to foo.h.
2038
2039 ** #line in output files
2040 - --no-line works properly.
2041
2042 ** Bison can no longer be built by a K&R C compiler; it requires C89 or
2043 later to be built. This change originally took place a few versions
2044 ago, but nobody noticed until we recently asked someone to try
2045 building Bison with a K&R C compiler.
2046 \f
2047 * Changes in version 1.75, 2002-10-14:
2048
2049 ** Bison should now work on 64-bit hosts.
2050
2051 ** Indonesian translation thanks to Tedi Heriyanto.
2052
2053 ** GLR parsers
2054 Fix spurious parse errors.
2055
2056 ** Pure parsers
2057 Some people redefine yyerror to steal yyparse' private variables.
2058 Reenable this trick until an official feature replaces it.
2059
2060 ** Type Clashes
2061 In agreement with POSIX and with other Yaccs, leaving a default
2062 action is valid when $$ is untyped, and $1 typed:
2063
2064 untyped: ... typed;
2065
2066 but the converse remains an error:
2067
2068 typed: ... untyped;
2069
2070 ** Values of mid-rule actions
2071 The following code:
2072
2073 foo: { ... } { $$ = $1; } ...
2074
2075 was incorrectly rejected: $1 is defined in the second mid-rule
2076 action, and is equal to the $$ of the first mid-rule action.
2077 \f
2078 * Changes in version 1.50, 2002-10-04:
2079
2080 ** GLR parsing
2081 The declaration
2082 %glr-parser
2083 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
2084 almost any context-free grammar, ambiguous or not. The new declarations
2085 %dprec and %merge on grammar rules allow parse-time resolution of
2086 ambiguities. Contributed by Paul Hilfinger.
2087
2088 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
2089 like the Alpha, so please stick to 32-bit hosts for now.
2090
2091 ** Output Directory
2092 When not in Yacc compatibility mode, when the output file was not
2093 specified, running "bison foo/bar.y" created "foo/bar.c". It
2094 now creates "bar.c".
2095
2096 ** Undefined token
2097 The undefined token was systematically mapped to 2 which prevented
2098 the use of 2 by the user. This is no longer the case.
2099
2100 ** Unknown token numbers
2101 If yylex returned an out of range value, yyparse could die. This is
2102 no longer the case.
2103
2104 ** Error token
2105 According to POSIX, the error token must be 256.
2106 Bison extends this requirement by making it a preference: *if* the
2107 user specified that one of her tokens is numbered 256, then error
2108 will be mapped onto another number.
2109
2110 ** Verbose error messages
2111 They no longer report "..., expecting error or..." for states where
2112 error recovery is possible.
2113
2114 ** End token
2115 Defaults to "$end" instead of "$".
2116
2117 ** Error recovery now conforms to documentation and to POSIX
2118 When a Bison-generated parser encounters a syntax error, it now pops
2119 the stack until it finds a state that allows shifting the error
2120 token. Formerly, it popped the stack until it found a state that
2121 allowed some non-error action other than a default reduction on the
2122 error token. The new behavior has long been the documented behavior,
2123 and has long been required by POSIX. For more details, please see
2124 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
2125 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
2126
2127 ** Traces
2128 Popped tokens and nonterminals are now reported.
2129
2130 ** Larger grammars
2131 Larger grammars are now supported (larger token numbers, larger grammar
2132 size (= sum of the LHS and RHS lengths), larger LALR tables).
2133 Formerly, many of these numbers ran afoul of 16-bit limits;
2134 now these limits are 32 bits on most hosts.
2135
2136 ** Explicit initial rule
2137 Bison used to play hacks with the initial rule, which the user does
2138 not write. It is now explicit, and visible in the reports and
2139 graphs as rule 0.
2140
2141 ** Useless rules
2142 Before, Bison reported the useless rules, but, although not used,
2143 included them in the parsers. They are now actually removed.
2144
2145 ** Useless rules, useless nonterminals
2146 They are now reported, as a warning, with their locations.
2147
2148 ** Rules never reduced
2149 Rules that can never be reduced because of conflicts are now
2150 reported.
2151
2152 ** Incorrect "Token not used"
2153 On a grammar such as
2154
2155 %token useless useful
2156 %%
2157 exp: '0' %prec useful;
2158
2159 where a token was used to set the precedence of the last rule,
2160 bison reported both "useful" and "useless" as useless tokens.
2161
2162 ** Revert the C++ namespace changes introduced in 1.31
2163 as they caused too many portability hassles.
2164
2165 ** Default locations
2166 By an accident of design, the default computation of @$ was
2167 performed after another default computation was performed: @$ = @1.
2168 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
2169 the computation of @$.
2170
2171 ** Token end-of-file
2172 The token end of file may be specified by the user, in which case,
2173 the user symbol is used in the reports, the graphs, and the verbose
2174 error messages instead of "$end", which remains being the default.
2175 For instance
2176 %token MYEOF 0
2177 or
2178 %token MYEOF 0 "end of file"
2179
2180 ** Semantic parser
2181 This old option, which has been broken for ages, is removed.
2182
2183 ** New translations
2184 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
2185 Croatian, thanks to Denis Lackovic.
2186
2187 ** Incorrect token definitions
2188 When given
2189 %token 'a' "A"
2190 bison used to output
2191 #define 'a' 65
2192
2193 ** Token definitions as enums
2194 Tokens are output both as the traditional #define's, and, provided
2195 the compiler supports ANSI C or is a C++ compiler, as enums.
2196 This lets debuggers display names instead of integers.
2197
2198 ** Reports
2199 In addition to --verbose, bison supports --report=THINGS, which
2200 produces additional information:
2201 - itemset
2202 complete the core item sets with their closure
2203 - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back]
2204 explicitly associate lookahead tokens to items
2205 - solved
2206 describe shift/reduce conflicts solving.
2207 Bison used to systematically output this information on top of
2208 the report. Solved conflicts are now attached to their states.
2209
2210 ** Type clashes
2211 Previous versions don't complain when there is a type clash on
2212 the default action if the rule has a mid-rule action, such as in:
2213
2214 %type <foo> bar
2215 %%
2216 bar: '0' {} '0';
2217
2218 This is fixed.
2219
2220 ** GNU M4 is now required when using Bison.
2221 \f
2222 * Changes in version 1.35, 2002-03-25:
2223
2224 ** C Skeleton
2225 Some projects use Bison's C parser with C++ compilers, and define
2226 YYSTYPE as a class. The recent adjustment of C parsers for data
2227 alignment and 64 bit architectures made this impossible.
2228
2229 Because for the time being no real solution for C++ parser
2230 generation exists, kludges were implemented in the parser to
2231 maintain this use. In the future, when Bison has C++ parsers, this
2232 kludge will be disabled.
2233
2234 This kludge also addresses some C++ problems when the stack was
2235 extended.
2236 \f
2237 * Changes in version 1.34, 2002-03-12:
2238
2239 ** File name clashes are detected
2240 $ bison foo.y -d -o foo.x
2241 fatal error: header and parser would both be named "foo.x"
2242
2243 ** A missing ";" at the end of a rule triggers a warning
2244 In accordance with POSIX, and in agreement with other
2245 Yacc implementations, Bison will mandate this semicolon in the near
2246 future. This eases the implementation of a Bison parser of Bison
2247 grammars by making this grammar LALR(1) instead of LR(2). To
2248 facilitate the transition, this release introduces a warning.
2249
2250 ** Revert the C++ namespace changes introduced in 1.31, as they caused too
2251 many portability hassles.
2252
2253 ** DJGPP support added.
2254
2255 ** Fix test suite portability problems.
2256 \f
2257 * Changes in version 1.33, 2002-02-07:
2258
2259 ** Fix C++ issues
2260 Groff could not be compiled for the definition of size_t was lacking
2261 under some conditions.
2262
2263 ** Catch invalid @n
2264 As is done with $n.
2265 \f
2266 * Changes in version 1.32, 2002-01-23:
2267
2268 ** Fix Yacc output file names
2269
2270 ** Portability fixes
2271
2272 ** Italian, Dutch translations
2273 \f
2274 * Changes in version 1.31, 2002-01-14:
2275
2276 ** Many Bug Fixes
2277
2278 ** GNU Gettext and %expect
2279 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
2280 Bison dies on incorrect %expectations, we fear there will be
2281 too many bug reports for Gettext, so _for the time being_, %expect
2282 does not trigger an error when the input file is named "plural.y".
2283
2284 ** Use of alloca in parsers
2285 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
2286 malloc exclusively. Since 1.29, but was not NEWS'ed.
2287
2288 alloca is used only when compiled with GCC, to avoid portability
2289 problems as on AIX.
2290
2291 ** yyparse now returns 2 if memory is exhausted; formerly it dumped core.
2292
2293 ** When the generated parser lacks debugging code, YYDEBUG is now 0
2294 (as POSIX requires) instead of being undefined.
2295
2296 ** User Actions
2297 Bison has always permitted actions such as { $$ = $1 }: it adds the
2298 ending semicolon. Now if in Yacc compatibility mode, the semicolon
2299 is no longer output: one has to write { $$ = $1; }.
2300
2301 ** Better C++ compliance
2302 The output parsers try to respect C++ namespaces.
2303 [This turned out to be a failed experiment, and it was reverted later.]
2304
2305 ** Reduced Grammars
2306 Fixed bugs when reporting useless nonterminals.
2307
2308 ** 64 bit hosts
2309 The parsers work properly on 64 bit hosts.
2310
2311 ** Error messages
2312 Some calls to strerror resulted in scrambled or missing error messages.
2313
2314 ** %expect
2315 When the number of shift/reduce conflicts is correct, don't issue
2316 any warning.
2317
2318 ** The verbose report includes the rule line numbers.
2319
2320 ** Rule line numbers are fixed in traces.
2321
2322 ** Swedish translation
2323
2324 ** Parse errors
2325 Verbose parse error messages from the parsers are better looking.
2326 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
2327 Now: parse error: unexpected '/', expecting "number" or '-' or '('
2328
2329 ** Fixed parser memory leaks.
2330 When the generated parser was using malloc to extend its stacks, the
2331 previous allocations were not freed.
2332
2333 ** Fixed verbose output file.
2334 Some newlines were missing.
2335 Some conflicts in state descriptions were missing.
2336
2337 ** Fixed conflict report.
2338 Option -v was needed to get the result.
2339
2340 ** %expect
2341 Was not used.
2342 Mismatches are errors, not warnings.
2343
2344 ** Fixed incorrect processing of some invalid input.
2345
2346 ** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
2347
2348 ** Fixed some typos in the documentation.
2349
2350 ** %token MY_EOF 0 is supported.
2351 Before, MY_EOF was silently renumbered as 257.
2352
2353 ** doc/refcard.tex is updated.
2354
2355 ** %output, %file-prefix, %name-prefix.
2356 New.
2357
2358 ** --output
2359 New, aliasing "--output-file".
2360 \f
2361 * Changes in version 1.30, 2001-10-26:
2362
2363 ** "--defines" and "--graph" have now an optional argument which is the
2364 output file name. "-d" and "-g" do not change; they do not take any
2365 argument.
2366
2367 ** "%source_extension" and "%header_extension" are removed, failed
2368 experiment.
2369
2370 ** Portability fixes.
2371 \f
2372 * Changes in version 1.29, 2001-09-07:
2373
2374 ** The output file does not define const, as this caused problems when used
2375 with common autoconfiguration schemes. If you still use ancient compilers
2376 that lack const, compile with the equivalent of the C compiler option
2377 "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this.
2378
2379 ** Added "-g" and "--graph".
2380
2381 ** The Bison manual is now distributed under the terms of the GNU FDL.
2382
2383 ** The input and the output files has automatically a similar extension.
2384
2385 ** Russian translation added.
2386
2387 ** NLS support updated; should hopefully be less troublesome.
2388
2389 ** Added the old Bison reference card.
2390
2391 ** Added "--locations" and "%locations".
2392
2393 ** Added "-S" and "--skeleton".
2394
2395 ** "%raw", "-r", "--raw" is disabled.
2396
2397 ** Special characters are escaped when output. This solves the problems
2398 of the #line lines with path names including backslashes.
2399
2400 ** New directives.
2401 "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose",
2402 "%debug", "%source_extension" and "%header_extension".
2403
2404 ** @$
2405 Automatic location tracking.
2406 \f
2407 * Changes in version 1.28, 1999-07-06:
2408
2409 ** Should compile better now with K&R compilers.
2410
2411 ** Added NLS.
2412
2413 ** Fixed a problem with escaping the double quote character.
2414
2415 ** There is now a FAQ.
2416 \f
2417 * Changes in version 1.27:
2418
2419 ** The make rule which prevented bison.simple from being created on
2420 some systems has been fixed.
2421 \f
2422 * Changes in version 1.26:
2423
2424 ** Bison now uses Automake.
2425
2426 ** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
2427
2428 ** Token numbers now start at 257 as previously documented, not 258.
2429
2430 ** Bison honors the TMPDIR environment variable.
2431
2432 ** A couple of buffer overruns have been fixed.
2433
2434 ** Problems when closing files should now be reported.
2435
2436 ** Generated parsers should now work even on operating systems which do
2437 not provide alloca().
2438 \f
2439 * Changes in version 1.25, 1995-10-16:
2440
2441 ** Errors in the input grammar are not fatal; Bison keeps reading
2442 the grammar file, and reports all the errors found in it.
2443
2444 ** Tokens can now be specified as multiple-character strings: for
2445 example, you could use "<=" for a token which looks like <=, instead
2446 of choosing a name like LESSEQ.
2447
2448 ** The %token_table declaration says to write a table of tokens (names
2449 and numbers) into the parser file. The yylex function can use this
2450 table to recognize multiple-character string tokens, or for other
2451 purposes.
2452
2453 ** The %no_lines declaration says not to generate any #line preprocessor
2454 directives in the parser file.
2455
2456 ** The %raw declaration says to use internal Bison token numbers, not
2457 Yacc-compatible token numbers, when token names are defined as macros.
2458
2459 ** The --no-parser option produces the parser tables without including
2460 the parser engine; a project can now use its own parser engine.
2461 The actions go into a separate file called NAME.act, in the form of
2462 a switch statement body.
2463 \f
2464 * Changes in version 1.23:
2465
2466 The user can define YYPARSE_PARAM as the name of an argument to be
2467 passed into yyparse. The argument should have type void *. It should
2468 actually point to an object. Grammar actions can access the variable
2469 by casting it to the proper pointer type.
2470
2471 Line numbers in output file corrected.
2472 \f
2473 * Changes in version 1.22:
2474
2475 --help option added.
2476 \f
2477 * Changes in version 1.20:
2478
2479 Output file does not redefine const for C++.
2480
2481 -----
2482
2483 Copyright (C) 1995-2013 Free Software Foundation, Inc.
2484
2485 This file is part of Bison, the GNU Parser Generator.
2486
2487 This program is free software: you can redistribute it and/or modify
2488 it under the terms of the GNU General Public License as published by
2489 the Free Software Foundation, either version 3 of the License, or
2490 (at your option) any later version.
2491
2492 This program is distributed in the hope that it will be useful,
2493 but WITHOUT ANY WARRANTY; without even the implied warranty of
2494 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2495 GNU General Public License for more details.
2496
2497 You should have received a copy of the GNU General Public License
2498 along with this program. If not, see <http://www.gnu.org/licenses/>.
2499
2500 LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr
2501 LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr
2502 LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar
2503 LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr
2504 LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks
2505 LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG
2506 LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush
2507 LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly
2508 LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's
2509 LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de
2510 LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto
2511 LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs
2512 LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF
2513 LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY
2514 LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL
2515 LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh
2516 LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf
2517 LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp Wother nterm arg init
2518 LocalWords: TOK calc yyo fval Wconflicts parsers yystackp yyval yynerrs
2519 LocalWords: Théophile Ranquet Santet fno fnone stype associativity Tolmer
2520 LocalWords: Wprecedence Rassoul Wempty Paolo Bonzini parser's Michiel loc
2521 LocalWords: redeclaration sval fcaret reentrant XSLT xsl Wmaybe yyvsp Tedi
2522 LocalWords: pragmas noreturn untyped Rozenman unexpanded Wojciech Polak
2523 LocalWords: Alexandre MERCHANTABILITY
2524
2525 Local Variables:
2526 mode: outline
2527 fill-column: 76
2528 End: