]> git.saurik.com Git - bison.git/blame - NEWS
maint: post-release administrivia
[bison.git] / NEWS
CommitLineData
ed7658fe 1GNU Bison NEWS
3af4feb2 2
11d82017
AD
3* Noteworthy changes in release ?.? (????-??-??) [?]
4
5
da6f0746 6* Noteworthy changes in release 3.0.3 (2015-01-15) [stable]
0bd5ee5f 7
461f1516
AD
8** Bug fixes
9
ee028dce
AD
10*** C++ with Variants (lalr1.cc)
11
12 Problems with %destructor and '%define parse.assert' have been fixed.
13
14*** Named %union support (yacc.c, glr.c)
827bc59c
AD
15
16 Bison 3.0 introduced a regression on named %union such as
17
18 %union foo { int ival; };
19
ee028dce 20 The possibility to use a name was introduced "for Yacc compatibility".
827bc59c
AD
21 It is however not required by POSIX Yacc, and its usefulness is not clear.
22
ee028dce 23*** %define api.value.type union with %defines (yacc.c, glr.c)
8d4dc896 24
ee028dce
AD
25 The C parsers were broken when %defines was used together with "%define
26 api.value.type union".
8d4dc896 27
461f1516
AD
28*** Redeclarations are reported in proper order
29
30 On
31
32 %token FOO "foo"
33 %printer {} "foo"
34 %printer {} FOO
35
36 bison used to report:
37
38 /tmp/foo.yy:2.10-11: error: %printer redeclaration for FOO
39 %printer {} "foo"
40 ^^
41 /tmp/foo.yy:3.10-11: previous declaration
42 %printer {} FOO
43 ^^
44
45 Now, the "previous" declaration is always the first one.
46
0bd5ee5f 47
b167e7ba
AD
48** Documentation
49
50 Bison now installs various files in its docdir (which defaults to
51 '/usr/local/share/doc/bison'), including the three fully blown examples
52 extracted from the documentation:
53
54 - rpcalc
55 Reverse polish calculator, a simple introductory example.
56 - mfcalc
57 Multi-function Calc, a calculator with memory and functions and located
58 error messages.
59 - calc++
60 a calculator in C++ using variant support and token constructors.
61
a3efd13e 62* Noteworthy changes in release 3.0.2 (2013-12-05) [stable]
6d94eebb 63
7b0ca050
AD
64** Bug fixes
65
184b42c8
AD
66*** Generated source files when errors are reported
67
68 When warnings are issued and -Werror is set, bison would still generate
69 the source files (*.c, *.h...). As a consequence, some runs of "make"
70 could fail the first time, but not the second (as the files were generated
71 anyway).
72
73 This is fixed: bison no longer generates this source files, but, of
74 course, still produces the various reports (*.output, *.xml, etc.).
75
be29c71d
AD
76*** %empty is used in reports
77
78 Empty right-hand sides are denoted by '%empty' in all the reports (text,
79 dot, XML and formats derived from it).
80
7b0ca050
AD
81*** YYERROR and variants
82
83 When C++ variant support is enabled, an error triggered via YYERROR, but
84 not caught via error recovery, resulted in a double deletion.
6d94eebb 85
c2ecada3 86* Noteworthy changes in release 3.0.1 (2013-11-12) [stable]
e386b50f 87
0a244a22
AD
88** Bug fixes
89
39bace5d
AD
90*** Errors in caret diagnostics
91
92 On some platforms, some errors could result in endless diagnostics.
e4678430
AD
93
94*** Fixes of the -Werror option
95
96 Options such as "-Werror -Wno-error=foo" were still turning "foo"
97 diagnostics into errors instead of warnings. This is fixed.
98
99 Actually, for consistency with GCC, "-Wno-error=foo -Werror" now also
100 leaves "foo" diagnostics as warnings. Similarly, with "-Werror=foo
101 -Wno-error", "foo" diagnostics are now errors.
e386b50f 102
071863b3
AD
103*** GLR Predicates
104
105 As demonstrated in the documentation, one can now leave spaces between
106 "%?" and its "{".
107
265640d5
AD
108*** Installation
109
110 The yacc.1 man page is no longer installed if --disable-yacc was
111 specified.
112
39bace5d
AD
113*** Fixes in the test suite
114
115 Bugs and portability issues.
116
534497f5 117* Noteworthy changes in release 3.0 (2013-07-25) [stable]
facb910c 118
8458a411 119** WARNING: Future backward-incompatibilities!
597afd73 120
597afd73
AD
121 Like other GNU packages, Bison will start using some of the C99 features
122 for its own code, especially the definition of variables after statements.
123 The generated C parsers still aim at C90.
124
8458a411 125** Backward incompatible changes
47db7ed1
AD
126
127*** Obsolete features
128
40bb6f78
AD
129 Support for YYFAIL is removed (deprecated in Bison 2.4.2): use YYERROR.
130
131 Support for yystype and yyltype is removed (deprecated in Bison 1.875):
132 use YYSTYPE and YYLTYPE.
133
134 Support for YYLEX_PARAM and YYPARSE_PARAM is removed (deprecated in Bison
135 1.875): use %lex-param, %parse-param, or %param.
47db7ed1 136
05e25f23
AD
137 Missing semicolons at the end of actions are no longer added (as announced
138 in the release 2.5).
139
1fa19a76
AD
140*** Use of YACC='bison -y'
141
142 TL;DR: With Autoconf <= 2.69, pass -Wno-yacc to (AM_)YFLAGS if you use
143 Bison extensions.
144
145 Traditional Yacc generates 'y.tab.c' whatever the name of the input file.
146 Therefore Makefiles written for Yacc expect 'y.tab.c' (and possibly
147 'y.tab.h' and 'y.outout') to be generated from 'foo.y'.
148
149 To this end, for ages, AC_PROG_YACC, Autoconf's macro to look for an
150 implementation of Yacc, was using Bison as 'bison -y'. While it does
151 ensure compatible output file names, it also enables warnings for
152 incompatibilities with POSIX Yacc. In other words, 'bison -y' triggers
153 warnings for Bison extensions.
154
155 Autoconf 2.70+ fixes this incompatibility by using YACC='bison -o y.tab.c'
156 (which also generates 'y.tab.h' and 'y.output' when needed).
157 Alternatively, disable Yacc warnings by passing '-Wno-yacc' to your Yacc
158 flags (YFLAGS, or AM_YFLAGS with Automake).
159
597afd73
AD
160** Bug fixes
161
c21e515e 162*** The epilogue is no longer affected by internal #defines (glr.c)
597afd73
AD
163
164 The glr.c skeleton uses defines such as #define yylval (yystackp->yyval) in
165 generated code. These weren't properly undefined before the inclusion of
166 the user epilogue, so functions such as the following were butchered by the
167 preprocessor expansion:
168
169 int yylex (YYSTYPE *yylval);
170
6c7022f7 171 This is fixed: yylval, yynerrs, yychar, and yylloc are now valid
597afd73
AD
172 identifiers for user-provided variables.
173
f0f95a50
AD
174*** stdio.h is no longer needed when locations are enabled (yacc.c)
175
176 Changes in Bison 2.7 introduced a dependency on FILE and fprintf when
177 locations are enabled. This is fixed.
178
bb4b189b
AD
179*** Warnings about useless %pure-parser/%define api.pure are restored
180
597afd73
AD
181** Diagnostics reported by Bison
182
183 Most of these features were contributed by Théophile Ranquet and Victor
184 Santet.
73370a9d 185
016426c1
TR
186*** Carets
187
188 Version 2.7 introduced caret errors, for a prettier output. These are now
189 activated by default. The old format can still be used by invoking Bison
190 with -fno-caret (or -fnone).
191
fec5f3c0
AD
192 Some error messages that reproduced excerpts of the grammar are now using
193 the caret information only. For instance on:
194
195 %%
196 exp: 'a' | 'a';
197
198 Bison 2.7 reports:
199
200 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
201 in.y:2.12-14: warning: rule useless in parser due to conflicts: exp: 'a' [-Wother]
202
203 Now bison reports:
204
205 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
206 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
207 exp: 'a' | 'a';
208 ^^^
209
210 and "bison -fno-caret" reports:
211
212 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
213 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
214
1048a1c9 215*** Enhancements of the -Werror option
518e8830 216
1048a1c9 217 The -Werror=CATEGORY option is now recognized, and will treat specified
d949eefd
AD
218 warnings as errors. The warnings need not have been explicitly activated
219 using the -W option, this is similar to what GCC 4.7 does.
1048a1c9
AD
220
221 For example, given the following command line, Bison will treat both
d949eefd 222 warnings related to POSIX Yacc incompatibilities and S/R conflicts as
1048a1c9
AD
223 errors (and only those):
224
225 $ bison -Werror=yacc,error=conflicts-sr input.y
226
227 If no categories are specified, -Werror will make all active warnings into
228 errors. For example, the following line does the same the previous example:
229
230 $ bison -Werror -Wnone -Wyacc -Wconflicts-sr input.y
231
232 (By default -Wconflicts-sr,conflicts-rr,deprecated,other is enabled.)
233
234 Note that the categories in this -Werror option may not be prefixed with
235 "no-". However, -Wno-error[=CATEGORY] is valid.
236
237 Note that -y enables -Werror=yacc. Therefore it is now possible to require
238 Yacc-like behavior (e.g., always generate y.tab.c), but to report
239 incompatibilities as warnings: "-y -Wno-error=yacc".
518e8830 240
46bdb8ec 241*** The display of warnings is now richer
73370a9d 242
46bdb8ec 243 The option that controls a given warning is now displayed:
73370a9d 244
46bdb8ec 245 foo.y:4.6: warning: type clash on default action: <foo> != <bar> [-Wother]
73370a9d 246
46bdb8ec
TR
247 In the case of warnings treated as errors, the prefix is changed from
248 "warning: " to "error: ", and the suffix is displayed, in a manner similar
d949eefd 249 to GCC, as [-Werror=CATEGORY].
1048a1c9 250
46bdb8ec
TR
251 For instance, where the previous version of Bison would report (and exit
252 with failure):
1048a1c9
AD
253
254 bison: warnings being treated as errors
46bdb8ec 255 input.y:1.1: warning: stray ',' treated as white space
1048a1c9 256
46bdb8ec 257 it now reports:
1048a1c9 258
1048a1c9
AD
259 input.y:1.1: error: stray ',' treated as white space [-Werror=other]
260
261*** Deprecated constructs
262
263 The new 'deprecated' warning category flags obsolete constructs whose
264 support will be discontinued. It is enabled by default. These warnings
265 used to be reported as 'other' warnings.
266
73370a9d 267*** Useless semantic types
9641b918
VS
268
269 Bison now warns about useless (uninhabited) semantic types. Since
270 semantic types are not declared to Bison (they are defined in the opaque
271 %union structure), it is %printer/%destructor directives about useless
272 types that trigger the warning:
273
274 %token <type1> term
275 %type <type2> nterm
276 %printer {} <type1> <type3>
277 %destructor {} <type2> <type4>
278 %%
279 nterm: term { $$ = $1; };
280
281 3.28-34: warning: type <type3> is used, but is not associated to any symbol
282 4.28-34: warning: type <type4> is used, but is not associated to any symbol
283
31557b9e 284*** Undefined but unused symbols
b921d92f 285
31557b9e
AD
286 Bison used to raise an error for undefined symbols that are not used in
287 the grammar. This is now only a warning.
b921d92f
VS
288
289 %printer {} symbol1
290 %destructor {} symbol2
31557b9e 291 %type <type> symbol3
b921d92f
VS
292 %%
293 exp: "a";
294
73370a9d 295*** Useless destructors or printers
ea9a35c6
VS
296
297 Bison now warns about useless destructors or printers. In the following
298 example, the printer for <type1>, and the destructor for <type2> are
299 useless: all symbols of <type1> (token1) already have a printer, and all
300 symbols of type <type2> (token2) already have a destructor.
301
302 %token <type1> token1
303 <type2> token2
304 <type3> token3
305 <type4> token4
306 %printer {} token1 <type1> <type3>
307 %destructor {} token2 <type2> <type4>
308
d87ea54c
AD
309*** Conflicts
310
311 The warnings and error messages about shift/reduce and reduce/reduce
312 conflicts have been normalized. For instance on the following foo.y file:
313
314 %glr-parser
315 %%
316 exp: exp '+' exp | '0' | '0';
317
318 compare the previous version of bison:
319
320 $ bison foo.y
321 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
322 $ bison -Werror foo.y
323 bison: warnings being treated as errors
324 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
325
326 with the new behavior:
327
328 $ bison foo.y
329 foo.y: warning: 1 shift/reduce conflict [-Wconflicts-sr]
330 foo.y: warning: 2 reduce/reduce conflicts [-Wconflicts-rr]
331 $ bison -Werror foo.y
9503b0a4
TR
332 foo.y: error: 1 shift/reduce conflict [-Werror=conflicts-sr]
333 foo.y: error: 2 reduce/reduce conflicts [-Werror=conflicts-rr]
d87ea54c
AD
334
335 When %expect or %expect-rr is used, such as with bar.y:
336
337 %expect 0
338 %glr-parser
339 %%
340 exp: exp '+' exp | '0' | '0';
341
342 Former behavior:
343
344 $ bison bar.y
345 bar.y: conflicts: 1 shift/reduce, 2 reduce/reduce
346 bar.y: expected 0 shift/reduce conflicts
347 bar.y: expected 0 reduce/reduce conflicts
348
349 New one:
350
351 $ bison bar.y
9503b0a4
TR
352 bar.y: error: shift/reduce conflicts: 1 found, 0 expected
353 bar.y: error: reduce/reduce conflicts: 2 found, 0 expected
d87ea54c 354
f24695ef
AD
355** Incompatibilities with POSIX Yacc
356
357 The 'yacc' category is no longer part of '-Wall', enable it explicitly
358 with '-Wyacc'.
359
2055a44e
AD
360** Additional yylex/yyparse arguments
361
6dc4663d
AD
362 The new directive %param declares additional arguments to both yylex and
363 yyparse. The %lex-param, %parse-param, and %param directives support one
364 or more arguments. Instead of
2055a44e 365
e436fa67
AD
366 %lex-param {arg1_type *arg1}
367 %lex-param {arg2_type *arg2}
368 %parse-param {arg1_type *arg1}
369 %parse-param {arg2_type *arg2}
2055a44e
AD
370
371 one may now declare
372
e436fa67 373 %param {arg1_type *arg1} {arg2_type *arg2}
2055a44e 374
630a0218
AD
375** Types of values for %define variables
376
377 Bison used to make no difference between '%define foo bar' and '%define
378 foo "bar"'. The former is now called a 'keyword value', and the latter a
379 'string value'. A third kind was added: 'code values', such as '%define
380 foo {bar}'.
381
382 Keyword variables are used for fixed value sets, e.g.,
383
384 %define lr.type lalr
385
386 Code variables are used for value in the target language, e.g.,
387
388 %define api.value.type {struct semantic_type}
389
390 String variables are used remaining cases, e.g. file names.
391
2a6b66c5 392** Variable api.token.prefix
99c08fb6 393
2a6b66c5 394 The variable api.token.prefix changes the way tokens are identified in
99c08fb6
AD
395 the generated files. This is especially useful to avoid collisions
396 with identifiers in the target language. For instance
397
e436fa67 398 %token FILE for ERROR
630a0218 399 %define api.token.prefix {TOK_}
e436fa67
AD
400 %%
401 start: FILE for ERROR;
99c08fb6
AD
402
403 will generate the definition of the symbols TOK_FILE, TOK_for, and
404 TOK_ERROR in the generated sources. In particular, the scanner must
405 use these prefixed token names, although the grammar itself still
406 uses the short names (as in the sample rule given above).
407
6574576c
AD
408** Variable api.value.type
409
410 This new %define variable supersedes the #define macro YYSTYPE. The use
411 of YYSTYPE is discouraged. In particular, #defining YYSTYPE *and* either
412 using %union or %defining api.value.type results in undefined behavior.
413
414 Either define api.value.type, or use "%union":
415
416 %union
417 {
418 int ival;
419 char *sval;
420 }
421 %token <ival> INT "integer"
422 %token <sval> STRING "string"
423 %printer { fprintf (yyo, "%d", $$); } <ival>
424 %destructor { free ($$); } <sval>
425
426 /* In yylex(). */
427 yylval.ival = 42; return INT;
428 yylval.sval = "42"; return STRING;
429
630a0218
AD
430 The %define variable api.value.type supports both keyword and code values.
431
432 The keyword value 'union' means that the user provides genuine types, not
435575cb 433 union member names such as "ival" and "sval" above (WARNING: will fail if
1fa19a76 434 -y/--yacc/%yacc is enabled).
6574576c 435
435575cb 436 %define api.value.type union
6574576c
AD
437 %token <int> INT "integer"
438 %token <char *> STRING "string"
439 %printer { fprintf (yyo, "%d", $$); } <int>
440 %destructor { free ($$); } <char *>
441
442 /* In yylex(). */
443 yylval.INT = 42; return INT;
444 yylval.STRING = "42"; return STRING;
445
435575cb
AD
446 The keyword value variant is somewhat equivalent, but for C++ special
447 provision is made to allow classes to be used (more about this below).
6574576c 448
435575cb 449 %define api.value.type variant
6574576c
AD
450 %token <int> INT "integer"
451 %token <std::string> STRING "string"
452
630a0218 453 Code values (in braces) denote user defined types. This is where YYSTYPE
435575cb 454 used to be used.
6574576c
AD
455
456 %code requires
457 {
458 struct my_value
459 {
460 enum
461 {
462 is_int, is_string
463 } kind;
464 union
465 {
466 int ival;
467 char *sval;
468 } u;
469 };
470 }
435575cb 471 %define api.value.type {struct my_value}
6574576c
AD
472 %token <u.ival> INT "integer"
473 %token <u.sval> STRING "string"
474 %printer { fprintf (yyo, "%d", $$); } <u.ival>
475 %destructor { free ($$); } <u.sval>
476
477 /* In yylex(). */
478 yylval.u.ival = 42; return INT;
479 yylval.u.sval = "42"; return STRING;
480
31b850d2
AD
481** Variable parse.error
482
1f77b2e0
AD
483 This variable controls the verbosity of error messages. The use of the
484 %error-verbose directive is deprecated in favor of "%define parse.error
485 verbose".
31b850d2 486
c21e515e
AD
487** Renamed %define variables
488
489 The following variables have been renamed for consistency. Backward
490 compatibility is ensured, but upgrading is recommended.
491
492 lr.default-reductions -> lr.default-reduction
493 lr.keep-unreachable-states -> lr.keep-unreachable-state
494 namespace -> api.namespace
495 stype -> api.value.type
496
ca2a6d15
PH
497** Semantic predicates
498
597afd73
AD
499 Contributed by Paul Hilfinger.
500
1f77b2e0
AD
501 The new, experimental, semantic-predicate feature allows actions of the
502 form "%?{ BOOLEAN-EXPRESSION }", which cause syntax errors (as for
ca2a6d15 503 YYERROR) if the expression evaluates to 0, and are evaluated immediately
1f77b2e0
AD
504 in GLR parsers, rather than being deferred. The result is that they allow
505 the programmer to prune possible parses based on the values of run-time
506 expressions.
ca2a6d15 507
d1400569
AD
508** The directive %expect-rr is now an error in non GLR mode
509
510 It used to be an error only if used in non GLR mode, _and_ if there are
511 reduce/reduce conflicts.
512
5202b6ac 513** Tokens are numbered in their order of appearance
93561c21 514
5202b6ac
VT
515 Contributed by Valentin Tolmer.
516
517 With '%token A B', A had a number less than the one of B. However,
518 precedence declarations used to generate a reversed order. This is now
519 fixed, and introducing tokens with any of %token, %left, %right,
520 %precedence, or %nonassoc yields the same result.
521
522 When mixing declarations of tokens with a litteral character (e.g., 'a')
523 or with an identifier (e.g., B) in a precedence declaration, Bison
524 numbered the litteral characters first. For example
525
526 %right A B 'c' 'd'
527
528 would lead to the tokens declared in this order: 'c' 'd' A B. Again, the
529 input order is now preserved.
530
531 These changes were made so that one can remove useless precedence and
532 associativity declarations (i.e., map %nonassoc, %left or %right to
533 %precedence, or to %token) and get exactly the same output.
93561c21 534
cc2235ac
VT
535** Useless precedence and associativity
536
d2f9ae18
AD
537 Contributed by Valentin Tolmer.
538
1282c124 539 When developing and maintaining a grammar, useless associativity and
cc2235ac
VT
540 precedence directives are common. They can be a nuisance: new ambiguities
541 arising are sometimes masked because their conflicts are resolved due to
542 the extra precedence or associativity information. Furthermore, it can
543 hinder the comprehension of a new grammar: one will wonder about the role
544 of a precedence, where in fact it is useless. The following changes aim
545 at detecting and reporting these extra directives.
546
547*** Precedence warning category
548
549 A new category of warning, -Wprecedence, was introduced. It flags the
550 useless precedence and associativity directives.
551
552*** Useless associativity
553
554 Bison now warns about symbols with a declared associativity that is never
555 used to resolve conflicts. In that case, using %precedence is sufficient;
556 the parsing tables will remain unchanged. Solving these warnings may raise
557 useless precedence warnings, as the symbols no longer have associativity.
558 For example:
559
560 %left '+'
561 %left '*'
562 %%
563 exp:
1282c124
AD
564 "number"
565 | exp '+' "number"
cc2235ac
VT
566 | exp '*' exp
567 ;
568
569 will produce a
570
571 warning: useless associativity for '+', use %precedence [-Wprecedence]
572 %left '+'
573 ^^^
574
575*** Useless precedence
576
577 Bison now warns about symbols with a declared precedence and no declared
578 associativity (i.e., declared with %precedence), and whose precedence is
579 never used. In that case, the symbol can be safely declared with %token
580 instead, without modifying the parsing tables. For example:
581
582 %precedence '='
583 %%
1282c124 584 exp: "var" '=' "number";
cc2235ac
VT
585
586 will produce a
587
588 warning: useless precedence for '=' [-Wprecedence]
589 %precedence '='
590 ^^^
591
592*** Useless precedence and associativity
593
594 In case of both useless precedence and associativity, the issue is flagged
595 as follows:
596
597 %nonassoc '='
598 %%
1282c124 599 exp: "var" '=' "number";
cc2235ac
VT
600
601 The warning is:
602
603 warning: useless precedence and associativity for '=' [-Wprecedence]
604 %nonassoc '='
605 ^^^
09add9c2
AD
606
607** Empty rules
608
6240346a
AD
609 With help from Joel E. Denny and Gabriel Rassoul.
610
09add9c2
AD
611 Empty rules (i.e., with an empty right-hand side) can now be explicitly
612 marked by the new %empty directive. Using %empty on a non-empty rule is
613 an error. The new -Wempty-rule warning reports empty rules without
614 %empty. On the following grammar:
615
616 %%
617 s: a b c;
618 a: ;
619 b: %empty;
620 c: 'a' %empty;
621
622 bison reports:
623
624 3.4-5: warning: empty rule without %empty [-Wempty-rule]
625 a: {}
626 ^^
627 5.8-13: error: %empty on non-empty rule
628 c: 'a' %empty {};
629 ^^^^^^
cc2235ac 630
c21e515e
AD
631** Java skeleton improvements
632
c21e515e
AD
633 The constants for token names were moved to the Lexer interface. Also, it
634 is possible to add code to the parser's constructors using "%code init"
635 and "%define init_throws".
aa94def1
DH
636 Contributed by Paolo Bonzini.
637
de1a2f20
AD
638 The Java skeleton now supports push parsing.
639 Contributed by Dennis Heimbigner.
640
c21e515e
AD
641** C++ skeletons improvements
642
643*** The parser header is no longer mandatory (lalr1.cc, glr.cc)
644
645 Using %defines is now optional. Without it, the needed support classes
646 are defined in the generated parser, instead of additional files (such as
647 location.hh, position.hh and stack.hh).
648
649*** Locations are no longer mandatory (lalr1.cc, glr.cc)
650
651 Both lalr1.cc and glr.cc no longer require %location.
652
653*** syntax_error exception (lalr1.cc)
654
655 The C++ parser features a syntax_error exception, which can be
656 thrown from the scanner or from user rules to raise syntax errors.
657 This facilitates reporting errors caught in sub-functions (e.g.,
658 rejecting too large integral literals from a conversion function
659 used by the scanner, or rejecting invalid combinations from a
660 factory invoked by the user actions).
661
662*** %define api.value.type variant
663
664 This is based on a submission from Michiel De Wilde. With help
665 from Théophile Ranquet.
666
667 In this mode, complex C++ objects can be used as semantic values. For
668 instance:
669
670 %token <::std::string> TEXT;
671 %token <int> NUMBER;
672 %token SEMICOLON ";"
673 %type <::std::string> item;
674 %type <::std::list<std::string>> list;
675 %%
676 result:
677 list { std::cout << $1 << std::endl; }
678 ;
679
680 list:
6240346a 681 %empty { /* Generates an empty string list. */ }
c21e515e
AD
682 | list item ";" { std::swap ($$, $1); $$.push_back ($2); }
683 ;
684
685 item:
686 TEXT { std::swap ($$, $1); }
687 | NUMBER { $$ = string_cast ($1); }
688 ;
689
690*** %define api.token.constructor
691
692 When variants are enabled, Bison can generate functions to build the
693 tokens. This guarantees that the token type (e.g., NUMBER) is consistent
694 with the semantic value (e.g., int):
695
696 parser::symbol_type yylex ()
697 {
698 parser::location_type loc = ...;
699 ...
700 return parser::make_TEXT ("Hello, world!", loc);
701 ...
702 return parser::make_NUMBER (42, loc);
703 ...
704 return parser::make_SEMICOLON (loc);
705 ...
706 }
707
75ae8299
AD
708*** C++ locations
709
710 There are operator- and operator-= for 'location'. Negative line/column
711 increments can no longer underflow the resulting value.
712
1f5542fe 713* Noteworthy changes in release 2.7.1 (2013-04-15) [stable]
cc8962bd 714
80a2826e
AD
715** Bug fixes
716
717*** Fix compiler attribute portability (yacc.c)
718
719 With locations enabled, __attribute__ was used unprotected.
0a7b8559 720
e83be476
AD
721*** Fix some compiler warnings (lalr1.cc)
722
c13bb348 723* Noteworthy changes in release 2.7 (2012-12-12) [stable]
effd30c0 724
edf9a06f 725** Bug fixes
7bada535 726
edf9a06f 727 Warnings about uninitialized yylloc in yyparse have been fixed.
7bada535 728
1127a75a
AD
729 Restored C90 compliance (yet no report was ever made).
730
d4fe9e88 731** Diagnostics are improved
7bada535 732
8458a411
AD
733 Contributed by Théophile Ranquet.
734
d4fe9e88 735*** Changes in the format of error messages
7bada535 736
d4fe9e88 737 This used to be the format of many error reports:
1f1bd572 738
d4fe9e88
AD
739 input.y:2.7-12: %type redeclaration for exp
740 input.y:1.7-12: previous declaration
1f1bd572 741
d4fe9e88 742 It is now:
1f1bd572 743
d4fe9e88
AD
744 input.y:2.7-12: error: %type redeclaration for exp
745 input.y:1.7-12: previous declaration
cbaea010 746
d4fe9e88 747*** New format for error reports: carets
cbaea010 748
d4fe9e88 749 Caret errors have been added to Bison:
cbaea010 750
d4fe9e88
AD
751 input.y:2.7-12: error: %type redeclaration for exp
752 %type <sval> exp
753 ^^^^^^
754 input.y:1.7-12: previous declaration
755 %type <ival> exp
756 ^^^^^^
cbaea010 757
d4fe9e88 758 or
cbaea010 759
7bada535 760 input.y:3.20-23: error: ambiguous reference: '$exp'
fb6040f0 761 exp: exp '+' exp { $exp = $1 + $3; };
7bada535 762 ^^^^
fb6040f0
TR
763 input.y:3.1-3: refers to: $exp at $$
764 exp: exp '+' exp { $exp = $1 + $3; };
765 ^^^
766 input.y:3.6-8: refers to: $exp at $1
767 exp: exp '+' exp { $exp = $1 + $3; };
768 ^^^
769 input.y:3.14-16: refers to: $exp at $3
770 exp: exp '+' exp { $exp = $1 + $3; };
771 ^^^
7bada535 772
1282c124
AD
773 The default behavior for now is still not to display these unless
774 explicitly asked with -fcaret (or -fall). However, in a later release, it
d4fe9e88
AD
775 will be made the default behavior (but may still be deactivated with
776 -fno-caret).
d3e4409a 777
1f1bd572 778** New value for %define variable: api.pure full
d3e4409a 779
1f1bd572 780 The %define variable api.pure requests a pure (reentrant) parser. However,
d4fe9e88
AD
781 for historical reasons, using it in a location-tracking Yacc parser
782 resulted in a yyerror function that did not take a location as a
783 parameter. With this new value, the user may request a better pure parser,
784 where yyerror does take a location as a parameter (in location-tracking
785 parsers).
1f1bd572
TR
786
787 The use of "%define api.pure true" is deprecated in favor of this new
788 "%define api.pure full".
d3e4409a 789
7287be84 790** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java)
db8ab2be
AD
791
792 The %define variable api.location.type defines the name of the type to use
793 for locations. When defined, Bison no longer generates the position.hh
794 and location.hh files, nor does the parser will include them: the user is
795 then responsible to define her type.
796
797 This can be used in programs with several parsers to factor their location
7287be84
AD
798 and position files: let one of them generate them, and the others just use
799 them.
db8ab2be
AD
800
801 This feature was actually introduced, but not documented, in Bison 2.5,
802 under the name "location_type" (which is maintained for backward
803 compatibility).
804
7287be84
AD
805 For consistency, lalr1.java's %define variables location_type and
806 position_type are deprecated in favor of api.location.type and
807 api.position.type.
808
d4fe9e88
AD
809** Exception safety (lalr1.cc)
810
811 The parse function now catches exceptions, uses the %destructors to
812 release memory (the lookahead symbol and the symbols pushed on the stack)
813 before re-throwing the exception.
814
815 This feature is somewhat experimental. User feedback would be
816 appreciated.
817
9c16d399 818** Graph improvements in DOT and XSLT
fc4fdd62 819
8458a411
AD
820 Contributed by Théophile Ranquet.
821
fc4fdd62
TR
822 The graphical presentation of the states is more readable: their shape is
823 now rectangular, the state number is clearly displayed, and the items are
824 numbered and left-justified.
825
826 The reductions are now explicitly represented as transitions to other
827 diamond shaped nodes.
828
9c16d399
TR
829 These changes are present in both --graph output and xml2dot.xsl XSLT
830 processing, with minor (documented) differences.
831
d4fe9e88 832** %language is no longer an experimental feature.
fb4c8a7c 833
d4fe9e88
AD
834 The introduction of this feature, in 2.4, was four years ago. The
835 --language option and the %language directive are no longer experimental.
fb4c8a7c 836
53e2cd1e
AD
837** Documentation
838
839 The sections about shift/reduce and reduce/reduce conflicts resolution
840 have been fixed and extended.
9d2423f5 841
d4fe9e88
AD
842 Although introduced more than four years ago, XML and Graphviz reports
843 were not properly documented.
844
be22823e
AD
845 The translation of mid-rule actions is now described.
846
9d3f7eaf 847* Noteworthy changes in release 2.6.5 (2012-11-07) [stable]
6f1360bd 848
a68b1f23
AD
849 We consider compiler warnings about Bison generated parsers to be bugs.
850 Rather than working around them in your own project, please consider
851 reporting them to us.
852
853** Bug fixes
854
855 Warnings about uninitialized yylval and/or yylloc for push parsers with a
856 pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to
857 3.2.
858
859 Other issues in the test suite have been addressed.
6f1360bd 860
1282c124 861 Null characters are correctly displayed in error messages.
95066e92 862
a1d1ab50
AD
863 When possible, yylloc is correctly initialized before calling yylex. It
864 is no longer necessary to initialize it in the %initial-action.
865
0ac15849 866* Noteworthy changes in release 2.6.4 (2012-10-23) [stable]
a4eb820f 867
468455e1 868 Bison 2.6.3's --version was incorrect. This release fixes this issue.
a4eb820f 869
6eb8f74f 870* Noteworthy changes in release 2.6.3 (2012-10-22) [stable]
933ec544 871
6b4cb804
AD
872** Bug fixes
873
a1a77e1f 874 Bugs and portability issues in the test suite have been fixed.
6b4cb804
AD
875
876 Some errors in translations have been addressed, and --help now directs
877 users to the appropriate place to report them.
878
879 Stray Info files shipped by accident are removed.
880
881 Incorrect definitions of YY_, issued by yacc.c when no parser header is
882 generated, are removed.
9c26b8fc 883
a2b3f101
TR
884 All the generated headers are self-contained.
885
c9d5bcc9
AD
886** Header guards (yacc.c, glr.c, glr.cc)
887
888 In order to avoid collisions, the header guards are now
889 YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>.
890 For instance the header generated from
891
892 %define api.prefix "calc"
893 %defines "lib/parse.h"
894
895 will use YY_CALC_LIB_PARSE_H_INCLUDED as guard.
896
c12c4c50 897** Fix compiler warnings in the generated parser (yacc.c, glr.c)
321d3e35
AD
898
899 The compilation of pure parsers (%define api.pure) can trigger GCC
900 warnings such as:
901
902 input.c: In function 'yyparse':
903 input.c:1503:12: warning: 'yylval' may be used uninitialized in this
904 function [-Wmaybe-uninitialized]
905 *++yyvsp = yylval;
906 ^
907
908 This is now fixed; pragmas to avoid these warnings are no longer needed.
909
c12c4c50
AD
910 Warnings from clang ("equality comparison with extraneous parentheses" and
911 "function declared 'noreturn' should not return") have also been
912 addressed.
913
e1eeecd3 914* Noteworthy changes in release 2.6.2 (2012-08-03) [stable]
9c26b8fc 915
43ca8040
AD
916** Bug fixes
917
918 Buffer overruns, complaints from Flex, and portability issues in the test
919 suite have been fixed.
920
c9d546b2
AD
921** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc)
922
923 Trailing end-of-lines in %parse-param or %lex-param would result in
924 invalid C++. This is fixed.
9c26b8fc 925
dcd5344d
AD
926** Spurious spaces and end-of-lines
927
928 The generated files no longer end (nor start) with empty lines.
929
77b214ef 930* Noteworthy changes in release 2.6.1 (2012-07-30) [stable]
a4107f24 931
8617d87e
AD
932 Bison no longer executes user-specified M4 code when processing a grammar.
933
e20e6a50
AD
934** Future Changes
935
936 In addition to the removal of the features announced in Bison 2.6, the
937 next major release will remove the "Temporary hack for adding a semicolon
938 to the user action", as announced in the release 2.5. Instead of:
939
940 exp: exp "+" exp { $$ = $1 + $3 };
941
942 write:
943
944 exp: exp "+" exp { $$ = $1 + $3; };
945
8617d87e
AD
946** Bug fixes
947
0e164d43
AD
948*** Type names are now properly escaped.
949
950*** glr.cc: set_debug_level and debug_level work as expected.
a4107f24 951
26313726
AD
952*** Stray @ or $ in actions
953
954 While Bison used to warn about stray $ or @ in action rules, it did not
955 for other actions such as printers, destructors, or initial actions. It
956 now does.
957
cd735a8c 958** Type names in actions
4982f078
AD
959
960 For consistency with rule actions, it is now possible to qualify $$ by a
cd735a8c 961 type-name in destructors, printers, and initial actions. For instance:
4982f078
AD
962
963 %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>;
964
965 will display two values for each typed and untyped symbol (provided
cd735a8c 966 that YYSTYPE has both "ival" and "fval" fields).
60aa04a2 967
1505e8bb 968* Noteworthy changes in release 2.6 (2012-07-19) [stable]
0f11eec2 969
d0a30438 970** Future changes
9553083c 971
55d1006f
AD
972 The next major release of Bison will drop support for the following
973 deprecated features. Please report disagreements to bug-bison@gnu.org.
0f11eec2 974
aaf61036 975*** K&R C parsers
55d1006f
AD
976
977 Support for generating parsers in K&R C will be removed. Parsers
242cc08e 978 generated for C support ISO C90, and are tested with ISO C99 and ISO C11
55d1006f
AD
979 compilers.
980
258cddbc 981*** Features deprecated since Bison 1.875
0f11eec2 982
258cddbc
AD
983 The definitions of yystype and yyltype will be removed; use YYSTYPE and
984 YYLTYPE.
0f11eec2 985
258cddbc
AD
986 YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and
987 %lex-param, will no longer be supported.
988
989 Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use
990 %error-verbose.
55d1006f
AD
991
992*** The generated header will be included (yacc.c)
0f11eec2
AD
993
994 Instead of duplicating the content of the generated header (definition of
55d1006f
AD
995 YYSTYPE, yyparse declaration etc.), the generated parser will include it,
996 as is already the case for GLR or C++ parsers. This change is deferred
997 because existing versions of ylwrap (e.g., Automake 1.12.1) do not support
998 it.
0f11eec2 999
c2425191 1000** Generated Parser Headers
56ca3d8f 1001
258cddbc 1002*** Guards (yacc.c, glr.c, glr.cc)
c3e9f08f
AD
1003
1004 The generated headers are now guarded, as is already the case for C++
242cc08e 1005 parsers (lalr1.cc). For instance, with --defines=foo.h:
c3e9f08f 1006
e29f0771
AD
1007 #ifndef YY_FOO_H
1008 # define YY_FOO_H
1009 ...
1010 #endif /* !YY_FOO_H */
c3e9f08f 1011
258cddbc 1012*** New declarations (yacc.c, glr.c)
56ca3d8f
AD
1013
1014 The generated header now declares yydebug and yyparse. Both honor
1015 --name-prefix=bar_, and yield
1016
e29f0771 1017 int bar_parse (void);
56ca3d8f
AD
1018
1019 rather than
1020
e29f0771
AD
1021 #define yyparse bar_parse
1022 int yyparse (void);
56ca3d8f
AD
1023
1024 in order to facilitate the inclusion of several parser headers inside a
1025 single compilation unit.
c3e9f08f 1026
258cddbc
AD
1027*** Exported symbols in C++
1028
1029 The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the
1030 header, are removed, as they prevent the possibility of including several
1031 generated headers from a single compilation unit.
1032
694af10c
AD
1033*** YYLSP_NEEDED
1034
1035 For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no
1036 longer defined.
1037
4b3847c3
AD
1038** New %define variable: api.prefix
1039
1040 Now that the generated headers are more complete and properly protected
1041 against multiple inclusions, constant names, such as YYSTYPE are a
1042 problem. While yyparse and others are properly renamed by %name-prefix,
1043 YYSTYPE, YYDEBUG and others have never been affected by it. Because it
1044 would introduce backward compatibility issues in projects not expecting
1045 YYSTYPE to be renamed, instead of changing the behavior of %name-prefix,
1046 it is deprecated in favor of a new %define variable: api.prefix.
1047
1048 The following examples compares both:
1049
1050 %name-prefix "bar_" | %define api.prefix "bar_"
1051 %token <ival> FOO %token <ival> FOO
1052 %union { int ival; } %union { int ival; }
1053 %% %%
1054 exp: 'a'; exp: 'a';
1055
1056 bison generates:
1057
1058 #ifndef BAR_FOO_H #ifndef BAR_FOO_H
1059 # define BAR_FOO_H # define BAR_FOO_H
1060
1061 /* Enabling traces. */ /* Enabling traces. */
5f108727
AD
1062 # ifndef YYDEBUG | # ifndef BAR_DEBUG
1063 > # if defined YYDEBUG
1064 > # if YYDEBUG
1065 > # define BAR_DEBUG 1
1066 > # else
1067 > # define BAR_DEBUG 0
1068 > # endif
1069 > # else
1070 # define YYDEBUG 0 | # define BAR_DEBUG 0
1071 > # endif
1072 # endif | # endif
1073
1074 # if YYDEBUG | # if BAR_DEBUG
4b3847c3
AD
1075 extern int bar_debug; extern int bar_debug;
1076 # endif # endif
1077
1078 /* Tokens. */ /* Tokens. */
1079 # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE
1080 # define YYTOKENTYPE | # define BAR_TOKENTYPE
1081 enum yytokentype { | enum bar_tokentype {
1082 FOO = 258 FOO = 258
1083 }; };
1084 # endif # endif
1085
1086 #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \
1087 && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED
1088 typedef union YYSTYPE | typedef union BAR_STYPE
1089 { {
1090 int ival; int ival;
1091 } YYSTYPE; | } BAR_STYPE;
1092 # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1
1093 #endif #endif
1094
1095 extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval;
1096
1097 int bar_parse (void); int bar_parse (void);
1098
1099 #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */
1100
dfaac272 1101* Noteworthy changes in release 2.5.1 (2012-06-05) [stable]
df6e3db0 1102
debe2c03 1103** Future changes:
765e1bd4 1104
e4ab1254 1105 The next major release will drop support for generating parsers in K&R C.
041308d0 1106
466b4cf2 1107** yacc.c: YYBACKUP works as expected.
ef51bfa7 1108
d834eca0 1109** glr.c improvements:
041308d0 1110
d834eca0 1111*** Location support is eliminated when not requested:
041308d0 1112
e4ab1254
AD
1113 GLR parsers used to include location-related code even when locations were
1114 not requested, and therefore not even usable.
378e917c 1115
d834eca0 1116*** __attribute__ is preserved:
d115aad9 1117
e4ab1254
AD
1118 __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e.,
1119 when -std is passed to GCC).
041308d0 1120
466b4cf2 1121** lalr1.java: several fixes:
041308d0 1122
e4ab1254
AD
1123 The Java parser no longer throws ArrayIndexOutOfBoundsException if the
1124 first token leads to a syntax error. Some minor clean ups.
041308d0 1125
22172d47 1126** Changes for C++:
ef51bfa7 1127
22172d47 1128*** C++11 compatibility:
ef51bfa7 1129
e4ab1254
AD
1130 C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L
1131 or higher.
936c88d1 1132
22172d47
AD
1133*** Header guards
1134
1135 The header files such as "parser.hh", "location.hh", etc. used a constant
1136 name for preprocessor guards, for instance:
1137
e29f0771
AD
1138 #ifndef BISON_LOCATION_HH
1139 # define BISON_LOCATION_HH
1140 ...
1141 #endif // !BISON_LOCATION_HH
22172d47
AD
1142
1143 The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower
1144 case characters are converted to upper case, and series of
1145 non-alphanumerical characters are converted to an underscore.
1146
1147 With "bison -o lang++/parser.cc", "location.hh" would now include:
1148
e29f0771
AD
1149 #ifndef YY_LANG_LOCATION_HH
1150 # define YY_LANG_LOCATION_HH
1151 ...
1152 #endif // !YY_LANG_LOCATION_HH
22172d47
AD
1153
1154*** C++ locations:
936c88d1 1155
e4ab1254
AD
1156 The position and location constructors (and their initialize methods)
1157 accept new arguments for line and column. Several issues in the
1158 documentation were fixed.
936c88d1 1159
466b4cf2
AD
1160** liby is no longer asking for "rpl_fprintf" on some platforms.
1161
7e508a2b
AD
1162** Changes in the manual:
1163
1164*** %printer is documented
1165
e4ab1254
AD
1166 The "%printer" directive, supported since at least Bison 1.50, is finally
1167 documented. The "mfcalc" example is extended to demonstrate it.
7e508a2b 1168
e4ab1254
AD
1169 For consistency with the C skeletons, the C++ parsers now also support
1170 "yyoutput" (as an alias to "debug_stream ()").
7e508a2b
AD
1171
1172*** Several improvements have been made:
466b4cf2 1173
e4ab1254
AD
1174 The layout for grammar excerpts was changed to a more compact scheme.
1175 Named references are motivated. The description of the automaton
1176 description file (*.output) is updated to the current format. Incorrect
1177 index entries were fixed. Some other errors were fixed.
466b4cf2 1178
86b08b49
AD
1179** Building bison:
1180
1181*** Conflicting prototypes with recent/modified Flex.
466b4cf2 1182
e4ab1254
AD
1183 Fixed build problems with the current, unreleased, version of Flex, and
1184 some modified versions of 2.5.35, which have modified function prototypes.
466b4cf2 1185
8ef26c2a
AD
1186*** Warnings during the build procedure have been eliminated.
1187
1188*** Several portability problems in the test suite have been fixed:
466b4cf2 1189
e4ab1254
AD
1190 This includes warnings with some compilers, unexpected behavior of tools
1191 such as diff, warning messages from the test suite itself, etc.
466b4cf2 1192
91aadcc7 1193*** The install-pdf target works properly:
8ef26c2a 1194
e4ab1254
AD
1195 Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer
1196 halts in the middle of its course.
8ef26c2a 1197
28801043 1198* Changes in version 2.5 (2011-05-14):
50cca368 1199
82f3355e
JD
1200** Grammar symbol names can now contain non-initial dashes:
1201
1202 Consistently with directives (such as %error-verbose) and with
1203 %define variables (e.g. push-pull), grammar symbol names may contain
1204 dashes in any position except the beginning. This is a GNU
1205 extension over POSIX Yacc. Thus, use of this extension is reported
1206 by -Wyacc and rejected in Yacc mode (--yacc).
1207
f1b238df 1208** Named references:
66381412
AR
1209
1210 Historically, Yacc and Bison have supported positional references
1211 ($n, $$) to allow access to symbol values from inside of semantic
1212 actions code.
1213
1214 Starting from this version, Bison can also accept named references.
1215 When no ambiguity is possible, original symbol names may be used
1216 as named references:
1217
4b568fc0 1218 if_stmt : "if" cond_expr "then" then_stmt ';'
66381412
AR
1219 { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }
1220
1221 In the more common case, explicit names may be declared:
1222
4b568fc0 1223 stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';'
66381412
AR
1224 { $res = mk_if_stmt($cond, $then, $else); }
1225
5b1ff423 1226 Location information is also accessible using @name syntax. When
66381412
AR
1227 accessing symbol names containing dots or dashes, explicit bracketing
1228 ($[sym.1]) must be used.
1229
5b1ff423 1230 These features are experimental in this version. More user feedback
66381412 1231 will help to stabilize them.
2bd435c3 1232 Contributed by Alex Rozenman.
66381412 1233
f1b238df 1234** IELR(1) and canonical LR(1):
eb45ef3b
JD
1235
1236 IELR(1) is a minimal LR(1) parser table generation algorithm. That
1237 is, given any context-free grammar, IELR(1) generates parser tables
7262f54f 1238 with the full language-recognition power of canonical LR(1) but with
f1b238df
JD
1239 nearly the same number of parser states as LALR(1). This reduction
1240 in parser states is often an order of magnitude. More importantly,
eb45ef3b
JD
1241 because canonical LR(1)'s extra parser states may contain duplicate
1242 conflicts in the case of non-LR(1) grammars, the number of conflicts
1243 for IELR(1) is often an order of magnitude less as well. This can
1244 significantly reduce the complexity of developing of a grammar.
1245
1246 Bison can now generate IELR(1) and canonical LR(1) parser tables in
1247 place of its traditional LALR(1) parser tables, which remain the
1248 default. You can specify the type of parser tables in the grammar
1249 file with these directives:
1250
cf499cff
JD
1251 %define lr.type lalr
1252 %define lr.type ielr
1253 %define lr.type canonical-lr
eb45ef3b 1254
7fceb615 1255 The default-reduction optimization in the parser tables can also be
e4ab1254
AD
1256 adjusted using "%define lr.default-reductions". For details on both
1257 of these features, see the new section "Tuning LR" in the Bison
7fceb615 1258 manual.
eb45ef3b
JD
1259
1260 These features are experimental. More user feedback will help to
1261 stabilize them.
1262
8458a411
AD
1263** LAC (Lookahead Correction) for syntax error handling
1264
1265 Contributed by Joel E. Denny.
fcf834f9
JD
1266
1267 Canonical LR, IELR, and LALR can suffer from a couple of problems
1268 upon encountering a syntax error. First, the parser might perform
1269 additional parser stack reductions before discovering the syntax
7fceb615 1270 error. Such reductions can perform user semantic actions that are
fcf834f9
JD
1271 unexpected because they are based on an invalid token, and they
1272 cause error recovery to begin in a different syntactic context than
1273 the one in which the invalid token was encountered. Second, when
7fceb615 1274 verbose error messages are enabled (with %error-verbose or the
e4ab1254 1275 obsolete "#define YYERROR_VERBOSE"), the expected token list in the
7fceb615
JD
1276 syntax error message can both contain invalid tokens and omit valid
1277 tokens.
fcf834f9
JD
1278
1279 The culprits for the above problems are %nonassoc, default
1280 reductions in inconsistent states, and parser state merging. Thus,
1281 IELR and LALR suffer the most. Canonical LR can suffer only if
1282 %nonassoc is used or if default reductions are enabled for
1283 inconsistent states.
1284
7fceb615
JD
1285 LAC is a new mechanism within the parsing algorithm that solves
1286 these problems for canonical LR, IELR, and LALR without sacrificing
1287 %nonassoc, default reductions, or state merging. When LAC is in
1288 use, canonical LR and IELR behave almost exactly the same for both
1289 syntactically acceptable and syntactically unacceptable input.
fcf834f9
JD
1290 While LALR still does not support the full language-recognition
1291 power of canonical LR and IELR, LAC at least enables LALR's syntax
1292 error handling to correctly reflect LALR's language-recognition
1293 power.
1294
1295 Currently, LAC is only supported for deterministic parsers in C.
1296 You can enable LAC with the following directive:
1297
1298 %define parse.lac full
1299
e4ab1254 1300 See the new section "LAC" in the Bison manual for additional
7fceb615 1301 details including a few caveats.
fcf834f9
JD
1302
1303 LAC is an experimental feature. More user feedback will help to
1304 stabilize it.
1305
d397d9f0 1306** %define improvements:
cf499cff 1307
f1b238df 1308*** Can now be invoked via the command line:
50cca368 1309
de5ab940 1310 Each of these command-line options
50cca368 1311
de5ab940
JD
1312 -D NAME[=VALUE]
1313 --define=NAME[=VALUE]
1314
1315 -F NAME[=VALUE]
1316 --force-define=NAME[=VALUE]
50cca368
JD
1317
1318 is equivalent to this grammar file declaration
1319
de5ab940 1320 %define NAME ["VALUE"]
50cca368 1321
de5ab940
JD
1322 except that the manner in which Bison processes multiple definitions
1323 for the same NAME differs. Most importantly, -F and --force-define
1324 quietly override %define, but -D and --define do not. For further
e4ab1254 1325 details, see the section "Bison Options" in the Bison manual.
50cca368 1326
f1b238df 1327*** Variables renamed:
67212941
JD
1328
1329 The following %define variables
1330
1331 api.push_pull
1332 lr.keep_unreachable_states
1333
1334 have been renamed to
1335
1336 api.push-pull
1337 lr.keep-unreachable-states
1338
1339 The old names are now deprecated but will be maintained indefinitely
1340 for backward compatibility.
1341
7262f54f 1342*** Values no longer need to be quoted in the grammar file:
cf499cff
JD
1343
1344 If a %define value is an identifier, it no longer needs to be placed
1345 within quotations marks. For example,
1346
1347 %define api.push-pull "push"
1348
1349 can be rewritten as
1350
1351 %define api.push-pull push
1352
d397d9f0 1353*** Unrecognized variables are now errors not warnings.
cdf3f113 1354
d397d9f0
JD
1355*** Multiple invocations for any variable is now an error not a warning.
1356
1357** Unrecognized %code qualifiers are now errors not warnings.
1358
1359** Character literals not of length one:
1360
1361 Previously, Bison quietly converted all character literals to length
1362 one. For example, without warning, Bison interpreted the operators in
1363 the following grammar to be the same token:
1364
1365 exp: exp '++'
1366 | exp '+' exp
1367 ;
1368
1369 Bison now warns when a character literal is not of length one. In
1370 some future release, Bison will start reporting an error instead.
1371
1372** Destructor calls fixed for lookaheads altered in semantic actions:
1373
1374 Previously for deterministic parsers in C, if a user semantic action
1375 altered yychar, the parser in some cases used the old yychar value to
1376 determine which destructor to call for the lookahead upon a syntax
1377 error or upon parser return. This bug has been fixed.
1378
1379** C++ parsers use YYRHSLOC:
1380
1381 Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
1382 macro and use it in the default YYLLOC_DEFAULT. You are encouraged
e4ab1254
AD
1383 to use it. If, for instance, your location structure has "first"
1384 and "last" members, instead of
d397d9f0 1385
e29f0771
AD
1386 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1387 do \
1388 if (N) \
1389 { \
1390 (Current).first = (Rhs)[1].location.first; \
1391 (Current).last = (Rhs)[N].location.last; \
1392 } \
1393 else \
1394 { \
1395 (Current).first = (Current).last = (Rhs)[0].location.last; \
1396 } \
1397 while (false)
d397d9f0
JD
1398
1399 use:
1400
e29f0771
AD
1401 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1402 do \
1403 if (N) \
1404 { \
1405 (Current).first = YYRHSLOC (Rhs, 1).first; \
1406 (Current).last = YYRHSLOC (Rhs, N).last; \
1407 } \
1408 else \
1409 { \
1410 (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \
1411 } \
1412 while (false)
d397d9f0
JD
1413
1414** YYLLOC_DEFAULT in C++:
1415
1416 The default implementation of YYLLOC_DEFAULT used to be issued in
1417 the header file. It is now output in the implementation file, after
1418 the user %code sections so that its #ifndef guard does not try to
1419 override the user's YYLLOC_DEFAULT if provided.
cdf3f113 1420
f1b238df 1421** YYFAIL now produces warnings and Java parsers no longer implement it:
4395a9ff
JD
1422
1423 YYFAIL has existed for many years as an undocumented feature of
1424 deterministic parsers in C generated by Bison. More recently, it was
1425 a documented feature of Bison's experimental Java parsers. As
1426 promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
1427 semantic action now produces a deprecation warning, and Java parsers
1428 no longer implement YYFAIL at all. For further details, including a
1429 discussion of how to suppress C preprocessor warnings about YYFAIL
1430 being unused, see the Bison 2.4.2 NEWS entry.
1431
f1b238df 1432** Temporary hack for adding a semicolon to the user action:
197b82ba
JD
1433
1434 Previously, Bison appended a semicolon to every user action for
1435 reductions when the output language defaulted to C (specifically, when
1436 neither %yacc, %language, %skeleton, or equivalent command-line
1437 options were specified). This allowed actions such as
1438
1439 exp: exp "+" exp { $$ = $1 + $3 };
1440
1441 instead of
1442
1443 exp: exp "+" exp { $$ = $1 + $3; };
1444
1445 As a first step in removing this misfeature, Bison now issues a
1446 warning when it appends a semicolon. Moreover, in cases where Bison
1447 cannot easily determine whether a semicolon is needed (for example, an
1448 action ending with a cpp directive or a braced compound initializer),
1449 it no longer appends one. Thus, the C compiler might now complain
1450 about a missing semicolon where it did not before. Future releases of
1451 Bison will cease to append semicolons entirely.
1452
d2060f06
JD
1453** Verbose syntax error message fixes:
1454
e4ab1254 1455 When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is
7fceb615
JD
1456 specified, syntax error messages produced by the generated parser
1457 include the unexpected token as well as a list of expected tokens.
1458 The effect of %nonassoc on these verbose messages has been corrected
1459 in two ways, but a more complete fix requires LAC, described above:
d2060f06
JD
1460
1461*** When %nonassoc is used, there can exist parser states that accept no
1462 tokens, and so the parser does not always require a lookahead token
1463 in order to detect a syntax error. Because no unexpected token or
1464 expected tokens can then be reported, the verbose syntax error
1465 message described above is suppressed, and the parser instead
e4ab1254 1466 reports the simpler message, "syntax error". Previously, this
d2060f06
JD
1467 suppression was sometimes erroneously triggered by %nonassoc when a
1468 lookahead was actually required. Now verbose messages are
1469 suppressed only when all previous lookaheads have already been
1470 shifted or discarded.
1471
1472*** Previously, the list of expected tokens erroneously included tokens
1473 that would actually induce a syntax error because conflicts for them
1474 were resolved with %nonassoc in the current parser state. Such
1475 tokens are now properly omitted from the list.
1476
1477*** Expected token lists are still often wrong due to state merging
fcf834f9
JD
1478 (from LALR or IELR) and default reductions, which can both add
1479 invalid tokens and subtract valid tokens. Canonical LR almost
1480 completely fixes this problem by eliminating state merging and
1481 default reductions. However, there is one minor problem left even
1482 when using canonical LR and even after the fixes above. That is,
1483 if the resolution of a conflict with %nonassoc appears in a later
1484 parser state than the one at which some syntax error is
1485 discovered, the conflicted token is still erroneously included in
1486 the expected token list. Bison's new LAC implementation,
1487 described above, eliminates this problem and the need for
1488 canonical LR. However, LAC is still experimental and is disabled
1489 by default.
53f036ce 1490
1a33f4f6
JD
1491** Java skeleton fixes:
1492
1493*** A location handling bug has been fixed.
1494
1495*** The top element of each of the value stack and location stack is now
1496 cleared when popped so that it can be garbage collected.
6771a463 1497
02803d55
JD
1498*** Parser traces now print the top element of the stack.
1499
86408959
JD
1500** -W/--warnings fixes:
1501
e4ab1254 1502*** Bison now properly recognizes the "no-" versions of categories:
86408959
JD
1503
1504 For example, given the following command line, Bison now enables all
1505 warnings except warnings for incompatibilities with POSIX Yacc:
1506
1507 bison -Wall,no-yacc gram.y
1508
786743d5
JD
1509*** Bison now treats S/R and R/R conflicts like other warnings:
1510
1511 Previously, conflict reports were independent of Bison's normal
1512 warning system. Now, Bison recognizes the warning categories
e4ab1254 1513 "conflicts-sr" and "conflicts-rr". This change has important
786743d5
JD
1514 consequences for the -W and --warnings command-line options. For
1515 example:
1516
1517 bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
1518 bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
1519 bison -Wnone gram.y # no conflicts are reported
1520 bison -Werror gram.y # any conflict is an error
1521
1522 However, as before, if the %expect or %expect-rr directive is
1523 specified, an unexpected number of conflicts is an error, and an
1524 expected number of conflicts is not reported, so -W and --warning
1525 then have no effect on the conflict report.
1526
e4ab1254 1527*** The "none" category no longer disables a preceding "error":
bf0e44e8
JD
1528
1529 For example, for the following command line, Bison now reports
1530 errors instead of warnings for incompatibilities with POSIX Yacc:
1531
1532 bison -Werror,none,yacc gram.y
1533
e4ab1254 1534*** The "none" category now disables all Bison warnings:
c39014ae 1535
e4ab1254 1536 Previously, the "none" category disabled only Bison warnings for
c39014ae
JD
1537 which there existed a specific -W/--warning category. However,
1538 given the following command line, Bison is now guaranteed to
1539 suppress all warnings:
1540
1541 bison -Wnone gram.y
1542
1f36f544
JD
1543** Precedence directives can now assign token number 0:
1544
1545 Since Bison 2.3b, which restored the ability of precedence
1546 directives to assign token numbers, doing so for token number 0 has
1547 produced an assertion failure. For example:
1548
1549 %left END 0
1550
1551 This bug has been fixed.
1552
64877e5e 1553* Changes in version 2.4.3 (2010-08-05):
8b9e021f 1554
2bfcac9a
JD
1555** Bison now obeys -Werror and --warnings=error for warnings about
1556 grammar rules that are useless in the parser due to conflicts.
1557
8b9e021f
JD
1558** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
1559 been fixed.
1560
4ad3921d
JD
1561** Failures in the test suite for GCC 4.5 have been fixed.
1562
06cb07d5
JD
1563** Failures in the test suite for some versions of Sun Studio C++ have
1564 been fixed.
1565
9b5049bd
JD
1566** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
1567 warnings about undefined %prec identifiers will not be converted to
1568 errors in Bison 2.5. They will remain warnings, which should be
1569 sufficient for POSIX while avoiding backward compatibility issues.
1570
93d7dde9
JD
1571** Minor documentation fixes.
1572
e19a049c 1573* Changes in version 2.4.2 (2010-03-20):
74553c98 1574
f39ab286
JD
1575** Some portability problems that resulted in failures and livelocks
1576 in the test suite on some versions of at least Solaris, AIX, HP-UX,
e19a049c
JD
1577 RHEL4, and Tru64 have been addressed. As a result, fatal Bison
1578 errors should no longer cause M4 to report a broken pipe on the
f39ab286
JD
1579 affected platforms.
1580
e4ab1254 1581** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately.
8bb3a2e7
JD
1582
1583 POSIX specifies that an error be reported for any identifier that does
1584 not appear on the LHS of a grammar rule and that is not defined by
1585 %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this
1586 error report for the case when an identifier appears only after a
1587 %prec directive. It is now restored. However, for backward
1588 compatibility with recent Bison releases, it is only a warning for
1589 now. In Bison 2.5 and later, it will return to being an error.
9b5049bd
JD
1590 [Between the 2.4.2 and 2.4.3 releases, it was decided that this
1591 warning will not be converted to an error in Bison 2.5.]
8bb3a2e7 1592
d8911864
EB
1593** Detection of GNU M4 1.4.6 or newer during configure is improved.
1594
a603c6e0
JD
1595** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
1596 YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
1597 avoided.
c938d650 1598
98a345a2
JD
1599** %code is now a permanent feature.
1600
1601 A traditional Yacc prologue directive is written in the form:
1602
1603 %{CODE%}
1604
1605 To provide a more flexible alternative, Bison 2.3b introduced the
1606 %code directive with the following forms for C/C++:
1607
1608 %code {CODE}
1609 %code requires {CODE}
1610 %code provides {CODE}
1611 %code top {CODE}
1612
1613 These forms are now considered permanent features of Bison. See the
1614 %code entries in the section "Bison Declaration Summary" in the Bison
1615 manual for a summary of their functionality. See the section
1616 "Prologue Alternatives" for a detailed discussion including the
1617 advantages of %code over the traditional Yacc prologue directive.
1618
1619 Bison's Java feature as a whole including its current usage of %code
1620 is still considered experimental.
1621
1625df5b
JD
1622** YYFAIL is deprecated and will eventually be removed.
1623
1624 YYFAIL has existed for many years as an undocumented feature of
1625 deterministic parsers in C generated by Bison. Previously, it was
1626 documented for Bison's experimental Java parsers. YYFAIL is no longer
1627 documented for Java parsers and is formally deprecated in both cases.
1628 Users are strongly encouraged to migrate to YYERROR, which is
1629 specified by POSIX.
1630
1631 Like YYERROR, you can invoke YYFAIL from a semantic action in order to
1632 induce a syntax error. The most obvious difference from YYERROR is
1633 that YYFAIL will automatically invoke yyerror to report the syntax
1634 error so that you don't have to. However, there are several other
1635 subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
e4ab1254 1636 inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is
1625df5b
JD
1637 used. For a more detailed discussion, see:
1638
1639 http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
1640
1641 The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
1642 deterministic parsers in C will continue to implement it. However,
1643 because YYFAIL is already flawed, it seems futile to try to make new
1644 Bison features compatible with it. Thus, during parser generation,
1645 Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
1646 rule action. In a later release, YYFAIL will be disabled for
e4ab1254 1647 %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will
1625df5b
JD
1648 be removed altogether.
1649
1650 There exists at least one case where Bison 2.5's YYFAIL warning will
1651 be a false positive. Some projects add phony uses of YYFAIL and other
1652 Bison-defined macros for the sole purpose of suppressing C
1653 preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
1654 To avoid Bison's future warning, such YYFAIL uses can be moved to the
e4ab1254 1655 epilogue (that is, after the second "%%") in the Bison input file. In
1625df5b
JD
1656 this release (2.4.2), Bison already generates its own code to suppress
1657 C preprocessor warnings for YYFAIL, so projects can remove their own
1658 phony uses of YYFAIL if compatibility with Bison releases prior to
1659 2.4.2 is not necessary.
1660
2755de8f
AD
1661** Internationalization.
1662
1663 Fix a regression introduced in Bison 2.4: Under some circumstances,
1664 message translations were not installed although supported by the
1665 host system.
1666
74553c98 1667* Changes in version 2.4.1 (2008-12-11):
c9ba9e59 1668
0ea583d2
AD
1669** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
1670 declarations have been fixed.
1979121c 1671
0ea583d2
AD
1672** Temporary hack for adding a semicolon to the user action.
1673
1674 Bison used to prepend a trailing semicolon at the end of the user
1675 action for reductions. This allowed actions such as
1676
1677 exp: exp "+" exp { $$ = $1 + $3 };
1678
1679 instead of
1680
1681 exp: exp "+" exp { $$ = $1 + $3; };
1682
e4ab1254 1683 Some grammars still depend on this "feature". Bison 2.4.1 restores
0ea583d2
AD
1684 the previous behavior in the case of C output (specifically, when
1685 neither %language or %skeleton or equivalent command-line options
1686 are used) to leave more time for grammars depending on the old
1687 behavior to be adjusted. Future releases of Bison will disable this
1688 feature.
1689
1690** A few minor improvements to the Bison manual.
c9ba9e59 1691
402b123d 1692* Changes in version 2.4 (2008-11-02):
7bd1665a 1693
402b123d 1694** %language is an experimental feature.
ed4d67dc
JD
1695
1696 We first introduced this feature in test release 2.3b as a cleaner
1697 alternative to %skeleton. Since then, we have discussed the possibility of
1698 modifying its effect on Bison's output file names. Thus, in this release,
1699 we consider %language to be an experimental feature that will likely evolve
1700 in future releases.
7bd1665a 1701
402b123d 1702** Forward compatibility with GNU M4 has been improved.
241fda7a 1703
402b123d 1704** Several bugs in the C++ skeleton and the experimental Java skeleton have been
241fda7a
JD
1705 fixed.
1706
402b123d 1707* Changes in version 2.3b (2008-05-27):
35fe0834 1708
402b123d 1709** The quotes around NAME that used to be required in the following directive
d9df47b6
JD
1710 are now deprecated:
1711
1712 %define NAME "VALUE"
1713
e4ab1254 1714** The directive "%pure-parser" is now deprecated in favor of:
d9df47b6
JD
1715
1716 %define api.pure
1717
1718 which has the same effect except that Bison is more careful to warn about
1719 unreasonable usage in the latter case.
1720
402b123d 1721** Push Parsing
c373bf8b
JD
1722
1723 Bison can now generate an LALR(1) parser in C with a push interface. That
e4ab1254
AD
1724 is, instead of invoking "yyparse", which pulls tokens from "yylex", you can
1725 push one token at a time to the parser using "yypush_parse", which will
c373bf8b
JD
1726 return to the caller after processing each token. By default, the push
1727 interface is disabled. Either of the following directives will enable it:
1728
1729 %define api.push_pull "push" // Just push; does not require yylex.
1730 %define api.push_pull "both" // Push and pull; requires yylex.
1731
e4ab1254 1732 See the new section "A Push Parser" in the Bison manual for details.
c373bf8b 1733
59da312b
JD
1734 The current push parsing interface is experimental and may evolve. More user
1735 feedback will help to stabilize it.
1736
402b123d 1737** The -g and --graph options now output graphs in Graphviz DOT format,
8e55b3aa
JD
1738 not VCG format. Like --graph, -g now also takes an optional FILE argument
1739 and thus cannot be bundled with other short options.
c373bf8b 1740
402b123d 1741** Java
59da312b
JD
1742
1743 Bison can now generate an LALR(1) parser in Java. The skeleton is
e4ab1254 1744 "data/lalr1.java". Consider using the new %language directive instead of
59da312b
JD
1745 %skeleton to select it.
1746
e4ab1254 1747 See the new section "Java Parsers" in the Bison manual for details.
59da312b
JD
1748
1749 The current Java interface is experimental and may evolve. More user
1750 feedback will help to stabilize it.
2bd435c3 1751 Contributed by Paolo Bonzini.
59da312b 1752
402b123d 1753** %language
59da312b
JD
1754
1755 This new directive specifies the programming language of the generated
d43f77e7
PB
1756 parser, which can be C (the default), C++, or Java. Besides the skeleton
1757 that Bison uses, the directive affects the names of the generated files if
1758 the grammar file's name ends in ".y".
59da312b 1759
402b123d 1760** XML Automaton Report
59da312b
JD
1761
1762 Bison can now generate an XML report of the LALR(1) automaton using the new
e4ab1254 1763 "--xml" option. The current XML schema is experimental and may evolve. More
59da312b 1764 user feedback will help to stabilize it.
2bd435c3 1765 Contributed by Wojciech Polak.
c373bf8b 1766
402b123d 1767** The grammar file may now specify the name of the parser header file using
c373bf8b
JD
1768 %defines. For example:
1769
1770 %defines "parser.h"
1771
402b123d 1772** When reporting useless rules, useless nonterminals, and unused terminals,
d80fb37a
JD
1773 Bison now employs the terms "useless in grammar" instead of "useless",
1774 "useless in parser" instead of "never reduced", and "unused in grammar"
1775 instead of "unused".
cff03fb2 1776
402b123d 1777** Unreachable State Removal
c373bf8b
JD
1778
1779 Previously, Bison sometimes generated parser tables containing unreachable
31984206
JD
1780 states. A state can become unreachable during conflict resolution if Bison
1781 disables a shift action leading to it from a predecessor state. Bison now:
75ad86ee
JD
1782
1783 1. Removes unreachable states.
1784
1785 2. Does not report any conflicts that appeared in unreachable states.
1786 WARNING: As a result, you may need to update %expect and %expect-rr
1787 directives in existing grammar files.
1788
1789 3. For any rule used only in such states, Bison now reports the rule as
cff03fb2 1790 "useless in parser due to conflicts".
75ad86ee 1791
31984206
JD
1792 This feature can be disabled with the following directive:
1793
1794 %define lr.keep_unreachable_states
1795
e4ab1254 1796 See the %define entry in the "Bison Declaration Summary" in the Bison manual
31984206
JD
1797 for further discussion.
1798
e4ab1254 1799** Lookahead Set Correction in the ".output" Report
b1cc23c4 1800
e4ab1254
AD
1801 When instructed to generate a ".output" file including lookahead sets
1802 (using "--report=lookahead", for example), Bison now prints each reduction's
88c78747
JD
1803 lookahead set only next to the associated state's one item that (1) is
1804 associated with the same rule as the reduction and (2) has its dot at the end
1805 of its RHS. Previously, Bison also erroneously printed the lookahead set
1806 next to all of the state's other items associated with the same rule. This
e4ab1254 1807 bug affected only the ".output" file and not the generated parser source
88c78747
JD
1808 code.
1809
e4ab1254 1810** --report-file=FILE is a new option to override the default ".output" file
59da312b 1811 name.
1bb2bd75 1812
e4ab1254 1813** The "=" that used to be required in the following directives is now
02975b9a
JD
1814 deprecated:
1815
1816 %file-prefix "parser"
1817 %name-prefix "c_"
1818 %output "parser.c"
1819
e4ab1254 1820** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}"
c373bf8b
JD
1821
1822 Bison 2.3a provided a new set of directives as a more flexible alternative to
8e0a5e9e
JD
1823 the traditional Yacc prologue blocks. Those have now been consolidated into
1824 a single %code directive with an optional qualifier field, which identifies
1825 the purpose of the code and thus the location(s) where Bison should generate
1826 it:
1827
e4ab1254
AD
1828 1. "%code {CODE}" replaces "%after-header {CODE}"
1829 2. "%code requires {CODE}" replaces "%start-header {CODE}"
1830 3. "%code provides {CODE}" replaces "%end-header {CODE}"
1831 4. "%code top {CODE}" replaces "%before-header {CODE}"
8e0a5e9e 1832
e4ab1254
AD
1833 See the %code entries in section "Bison Declaration Summary" in the Bison
1834 manual for a summary of the new functionality. See the new section "Prologue
1835 Alternatives" for a detailed discussion including the advantages of %code
8e0a5e9e
JD
1836 over the traditional Yacc prologues.
1837
1838 The prologue alternatives are experimental. More user feedback will help to
1839 determine whether they should become permanent features.
1840
402b123d 1841** Revised warning: unset or unused mid-rule values
17bd8a73
JD
1842
1843 Since Bison 2.2, Bison has warned about mid-rule values that are set but not
1844 used within any of the actions of the parent rule. For example, Bison warns
1845 about unused $2 in:
1846
1847 exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };
1848
1849 Now, Bison also warns about mid-rule values that are used but not set. For
1850 example, Bison warns about unset $$ in the mid-rule action in:
1851
1852 exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };
1853
1854 However, Bison now disables both of these warnings by default since they
1855 sometimes prove to be false alarms in existing grammars employing the Yacc
1856 constructs $0 or $-N (where N is some positive integer).
1857
e4ab1254
AD
1858 To enable these warnings, specify the option "--warnings=midrule-values" or
1859 "-W", which is a synonym for "--warnings=all".
17bd8a73 1860
e4ab1254 1861** Default %destructor or %printer with "<*>" or "<>"
c373bf8b
JD
1862
1863 Bison now recognizes two separate kinds of default %destructor's and
12e35840
JD
1864 %printer's:
1865
e4ab1254 1866 1. Place "<*>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1867 %destructor/%printer for all grammar symbols for which you have formally
1868 declared semantic type tags.
1869
e4ab1254 1870 2. Place "<>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1871 %destructor/%printer for all grammar symbols without declared semantic
1872 type tags.
1873
e4ab1254
AD
1874 Bison no longer supports the "%symbol-default" notation from Bison 2.3a.
1875 "<*>" and "<>" combined achieve the same effect with one exception: Bison no
12e35840
JD
1876 longer applies any %destructor to a mid-rule value if that mid-rule value is
1877 not actually ever referenced using either $$ or $n in a semantic action.
1878
85894313
JD
1879 The default %destructor's and %printer's are experimental. More user
1880 feedback will help to determine whether they should become permanent
1881 features.
1882
e4ab1254 1883 See the section "Freeing Discarded Symbols" in the Bison manual for further
12e35840
JD
1884 details.
1885
402b123d 1886** %left, %right, and %nonassoc can now declare token numbers. This is required
e4ab1254 1887 by POSIX. However, see the end of section "Operator Precedence" in the Bison
ab7f29f8
JD
1888 manual for a caveat concerning the treatment of literal strings.
1889
402b123d 1890** The nonfunctional --no-parser, -n, and %no-parser options have been
b1cc23c4
JD
1891 completely removed from Bison.
1892
402b123d 1893* Changes in version 2.3a, 2006-09-13:
742e4900 1894
402b123d 1895** Instead of %union, you can define and use your own union type
ddc8ede1
PE
1896 YYSTYPE if your grammar contains at least one <type> tag.
1897 Your YYSTYPE need not be a macro; it can be a typedef.
1898 This change is for compatibility with other Yacc implementations,
1899 and is required by POSIX.
1900
402b123d 1901** Locations columns and lines start at 1.
cd48d21d
AD
1902 In accordance with the GNU Coding Standards and Emacs.
1903
402b123d 1904** You may now declare per-type and default %destructor's and %printer's:
ec5479ce
JD
1905
1906 For example:
1907
b2a0b7ca
JD
1908 %union { char *string; }
1909 %token <string> STRING1
1910 %token <string> STRING2
1911 %type <string> string1
1912 %type <string> string2
1913 %union { char character; }
1914 %token <character> CHR
1915 %type <character> chr
1916 %destructor { free ($$); } %symbol-default
1917 %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
1918 %destructor { } <character>
1919
1920 guarantees that, when the parser discards any user-defined symbol that has a
e4ab1254
AD
1921 semantic type tag other than "<character>", it passes its semantic value to
1922 "free". However, when the parser discards a "STRING1" or a "string1", it
1923 also prints its line number to "stdout". It performs only the second
1924 "%destructor" in this case, so it invokes "free" only once.
ec5479ce 1925
85894313
JD
1926 [Although we failed to mention this here in the 2.3a release, the default
1927 %destructor's and %printer's were experimental, and they were rewritten in
1928 future versions.]
1929
e4ab1254
AD
1930** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y",
1931 "--yacc", or "%yacc"), Bison no longer generates #define statements for
b931235e
JD
1932 associating token numbers with token names. Removing the #define statements
1933 helps to sanitize the global namespace during preprocessing, but POSIX Yacc
1934 requires them. Bison still generates an enum for token names in all cases.
1935
402b123d 1936** Handling of traditional Yacc prologue blocks is now more consistent but
34f98f46 1937 potentially incompatible with previous releases of Bison.
9bc0dd67
JD
1938
1939 As before, you declare prologue blocks in your grammar file with the
e4ab1254 1940 "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all
34f98f46
JD
1941 prologue blocks that you've declared before the first %union. To generate
1942 the post-prologue, Bison concatenates all prologue blocks that you've
ddc8ede1 1943 declared after the first %union.
9bc0dd67 1944
34f98f46 1945 Previous releases of Bison inserted the pre-prologue into both the header
9bc0dd67
JD
1946 file and the code file in all cases except for LALR(1) parsers in C. In the
1947 latter case, Bison inserted it only into the code file. For parsers in C++,
1948 the point of insertion was before any token definitions (which associate
1949 token numbers with names). For parsers in C, the point of insertion was
1950 after the token definitions.
1951
1952 Now, Bison never inserts the pre-prologue into the header file. In the code
1953 file, it always inserts it before the token definitions.
1954
402b123d 1955** Bison now provides a more flexible alternative to the traditional Yacc
34f98f46
JD
1956 prologue blocks: %before-header, %start-header, %end-header, and
1957 %after-header.
1958
1959 For example, the following declaration order in the grammar file reflects the
1960 order in which Bison will output these code blocks. However, you are free to
1961 declare these code blocks in your grammar file in whatever order is most
1962 convenient for you:
1963
1964 %before-header {
1965 /* Bison treats this block like a pre-prologue block: it inserts it into
1966 * the code file before the contents of the header file. It does *not*
1967 * insert it into the header file. This is a good place to put
1968 * #include's that you want at the top of your code file. A common
e4ab1254 1969 * example is '#include "system.h"'. */
34f98f46
JD
1970 }
1971 %start-header {
1972 /* Bison inserts this block into both the header file and the code file.
1973 * In both files, the point of insertion is before any Bison-generated
1974 * token, semantic type, location type, and class definitions. This is a
1975 * good place to define %union dependencies, for example. */
9bc0dd67
JD
1976 }
1977 %union {
34f98f46
JD
1978 /* Unlike the traditional Yacc prologue blocks, the output order for the
1979 * new %*-header blocks is not affected by their declaration position
1980 * relative to any %union in the grammar file. */
9bc0dd67 1981 }
34f98f46
JD
1982 %end-header {
1983 /* Bison inserts this block into both the header file and the code file.
1984 * In both files, the point of insertion is after the Bison-generated
1985 * definitions. This is a good place to declare or define public
1986 * functions or data structures that depend on the Bison-generated
1987 * definitions. */
9bc0dd67 1988 }
34f98f46
JD
1989 %after-header {
1990 /* Bison treats this block like a post-prologue block: it inserts it into
1991 * the code file after the contents of the header file. It does *not*
1992 * insert it into the header file. This is a good place to declare or
1993 * define internal functions or data structures that depend on the
1994 * Bison-generated definitions. */
1995 }
1996
1997 If you have multiple occurrences of any one of the above declarations, Bison
1998 will concatenate the contents in declaration order.
9bc0dd67 1999
85894313
JD
2000 [Although we failed to mention this here in the 2.3a release, the prologue
2001 alternatives were experimental, and they were rewritten in future versions.]
2002
e4ab1254 2003** The option "--report=look-ahead" has been changed to "--report=lookahead".
9e6e7ed2
PE
2004 The old spelling still works, but is not documented and may be removed
2005 in a future release.
742e4900 2006
402b123d 2007* Changes in version 2.3, 2006-06-05:
4ad3ed84 2008
e4ab1254 2009** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING",
4ad3ed84
PE
2010 for compatibility with LALR(1) grammars.
2011
402b123d 2012** It is now documented that any definition of YYSTYPE or YYLTYPE should
4ad3ed84
PE
2013 be to a type name that does not contain parentheses or brackets.
2014
402b123d 2015* Changes in version 2.2, 2006-05-19:
193d7c70 2016
402b123d 2017** The distribution terms for all Bison-generated parsers now permit
193d7c70
PE
2018 using the parsers in nonfree programs. Previously, this permission
2019 was granted only for Bison-generated LALR(1) parsers in C.
5f4236a0 2020
402b123d 2021** %name-prefix changes the namespace name in C++ outputs.
aa08666d 2022
402b123d 2023** The C++ parsers export their token_type.
5f4236a0 2024
402b123d 2025** Bison now allows multiple %union declarations, and concatenates
d6ca7905
PE
2026 their contents together.
2027
402b123d 2028** New warning: unused values
4d7bc38c
PE
2029 Right-hand side symbols whose values are not used are reported,
2030 if the symbols have destructors. For instance:
affac613 2031
8f3596a6 2032 exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
e9690142
JD
2033 | exp "+" exp
2034 ;
affac613 2035
8f3596a6
AD
2036 will trigger a warning about $$ and $5 in the first rule, and $3 in
2037 the second ($1 is copied to $$ by the default rule). This example
4e26c69e 2038 most likely contains three errors, and could be rewritten as:
affac613 2039
4e26c69e 2040 exp: exp "?" exp ":" exp
e9690142
JD
2041 { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
2042 | exp "+" exp
2043 { $$ = $1 ? $1 : $3; if ($1) free ($3); }
2044 ;
affac613 2045
4e26c69e
PE
2046 However, if the original actions were really intended, memory leaks
2047 and all, the warnings can be suppressed by letting Bison believe the
2048 values are used, e.g.:
721be13c 2049
8f3596a6 2050 exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
e9690142
JD
2051 | exp "+" exp { $$ = $1; (void) $3; }
2052 ;
721be13c 2053
84866159
AD
2054 If there are mid-rule actions, the warning is issued if no action
2055 uses it. The following triggers no warning: $1 and $3 are used.
2056
2057 exp: exp { push ($1); } '+' exp { push ($3); sum (); };
2058
721be13c
PE
2059 The warning is intended to help catching lost values and memory leaks.
2060 If a value is ignored, its associated memory typically is not reclaimed.
affac613 2061
402b123d 2062** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
9d9b8b70
PE
2063 Destructors are now called when user code invokes YYABORT, YYACCEPT,
2064 and YYERROR, for all objects on the stack, other than objects
2065 corresponding to the right-hand side of the current rule.
a85284cf 2066
402b123d 2067** %expect, %expect-rr
035aa4a0
PE
2068 Incorrect numbers of expected conflicts are now actual errors,
2069 instead of warnings.
2070
402b123d 2071** GLR, YACC parsers.
4e26c69e
PE
2072 The %parse-params are available in the destructors (and the
2073 experimental printers) as per the documentation.
4b367315 2074
e4ab1254 2075** Bison now warns if it finds a stray "$" or "@" in an action.
ad6a9b97 2076
402b123d 2077** %require "VERSION"
4e26c69e
PE
2078 This specifies that the grammar file depends on features implemented
2079 in Bison version VERSION or higher.
b50d2359 2080
402b123d 2081** lalr1.cc: The token and value types are now class members.
e14d0ab6
AD
2082 The tokens were defined as free form enums and cpp macros. YYSTYPE
2083 was defined as a free form union. They are now class members:
e4ab1254
AD
2084 tokens are enumerations of the "yy::parser::token" struct, and the
2085 semantic values have the "yy::parser::semantic_type" type.
fb9712a9
AD
2086
2087 If you do not want or can update to this scheme, the directive
e4ab1254 2088 '%define "global_tokens_and_yystype" "1"' triggers the global
b50d2359
AD
2089 definition of tokens and YYSTYPE. This change is suitable both
2090 for previous releases of Bison, and this one.
fb9712a9 2091
b50d2359 2092 If you wish to update, then make sure older version of Bison will
e4ab1254 2093 fail using '%require "2.2"'.
fb9712a9 2094
402b123d 2095** DJGPP support added.
193d7c70 2096\f
402b123d 2097* Changes in version 2.1, 2005-09-16:
1ce59070 2098
402b123d 2099** The C++ lalr1.cc skeleton supports %lex-param.
e14d0ab6 2100
402b123d 2101** Bison-generated parsers now support the translation of diagnostics like
baf785db
PE
2102 "syntax error" into languages other than English. The default
2103 language is still English. For details, please see the new
0410a6e0
PE
2104 Internationalization section of the Bison manual. Software
2105 distributors should also see the new PACKAGING file. Thanks to
2106 Bruno Haible for this new feature.
1ce59070 2107
402b123d 2108** Wording in the Bison-generated parsers has been changed slightly to
1a059451
PE
2109 simplify translation. In particular, the message "memory exhausted"
2110 has replaced "parser stack overflow", as the old message was not
2111 always accurate for modern Bison-generated parsers.
2112
402b123d 2113** Destructors are now called when the parser aborts, for all symbols left
258b75ca
PE
2114 behind on the stack. Also, the start symbol is now destroyed after a
2115 successful parse. In both cases, the behavior was formerly inconsistent.
2116
402b123d 2117** When generating verbose diagnostics, Bison-generated parsers no longer
72f000b0
PE
2118 quote the literal strings associated with tokens. For example, for
2119 a syntax error associated with '%token NUM "number"' they might
2120 print 'syntax error, unexpected number' instead of 'syntax error,
2121 unexpected "number"'.
193d7c70 2122\f
402b123d 2123* Changes in version 2.0, 2004-12-25:
efeed023 2124
402b123d 2125** Possibly-incompatible changes
d7e14fc0 2126
82de6b0d
PE
2127 - Bison-generated parsers no longer default to using the alloca function
2128 (when available) to extend the parser stack, due to widespread
2129 problems in unchecked stack-overflow detection. You can "#define
2130 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
2131 the manual to determine safe values for YYMAXDEPTH in that case.
8dd162d3 2132
82de6b0d
PE
2133 - Error token location.
2134 During error recovery, the location of the syntax error is updated
2135 to cover the whole sequence covered by the error token: it includes
2136 the shifted symbols thrown away during the first part of the error
2137 recovery, and the lookahead rejected during the second part.
18d192f0 2138
82de6b0d
PE
2139 - Semicolon changes:
2140 . Stray semicolons are no longer allowed at the start of a grammar.
2141 . Semicolons are now required after in-grammar declarations.
e342c3be 2142
82de6b0d
PE
2143 - Unescaped newlines are no longer allowed in character constants or
2144 string literals. They were never portable, and GCC 3.4.0 has
2145 dropped support for them. Better diagnostics are now generated if
2146 forget a closing quote.
8dd162d3 2147
82de6b0d 2148 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
f74b6f91 2149
402b123d 2150** New features
1452af69 2151
82de6b0d 2152 - GLR grammars now support locations.
4febdd96 2153
82de6b0d
PE
2154 - New directive: %initial-action.
2155 This directive allows the user to run arbitrary code (including
2156 initializing @$) from yyparse before parsing starts.
1452af69 2157
82de6b0d
PE
2158 - A new directive "%expect-rr N" specifies the expected number of
2159 reduce/reduce conflicts in GLR parsers.
1452af69 2160
e4ab1254 2161 - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d".
82de6b0d 2162 This is a GNU extension.
4febdd96 2163
e4ab1254 2164 - The option "--report=lookahead" was changed to "--report=look-ahead".
9e6e7ed2 2165 [However, this was changed back after 2.3.]
1452af69 2166
82de6b0d 2167 - Experimental %destructor support has been added to lalr1.cc.
1452af69 2168
82de6b0d
PE
2169 - New configure option --disable-yacc, to disable installation of the
2170 yacc command and -ly library introduced in 1.875 for POSIX conformance.
6040d338 2171
402b123d 2172** Bug fixes
d5a3fe37 2173
82de6b0d
PE
2174 - For now, %expect-count violations are now just warnings, not errors.
2175 This is for compatibility with Bison 1.75 and earlier (when there are
2176 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
2177 are too many or too few shift/reduce conflicts). However, in future
2178 versions of Bison we plan to improve the %expect machinery so that
2179 these violations will become errors again.
3473d0f8 2180
82de6b0d
PE
2181 - Within Bison itself, numbers (e.g., goto numbers) are no longer
2182 arbitrarily limited to 16-bit counts.
d600ee67 2183
82de6b0d 2184 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
d600ee67 2185\f
402b123d 2186* Changes in version 1.875, 2003-01-01:
963fcc17 2187
402b123d 2188** The documentation license has been upgraded to version 1.2
dc546b0f 2189 of the GNU Free Documentation License.
75eb3bc4 2190
402b123d 2191** syntax error processing
75eb3bc4 2192
dc546b0f
PE
2193 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
2194 locations too. This fixes bugs in error-location computation.
75eb3bc4 2195
dc546b0f
PE
2196 - %destructor
2197 It is now possible to reclaim the memory associated to symbols
2198 discarded during error recovery. This feature is still experimental.
20daca06 2199
dc546b0f
PE
2200 - %error-verbose
2201 This new directive is preferred over YYERROR_VERBOSE.
74724a70 2202
dc546b0f
PE
2203 - #defining yyerror to steal internal variables is discouraged.
2204 It is not guaranteed to work forever.
d1de5372 2205
402b123d 2206** POSIX conformance
d1de5372 2207
dc546b0f
PE
2208 - Semicolons are once again optional at the end of grammar rules.
2209 This reverts to the behavior of Bison 1.33 and earlier, and improves
2210 compatibility with Yacc.
74724a70 2211
e4ab1254
AD
2212 - "parse error" -> "syntax error"
2213 Bison now uniformly uses the term "syntax error"; formerly, the code
2214 and manual sometimes used the term "parse error" instead. POSIX
2215 requires "syntax error" in diagnostics, and it was thought better to
dc546b0f 2216 be consistent.
74724a70 2217
dc546b0f
PE
2218 - The documentation now emphasizes that yylex and yyerror must be
2219 declared before use. C99 requires this.
d1de5372 2220
dc546b0f
PE
2221 - Bison now parses C99 lexical constructs like UCNs and
2222 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
d1de5372 2223
dc546b0f
PE
2224 - File names are properly escaped in C output. E.g., foo\bar.y is
2225 output as "foo\\bar.y".
6780ca7a 2226
dc546b0f 2227 - Yacc command and library now available
e4ab1254 2228 The Bison distribution now installs a "yacc" command, as POSIX requires.
dc546b0f
PE
2229 Also, Bison now installs a small library liby.a containing
2230 implementations of Yacc-compatible yyerror and main functions.
2231 This library is normally not useful, but POSIX requires it.
6e649e65 2232
dc546b0f 2233 - Type clashes now generate warnings, not errors.
6e649e65 2234
dc546b0f
PE
2235 - If the user does not define YYSTYPE as a macro, Bison now declares it
2236 using typedef instead of defining it as a macro.
2237 For consistency, YYLTYPE is also declared instead of defined.
9501dc6e 2238
402b123d 2239** Other compatibility issues
886a425c 2240
e4ab1254
AD
2241 - %union directives can now have a tag before the "{", e.g., the
2242 directive "%union foo {...}" now generates the C code
2243 "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility.
2244 The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc.
2245 For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype".
dc546b0f 2246 This is for compatibility with both Yacc and Bison 1.35.
72f889cc 2247
e4ab1254 2248 - ";" is output before the terminating "}" of an action, for
dc546b0f 2249 compatibility with Bison 1.35.
886a425c 2250
dc546b0f 2251 - Bison now uses a Yacc-style format for conflict reports, e.g.,
e4ab1254 2252 "conflicts: 2 shift/reduce, 1 reduce/reduce".
437c2d80 2253
e4ab1254 2254 - "yystype" and "yyltype" are now obsolescent macros instead of being
dc546b0f
PE
2255 typedefs or tags; they are no longer documented and are planned to be
2256 withdrawn in a future release.
2a8d363a 2257
402b123d 2258** GLR parser notes
2a8d363a 2259
dc546b0f
PE
2260 - GLR and inline
2261 Users of Bison have to decide how they handle the portability of the
e4ab1254 2262 C keyword "inline".
959e5f51 2263
e4ab1254
AD
2264 - "parsing stack overflow..." -> "parser stack overflow"
2265 GLR parsers now report "parser stack overflow" as per the Bison manual.
900c5db5 2266
18ad57b3
AD
2267** %parse-param and %lex-param
2268 The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass
2269 additional context to yyparse and yylex. They suffer from several
2270 shortcomings:
2271
2272 - a single argument only can be added,
2273 - their types are weak (void *),
242cc08e 2274 - this context is not passed to ancillary functions such as yyerror,
18ad57b3
AD
2275 - only yacc.c parsers support them.
2276
2277 The new %parse-param/%lex-param directives provide a more precise control.
2278 For instance:
2279
2280 %parse-param {int *nastiness}
2281 %lex-param {int *nastiness}
2282 %parse-param {int *randomness}
2283
2284 results in the following signatures:
2285
2286 int yylex (int *nastiness);
2287 int yyparse (int *nastiness, int *randomness);
2288
2289 or, if both %pure-parser and %locations are used:
2290
2291 int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness);
2292 int yyparse (int *nastiness, int *randomness);
2293
402b123d 2294** Bison now warns if it detects conflicting outputs to the same file,
e4ab1254 2295 e.g., it generates a warning for "bison -d -o foo.h foo.y" since
dc546b0f 2296 that command outputs both code and header to foo.h.
6e40b4eb 2297
402b123d 2298** #line in output files
dc546b0f 2299 - --no-line works properly.
6e40b4eb 2300
402b123d 2301** Bison can no longer be built by a K&R C compiler; it requires C89 or
6e40b4eb
AD
2302 later to be built. This change originally took place a few versions
2303 ago, but nobody noticed until we recently asked someone to try
2304 building Bison with a K&R C compiler.
d600ee67 2305\f
402b123d 2306* Changes in version 1.75, 2002-10-14:
7933f2b5 2307
402b123d 2308** Bison should now work on 64-bit hosts.
7933f2b5 2309
402b123d 2310** Indonesian translation thanks to Tedi Heriyanto.
7933f2b5 2311
402b123d 2312** GLR parsers
f50adbbd
AD
2313 Fix spurious parse errors.
2314
402b123d 2315** Pure parsers
f50adbbd
AD
2316 Some people redefine yyerror to steal yyparse' private variables.
2317 Reenable this trick until an official feature replaces it.
2318
402b123d 2319** Type Clashes
d90c934c
AD
2320 In agreement with POSIX and with other Yaccs, leaving a default
2321 action is valid when $$ is untyped, and $1 typed:
2322
e9690142 2323 untyped: ... typed;
d90c934c
AD
2324
2325 but the converse remains an error:
2326
e9690142 2327 typed: ... untyped;
d90c934c 2328
402b123d 2329** Values of mid-rule actions
d90c934c
AD
2330 The following code:
2331
e9690142 2332 foo: { ... } { $$ = $1; } ...
d90c934c
AD
2333
2334 was incorrectly rejected: $1 is defined in the second mid-rule
2335 action, and is equal to the $$ of the first mid-rule action.
d600ee67 2336\f
402b123d 2337* Changes in version 1.50, 2002-10-04:
adc8c848 2338
402b123d 2339** GLR parsing
676385e2
PH
2340 The declaration
2341 %glr-parser
2342 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
2343 almost any context-free grammar, ambiguous or not. The new declarations
e8832397 2344 %dprec and %merge on grammar rules allow parse-time resolution of
676385e2
PH
2345 ambiguities. Contributed by Paul Hilfinger.
2346
7933f2b5 2347 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
420f93c8
PE
2348 like the Alpha, so please stick to 32-bit hosts for now.
2349
402b123d 2350** Output Directory
8c165d89 2351 When not in Yacc compatibility mode, when the output file was not
e4ab1254
AD
2352 specified, running "bison foo/bar.y" created "foo/bar.c". It
2353 now creates "bar.c".
8c165d89 2354
402b123d 2355** Undefined token
007a50a4 2356 The undefined token was systematically mapped to 2 which prevented
e88dbdbf 2357 the use of 2 by the user. This is no longer the case.
007a50a4 2358
402b123d 2359** Unknown token numbers
e88dbdbf 2360 If yylex returned an out of range value, yyparse could die. This is
007a50a4
AD
2361 no longer the case.
2362
402b123d 2363** Error token
e88dbdbf 2364 According to POSIX, the error token must be 256.
23c5a174
AD
2365 Bison extends this requirement by making it a preference: *if* the
2366 user specified that one of her tokens is numbered 256, then error
2367 will be mapped onto another number.
2368
402b123d 2369** Verbose error messages
e4ab1254 2370 They no longer report "..., expecting error or..." for states where
217598da
AD
2371 error recovery is possible.
2372
402b123d 2373** End token
e4ab1254 2374 Defaults to "$end" instead of "$".
217598da 2375
402b123d 2376** Error recovery now conforms to documentation and to POSIX
68cd8af3
PE
2377 When a Bison-generated parser encounters a syntax error, it now pops
2378 the stack until it finds a state that allows shifting the error
2379 token. Formerly, it popped the stack until it found a state that
2380 allowed some non-error action other than a default reduction on the
2381 error token. The new behavior has long been the documented behavior,
2382 and has long been required by POSIX. For more details, please see
337116ba
PE
2383 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
2384 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
68cd8af3 2385
402b123d 2386** Traces
5504898e
AD
2387 Popped tokens and nonterminals are now reported.
2388
402b123d 2389** Larger grammars
a861a339
PE
2390 Larger grammars are now supported (larger token numbers, larger grammar
2391 size (= sum of the LHS and RHS lengths), larger LALR tables).
2392 Formerly, many of these numbers ran afoul of 16-bit limits;
2393 now these limits are 32 bits on most hosts.
355e7c1c 2394
402b123d 2395** Explicit initial rule
643a5994
AD
2396 Bison used to play hacks with the initial rule, which the user does
2397 not write. It is now explicit, and visible in the reports and
2398 graphs as rule 0.
23c5a174 2399
402b123d 2400** Useless rules
643a5994 2401 Before, Bison reported the useless rules, but, although not used,
77714df2 2402 included them in the parsers. They are now actually removed.
23c5a174 2403
402b123d 2404** Useless rules, useless nonterminals
6b98e4b5
AD
2405 They are now reported, as a warning, with their locations.
2406
402b123d 2407** Rules never reduced
e8832397
AD
2408 Rules that can never be reduced because of conflicts are now
2409 reported.
2410
e4ab1254 2411** Incorrect "Token not used"
11652ab3
AD
2412 On a grammar such as
2413
e29f0771
AD
2414 %token useless useful
2415 %%
2416 exp: '0' %prec useful;
11652ab3
AD
2417
2418 where a token was used to set the precedence of the last rule,
e4ab1254 2419 bison reported both "useful" and "useless" as useless tokens.
11652ab3 2420
402b123d 2421** Revert the C++ namespace changes introduced in 1.31
77714df2 2422 as they caused too many portability hassles.
0179dd65 2423
402b123d 2424** Default locations
b2d52318
AD
2425 By an accident of design, the default computation of @$ was
2426 performed after another default computation was performed: @$ = @1.
2427 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
2428 the computation of @$.
adc8c848 2429
402b123d 2430** Token end-of-file
b7c49edf
AD
2431 The token end of file may be specified by the user, in which case,
2432 the user symbol is used in the reports, the graphs, and the verbose
e4ab1254 2433 error messages instead of "$end", which remains being the default.
b7c49edf 2434 For instance
e29f0771 2435 %token MYEOF 0
b7c49edf 2436 or
e29f0771 2437 %token MYEOF 0 "end of file"
fdbcd8e2 2438
402b123d 2439** Semantic parser
fdbcd8e2
AD
2440 This old option, which has been broken for ages, is removed.
2441
402b123d 2442** New translations
a861a339 2443 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
84614e13
AD
2444 Croatian, thanks to Denis Lackovic.
2445
402b123d 2446** Incorrect token definitions
e4ab1254
AD
2447 When given
2448 %token 'a' "A"
2449 bison used to output
2450 #define 'a' 65
b87f8b21 2451
402b123d 2452** Token definitions as enums
77714df2
AD
2453 Tokens are output both as the traditional #define's, and, provided
2454 the compiler supports ANSI C or is a C++ compiler, as enums.
e88dbdbf 2455 This lets debuggers display names instead of integers.
77714df2 2456
402b123d 2457** Reports
ec3bc396
AD
2458 In addition to --verbose, bison supports --report=THINGS, which
2459 produces additional information:
b408954b
AD
2460 - itemset
2461 complete the core item sets with their closure
e4ab1254 2462 - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back]
9e6e7ed2 2463 explicitly associate lookahead tokens to items
b408954b
AD
2464 - solved
2465 describe shift/reduce conflicts solving.
2466 Bison used to systematically output this information on top of
2467 the report. Solved conflicts are now attached to their states.
ec3bc396 2468
402b123d 2469** Type clashes
9af3fbce
AD
2470 Previous versions don't complain when there is a type clash on
2471 the default action if the rule has a mid-rule action, such as in:
2472
e29f0771
AD
2473 %type <foo> bar
2474 %%
2475 bar: '0' {} '0';
9af3fbce
AD
2476
2477 This is fixed.
a861a339 2478
402b123d 2479** GNU M4 is now required when using Bison.
f987e9d2 2480\f
402b123d 2481* Changes in version 1.35, 2002-03-25:
76551463 2482
402b123d 2483** C Skeleton
76551463
AD
2484 Some projects use Bison's C parser with C++ compilers, and define
2485 YYSTYPE as a class. The recent adjustment of C parsers for data
2486 alignment and 64 bit architectures made this impossible.
2487
2488 Because for the time being no real solution for C++ parser
2489 generation exists, kludges were implemented in the parser to
2490 maintain this use. In the future, when Bison has C++ parsers, this
2491 kludge will be disabled.
2492
2493 This kludge also addresses some C++ problems when the stack was
2494 extended.
76551463 2495\f
402b123d 2496* Changes in version 1.34, 2002-03-12:
76551463 2497
402b123d 2498** File name clashes are detected
76551463 2499 $ bison foo.y -d -o foo.x
e4ab1254 2500 fatal error: header and parser would both be named "foo.x"
76551463 2501
e4ab1254 2502** A missing ";" at the end of a rule triggers a warning
76551463
AD
2503 In accordance with POSIX, and in agreement with other
2504 Yacc implementations, Bison will mandate this semicolon in the near
2505 future. This eases the implementation of a Bison parser of Bison
2506 grammars by making this grammar LALR(1) instead of LR(2). To
2507 facilitate the transition, this release introduces a warning.
2508
402b123d 2509** Revert the C++ namespace changes introduced in 1.31, as they caused too
76551463
AD
2510 many portability hassles.
2511
402b123d 2512** DJGPP support added.
76551463 2513
402b123d 2514** Fix test suite portability problems.
76551463 2515\f
402b123d 2516* Changes in version 1.33, 2002-02-07:
76551463 2517
402b123d 2518** Fix C++ issues
76551463
AD
2519 Groff could not be compiled for the definition of size_t was lacking
2520 under some conditions.
2521
402b123d 2522** Catch invalid @n
76551463
AD
2523 As is done with $n.
2524\f
402b123d 2525* Changes in version 1.32, 2002-01-23:
76551463 2526
402b123d 2527** Fix Yacc output file names
76551463 2528
402b123d 2529** Portability fixes
76551463 2530
402b123d 2531** Italian, Dutch translations
76551463 2532\f
402b123d 2533* Changes in version 1.31, 2002-01-14:
52d1aeee 2534
402b123d 2535** Many Bug Fixes
52d1aeee 2536
402b123d 2537** GNU Gettext and %expect
52d1aeee
MA
2538 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
2539 Bison dies on incorrect %expectations, we fear there will be
2540 too many bug reports for Gettext, so _for the time being_, %expect
e4ab1254 2541 does not trigger an error when the input file is named "plural.y".
52d1aeee 2542
402b123d 2543** Use of alloca in parsers
52d1aeee
MA
2544 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
2545 malloc exclusively. Since 1.29, but was not NEWS'ed.
2546
2547 alloca is used only when compiled with GCC, to avoid portability
2548 problems as on AIX.
2549
402b123d 2550** yyparse now returns 2 if memory is exhausted; formerly it dumped core.
b47dbebe 2551
402b123d 2552** When the generated parser lacks debugging code, YYDEBUG is now 0
52d1aeee
MA
2553 (as POSIX requires) instead of being undefined.
2554
402b123d 2555** User Actions
52d1aeee
MA
2556 Bison has always permitted actions such as { $$ = $1 }: it adds the
2557 ending semicolon. Now if in Yacc compatibility mode, the semicolon
2558 is no longer output: one has to write { $$ = $1; }.
2559
402b123d 2560** Better C++ compliance
52d1aeee 2561 The output parsers try to respect C++ namespaces.
76551463 2562 [This turned out to be a failed experiment, and it was reverted later.]
52d1aeee 2563
402b123d 2564** Reduced Grammars
52d1aeee
MA
2565 Fixed bugs when reporting useless nonterminals.
2566
402b123d 2567** 64 bit hosts
52d1aeee
MA
2568 The parsers work properly on 64 bit hosts.
2569
402b123d 2570** Error messages
52d1aeee
MA
2571 Some calls to strerror resulted in scrambled or missing error messages.
2572
402b123d 2573** %expect
52d1aeee
MA
2574 When the number of shift/reduce conflicts is correct, don't issue
2575 any warning.
2576
402b123d 2577** The verbose report includes the rule line numbers.
52d1aeee 2578
402b123d 2579** Rule line numbers are fixed in traces.
52d1aeee 2580
402b123d 2581** Swedish translation
52d1aeee 2582
402b123d 2583** Parse errors
52d1aeee
MA
2584 Verbose parse error messages from the parsers are better looking.
2585 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
2586 Now: parse error: unexpected '/', expecting "number" or '-' or '('
2587
402b123d 2588** Fixed parser memory leaks.
52d1aeee
MA
2589 When the generated parser was using malloc to extend its stacks, the
2590 previous allocations were not freed.
2591
402b123d 2592** Fixed verbose output file.
52d1aeee
MA
2593 Some newlines were missing.
2594 Some conflicts in state descriptions were missing.
2595
402b123d 2596** Fixed conflict report.
52d1aeee
MA
2597 Option -v was needed to get the result.
2598
402b123d 2599** %expect
52d1aeee
MA
2600 Was not used.
2601 Mismatches are errors, not warnings.
2602
402b123d 2603** Fixed incorrect processing of some invalid input.
52d1aeee 2604
402b123d 2605** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
52d1aeee 2606
402b123d 2607** Fixed some typos in the documentation.
52d1aeee 2608
402b123d 2609** %token MY_EOF 0 is supported.
52d1aeee
MA
2610 Before, MY_EOF was silently renumbered as 257.
2611
402b123d 2612** doc/refcard.tex is updated.
52d1aeee 2613
402b123d 2614** %output, %file-prefix, %name-prefix.
52d1aeee
MA
2615 New.
2616
402b123d 2617** --output
e4ab1254 2618 New, aliasing "--output-file".
52d1aeee 2619\f
402b123d 2620* Changes in version 1.30, 2001-10-26:
342b8b6e 2621
e4ab1254
AD
2622** "--defines" and "--graph" have now an optional argument which is the
2623 output file name. "-d" and "-g" do not change; they do not take any
342b8b6e
AD
2624 argument.
2625
e4ab1254 2626** "%source_extension" and "%header_extension" are removed, failed
342b8b6e
AD
2627 experiment.
2628
402b123d 2629** Portability fixes.
f987e9d2 2630\f
402b123d 2631* Changes in version 1.29, 2001-09-07:
342b8b6e 2632
402b123d 2633** The output file does not define const, as this caused problems when used
342b8b6e
AD
2634 with common autoconfiguration schemes. If you still use ancient compilers
2635 that lack const, compile with the equivalent of the C compiler option
e4ab1254 2636 "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this.
342b8b6e 2637
e4ab1254 2638** Added "-g" and "--graph".
f87a2205 2639
402b123d 2640** The Bison manual is now distributed under the terms of the GNU FDL.
f2b5126e 2641
402b123d 2642** The input and the output files has automatically a similar extension.
234a3be3 2643
402b123d 2644** Russian translation added.
f87a2205 2645
402b123d 2646** NLS support updated; should hopefully be less troublesome.
f87a2205 2647
402b123d 2648** Added the old Bison reference card.
c33638bb 2649
e4ab1254 2650** Added "--locations" and "%locations".
6deb4447 2651
e4ab1254 2652** Added "-S" and "--skeleton".
cd5bd6ac 2653
e4ab1254 2654** "%raw", "-r", "--raw" is disabled.
62ab6972 2655
402b123d 2656** Special characters are escaped when output. This solves the problems
cd5bd6ac
AD
2657 of the #line lines with path names including backslashes.
2658
402b123d 2659** New directives.
e4ab1254
AD
2660 "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose",
2661 "%debug", "%source_extension" and "%header_extension".
f987e9d2 2662
402b123d 2663** @$
f987e9d2 2664 Automatic location tracking.
f87a2205 2665\f
402b123d 2666* Changes in version 1.28, 1999-07-06:
d2e00347 2667
402b123d 2668** Should compile better now with K&R compilers.
d2e00347 2669
402b123d 2670** Added NLS.
d2e00347 2671
402b123d 2672** Fixed a problem with escaping the double quote character.
d2e00347 2673
402b123d 2674** There is now a FAQ.
d2e00347 2675\f
402b123d 2676* Changes in version 1.27:
5c31c3c2 2677
402b123d 2678** The make rule which prevented bison.simple from being created on
5c31c3c2
JT
2679 some systems has been fixed.
2680\f
402b123d 2681* Changes in version 1.26:
4be07551 2682
7e508a2b 2683** Bison now uses Automake.
4be07551 2684
402b123d 2685** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
4be07551 2686
402b123d 2687** Token numbers now start at 257 as previously documented, not 258.
4be07551 2688
402b123d 2689** Bison honors the TMPDIR environment variable.
4be07551 2690
402b123d 2691** A couple of buffer overruns have been fixed.
f51dbca1 2692
402b123d 2693** Problems when closing files should now be reported.
f51dbca1 2694
402b123d 2695** Generated parsers should now work even on operating systems which do
f51dbca1 2696 not provide alloca().
4be07551 2697\f
402b123d 2698* Changes in version 1.25, 1995-10-16:
df8878c5 2699
402b123d 2700** Errors in the input grammar are not fatal; Bison keeps reading
df8878c5 2701the grammar file, and reports all the errors found in it.
8c44d3ec 2702
402b123d 2703** Tokens can now be specified as multiple-character strings: for
df8878c5 2704example, you could use "<=" for a token which looks like <=, instead
7e508a2b 2705of choosing a name like LESSEQ.
df8878c5 2706
402b123d 2707** The %token_table declaration says to write a table of tokens (names
df8878c5
RS
2708and numbers) into the parser file. The yylex function can use this
2709table to recognize multiple-character string tokens, or for other
2710purposes.
2711
402b123d 2712** The %no_lines declaration says not to generate any #line preprocessor
df8878c5
RS
2713directives in the parser file.
2714
402b123d 2715** The %raw declaration says to use internal Bison token numbers, not
df8878c5
RS
2716Yacc-compatible token numbers, when token names are defined as macros.
2717
402b123d 2718** The --no-parser option produces the parser tables without including
df8878c5
RS
2719the parser engine; a project can now use its own parser engine.
2720The actions go into a separate file called NAME.act, in the form of
2721a switch statement body.
2722\f
402b123d 2723* Changes in version 1.23:
6780ca7a 2724
4d019228
DM
2725The user can define YYPARSE_PARAM as the name of an argument to be
2726passed into yyparse. The argument should have type void *. It should
2727actually point to an object. Grammar actions can access the variable
2728by casting it to the proper pointer type.
6780ca7a 2729
6780ca7a 2730Line numbers in output file corrected.
6780ca7a 2731\f
402b123d 2732* Changes in version 1.22:
6780ca7a
DM
2733
2734--help option added.
6780ca7a 2735\f
402b123d 2736* Changes in version 1.20:
6780ca7a
DM
2737
2738Output file does not redefine const for C++.
9f4503d6 2739
76551463
AD
2740-----
2741
3209eb1c 2742Copyright (C) 1995-2015 Free Software Foundation, Inc.
76551463 2743
74553c98 2744This file is part of Bison, the GNU Parser Generator.
76551463 2745
f16b0819 2746This program is free software: you can redistribute it and/or modify
76551463 2747it under the terms of the GNU General Public License as published by
f16b0819
PE
2748the Free Software Foundation, either version 3 of the License, or
2749(at your option) any later version.
76551463 2750
f16b0819 2751This program is distributed in the hope that it will be useful,
76551463
AD
2752but WITHOUT ANY WARRANTY; without even the implied warranty of
2753MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2754GNU General Public License for more details.
2755
2756You should have received a copy of the GNU General Public License
f16b0819 2757along with this program. If not, see <http://www.gnu.org/licenses/>.
7e508a2b
AD
2758
2759 LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr
2760 LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr
2761 LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar
2762 LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr
2763 LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks
2764 LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG
2765 LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush
2766 LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly
2767 LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's
2768 LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de
2769 LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto
2770 LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs
2771 LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF
242cc08e 2772 LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY
7e508a2b 2773 LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL
242cc08e
AD
2774 LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh
2775 LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf
dcb366b1 2776 LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp Wother nterm arg init
1282c124
AD
2777 LocalWords: TOK calc yyo fval Wconflicts parsers yystackp yyval yynerrs
2778 LocalWords: Théophile Ranquet Santet fno fnone stype associativity Tolmer
2779 LocalWords: Wprecedence Rassoul Wempty Paolo Bonzini parser's Michiel loc
2780 LocalWords: redeclaration sval fcaret reentrant XSLT xsl Wmaybe yyvsp Tedi
2781 LocalWords: pragmas noreturn untyped Rozenman unexpanded Wojciech Polak
6574576c 2782 LocalWords: Alexandre MERCHANTABILITY yytype
7e508a2b
AD
2783
2784Local Variables:
2785mode: outline
e4ab1254 2786fill-column: 76
7e508a2b 2787End: