]> git.saurik.com Git - bison.git/blame - NEWS
news: YYERROR vs variants
[bison.git] / NEWS
CommitLineData
ed7658fe 1GNU Bison NEWS
3af4feb2 2
6d94eebb
AD
3* Noteworthy changes in release ?.? (????-??-??) [?]
4
7b0ca050
AD
5** Bug fixes
6
7*** YYERROR and variants
8
9 When C++ variant support is enabled, an error triggered via YYERROR, but
10 not caught via error recovery, resulted in a double deletion.
6d94eebb 11
c2ecada3 12* Noteworthy changes in release 3.0.1 (2013-11-12) [stable]
e386b50f 13
0a244a22
AD
14** Bug fixes
15
39bace5d
AD
16*** Errors in caret diagnostics
17
18 On some platforms, some errors could result in endless diagnostics.
e4678430
AD
19
20*** Fixes of the -Werror option
21
22 Options such as "-Werror -Wno-error=foo" were still turning "foo"
23 diagnostics into errors instead of warnings. This is fixed.
24
25 Actually, for consistency with GCC, "-Wno-error=foo -Werror" now also
26 leaves "foo" diagnostics as warnings. Similarly, with "-Werror=foo
27 -Wno-error", "foo" diagnostics are now errors.
e386b50f 28
071863b3
AD
29*** GLR Predicates
30
31 As demonstrated in the documentation, one can now leave spaces between
32 "%?" and its "{".
33
265640d5
AD
34*** Installation
35
36 The yacc.1 man page is no longer installed if --disable-yacc was
37 specified.
38
39bace5d
AD
39*** Fixes in the test suite
40
41 Bugs and portability issues.
42
534497f5 43* Noteworthy changes in release 3.0 (2013-07-25) [stable]
facb910c 44
8458a411 45** WARNING: Future backward-incompatibilities!
597afd73 46
597afd73
AD
47 Like other GNU packages, Bison will start using some of the C99 features
48 for its own code, especially the definition of variables after statements.
49 The generated C parsers still aim at C90.
50
8458a411 51** Backward incompatible changes
47db7ed1
AD
52
53*** Obsolete features
54
40bb6f78
AD
55 Support for YYFAIL is removed (deprecated in Bison 2.4.2): use YYERROR.
56
57 Support for yystype and yyltype is removed (deprecated in Bison 1.875):
58 use YYSTYPE and YYLTYPE.
59
60 Support for YYLEX_PARAM and YYPARSE_PARAM is removed (deprecated in Bison
61 1.875): use %lex-param, %parse-param, or %param.
47db7ed1 62
05e25f23
AD
63 Missing semicolons at the end of actions are no longer added (as announced
64 in the release 2.5).
65
1fa19a76
AD
66*** Use of YACC='bison -y'
67
68 TL;DR: With Autoconf <= 2.69, pass -Wno-yacc to (AM_)YFLAGS if you use
69 Bison extensions.
70
71 Traditional Yacc generates 'y.tab.c' whatever the name of the input file.
72 Therefore Makefiles written for Yacc expect 'y.tab.c' (and possibly
73 'y.tab.h' and 'y.outout') to be generated from 'foo.y'.
74
75 To this end, for ages, AC_PROG_YACC, Autoconf's macro to look for an
76 implementation of Yacc, was using Bison as 'bison -y'. While it does
77 ensure compatible output file names, it also enables warnings for
78 incompatibilities with POSIX Yacc. In other words, 'bison -y' triggers
79 warnings for Bison extensions.
80
81 Autoconf 2.70+ fixes this incompatibility by using YACC='bison -o y.tab.c'
82 (which also generates 'y.tab.h' and 'y.output' when needed).
83 Alternatively, disable Yacc warnings by passing '-Wno-yacc' to your Yacc
84 flags (YFLAGS, or AM_YFLAGS with Automake).
85
597afd73
AD
86** Bug fixes
87
c21e515e 88*** The epilogue is no longer affected by internal #defines (glr.c)
597afd73
AD
89
90 The glr.c skeleton uses defines such as #define yylval (yystackp->yyval) in
91 generated code. These weren't properly undefined before the inclusion of
92 the user epilogue, so functions such as the following were butchered by the
93 preprocessor expansion:
94
95 int yylex (YYSTYPE *yylval);
96
6c7022f7 97 This is fixed: yylval, yynerrs, yychar, and yylloc are now valid
597afd73
AD
98 identifiers for user-provided variables.
99
f0f95a50
AD
100*** stdio.h is no longer needed when locations are enabled (yacc.c)
101
102 Changes in Bison 2.7 introduced a dependency on FILE and fprintf when
103 locations are enabled. This is fixed.
104
bb4b189b
AD
105*** Warnings about useless %pure-parser/%define api.pure are restored
106
597afd73
AD
107** Diagnostics reported by Bison
108
109 Most of these features were contributed by Théophile Ranquet and Victor
110 Santet.
73370a9d 111
016426c1
TR
112*** Carets
113
114 Version 2.7 introduced caret errors, for a prettier output. These are now
115 activated by default. The old format can still be used by invoking Bison
116 with -fno-caret (or -fnone).
117
fec5f3c0
AD
118 Some error messages that reproduced excerpts of the grammar are now using
119 the caret information only. For instance on:
120
121 %%
122 exp: 'a' | 'a';
123
124 Bison 2.7 reports:
125
126 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
127 in.y:2.12-14: warning: rule useless in parser due to conflicts: exp: 'a' [-Wother]
128
129 Now bison reports:
130
131 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
132 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
133 exp: 'a' | 'a';
134 ^^^
135
136 and "bison -fno-caret" reports:
137
138 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
139 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
140
1048a1c9 141*** Enhancements of the -Werror option
518e8830 142
1048a1c9 143 The -Werror=CATEGORY option is now recognized, and will treat specified
d949eefd
AD
144 warnings as errors. The warnings need not have been explicitly activated
145 using the -W option, this is similar to what GCC 4.7 does.
1048a1c9
AD
146
147 For example, given the following command line, Bison will treat both
d949eefd 148 warnings related to POSIX Yacc incompatibilities and S/R conflicts as
1048a1c9
AD
149 errors (and only those):
150
151 $ bison -Werror=yacc,error=conflicts-sr input.y
152
153 If no categories are specified, -Werror will make all active warnings into
154 errors. For example, the following line does the same the previous example:
155
156 $ bison -Werror -Wnone -Wyacc -Wconflicts-sr input.y
157
158 (By default -Wconflicts-sr,conflicts-rr,deprecated,other is enabled.)
159
160 Note that the categories in this -Werror option may not be prefixed with
161 "no-". However, -Wno-error[=CATEGORY] is valid.
162
163 Note that -y enables -Werror=yacc. Therefore it is now possible to require
164 Yacc-like behavior (e.g., always generate y.tab.c), but to report
165 incompatibilities as warnings: "-y -Wno-error=yacc".
518e8830 166
46bdb8ec 167*** The display of warnings is now richer
73370a9d 168
46bdb8ec 169 The option that controls a given warning is now displayed:
73370a9d 170
46bdb8ec 171 foo.y:4.6: warning: type clash on default action: <foo> != <bar> [-Wother]
73370a9d 172
46bdb8ec
TR
173 In the case of warnings treated as errors, the prefix is changed from
174 "warning: " to "error: ", and the suffix is displayed, in a manner similar
d949eefd 175 to GCC, as [-Werror=CATEGORY].
1048a1c9 176
46bdb8ec
TR
177 For instance, where the previous version of Bison would report (and exit
178 with failure):
1048a1c9
AD
179
180 bison: warnings being treated as errors
46bdb8ec 181 input.y:1.1: warning: stray ',' treated as white space
1048a1c9 182
46bdb8ec 183 it now reports:
1048a1c9 184
1048a1c9
AD
185 input.y:1.1: error: stray ',' treated as white space [-Werror=other]
186
187*** Deprecated constructs
188
189 The new 'deprecated' warning category flags obsolete constructs whose
190 support will be discontinued. It is enabled by default. These warnings
191 used to be reported as 'other' warnings.
192
73370a9d 193*** Useless semantic types
9641b918
VS
194
195 Bison now warns about useless (uninhabited) semantic types. Since
196 semantic types are not declared to Bison (they are defined in the opaque
197 %union structure), it is %printer/%destructor directives about useless
198 types that trigger the warning:
199
200 %token <type1> term
201 %type <type2> nterm
202 %printer {} <type1> <type3>
203 %destructor {} <type2> <type4>
204 %%
205 nterm: term { $$ = $1; };
206
207 3.28-34: warning: type <type3> is used, but is not associated to any symbol
208 4.28-34: warning: type <type4> is used, but is not associated to any symbol
209
31557b9e 210*** Undefined but unused symbols
b921d92f 211
31557b9e
AD
212 Bison used to raise an error for undefined symbols that are not used in
213 the grammar. This is now only a warning.
b921d92f
VS
214
215 %printer {} symbol1
216 %destructor {} symbol2
31557b9e 217 %type <type> symbol3
b921d92f
VS
218 %%
219 exp: "a";
220
73370a9d 221*** Useless destructors or printers
ea9a35c6
VS
222
223 Bison now warns about useless destructors or printers. In the following
224 example, the printer for <type1>, and the destructor for <type2> are
225 useless: all symbols of <type1> (token1) already have a printer, and all
226 symbols of type <type2> (token2) already have a destructor.
227
228 %token <type1> token1
229 <type2> token2
230 <type3> token3
231 <type4> token4
232 %printer {} token1 <type1> <type3>
233 %destructor {} token2 <type2> <type4>
234
d87ea54c
AD
235*** Conflicts
236
237 The warnings and error messages about shift/reduce and reduce/reduce
238 conflicts have been normalized. For instance on the following foo.y file:
239
240 %glr-parser
241 %%
242 exp: exp '+' exp | '0' | '0';
243
244 compare the previous version of bison:
245
246 $ bison foo.y
247 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
248 $ bison -Werror foo.y
249 bison: warnings being treated as errors
250 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
251
252 with the new behavior:
253
254 $ bison foo.y
255 foo.y: warning: 1 shift/reduce conflict [-Wconflicts-sr]
256 foo.y: warning: 2 reduce/reduce conflicts [-Wconflicts-rr]
257 $ bison -Werror foo.y
9503b0a4
TR
258 foo.y: error: 1 shift/reduce conflict [-Werror=conflicts-sr]
259 foo.y: error: 2 reduce/reduce conflicts [-Werror=conflicts-rr]
d87ea54c
AD
260
261 When %expect or %expect-rr is used, such as with bar.y:
262
263 %expect 0
264 %glr-parser
265 %%
266 exp: exp '+' exp | '0' | '0';
267
268 Former behavior:
269
270 $ bison bar.y
271 bar.y: conflicts: 1 shift/reduce, 2 reduce/reduce
272 bar.y: expected 0 shift/reduce conflicts
273 bar.y: expected 0 reduce/reduce conflicts
274
275 New one:
276
277 $ bison bar.y
9503b0a4
TR
278 bar.y: error: shift/reduce conflicts: 1 found, 0 expected
279 bar.y: error: reduce/reduce conflicts: 2 found, 0 expected
d87ea54c 280
f24695ef
AD
281** Incompatibilities with POSIX Yacc
282
283 The 'yacc' category is no longer part of '-Wall', enable it explicitly
284 with '-Wyacc'.
285
2055a44e
AD
286** Additional yylex/yyparse arguments
287
6dc4663d
AD
288 The new directive %param declares additional arguments to both yylex and
289 yyparse. The %lex-param, %parse-param, and %param directives support one
290 or more arguments. Instead of
2055a44e 291
e436fa67
AD
292 %lex-param {arg1_type *arg1}
293 %lex-param {arg2_type *arg2}
294 %parse-param {arg1_type *arg1}
295 %parse-param {arg2_type *arg2}
2055a44e
AD
296
297 one may now declare
298
e436fa67 299 %param {arg1_type *arg1} {arg2_type *arg2}
2055a44e 300
630a0218
AD
301** Types of values for %define variables
302
303 Bison used to make no difference between '%define foo bar' and '%define
304 foo "bar"'. The former is now called a 'keyword value', and the latter a
305 'string value'. A third kind was added: 'code values', such as '%define
306 foo {bar}'.
307
308 Keyword variables are used for fixed value sets, e.g.,
309
310 %define lr.type lalr
311
312 Code variables are used for value in the target language, e.g.,
313
314 %define api.value.type {struct semantic_type}
315
316 String variables are used remaining cases, e.g. file names.
317
2a6b66c5 318** Variable api.token.prefix
99c08fb6 319
2a6b66c5 320 The variable api.token.prefix changes the way tokens are identified in
99c08fb6
AD
321 the generated files. This is especially useful to avoid collisions
322 with identifiers in the target language. For instance
323
e436fa67 324 %token FILE for ERROR
630a0218 325 %define api.token.prefix {TOK_}
e436fa67
AD
326 %%
327 start: FILE for ERROR;
99c08fb6
AD
328
329 will generate the definition of the symbols TOK_FILE, TOK_for, and
330 TOK_ERROR in the generated sources. In particular, the scanner must
331 use these prefixed token names, although the grammar itself still
332 uses the short names (as in the sample rule given above).
333
6574576c
AD
334** Variable api.value.type
335
336 This new %define variable supersedes the #define macro YYSTYPE. The use
337 of YYSTYPE is discouraged. In particular, #defining YYSTYPE *and* either
338 using %union or %defining api.value.type results in undefined behavior.
339
340 Either define api.value.type, or use "%union":
341
342 %union
343 {
344 int ival;
345 char *sval;
346 }
347 %token <ival> INT "integer"
348 %token <sval> STRING "string"
349 %printer { fprintf (yyo, "%d", $$); } <ival>
350 %destructor { free ($$); } <sval>
351
352 /* In yylex(). */
353 yylval.ival = 42; return INT;
354 yylval.sval = "42"; return STRING;
355
630a0218
AD
356 The %define variable api.value.type supports both keyword and code values.
357
358 The keyword value 'union' means that the user provides genuine types, not
435575cb 359 union member names such as "ival" and "sval" above (WARNING: will fail if
1fa19a76 360 -y/--yacc/%yacc is enabled).
6574576c 361
435575cb 362 %define api.value.type union
6574576c
AD
363 %token <int> INT "integer"
364 %token <char *> STRING "string"
365 %printer { fprintf (yyo, "%d", $$); } <int>
366 %destructor { free ($$); } <char *>
367
368 /* In yylex(). */
369 yylval.INT = 42; return INT;
370 yylval.STRING = "42"; return STRING;
371
435575cb
AD
372 The keyword value variant is somewhat equivalent, but for C++ special
373 provision is made to allow classes to be used (more about this below).
6574576c 374
435575cb 375 %define api.value.type variant
6574576c
AD
376 %token <int> INT "integer"
377 %token <std::string> STRING "string"
378
630a0218 379 Code values (in braces) denote user defined types. This is where YYSTYPE
435575cb 380 used to be used.
6574576c
AD
381
382 %code requires
383 {
384 struct my_value
385 {
386 enum
387 {
388 is_int, is_string
389 } kind;
390 union
391 {
392 int ival;
393 char *sval;
394 } u;
395 };
396 }
435575cb 397 %define api.value.type {struct my_value}
6574576c
AD
398 %token <u.ival> INT "integer"
399 %token <u.sval> STRING "string"
400 %printer { fprintf (yyo, "%d", $$); } <u.ival>
401 %destructor { free ($$); } <u.sval>
402
403 /* In yylex(). */
404 yylval.u.ival = 42; return INT;
405 yylval.u.sval = "42"; return STRING;
406
31b850d2
AD
407** Variable parse.error
408
1f77b2e0
AD
409 This variable controls the verbosity of error messages. The use of the
410 %error-verbose directive is deprecated in favor of "%define parse.error
411 verbose".
31b850d2 412
c21e515e
AD
413** Renamed %define variables
414
415 The following variables have been renamed for consistency. Backward
416 compatibility is ensured, but upgrading is recommended.
417
418 lr.default-reductions -> lr.default-reduction
419 lr.keep-unreachable-states -> lr.keep-unreachable-state
420 namespace -> api.namespace
421 stype -> api.value.type
422
ca2a6d15
PH
423** Semantic predicates
424
597afd73
AD
425 Contributed by Paul Hilfinger.
426
1f77b2e0
AD
427 The new, experimental, semantic-predicate feature allows actions of the
428 form "%?{ BOOLEAN-EXPRESSION }", which cause syntax errors (as for
ca2a6d15 429 YYERROR) if the expression evaluates to 0, and are evaluated immediately
1f77b2e0
AD
430 in GLR parsers, rather than being deferred. The result is that they allow
431 the programmer to prune possible parses based on the values of run-time
432 expressions.
ca2a6d15 433
d1400569
AD
434** The directive %expect-rr is now an error in non GLR mode
435
436 It used to be an error only if used in non GLR mode, _and_ if there are
437 reduce/reduce conflicts.
438
5202b6ac 439** Tokens are numbered in their order of appearance
93561c21 440
5202b6ac
VT
441 Contributed by Valentin Tolmer.
442
443 With '%token A B', A had a number less than the one of B. However,
444 precedence declarations used to generate a reversed order. This is now
445 fixed, and introducing tokens with any of %token, %left, %right,
446 %precedence, or %nonassoc yields the same result.
447
448 When mixing declarations of tokens with a litteral character (e.g., 'a')
449 or with an identifier (e.g., B) in a precedence declaration, Bison
450 numbered the litteral characters first. For example
451
452 %right A B 'c' 'd'
453
454 would lead to the tokens declared in this order: 'c' 'd' A B. Again, the
455 input order is now preserved.
456
457 These changes were made so that one can remove useless precedence and
458 associativity declarations (i.e., map %nonassoc, %left or %right to
459 %precedence, or to %token) and get exactly the same output.
93561c21 460
cc2235ac
VT
461** Useless precedence and associativity
462
d2f9ae18
AD
463 Contributed by Valentin Tolmer.
464
1282c124 465 When developing and maintaining a grammar, useless associativity and
cc2235ac
VT
466 precedence directives are common. They can be a nuisance: new ambiguities
467 arising are sometimes masked because their conflicts are resolved due to
468 the extra precedence or associativity information. Furthermore, it can
469 hinder the comprehension of a new grammar: one will wonder about the role
470 of a precedence, where in fact it is useless. The following changes aim
471 at detecting and reporting these extra directives.
472
473*** Precedence warning category
474
475 A new category of warning, -Wprecedence, was introduced. It flags the
476 useless precedence and associativity directives.
477
478*** Useless associativity
479
480 Bison now warns about symbols with a declared associativity that is never
481 used to resolve conflicts. In that case, using %precedence is sufficient;
482 the parsing tables will remain unchanged. Solving these warnings may raise
483 useless precedence warnings, as the symbols no longer have associativity.
484 For example:
485
486 %left '+'
487 %left '*'
488 %%
489 exp:
1282c124
AD
490 "number"
491 | exp '+' "number"
cc2235ac
VT
492 | exp '*' exp
493 ;
494
495 will produce a
496
497 warning: useless associativity for '+', use %precedence [-Wprecedence]
498 %left '+'
499 ^^^
500
501*** Useless precedence
502
503 Bison now warns about symbols with a declared precedence and no declared
504 associativity (i.e., declared with %precedence), and whose precedence is
505 never used. In that case, the symbol can be safely declared with %token
506 instead, without modifying the parsing tables. For example:
507
508 %precedence '='
509 %%
1282c124 510 exp: "var" '=' "number";
cc2235ac
VT
511
512 will produce a
513
514 warning: useless precedence for '=' [-Wprecedence]
515 %precedence '='
516 ^^^
517
518*** Useless precedence and associativity
519
520 In case of both useless precedence and associativity, the issue is flagged
521 as follows:
522
523 %nonassoc '='
524 %%
1282c124 525 exp: "var" '=' "number";
cc2235ac
VT
526
527 The warning is:
528
529 warning: useless precedence and associativity for '=' [-Wprecedence]
530 %nonassoc '='
531 ^^^
09add9c2
AD
532
533** Empty rules
534
6240346a
AD
535 With help from Joel E. Denny and Gabriel Rassoul.
536
09add9c2
AD
537 Empty rules (i.e., with an empty right-hand side) can now be explicitly
538 marked by the new %empty directive. Using %empty on a non-empty rule is
539 an error. The new -Wempty-rule warning reports empty rules without
540 %empty. On the following grammar:
541
542 %%
543 s: a b c;
544 a: ;
545 b: %empty;
546 c: 'a' %empty;
547
548 bison reports:
549
550 3.4-5: warning: empty rule without %empty [-Wempty-rule]
551 a: {}
552 ^^
553 5.8-13: error: %empty on non-empty rule
554 c: 'a' %empty {};
555 ^^^^^^
cc2235ac 556
c21e515e
AD
557** Java skeleton improvements
558
c21e515e
AD
559 The constants for token names were moved to the Lexer interface. Also, it
560 is possible to add code to the parser's constructors using "%code init"
561 and "%define init_throws".
aa94def1
DH
562 Contributed by Paolo Bonzini.
563
de1a2f20
AD
564 The Java skeleton now supports push parsing.
565 Contributed by Dennis Heimbigner.
566
c21e515e
AD
567** C++ skeletons improvements
568
569*** The parser header is no longer mandatory (lalr1.cc, glr.cc)
570
571 Using %defines is now optional. Without it, the needed support classes
572 are defined in the generated parser, instead of additional files (such as
573 location.hh, position.hh and stack.hh).
574
575*** Locations are no longer mandatory (lalr1.cc, glr.cc)
576
577 Both lalr1.cc and glr.cc no longer require %location.
578
579*** syntax_error exception (lalr1.cc)
580
581 The C++ parser features a syntax_error exception, which can be
582 thrown from the scanner or from user rules to raise syntax errors.
583 This facilitates reporting errors caught in sub-functions (e.g.,
584 rejecting too large integral literals from a conversion function
585 used by the scanner, or rejecting invalid combinations from a
586 factory invoked by the user actions).
587
588*** %define api.value.type variant
589
590 This is based on a submission from Michiel De Wilde. With help
591 from Théophile Ranquet.
592
593 In this mode, complex C++ objects can be used as semantic values. For
594 instance:
595
596 %token <::std::string> TEXT;
597 %token <int> NUMBER;
598 %token SEMICOLON ";"
599 %type <::std::string> item;
600 %type <::std::list<std::string>> list;
601 %%
602 result:
603 list { std::cout << $1 << std::endl; }
604 ;
605
606 list:
6240346a 607 %empty { /* Generates an empty string list. */ }
c21e515e
AD
608 | list item ";" { std::swap ($$, $1); $$.push_back ($2); }
609 ;
610
611 item:
612 TEXT { std::swap ($$, $1); }
613 | NUMBER { $$ = string_cast ($1); }
614 ;
615
616*** %define api.token.constructor
617
618 When variants are enabled, Bison can generate functions to build the
619 tokens. This guarantees that the token type (e.g., NUMBER) is consistent
620 with the semantic value (e.g., int):
621
622 parser::symbol_type yylex ()
623 {
624 parser::location_type loc = ...;
625 ...
626 return parser::make_TEXT ("Hello, world!", loc);
627 ...
628 return parser::make_NUMBER (42, loc);
629 ...
630 return parser::make_SEMICOLON (loc);
631 ...
632 }
633
75ae8299
AD
634*** C++ locations
635
636 There are operator- and operator-= for 'location'. Negative line/column
637 increments can no longer underflow the resulting value.
638
1f5542fe 639* Noteworthy changes in release 2.7.1 (2013-04-15) [stable]
cc8962bd 640
80a2826e
AD
641** Bug fixes
642
643*** Fix compiler attribute portability (yacc.c)
644
645 With locations enabled, __attribute__ was used unprotected.
0a7b8559 646
e83be476
AD
647*** Fix some compiler warnings (lalr1.cc)
648
c13bb348 649* Noteworthy changes in release 2.7 (2012-12-12) [stable]
effd30c0 650
edf9a06f 651** Bug fixes
7bada535 652
edf9a06f 653 Warnings about uninitialized yylloc in yyparse have been fixed.
7bada535 654
1127a75a
AD
655 Restored C90 compliance (yet no report was ever made).
656
d4fe9e88 657** Diagnostics are improved
7bada535 658
8458a411
AD
659 Contributed by Théophile Ranquet.
660
d4fe9e88 661*** Changes in the format of error messages
7bada535 662
d4fe9e88 663 This used to be the format of many error reports:
1f1bd572 664
d4fe9e88
AD
665 input.y:2.7-12: %type redeclaration for exp
666 input.y:1.7-12: previous declaration
1f1bd572 667
d4fe9e88 668 It is now:
1f1bd572 669
d4fe9e88
AD
670 input.y:2.7-12: error: %type redeclaration for exp
671 input.y:1.7-12: previous declaration
cbaea010 672
d4fe9e88 673*** New format for error reports: carets
cbaea010 674
d4fe9e88 675 Caret errors have been added to Bison:
cbaea010 676
d4fe9e88
AD
677 input.y:2.7-12: error: %type redeclaration for exp
678 %type <sval> exp
679 ^^^^^^
680 input.y:1.7-12: previous declaration
681 %type <ival> exp
682 ^^^^^^
cbaea010 683
d4fe9e88 684 or
cbaea010 685
7bada535 686 input.y:3.20-23: error: ambiguous reference: '$exp'
fb6040f0 687 exp: exp '+' exp { $exp = $1 + $3; };
7bada535 688 ^^^^
fb6040f0
TR
689 input.y:3.1-3: refers to: $exp at $$
690 exp: exp '+' exp { $exp = $1 + $3; };
691 ^^^
692 input.y:3.6-8: refers to: $exp at $1
693 exp: exp '+' exp { $exp = $1 + $3; };
694 ^^^
695 input.y:3.14-16: refers to: $exp at $3
696 exp: exp '+' exp { $exp = $1 + $3; };
697 ^^^
7bada535 698
1282c124
AD
699 The default behavior for now is still not to display these unless
700 explicitly asked with -fcaret (or -fall). However, in a later release, it
d4fe9e88
AD
701 will be made the default behavior (but may still be deactivated with
702 -fno-caret).
d3e4409a 703
1f1bd572 704** New value for %define variable: api.pure full
d3e4409a 705
1f1bd572 706 The %define variable api.pure requests a pure (reentrant) parser. However,
d4fe9e88
AD
707 for historical reasons, using it in a location-tracking Yacc parser
708 resulted in a yyerror function that did not take a location as a
709 parameter. With this new value, the user may request a better pure parser,
710 where yyerror does take a location as a parameter (in location-tracking
711 parsers).
1f1bd572
TR
712
713 The use of "%define api.pure true" is deprecated in favor of this new
714 "%define api.pure full".
d3e4409a 715
7287be84 716** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java)
db8ab2be
AD
717
718 The %define variable api.location.type defines the name of the type to use
719 for locations. When defined, Bison no longer generates the position.hh
720 and location.hh files, nor does the parser will include them: the user is
721 then responsible to define her type.
722
723 This can be used in programs with several parsers to factor their location
7287be84
AD
724 and position files: let one of them generate them, and the others just use
725 them.
db8ab2be
AD
726
727 This feature was actually introduced, but not documented, in Bison 2.5,
728 under the name "location_type" (which is maintained for backward
729 compatibility).
730
7287be84
AD
731 For consistency, lalr1.java's %define variables location_type and
732 position_type are deprecated in favor of api.location.type and
733 api.position.type.
734
d4fe9e88
AD
735** Exception safety (lalr1.cc)
736
737 The parse function now catches exceptions, uses the %destructors to
738 release memory (the lookahead symbol and the symbols pushed on the stack)
739 before re-throwing the exception.
740
741 This feature is somewhat experimental. User feedback would be
742 appreciated.
743
9c16d399 744** Graph improvements in DOT and XSLT
fc4fdd62 745
8458a411
AD
746 Contributed by Théophile Ranquet.
747
fc4fdd62
TR
748 The graphical presentation of the states is more readable: their shape is
749 now rectangular, the state number is clearly displayed, and the items are
750 numbered and left-justified.
751
752 The reductions are now explicitly represented as transitions to other
753 diamond shaped nodes.
754
9c16d399
TR
755 These changes are present in both --graph output and xml2dot.xsl XSLT
756 processing, with minor (documented) differences.
757
d4fe9e88 758** %language is no longer an experimental feature.
fb4c8a7c 759
d4fe9e88
AD
760 The introduction of this feature, in 2.4, was four years ago. The
761 --language option and the %language directive are no longer experimental.
fb4c8a7c 762
53e2cd1e
AD
763** Documentation
764
765 The sections about shift/reduce and reduce/reduce conflicts resolution
766 have been fixed and extended.
9d2423f5 767
d4fe9e88
AD
768 Although introduced more than four years ago, XML and Graphviz reports
769 were not properly documented.
770
be22823e
AD
771 The translation of mid-rule actions is now described.
772
9d3f7eaf 773* Noteworthy changes in release 2.6.5 (2012-11-07) [stable]
6f1360bd 774
a68b1f23
AD
775 We consider compiler warnings about Bison generated parsers to be bugs.
776 Rather than working around them in your own project, please consider
777 reporting them to us.
778
779** Bug fixes
780
781 Warnings about uninitialized yylval and/or yylloc for push parsers with a
782 pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to
783 3.2.
784
785 Other issues in the test suite have been addressed.
6f1360bd 786
1282c124 787 Null characters are correctly displayed in error messages.
95066e92 788
a1d1ab50
AD
789 When possible, yylloc is correctly initialized before calling yylex. It
790 is no longer necessary to initialize it in the %initial-action.
791
0ac15849 792* Noteworthy changes in release 2.6.4 (2012-10-23) [stable]
a4eb820f 793
468455e1 794 Bison 2.6.3's --version was incorrect. This release fixes this issue.
a4eb820f 795
6eb8f74f 796* Noteworthy changes in release 2.6.3 (2012-10-22) [stable]
933ec544 797
6b4cb804
AD
798** Bug fixes
799
a1a77e1f 800 Bugs and portability issues in the test suite have been fixed.
6b4cb804
AD
801
802 Some errors in translations have been addressed, and --help now directs
803 users to the appropriate place to report them.
804
805 Stray Info files shipped by accident are removed.
806
807 Incorrect definitions of YY_, issued by yacc.c when no parser header is
808 generated, are removed.
9c26b8fc 809
a2b3f101
TR
810 All the generated headers are self-contained.
811
c9d5bcc9
AD
812** Header guards (yacc.c, glr.c, glr.cc)
813
814 In order to avoid collisions, the header guards are now
815 YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>.
816 For instance the header generated from
817
818 %define api.prefix "calc"
819 %defines "lib/parse.h"
820
821 will use YY_CALC_LIB_PARSE_H_INCLUDED as guard.
822
c12c4c50 823** Fix compiler warnings in the generated parser (yacc.c, glr.c)
321d3e35
AD
824
825 The compilation of pure parsers (%define api.pure) can trigger GCC
826 warnings such as:
827
828 input.c: In function 'yyparse':
829 input.c:1503:12: warning: 'yylval' may be used uninitialized in this
830 function [-Wmaybe-uninitialized]
831 *++yyvsp = yylval;
832 ^
833
834 This is now fixed; pragmas to avoid these warnings are no longer needed.
835
c12c4c50
AD
836 Warnings from clang ("equality comparison with extraneous parentheses" and
837 "function declared 'noreturn' should not return") have also been
838 addressed.
839
e1eeecd3 840* Noteworthy changes in release 2.6.2 (2012-08-03) [stable]
9c26b8fc 841
43ca8040
AD
842** Bug fixes
843
844 Buffer overruns, complaints from Flex, and portability issues in the test
845 suite have been fixed.
846
c9d546b2
AD
847** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc)
848
849 Trailing end-of-lines in %parse-param or %lex-param would result in
850 invalid C++. This is fixed.
9c26b8fc 851
dcd5344d
AD
852** Spurious spaces and end-of-lines
853
854 The generated files no longer end (nor start) with empty lines.
855
77b214ef 856* Noteworthy changes in release 2.6.1 (2012-07-30) [stable]
a4107f24 857
8617d87e
AD
858 Bison no longer executes user-specified M4 code when processing a grammar.
859
e20e6a50
AD
860** Future Changes
861
862 In addition to the removal of the features announced in Bison 2.6, the
863 next major release will remove the "Temporary hack for adding a semicolon
864 to the user action", as announced in the release 2.5. Instead of:
865
866 exp: exp "+" exp { $$ = $1 + $3 };
867
868 write:
869
870 exp: exp "+" exp { $$ = $1 + $3; };
871
8617d87e
AD
872** Bug fixes
873
0e164d43
AD
874*** Type names are now properly escaped.
875
876*** glr.cc: set_debug_level and debug_level work as expected.
a4107f24 877
26313726
AD
878*** Stray @ or $ in actions
879
880 While Bison used to warn about stray $ or @ in action rules, it did not
881 for other actions such as printers, destructors, or initial actions. It
882 now does.
883
cd735a8c 884** Type names in actions
4982f078
AD
885
886 For consistency with rule actions, it is now possible to qualify $$ by a
cd735a8c 887 type-name in destructors, printers, and initial actions. For instance:
4982f078
AD
888
889 %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>;
890
891 will display two values for each typed and untyped symbol (provided
cd735a8c 892 that YYSTYPE has both "ival" and "fval" fields).
60aa04a2 893
1505e8bb 894* Noteworthy changes in release 2.6 (2012-07-19) [stable]
0f11eec2 895
d0a30438 896** Future changes
9553083c 897
55d1006f
AD
898 The next major release of Bison will drop support for the following
899 deprecated features. Please report disagreements to bug-bison@gnu.org.
0f11eec2 900
aaf61036 901*** K&R C parsers
55d1006f
AD
902
903 Support for generating parsers in K&R C will be removed. Parsers
242cc08e 904 generated for C support ISO C90, and are tested with ISO C99 and ISO C11
55d1006f
AD
905 compilers.
906
258cddbc 907*** Features deprecated since Bison 1.875
0f11eec2 908
258cddbc
AD
909 The definitions of yystype and yyltype will be removed; use YYSTYPE and
910 YYLTYPE.
0f11eec2 911
258cddbc
AD
912 YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and
913 %lex-param, will no longer be supported.
914
915 Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use
916 %error-verbose.
55d1006f
AD
917
918*** The generated header will be included (yacc.c)
0f11eec2
AD
919
920 Instead of duplicating the content of the generated header (definition of
55d1006f
AD
921 YYSTYPE, yyparse declaration etc.), the generated parser will include it,
922 as is already the case for GLR or C++ parsers. This change is deferred
923 because existing versions of ylwrap (e.g., Automake 1.12.1) do not support
924 it.
0f11eec2 925
c2425191 926** Generated Parser Headers
56ca3d8f 927
258cddbc 928*** Guards (yacc.c, glr.c, glr.cc)
c3e9f08f
AD
929
930 The generated headers are now guarded, as is already the case for C++
242cc08e 931 parsers (lalr1.cc). For instance, with --defines=foo.h:
c3e9f08f 932
e29f0771
AD
933 #ifndef YY_FOO_H
934 # define YY_FOO_H
935 ...
936 #endif /* !YY_FOO_H */
c3e9f08f 937
258cddbc 938*** New declarations (yacc.c, glr.c)
56ca3d8f
AD
939
940 The generated header now declares yydebug and yyparse. Both honor
941 --name-prefix=bar_, and yield
942
e29f0771 943 int bar_parse (void);
56ca3d8f
AD
944
945 rather than
946
e29f0771
AD
947 #define yyparse bar_parse
948 int yyparse (void);
56ca3d8f
AD
949
950 in order to facilitate the inclusion of several parser headers inside a
951 single compilation unit.
c3e9f08f 952
258cddbc
AD
953*** Exported symbols in C++
954
955 The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the
956 header, are removed, as they prevent the possibility of including several
957 generated headers from a single compilation unit.
958
694af10c
AD
959*** YYLSP_NEEDED
960
961 For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no
962 longer defined.
963
4b3847c3
AD
964** New %define variable: api.prefix
965
966 Now that the generated headers are more complete and properly protected
967 against multiple inclusions, constant names, such as YYSTYPE are a
968 problem. While yyparse and others are properly renamed by %name-prefix,
969 YYSTYPE, YYDEBUG and others have never been affected by it. Because it
970 would introduce backward compatibility issues in projects not expecting
971 YYSTYPE to be renamed, instead of changing the behavior of %name-prefix,
972 it is deprecated in favor of a new %define variable: api.prefix.
973
974 The following examples compares both:
975
976 %name-prefix "bar_" | %define api.prefix "bar_"
977 %token <ival> FOO %token <ival> FOO
978 %union { int ival; } %union { int ival; }
979 %% %%
980 exp: 'a'; exp: 'a';
981
982 bison generates:
983
984 #ifndef BAR_FOO_H #ifndef BAR_FOO_H
985 # define BAR_FOO_H # define BAR_FOO_H
986
987 /* Enabling traces. */ /* Enabling traces. */
5f108727
AD
988 # ifndef YYDEBUG | # ifndef BAR_DEBUG
989 > # if defined YYDEBUG
990 > # if YYDEBUG
991 > # define BAR_DEBUG 1
992 > # else
993 > # define BAR_DEBUG 0
994 > # endif
995 > # else
996 # define YYDEBUG 0 | # define BAR_DEBUG 0
997 > # endif
998 # endif | # endif
999
1000 # if YYDEBUG | # if BAR_DEBUG
4b3847c3
AD
1001 extern int bar_debug; extern int bar_debug;
1002 # endif # endif
1003
1004 /* Tokens. */ /* Tokens. */
1005 # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE
1006 # define YYTOKENTYPE | # define BAR_TOKENTYPE
1007 enum yytokentype { | enum bar_tokentype {
1008 FOO = 258 FOO = 258
1009 }; };
1010 # endif # endif
1011
1012 #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \
1013 && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED
1014 typedef union YYSTYPE | typedef union BAR_STYPE
1015 { {
1016 int ival; int ival;
1017 } YYSTYPE; | } BAR_STYPE;
1018 # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1
1019 #endif #endif
1020
1021 extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval;
1022
1023 int bar_parse (void); int bar_parse (void);
1024
1025 #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */
1026
dfaac272 1027* Noteworthy changes in release 2.5.1 (2012-06-05) [stable]
df6e3db0 1028
debe2c03 1029** Future changes:
765e1bd4 1030
e4ab1254 1031 The next major release will drop support for generating parsers in K&R C.
041308d0 1032
466b4cf2 1033** yacc.c: YYBACKUP works as expected.
ef51bfa7 1034
d834eca0 1035** glr.c improvements:
041308d0 1036
d834eca0 1037*** Location support is eliminated when not requested:
041308d0 1038
e4ab1254
AD
1039 GLR parsers used to include location-related code even when locations were
1040 not requested, and therefore not even usable.
378e917c 1041
d834eca0 1042*** __attribute__ is preserved:
d115aad9 1043
e4ab1254
AD
1044 __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e.,
1045 when -std is passed to GCC).
041308d0 1046
466b4cf2 1047** lalr1.java: several fixes:
041308d0 1048
e4ab1254
AD
1049 The Java parser no longer throws ArrayIndexOutOfBoundsException if the
1050 first token leads to a syntax error. Some minor clean ups.
041308d0 1051
22172d47 1052** Changes for C++:
ef51bfa7 1053
22172d47 1054*** C++11 compatibility:
ef51bfa7 1055
e4ab1254
AD
1056 C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L
1057 or higher.
936c88d1 1058
22172d47
AD
1059*** Header guards
1060
1061 The header files such as "parser.hh", "location.hh", etc. used a constant
1062 name for preprocessor guards, for instance:
1063
e29f0771
AD
1064 #ifndef BISON_LOCATION_HH
1065 # define BISON_LOCATION_HH
1066 ...
1067 #endif // !BISON_LOCATION_HH
22172d47
AD
1068
1069 The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower
1070 case characters are converted to upper case, and series of
1071 non-alphanumerical characters are converted to an underscore.
1072
1073 With "bison -o lang++/parser.cc", "location.hh" would now include:
1074
e29f0771
AD
1075 #ifndef YY_LANG_LOCATION_HH
1076 # define YY_LANG_LOCATION_HH
1077 ...
1078 #endif // !YY_LANG_LOCATION_HH
22172d47
AD
1079
1080*** C++ locations:
936c88d1 1081
e4ab1254
AD
1082 The position and location constructors (and their initialize methods)
1083 accept new arguments for line and column. Several issues in the
1084 documentation were fixed.
936c88d1 1085
466b4cf2
AD
1086** liby is no longer asking for "rpl_fprintf" on some platforms.
1087
7e508a2b
AD
1088** Changes in the manual:
1089
1090*** %printer is documented
1091
e4ab1254
AD
1092 The "%printer" directive, supported since at least Bison 1.50, is finally
1093 documented. The "mfcalc" example is extended to demonstrate it.
7e508a2b 1094
e4ab1254
AD
1095 For consistency with the C skeletons, the C++ parsers now also support
1096 "yyoutput" (as an alias to "debug_stream ()").
7e508a2b
AD
1097
1098*** Several improvements have been made:
466b4cf2 1099
e4ab1254
AD
1100 The layout for grammar excerpts was changed to a more compact scheme.
1101 Named references are motivated. The description of the automaton
1102 description file (*.output) is updated to the current format. Incorrect
1103 index entries were fixed. Some other errors were fixed.
466b4cf2 1104
86b08b49
AD
1105** Building bison:
1106
1107*** Conflicting prototypes with recent/modified Flex.
466b4cf2 1108
e4ab1254
AD
1109 Fixed build problems with the current, unreleased, version of Flex, and
1110 some modified versions of 2.5.35, which have modified function prototypes.
466b4cf2 1111
8ef26c2a
AD
1112*** Warnings during the build procedure have been eliminated.
1113
1114*** Several portability problems in the test suite have been fixed:
466b4cf2 1115
e4ab1254
AD
1116 This includes warnings with some compilers, unexpected behavior of tools
1117 such as diff, warning messages from the test suite itself, etc.
466b4cf2 1118
91aadcc7 1119*** The install-pdf target works properly:
8ef26c2a 1120
e4ab1254
AD
1121 Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer
1122 halts in the middle of its course.
8ef26c2a 1123
28801043 1124* Changes in version 2.5 (2011-05-14):
50cca368 1125
82f3355e
JD
1126** Grammar symbol names can now contain non-initial dashes:
1127
1128 Consistently with directives (such as %error-verbose) and with
1129 %define variables (e.g. push-pull), grammar symbol names may contain
1130 dashes in any position except the beginning. This is a GNU
1131 extension over POSIX Yacc. Thus, use of this extension is reported
1132 by -Wyacc and rejected in Yacc mode (--yacc).
1133
f1b238df 1134** Named references:
66381412
AR
1135
1136 Historically, Yacc and Bison have supported positional references
1137 ($n, $$) to allow access to symbol values from inside of semantic
1138 actions code.
1139
1140 Starting from this version, Bison can also accept named references.
1141 When no ambiguity is possible, original symbol names may be used
1142 as named references:
1143
4b568fc0 1144 if_stmt : "if" cond_expr "then" then_stmt ';'
66381412
AR
1145 { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }
1146
1147 In the more common case, explicit names may be declared:
1148
4b568fc0 1149 stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';'
66381412
AR
1150 { $res = mk_if_stmt($cond, $then, $else); }
1151
5b1ff423 1152 Location information is also accessible using @name syntax. When
66381412
AR
1153 accessing symbol names containing dots or dashes, explicit bracketing
1154 ($[sym.1]) must be used.
1155
5b1ff423 1156 These features are experimental in this version. More user feedback
66381412 1157 will help to stabilize them.
2bd435c3 1158 Contributed by Alex Rozenman.
66381412 1159
f1b238df 1160** IELR(1) and canonical LR(1):
eb45ef3b
JD
1161
1162 IELR(1) is a minimal LR(1) parser table generation algorithm. That
1163 is, given any context-free grammar, IELR(1) generates parser tables
7262f54f 1164 with the full language-recognition power of canonical LR(1) but with
f1b238df
JD
1165 nearly the same number of parser states as LALR(1). This reduction
1166 in parser states is often an order of magnitude. More importantly,
eb45ef3b
JD
1167 because canonical LR(1)'s extra parser states may contain duplicate
1168 conflicts in the case of non-LR(1) grammars, the number of conflicts
1169 for IELR(1) is often an order of magnitude less as well. This can
1170 significantly reduce the complexity of developing of a grammar.
1171
1172 Bison can now generate IELR(1) and canonical LR(1) parser tables in
1173 place of its traditional LALR(1) parser tables, which remain the
1174 default. You can specify the type of parser tables in the grammar
1175 file with these directives:
1176
cf499cff
JD
1177 %define lr.type lalr
1178 %define lr.type ielr
1179 %define lr.type canonical-lr
eb45ef3b 1180
7fceb615 1181 The default-reduction optimization in the parser tables can also be
e4ab1254
AD
1182 adjusted using "%define lr.default-reductions". For details on both
1183 of these features, see the new section "Tuning LR" in the Bison
7fceb615 1184 manual.
eb45ef3b
JD
1185
1186 These features are experimental. More user feedback will help to
1187 stabilize them.
1188
8458a411
AD
1189** LAC (Lookahead Correction) for syntax error handling
1190
1191 Contributed by Joel E. Denny.
fcf834f9
JD
1192
1193 Canonical LR, IELR, and LALR can suffer from a couple of problems
1194 upon encountering a syntax error. First, the parser might perform
1195 additional parser stack reductions before discovering the syntax
7fceb615 1196 error. Such reductions can perform user semantic actions that are
fcf834f9
JD
1197 unexpected because they are based on an invalid token, and they
1198 cause error recovery to begin in a different syntactic context than
1199 the one in which the invalid token was encountered. Second, when
7fceb615 1200 verbose error messages are enabled (with %error-verbose or the
e4ab1254 1201 obsolete "#define YYERROR_VERBOSE"), the expected token list in the
7fceb615
JD
1202 syntax error message can both contain invalid tokens and omit valid
1203 tokens.
fcf834f9
JD
1204
1205 The culprits for the above problems are %nonassoc, default
1206 reductions in inconsistent states, and parser state merging. Thus,
1207 IELR and LALR suffer the most. Canonical LR can suffer only if
1208 %nonassoc is used or if default reductions are enabled for
1209 inconsistent states.
1210
7fceb615
JD
1211 LAC is a new mechanism within the parsing algorithm that solves
1212 these problems for canonical LR, IELR, and LALR without sacrificing
1213 %nonassoc, default reductions, or state merging. When LAC is in
1214 use, canonical LR and IELR behave almost exactly the same for both
1215 syntactically acceptable and syntactically unacceptable input.
fcf834f9
JD
1216 While LALR still does not support the full language-recognition
1217 power of canonical LR and IELR, LAC at least enables LALR's syntax
1218 error handling to correctly reflect LALR's language-recognition
1219 power.
1220
1221 Currently, LAC is only supported for deterministic parsers in C.
1222 You can enable LAC with the following directive:
1223
1224 %define parse.lac full
1225
e4ab1254 1226 See the new section "LAC" in the Bison manual for additional
7fceb615 1227 details including a few caveats.
fcf834f9
JD
1228
1229 LAC is an experimental feature. More user feedback will help to
1230 stabilize it.
1231
d397d9f0 1232** %define improvements:
cf499cff 1233
f1b238df 1234*** Can now be invoked via the command line:
50cca368 1235
de5ab940 1236 Each of these command-line options
50cca368 1237
de5ab940
JD
1238 -D NAME[=VALUE]
1239 --define=NAME[=VALUE]
1240
1241 -F NAME[=VALUE]
1242 --force-define=NAME[=VALUE]
50cca368
JD
1243
1244 is equivalent to this grammar file declaration
1245
de5ab940 1246 %define NAME ["VALUE"]
50cca368 1247
de5ab940
JD
1248 except that the manner in which Bison processes multiple definitions
1249 for the same NAME differs. Most importantly, -F and --force-define
1250 quietly override %define, but -D and --define do not. For further
e4ab1254 1251 details, see the section "Bison Options" in the Bison manual.
50cca368 1252
f1b238df 1253*** Variables renamed:
67212941
JD
1254
1255 The following %define variables
1256
1257 api.push_pull
1258 lr.keep_unreachable_states
1259
1260 have been renamed to
1261
1262 api.push-pull
1263 lr.keep-unreachable-states
1264
1265 The old names are now deprecated but will be maintained indefinitely
1266 for backward compatibility.
1267
7262f54f 1268*** Values no longer need to be quoted in the grammar file:
cf499cff
JD
1269
1270 If a %define value is an identifier, it no longer needs to be placed
1271 within quotations marks. For example,
1272
1273 %define api.push-pull "push"
1274
1275 can be rewritten as
1276
1277 %define api.push-pull push
1278
d397d9f0 1279*** Unrecognized variables are now errors not warnings.
cdf3f113 1280
d397d9f0
JD
1281*** Multiple invocations for any variable is now an error not a warning.
1282
1283** Unrecognized %code qualifiers are now errors not warnings.
1284
1285** Character literals not of length one:
1286
1287 Previously, Bison quietly converted all character literals to length
1288 one. For example, without warning, Bison interpreted the operators in
1289 the following grammar to be the same token:
1290
1291 exp: exp '++'
1292 | exp '+' exp
1293 ;
1294
1295 Bison now warns when a character literal is not of length one. In
1296 some future release, Bison will start reporting an error instead.
1297
1298** Destructor calls fixed for lookaheads altered in semantic actions:
1299
1300 Previously for deterministic parsers in C, if a user semantic action
1301 altered yychar, the parser in some cases used the old yychar value to
1302 determine which destructor to call for the lookahead upon a syntax
1303 error or upon parser return. This bug has been fixed.
1304
1305** C++ parsers use YYRHSLOC:
1306
1307 Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
1308 macro and use it in the default YYLLOC_DEFAULT. You are encouraged
e4ab1254
AD
1309 to use it. If, for instance, your location structure has "first"
1310 and "last" members, instead of
d397d9f0 1311
e29f0771
AD
1312 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1313 do \
1314 if (N) \
1315 { \
1316 (Current).first = (Rhs)[1].location.first; \
1317 (Current).last = (Rhs)[N].location.last; \
1318 } \
1319 else \
1320 { \
1321 (Current).first = (Current).last = (Rhs)[0].location.last; \
1322 } \
1323 while (false)
d397d9f0
JD
1324
1325 use:
1326
e29f0771
AD
1327 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1328 do \
1329 if (N) \
1330 { \
1331 (Current).first = YYRHSLOC (Rhs, 1).first; \
1332 (Current).last = YYRHSLOC (Rhs, N).last; \
1333 } \
1334 else \
1335 { \
1336 (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \
1337 } \
1338 while (false)
d397d9f0
JD
1339
1340** YYLLOC_DEFAULT in C++:
1341
1342 The default implementation of YYLLOC_DEFAULT used to be issued in
1343 the header file. It is now output in the implementation file, after
1344 the user %code sections so that its #ifndef guard does not try to
1345 override the user's YYLLOC_DEFAULT if provided.
cdf3f113 1346
f1b238df 1347** YYFAIL now produces warnings and Java parsers no longer implement it:
4395a9ff
JD
1348
1349 YYFAIL has existed for many years as an undocumented feature of
1350 deterministic parsers in C generated by Bison. More recently, it was
1351 a documented feature of Bison's experimental Java parsers. As
1352 promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
1353 semantic action now produces a deprecation warning, and Java parsers
1354 no longer implement YYFAIL at all. For further details, including a
1355 discussion of how to suppress C preprocessor warnings about YYFAIL
1356 being unused, see the Bison 2.4.2 NEWS entry.
1357
f1b238df 1358** Temporary hack for adding a semicolon to the user action:
197b82ba
JD
1359
1360 Previously, Bison appended a semicolon to every user action for
1361 reductions when the output language defaulted to C (specifically, when
1362 neither %yacc, %language, %skeleton, or equivalent command-line
1363 options were specified). This allowed actions such as
1364
1365 exp: exp "+" exp { $$ = $1 + $3 };
1366
1367 instead of
1368
1369 exp: exp "+" exp { $$ = $1 + $3; };
1370
1371 As a first step in removing this misfeature, Bison now issues a
1372 warning when it appends a semicolon. Moreover, in cases where Bison
1373 cannot easily determine whether a semicolon is needed (for example, an
1374 action ending with a cpp directive or a braced compound initializer),
1375 it no longer appends one. Thus, the C compiler might now complain
1376 about a missing semicolon where it did not before. Future releases of
1377 Bison will cease to append semicolons entirely.
1378
d2060f06
JD
1379** Verbose syntax error message fixes:
1380
e4ab1254 1381 When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is
7fceb615
JD
1382 specified, syntax error messages produced by the generated parser
1383 include the unexpected token as well as a list of expected tokens.
1384 The effect of %nonassoc on these verbose messages has been corrected
1385 in two ways, but a more complete fix requires LAC, described above:
d2060f06
JD
1386
1387*** When %nonassoc is used, there can exist parser states that accept no
1388 tokens, and so the parser does not always require a lookahead token
1389 in order to detect a syntax error. Because no unexpected token or
1390 expected tokens can then be reported, the verbose syntax error
1391 message described above is suppressed, and the parser instead
e4ab1254 1392 reports the simpler message, "syntax error". Previously, this
d2060f06
JD
1393 suppression was sometimes erroneously triggered by %nonassoc when a
1394 lookahead was actually required. Now verbose messages are
1395 suppressed only when all previous lookaheads have already been
1396 shifted or discarded.
1397
1398*** Previously, the list of expected tokens erroneously included tokens
1399 that would actually induce a syntax error because conflicts for them
1400 were resolved with %nonassoc in the current parser state. Such
1401 tokens are now properly omitted from the list.
1402
1403*** Expected token lists are still often wrong due to state merging
fcf834f9
JD
1404 (from LALR or IELR) and default reductions, which can both add
1405 invalid tokens and subtract valid tokens. Canonical LR almost
1406 completely fixes this problem by eliminating state merging and
1407 default reductions. However, there is one minor problem left even
1408 when using canonical LR and even after the fixes above. That is,
1409 if the resolution of a conflict with %nonassoc appears in a later
1410 parser state than the one at which some syntax error is
1411 discovered, the conflicted token is still erroneously included in
1412 the expected token list. Bison's new LAC implementation,
1413 described above, eliminates this problem and the need for
1414 canonical LR. However, LAC is still experimental and is disabled
1415 by default.
53f036ce 1416
1a33f4f6
JD
1417** Java skeleton fixes:
1418
1419*** A location handling bug has been fixed.
1420
1421*** The top element of each of the value stack and location stack is now
1422 cleared when popped so that it can be garbage collected.
6771a463 1423
02803d55
JD
1424*** Parser traces now print the top element of the stack.
1425
86408959
JD
1426** -W/--warnings fixes:
1427
e4ab1254 1428*** Bison now properly recognizes the "no-" versions of categories:
86408959
JD
1429
1430 For example, given the following command line, Bison now enables all
1431 warnings except warnings for incompatibilities with POSIX Yacc:
1432
1433 bison -Wall,no-yacc gram.y
1434
786743d5
JD
1435*** Bison now treats S/R and R/R conflicts like other warnings:
1436
1437 Previously, conflict reports were independent of Bison's normal
1438 warning system. Now, Bison recognizes the warning categories
e4ab1254 1439 "conflicts-sr" and "conflicts-rr". This change has important
786743d5
JD
1440 consequences for the -W and --warnings command-line options. For
1441 example:
1442
1443 bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
1444 bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
1445 bison -Wnone gram.y # no conflicts are reported
1446 bison -Werror gram.y # any conflict is an error
1447
1448 However, as before, if the %expect or %expect-rr directive is
1449 specified, an unexpected number of conflicts is an error, and an
1450 expected number of conflicts is not reported, so -W and --warning
1451 then have no effect on the conflict report.
1452
e4ab1254 1453*** The "none" category no longer disables a preceding "error":
bf0e44e8
JD
1454
1455 For example, for the following command line, Bison now reports
1456 errors instead of warnings for incompatibilities with POSIX Yacc:
1457
1458 bison -Werror,none,yacc gram.y
1459
e4ab1254 1460*** The "none" category now disables all Bison warnings:
c39014ae 1461
e4ab1254 1462 Previously, the "none" category disabled only Bison warnings for
c39014ae
JD
1463 which there existed a specific -W/--warning category. However,
1464 given the following command line, Bison is now guaranteed to
1465 suppress all warnings:
1466
1467 bison -Wnone gram.y
1468
1f36f544
JD
1469** Precedence directives can now assign token number 0:
1470
1471 Since Bison 2.3b, which restored the ability of precedence
1472 directives to assign token numbers, doing so for token number 0 has
1473 produced an assertion failure. For example:
1474
1475 %left END 0
1476
1477 This bug has been fixed.
1478
64877e5e 1479* Changes in version 2.4.3 (2010-08-05):
8b9e021f 1480
2bfcac9a
JD
1481** Bison now obeys -Werror and --warnings=error for warnings about
1482 grammar rules that are useless in the parser due to conflicts.
1483
8b9e021f
JD
1484** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
1485 been fixed.
1486
4ad3921d
JD
1487** Failures in the test suite for GCC 4.5 have been fixed.
1488
06cb07d5
JD
1489** Failures in the test suite for some versions of Sun Studio C++ have
1490 been fixed.
1491
9b5049bd
JD
1492** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
1493 warnings about undefined %prec identifiers will not be converted to
1494 errors in Bison 2.5. They will remain warnings, which should be
1495 sufficient for POSIX while avoiding backward compatibility issues.
1496
93d7dde9
JD
1497** Minor documentation fixes.
1498
e19a049c 1499* Changes in version 2.4.2 (2010-03-20):
74553c98 1500
f39ab286
JD
1501** Some portability problems that resulted in failures and livelocks
1502 in the test suite on some versions of at least Solaris, AIX, HP-UX,
e19a049c
JD
1503 RHEL4, and Tru64 have been addressed. As a result, fatal Bison
1504 errors should no longer cause M4 to report a broken pipe on the
f39ab286
JD
1505 affected platforms.
1506
e4ab1254 1507** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately.
8bb3a2e7
JD
1508
1509 POSIX specifies that an error be reported for any identifier that does
1510 not appear on the LHS of a grammar rule and that is not defined by
1511 %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this
1512 error report for the case when an identifier appears only after a
1513 %prec directive. It is now restored. However, for backward
1514 compatibility with recent Bison releases, it is only a warning for
1515 now. In Bison 2.5 and later, it will return to being an error.
9b5049bd
JD
1516 [Between the 2.4.2 and 2.4.3 releases, it was decided that this
1517 warning will not be converted to an error in Bison 2.5.]
8bb3a2e7 1518
d8911864
EB
1519** Detection of GNU M4 1.4.6 or newer during configure is improved.
1520
a603c6e0
JD
1521** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
1522 YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
1523 avoided.
c938d650 1524
98a345a2
JD
1525** %code is now a permanent feature.
1526
1527 A traditional Yacc prologue directive is written in the form:
1528
1529 %{CODE%}
1530
1531 To provide a more flexible alternative, Bison 2.3b introduced the
1532 %code directive with the following forms for C/C++:
1533
1534 %code {CODE}
1535 %code requires {CODE}
1536 %code provides {CODE}
1537 %code top {CODE}
1538
1539 These forms are now considered permanent features of Bison. See the
1540 %code entries in the section "Bison Declaration Summary" in the Bison
1541 manual for a summary of their functionality. See the section
1542 "Prologue Alternatives" for a detailed discussion including the
1543 advantages of %code over the traditional Yacc prologue directive.
1544
1545 Bison's Java feature as a whole including its current usage of %code
1546 is still considered experimental.
1547
1625df5b
JD
1548** YYFAIL is deprecated and will eventually be removed.
1549
1550 YYFAIL has existed for many years as an undocumented feature of
1551 deterministic parsers in C generated by Bison. Previously, it was
1552 documented for Bison's experimental Java parsers. YYFAIL is no longer
1553 documented for Java parsers and is formally deprecated in both cases.
1554 Users are strongly encouraged to migrate to YYERROR, which is
1555 specified by POSIX.
1556
1557 Like YYERROR, you can invoke YYFAIL from a semantic action in order to
1558 induce a syntax error. The most obvious difference from YYERROR is
1559 that YYFAIL will automatically invoke yyerror to report the syntax
1560 error so that you don't have to. However, there are several other
1561 subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
e4ab1254 1562 inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is
1625df5b
JD
1563 used. For a more detailed discussion, see:
1564
1565 http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
1566
1567 The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
1568 deterministic parsers in C will continue to implement it. However,
1569 because YYFAIL is already flawed, it seems futile to try to make new
1570 Bison features compatible with it. Thus, during parser generation,
1571 Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
1572 rule action. In a later release, YYFAIL will be disabled for
e4ab1254 1573 %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will
1625df5b
JD
1574 be removed altogether.
1575
1576 There exists at least one case where Bison 2.5's YYFAIL warning will
1577 be a false positive. Some projects add phony uses of YYFAIL and other
1578 Bison-defined macros for the sole purpose of suppressing C
1579 preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
1580 To avoid Bison's future warning, such YYFAIL uses can be moved to the
e4ab1254 1581 epilogue (that is, after the second "%%") in the Bison input file. In
1625df5b
JD
1582 this release (2.4.2), Bison already generates its own code to suppress
1583 C preprocessor warnings for YYFAIL, so projects can remove their own
1584 phony uses of YYFAIL if compatibility with Bison releases prior to
1585 2.4.2 is not necessary.
1586
2755de8f
AD
1587** Internationalization.
1588
1589 Fix a regression introduced in Bison 2.4: Under some circumstances,
1590 message translations were not installed although supported by the
1591 host system.
1592
74553c98 1593* Changes in version 2.4.1 (2008-12-11):
c9ba9e59 1594
0ea583d2
AD
1595** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
1596 declarations have been fixed.
1979121c 1597
0ea583d2
AD
1598** Temporary hack for adding a semicolon to the user action.
1599
1600 Bison used to prepend a trailing semicolon at the end of the user
1601 action for reductions. This allowed actions such as
1602
1603 exp: exp "+" exp { $$ = $1 + $3 };
1604
1605 instead of
1606
1607 exp: exp "+" exp { $$ = $1 + $3; };
1608
e4ab1254 1609 Some grammars still depend on this "feature". Bison 2.4.1 restores
0ea583d2
AD
1610 the previous behavior in the case of C output (specifically, when
1611 neither %language or %skeleton or equivalent command-line options
1612 are used) to leave more time for grammars depending on the old
1613 behavior to be adjusted. Future releases of Bison will disable this
1614 feature.
1615
1616** A few minor improvements to the Bison manual.
c9ba9e59 1617
402b123d 1618* Changes in version 2.4 (2008-11-02):
7bd1665a 1619
402b123d 1620** %language is an experimental feature.
ed4d67dc
JD
1621
1622 We first introduced this feature in test release 2.3b as a cleaner
1623 alternative to %skeleton. Since then, we have discussed the possibility of
1624 modifying its effect on Bison's output file names. Thus, in this release,
1625 we consider %language to be an experimental feature that will likely evolve
1626 in future releases.
7bd1665a 1627
402b123d 1628** Forward compatibility with GNU M4 has been improved.
241fda7a 1629
402b123d 1630** Several bugs in the C++ skeleton and the experimental Java skeleton have been
241fda7a
JD
1631 fixed.
1632
402b123d 1633* Changes in version 2.3b (2008-05-27):
35fe0834 1634
402b123d 1635** The quotes around NAME that used to be required in the following directive
d9df47b6
JD
1636 are now deprecated:
1637
1638 %define NAME "VALUE"
1639
e4ab1254 1640** The directive "%pure-parser" is now deprecated in favor of:
d9df47b6
JD
1641
1642 %define api.pure
1643
1644 which has the same effect except that Bison is more careful to warn about
1645 unreasonable usage in the latter case.
1646
402b123d 1647** Push Parsing
c373bf8b
JD
1648
1649 Bison can now generate an LALR(1) parser in C with a push interface. That
e4ab1254
AD
1650 is, instead of invoking "yyparse", which pulls tokens from "yylex", you can
1651 push one token at a time to the parser using "yypush_parse", which will
c373bf8b
JD
1652 return to the caller after processing each token. By default, the push
1653 interface is disabled. Either of the following directives will enable it:
1654
1655 %define api.push_pull "push" // Just push; does not require yylex.
1656 %define api.push_pull "both" // Push and pull; requires yylex.
1657
e4ab1254 1658 See the new section "A Push Parser" in the Bison manual for details.
c373bf8b 1659
59da312b
JD
1660 The current push parsing interface is experimental and may evolve. More user
1661 feedback will help to stabilize it.
1662
402b123d 1663** The -g and --graph options now output graphs in Graphviz DOT format,
8e55b3aa
JD
1664 not VCG format. Like --graph, -g now also takes an optional FILE argument
1665 and thus cannot be bundled with other short options.
c373bf8b 1666
402b123d 1667** Java
59da312b
JD
1668
1669 Bison can now generate an LALR(1) parser in Java. The skeleton is
e4ab1254 1670 "data/lalr1.java". Consider using the new %language directive instead of
59da312b
JD
1671 %skeleton to select it.
1672
e4ab1254 1673 See the new section "Java Parsers" in the Bison manual for details.
59da312b
JD
1674
1675 The current Java interface is experimental and may evolve. More user
1676 feedback will help to stabilize it.
2bd435c3 1677 Contributed by Paolo Bonzini.
59da312b 1678
402b123d 1679** %language
59da312b
JD
1680
1681 This new directive specifies the programming language of the generated
d43f77e7
PB
1682 parser, which can be C (the default), C++, or Java. Besides the skeleton
1683 that Bison uses, the directive affects the names of the generated files if
1684 the grammar file's name ends in ".y".
59da312b 1685
402b123d 1686** XML Automaton Report
59da312b
JD
1687
1688 Bison can now generate an XML report of the LALR(1) automaton using the new
e4ab1254 1689 "--xml" option. The current XML schema is experimental and may evolve. More
59da312b 1690 user feedback will help to stabilize it.
2bd435c3 1691 Contributed by Wojciech Polak.
c373bf8b 1692
402b123d 1693** The grammar file may now specify the name of the parser header file using
c373bf8b
JD
1694 %defines. For example:
1695
1696 %defines "parser.h"
1697
402b123d 1698** When reporting useless rules, useless nonterminals, and unused terminals,
d80fb37a
JD
1699 Bison now employs the terms "useless in grammar" instead of "useless",
1700 "useless in parser" instead of "never reduced", and "unused in grammar"
1701 instead of "unused".
cff03fb2 1702
402b123d 1703** Unreachable State Removal
c373bf8b
JD
1704
1705 Previously, Bison sometimes generated parser tables containing unreachable
31984206
JD
1706 states. A state can become unreachable during conflict resolution if Bison
1707 disables a shift action leading to it from a predecessor state. Bison now:
75ad86ee
JD
1708
1709 1. Removes unreachable states.
1710
1711 2. Does not report any conflicts that appeared in unreachable states.
1712 WARNING: As a result, you may need to update %expect and %expect-rr
1713 directives in existing grammar files.
1714
1715 3. For any rule used only in such states, Bison now reports the rule as
cff03fb2 1716 "useless in parser due to conflicts".
75ad86ee 1717
31984206
JD
1718 This feature can be disabled with the following directive:
1719
1720 %define lr.keep_unreachable_states
1721
e4ab1254 1722 See the %define entry in the "Bison Declaration Summary" in the Bison manual
31984206
JD
1723 for further discussion.
1724
e4ab1254 1725** Lookahead Set Correction in the ".output" Report
b1cc23c4 1726
e4ab1254
AD
1727 When instructed to generate a ".output" file including lookahead sets
1728 (using "--report=lookahead", for example), Bison now prints each reduction's
88c78747
JD
1729 lookahead set only next to the associated state's one item that (1) is
1730 associated with the same rule as the reduction and (2) has its dot at the end
1731 of its RHS. Previously, Bison also erroneously printed the lookahead set
1732 next to all of the state's other items associated with the same rule. This
e4ab1254 1733 bug affected only the ".output" file and not the generated parser source
88c78747
JD
1734 code.
1735
e4ab1254 1736** --report-file=FILE is a new option to override the default ".output" file
59da312b 1737 name.
1bb2bd75 1738
e4ab1254 1739** The "=" that used to be required in the following directives is now
02975b9a
JD
1740 deprecated:
1741
1742 %file-prefix "parser"
1743 %name-prefix "c_"
1744 %output "parser.c"
1745
e4ab1254 1746** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}"
c373bf8b
JD
1747
1748 Bison 2.3a provided a new set of directives as a more flexible alternative to
8e0a5e9e
JD
1749 the traditional Yacc prologue blocks. Those have now been consolidated into
1750 a single %code directive with an optional qualifier field, which identifies
1751 the purpose of the code and thus the location(s) where Bison should generate
1752 it:
1753
e4ab1254
AD
1754 1. "%code {CODE}" replaces "%after-header {CODE}"
1755 2. "%code requires {CODE}" replaces "%start-header {CODE}"
1756 3. "%code provides {CODE}" replaces "%end-header {CODE}"
1757 4. "%code top {CODE}" replaces "%before-header {CODE}"
8e0a5e9e 1758
e4ab1254
AD
1759 See the %code entries in section "Bison Declaration Summary" in the Bison
1760 manual for a summary of the new functionality. See the new section "Prologue
1761 Alternatives" for a detailed discussion including the advantages of %code
8e0a5e9e
JD
1762 over the traditional Yacc prologues.
1763
1764 The prologue alternatives are experimental. More user feedback will help to
1765 determine whether they should become permanent features.
1766
402b123d 1767** Revised warning: unset or unused mid-rule values
17bd8a73
JD
1768
1769 Since Bison 2.2, Bison has warned about mid-rule values that are set but not
1770 used within any of the actions of the parent rule. For example, Bison warns
1771 about unused $2 in:
1772
1773 exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };
1774
1775 Now, Bison also warns about mid-rule values that are used but not set. For
1776 example, Bison warns about unset $$ in the mid-rule action in:
1777
1778 exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };
1779
1780 However, Bison now disables both of these warnings by default since they
1781 sometimes prove to be false alarms in existing grammars employing the Yacc
1782 constructs $0 or $-N (where N is some positive integer).
1783
e4ab1254
AD
1784 To enable these warnings, specify the option "--warnings=midrule-values" or
1785 "-W", which is a synonym for "--warnings=all".
17bd8a73 1786
e4ab1254 1787** Default %destructor or %printer with "<*>" or "<>"
c373bf8b
JD
1788
1789 Bison now recognizes two separate kinds of default %destructor's and
12e35840
JD
1790 %printer's:
1791
e4ab1254 1792 1. Place "<*>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1793 %destructor/%printer for all grammar symbols for which you have formally
1794 declared semantic type tags.
1795
e4ab1254 1796 2. Place "<>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1797 %destructor/%printer for all grammar symbols without declared semantic
1798 type tags.
1799
e4ab1254
AD
1800 Bison no longer supports the "%symbol-default" notation from Bison 2.3a.
1801 "<*>" and "<>" combined achieve the same effect with one exception: Bison no
12e35840
JD
1802 longer applies any %destructor to a mid-rule value if that mid-rule value is
1803 not actually ever referenced using either $$ or $n in a semantic action.
1804
85894313
JD
1805 The default %destructor's and %printer's are experimental. More user
1806 feedback will help to determine whether they should become permanent
1807 features.
1808
e4ab1254 1809 See the section "Freeing Discarded Symbols" in the Bison manual for further
12e35840
JD
1810 details.
1811
402b123d 1812** %left, %right, and %nonassoc can now declare token numbers. This is required
e4ab1254 1813 by POSIX. However, see the end of section "Operator Precedence" in the Bison
ab7f29f8
JD
1814 manual for a caveat concerning the treatment of literal strings.
1815
402b123d 1816** The nonfunctional --no-parser, -n, and %no-parser options have been
b1cc23c4
JD
1817 completely removed from Bison.
1818
402b123d 1819* Changes in version 2.3a, 2006-09-13:
742e4900 1820
402b123d 1821** Instead of %union, you can define and use your own union type
ddc8ede1
PE
1822 YYSTYPE if your grammar contains at least one <type> tag.
1823 Your YYSTYPE need not be a macro; it can be a typedef.
1824 This change is for compatibility with other Yacc implementations,
1825 and is required by POSIX.
1826
402b123d 1827** Locations columns and lines start at 1.
cd48d21d
AD
1828 In accordance with the GNU Coding Standards and Emacs.
1829
402b123d 1830** You may now declare per-type and default %destructor's and %printer's:
ec5479ce
JD
1831
1832 For example:
1833
b2a0b7ca
JD
1834 %union { char *string; }
1835 %token <string> STRING1
1836 %token <string> STRING2
1837 %type <string> string1
1838 %type <string> string2
1839 %union { char character; }
1840 %token <character> CHR
1841 %type <character> chr
1842 %destructor { free ($$); } %symbol-default
1843 %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
1844 %destructor { } <character>
1845
1846 guarantees that, when the parser discards any user-defined symbol that has a
e4ab1254
AD
1847 semantic type tag other than "<character>", it passes its semantic value to
1848 "free". However, when the parser discards a "STRING1" or a "string1", it
1849 also prints its line number to "stdout". It performs only the second
1850 "%destructor" in this case, so it invokes "free" only once.
ec5479ce 1851
85894313
JD
1852 [Although we failed to mention this here in the 2.3a release, the default
1853 %destructor's and %printer's were experimental, and they were rewritten in
1854 future versions.]
1855
e4ab1254
AD
1856** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y",
1857 "--yacc", or "%yacc"), Bison no longer generates #define statements for
b931235e
JD
1858 associating token numbers with token names. Removing the #define statements
1859 helps to sanitize the global namespace during preprocessing, but POSIX Yacc
1860 requires them. Bison still generates an enum for token names in all cases.
1861
402b123d 1862** Handling of traditional Yacc prologue blocks is now more consistent but
34f98f46 1863 potentially incompatible with previous releases of Bison.
9bc0dd67
JD
1864
1865 As before, you declare prologue blocks in your grammar file with the
e4ab1254 1866 "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all
34f98f46
JD
1867 prologue blocks that you've declared before the first %union. To generate
1868 the post-prologue, Bison concatenates all prologue blocks that you've
ddc8ede1 1869 declared after the first %union.
9bc0dd67 1870
34f98f46 1871 Previous releases of Bison inserted the pre-prologue into both the header
9bc0dd67
JD
1872 file and the code file in all cases except for LALR(1) parsers in C. In the
1873 latter case, Bison inserted it only into the code file. For parsers in C++,
1874 the point of insertion was before any token definitions (which associate
1875 token numbers with names). For parsers in C, the point of insertion was
1876 after the token definitions.
1877
1878 Now, Bison never inserts the pre-prologue into the header file. In the code
1879 file, it always inserts it before the token definitions.
1880
402b123d 1881** Bison now provides a more flexible alternative to the traditional Yacc
34f98f46
JD
1882 prologue blocks: %before-header, %start-header, %end-header, and
1883 %after-header.
1884
1885 For example, the following declaration order in the grammar file reflects the
1886 order in which Bison will output these code blocks. However, you are free to
1887 declare these code blocks in your grammar file in whatever order is most
1888 convenient for you:
1889
1890 %before-header {
1891 /* Bison treats this block like a pre-prologue block: it inserts it into
1892 * the code file before the contents of the header file. It does *not*
1893 * insert it into the header file. This is a good place to put
1894 * #include's that you want at the top of your code file. A common
e4ab1254 1895 * example is '#include "system.h"'. */
34f98f46
JD
1896 }
1897 %start-header {
1898 /* Bison inserts this block into both the header file and the code file.
1899 * In both files, the point of insertion is before any Bison-generated
1900 * token, semantic type, location type, and class definitions. This is a
1901 * good place to define %union dependencies, for example. */
9bc0dd67
JD
1902 }
1903 %union {
34f98f46
JD
1904 /* Unlike the traditional Yacc prologue blocks, the output order for the
1905 * new %*-header blocks is not affected by their declaration position
1906 * relative to any %union in the grammar file. */
9bc0dd67 1907 }
34f98f46
JD
1908 %end-header {
1909 /* Bison inserts this block into both the header file and the code file.
1910 * In both files, the point of insertion is after the Bison-generated
1911 * definitions. This is a good place to declare or define public
1912 * functions or data structures that depend on the Bison-generated
1913 * definitions. */
9bc0dd67 1914 }
34f98f46
JD
1915 %after-header {
1916 /* Bison treats this block like a post-prologue block: it inserts it into
1917 * the code file after the contents of the header file. It does *not*
1918 * insert it into the header file. This is a good place to declare or
1919 * define internal functions or data structures that depend on the
1920 * Bison-generated definitions. */
1921 }
1922
1923 If you have multiple occurrences of any one of the above declarations, Bison
1924 will concatenate the contents in declaration order.
9bc0dd67 1925
85894313
JD
1926 [Although we failed to mention this here in the 2.3a release, the prologue
1927 alternatives were experimental, and they were rewritten in future versions.]
1928
e4ab1254 1929** The option "--report=look-ahead" has been changed to "--report=lookahead".
9e6e7ed2
PE
1930 The old spelling still works, but is not documented and may be removed
1931 in a future release.
742e4900 1932
402b123d 1933* Changes in version 2.3, 2006-06-05:
4ad3ed84 1934
e4ab1254 1935** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING",
4ad3ed84
PE
1936 for compatibility with LALR(1) grammars.
1937
402b123d 1938** It is now documented that any definition of YYSTYPE or YYLTYPE should
4ad3ed84
PE
1939 be to a type name that does not contain parentheses or brackets.
1940
402b123d 1941* Changes in version 2.2, 2006-05-19:
193d7c70 1942
402b123d 1943** The distribution terms for all Bison-generated parsers now permit
193d7c70
PE
1944 using the parsers in nonfree programs. Previously, this permission
1945 was granted only for Bison-generated LALR(1) parsers in C.
5f4236a0 1946
402b123d 1947** %name-prefix changes the namespace name in C++ outputs.
aa08666d 1948
402b123d 1949** The C++ parsers export their token_type.
5f4236a0 1950
402b123d 1951** Bison now allows multiple %union declarations, and concatenates
d6ca7905
PE
1952 their contents together.
1953
402b123d 1954** New warning: unused values
4d7bc38c
PE
1955 Right-hand side symbols whose values are not used are reported,
1956 if the symbols have destructors. For instance:
affac613 1957
8f3596a6 1958 exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
e9690142
JD
1959 | exp "+" exp
1960 ;
affac613 1961
8f3596a6
AD
1962 will trigger a warning about $$ and $5 in the first rule, and $3 in
1963 the second ($1 is copied to $$ by the default rule). This example
4e26c69e 1964 most likely contains three errors, and could be rewritten as:
affac613 1965
4e26c69e 1966 exp: exp "?" exp ":" exp
e9690142
JD
1967 { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
1968 | exp "+" exp
1969 { $$ = $1 ? $1 : $3; if ($1) free ($3); }
1970 ;
affac613 1971
4e26c69e
PE
1972 However, if the original actions were really intended, memory leaks
1973 and all, the warnings can be suppressed by letting Bison believe the
1974 values are used, e.g.:
721be13c 1975
8f3596a6 1976 exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
e9690142
JD
1977 | exp "+" exp { $$ = $1; (void) $3; }
1978 ;
721be13c 1979
84866159
AD
1980 If there are mid-rule actions, the warning is issued if no action
1981 uses it. The following triggers no warning: $1 and $3 are used.
1982
1983 exp: exp { push ($1); } '+' exp { push ($3); sum (); };
1984
721be13c
PE
1985 The warning is intended to help catching lost values and memory leaks.
1986 If a value is ignored, its associated memory typically is not reclaimed.
affac613 1987
402b123d 1988** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
9d9b8b70
PE
1989 Destructors are now called when user code invokes YYABORT, YYACCEPT,
1990 and YYERROR, for all objects on the stack, other than objects
1991 corresponding to the right-hand side of the current rule.
a85284cf 1992
402b123d 1993** %expect, %expect-rr
035aa4a0
PE
1994 Incorrect numbers of expected conflicts are now actual errors,
1995 instead of warnings.
1996
402b123d 1997** GLR, YACC parsers.
4e26c69e
PE
1998 The %parse-params are available in the destructors (and the
1999 experimental printers) as per the documentation.
4b367315 2000
e4ab1254 2001** Bison now warns if it finds a stray "$" or "@" in an action.
ad6a9b97 2002
402b123d 2003** %require "VERSION"
4e26c69e
PE
2004 This specifies that the grammar file depends on features implemented
2005 in Bison version VERSION or higher.
b50d2359 2006
402b123d 2007** lalr1.cc: The token and value types are now class members.
e14d0ab6
AD
2008 The tokens were defined as free form enums and cpp macros. YYSTYPE
2009 was defined as a free form union. They are now class members:
e4ab1254
AD
2010 tokens are enumerations of the "yy::parser::token" struct, and the
2011 semantic values have the "yy::parser::semantic_type" type.
fb9712a9
AD
2012
2013 If you do not want or can update to this scheme, the directive
e4ab1254 2014 '%define "global_tokens_and_yystype" "1"' triggers the global
b50d2359
AD
2015 definition of tokens and YYSTYPE. This change is suitable both
2016 for previous releases of Bison, and this one.
fb9712a9 2017
b50d2359 2018 If you wish to update, then make sure older version of Bison will
e4ab1254 2019 fail using '%require "2.2"'.
fb9712a9 2020
402b123d 2021** DJGPP support added.
193d7c70 2022\f
402b123d 2023* Changes in version 2.1, 2005-09-16:
1ce59070 2024
402b123d 2025** The C++ lalr1.cc skeleton supports %lex-param.
e14d0ab6 2026
402b123d 2027** Bison-generated parsers now support the translation of diagnostics like
baf785db
PE
2028 "syntax error" into languages other than English. The default
2029 language is still English. For details, please see the new
0410a6e0
PE
2030 Internationalization section of the Bison manual. Software
2031 distributors should also see the new PACKAGING file. Thanks to
2032 Bruno Haible for this new feature.
1ce59070 2033
402b123d 2034** Wording in the Bison-generated parsers has been changed slightly to
1a059451
PE
2035 simplify translation. In particular, the message "memory exhausted"
2036 has replaced "parser stack overflow", as the old message was not
2037 always accurate for modern Bison-generated parsers.
2038
402b123d 2039** Destructors are now called when the parser aborts, for all symbols left
258b75ca
PE
2040 behind on the stack. Also, the start symbol is now destroyed after a
2041 successful parse. In both cases, the behavior was formerly inconsistent.
2042
402b123d 2043** When generating verbose diagnostics, Bison-generated parsers no longer
72f000b0
PE
2044 quote the literal strings associated with tokens. For example, for
2045 a syntax error associated with '%token NUM "number"' they might
2046 print 'syntax error, unexpected number' instead of 'syntax error,
2047 unexpected "number"'.
193d7c70 2048\f
402b123d 2049* Changes in version 2.0, 2004-12-25:
efeed023 2050
402b123d 2051** Possibly-incompatible changes
d7e14fc0 2052
82de6b0d
PE
2053 - Bison-generated parsers no longer default to using the alloca function
2054 (when available) to extend the parser stack, due to widespread
2055 problems in unchecked stack-overflow detection. You can "#define
2056 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
2057 the manual to determine safe values for YYMAXDEPTH in that case.
8dd162d3 2058
82de6b0d
PE
2059 - Error token location.
2060 During error recovery, the location of the syntax error is updated
2061 to cover the whole sequence covered by the error token: it includes
2062 the shifted symbols thrown away during the first part of the error
2063 recovery, and the lookahead rejected during the second part.
18d192f0 2064
82de6b0d
PE
2065 - Semicolon changes:
2066 . Stray semicolons are no longer allowed at the start of a grammar.
2067 . Semicolons are now required after in-grammar declarations.
e342c3be 2068
82de6b0d
PE
2069 - Unescaped newlines are no longer allowed in character constants or
2070 string literals. They were never portable, and GCC 3.4.0 has
2071 dropped support for them. Better diagnostics are now generated if
2072 forget a closing quote.
8dd162d3 2073
82de6b0d 2074 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
f74b6f91 2075
402b123d 2076** New features
1452af69 2077
82de6b0d 2078 - GLR grammars now support locations.
4febdd96 2079
82de6b0d
PE
2080 - New directive: %initial-action.
2081 This directive allows the user to run arbitrary code (including
2082 initializing @$) from yyparse before parsing starts.
1452af69 2083
82de6b0d
PE
2084 - A new directive "%expect-rr N" specifies the expected number of
2085 reduce/reduce conflicts in GLR parsers.
1452af69 2086
e4ab1254 2087 - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d".
82de6b0d 2088 This is a GNU extension.
4febdd96 2089
e4ab1254 2090 - The option "--report=lookahead" was changed to "--report=look-ahead".
9e6e7ed2 2091 [However, this was changed back after 2.3.]
1452af69 2092
82de6b0d 2093 - Experimental %destructor support has been added to lalr1.cc.
1452af69 2094
82de6b0d
PE
2095 - New configure option --disable-yacc, to disable installation of the
2096 yacc command and -ly library introduced in 1.875 for POSIX conformance.
6040d338 2097
402b123d 2098** Bug fixes
d5a3fe37 2099
82de6b0d
PE
2100 - For now, %expect-count violations are now just warnings, not errors.
2101 This is for compatibility with Bison 1.75 and earlier (when there are
2102 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
2103 are too many or too few shift/reduce conflicts). However, in future
2104 versions of Bison we plan to improve the %expect machinery so that
2105 these violations will become errors again.
3473d0f8 2106
82de6b0d
PE
2107 - Within Bison itself, numbers (e.g., goto numbers) are no longer
2108 arbitrarily limited to 16-bit counts.
d600ee67 2109
82de6b0d 2110 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
d600ee67 2111\f
402b123d 2112* Changes in version 1.875, 2003-01-01:
963fcc17 2113
402b123d 2114** The documentation license has been upgraded to version 1.2
dc546b0f 2115 of the GNU Free Documentation License.
75eb3bc4 2116
402b123d 2117** syntax error processing
75eb3bc4 2118
dc546b0f
PE
2119 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
2120 locations too. This fixes bugs in error-location computation.
75eb3bc4 2121
dc546b0f
PE
2122 - %destructor
2123 It is now possible to reclaim the memory associated to symbols
2124 discarded during error recovery. This feature is still experimental.
20daca06 2125
dc546b0f
PE
2126 - %error-verbose
2127 This new directive is preferred over YYERROR_VERBOSE.
74724a70 2128
dc546b0f
PE
2129 - #defining yyerror to steal internal variables is discouraged.
2130 It is not guaranteed to work forever.
d1de5372 2131
402b123d 2132** POSIX conformance
d1de5372 2133
dc546b0f
PE
2134 - Semicolons are once again optional at the end of grammar rules.
2135 This reverts to the behavior of Bison 1.33 and earlier, and improves
2136 compatibility with Yacc.
74724a70 2137
e4ab1254
AD
2138 - "parse error" -> "syntax error"
2139 Bison now uniformly uses the term "syntax error"; formerly, the code
2140 and manual sometimes used the term "parse error" instead. POSIX
2141 requires "syntax error" in diagnostics, and it was thought better to
dc546b0f 2142 be consistent.
74724a70 2143
dc546b0f
PE
2144 - The documentation now emphasizes that yylex and yyerror must be
2145 declared before use. C99 requires this.
d1de5372 2146
dc546b0f
PE
2147 - Bison now parses C99 lexical constructs like UCNs and
2148 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
d1de5372 2149
dc546b0f
PE
2150 - File names are properly escaped in C output. E.g., foo\bar.y is
2151 output as "foo\\bar.y".
6780ca7a 2152
dc546b0f 2153 - Yacc command and library now available
e4ab1254 2154 The Bison distribution now installs a "yacc" command, as POSIX requires.
dc546b0f
PE
2155 Also, Bison now installs a small library liby.a containing
2156 implementations of Yacc-compatible yyerror and main functions.
2157 This library is normally not useful, but POSIX requires it.
6e649e65 2158
dc546b0f 2159 - Type clashes now generate warnings, not errors.
6e649e65 2160
dc546b0f
PE
2161 - If the user does not define YYSTYPE as a macro, Bison now declares it
2162 using typedef instead of defining it as a macro.
2163 For consistency, YYLTYPE is also declared instead of defined.
9501dc6e 2164
402b123d 2165** Other compatibility issues
886a425c 2166
e4ab1254
AD
2167 - %union directives can now have a tag before the "{", e.g., the
2168 directive "%union foo {...}" now generates the C code
2169 "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility.
2170 The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc.
2171 For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype".
dc546b0f 2172 This is for compatibility with both Yacc and Bison 1.35.
72f889cc 2173
e4ab1254 2174 - ";" is output before the terminating "}" of an action, for
dc546b0f 2175 compatibility with Bison 1.35.
886a425c 2176
dc546b0f 2177 - Bison now uses a Yacc-style format for conflict reports, e.g.,
e4ab1254 2178 "conflicts: 2 shift/reduce, 1 reduce/reduce".
437c2d80 2179
e4ab1254 2180 - "yystype" and "yyltype" are now obsolescent macros instead of being
dc546b0f
PE
2181 typedefs or tags; they are no longer documented and are planned to be
2182 withdrawn in a future release.
2a8d363a 2183
402b123d 2184** GLR parser notes
2a8d363a 2185
dc546b0f
PE
2186 - GLR and inline
2187 Users of Bison have to decide how they handle the portability of the
e4ab1254 2188 C keyword "inline".
959e5f51 2189
e4ab1254
AD
2190 - "parsing stack overflow..." -> "parser stack overflow"
2191 GLR parsers now report "parser stack overflow" as per the Bison manual.
900c5db5 2192
18ad57b3
AD
2193** %parse-param and %lex-param
2194 The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass
2195 additional context to yyparse and yylex. They suffer from several
2196 shortcomings:
2197
2198 - a single argument only can be added,
2199 - their types are weak (void *),
242cc08e 2200 - this context is not passed to ancillary functions such as yyerror,
18ad57b3
AD
2201 - only yacc.c parsers support them.
2202
2203 The new %parse-param/%lex-param directives provide a more precise control.
2204 For instance:
2205
2206 %parse-param {int *nastiness}
2207 %lex-param {int *nastiness}
2208 %parse-param {int *randomness}
2209
2210 results in the following signatures:
2211
2212 int yylex (int *nastiness);
2213 int yyparse (int *nastiness, int *randomness);
2214
2215 or, if both %pure-parser and %locations are used:
2216
2217 int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness);
2218 int yyparse (int *nastiness, int *randomness);
2219
402b123d 2220** Bison now warns if it detects conflicting outputs to the same file,
e4ab1254 2221 e.g., it generates a warning for "bison -d -o foo.h foo.y" since
dc546b0f 2222 that command outputs both code and header to foo.h.
6e40b4eb 2223
402b123d 2224** #line in output files
dc546b0f 2225 - --no-line works properly.
6e40b4eb 2226
402b123d 2227** Bison can no longer be built by a K&R C compiler; it requires C89 or
6e40b4eb
AD
2228 later to be built. This change originally took place a few versions
2229 ago, but nobody noticed until we recently asked someone to try
2230 building Bison with a K&R C compiler.
d600ee67 2231\f
402b123d 2232* Changes in version 1.75, 2002-10-14:
7933f2b5 2233
402b123d 2234** Bison should now work on 64-bit hosts.
7933f2b5 2235
402b123d 2236** Indonesian translation thanks to Tedi Heriyanto.
7933f2b5 2237
402b123d 2238** GLR parsers
f50adbbd
AD
2239 Fix spurious parse errors.
2240
402b123d 2241** Pure parsers
f50adbbd
AD
2242 Some people redefine yyerror to steal yyparse' private variables.
2243 Reenable this trick until an official feature replaces it.
2244
402b123d 2245** Type Clashes
d90c934c
AD
2246 In agreement with POSIX and with other Yaccs, leaving a default
2247 action is valid when $$ is untyped, and $1 typed:
2248
e9690142 2249 untyped: ... typed;
d90c934c
AD
2250
2251 but the converse remains an error:
2252
e9690142 2253 typed: ... untyped;
d90c934c 2254
402b123d 2255** Values of mid-rule actions
d90c934c
AD
2256 The following code:
2257
e9690142 2258 foo: { ... } { $$ = $1; } ...
d90c934c
AD
2259
2260 was incorrectly rejected: $1 is defined in the second mid-rule
2261 action, and is equal to the $$ of the first mid-rule action.
d600ee67 2262\f
402b123d 2263* Changes in version 1.50, 2002-10-04:
adc8c848 2264
402b123d 2265** GLR parsing
676385e2
PH
2266 The declaration
2267 %glr-parser
2268 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
2269 almost any context-free grammar, ambiguous or not. The new declarations
e8832397 2270 %dprec and %merge on grammar rules allow parse-time resolution of
676385e2
PH
2271 ambiguities. Contributed by Paul Hilfinger.
2272
7933f2b5 2273 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
420f93c8
PE
2274 like the Alpha, so please stick to 32-bit hosts for now.
2275
402b123d 2276** Output Directory
8c165d89 2277 When not in Yacc compatibility mode, when the output file was not
e4ab1254
AD
2278 specified, running "bison foo/bar.y" created "foo/bar.c". It
2279 now creates "bar.c".
8c165d89 2280
402b123d 2281** Undefined token
007a50a4 2282 The undefined token was systematically mapped to 2 which prevented
e88dbdbf 2283 the use of 2 by the user. This is no longer the case.
007a50a4 2284
402b123d 2285** Unknown token numbers
e88dbdbf 2286 If yylex returned an out of range value, yyparse could die. This is
007a50a4
AD
2287 no longer the case.
2288
402b123d 2289** Error token
e88dbdbf 2290 According to POSIX, the error token must be 256.
23c5a174
AD
2291 Bison extends this requirement by making it a preference: *if* the
2292 user specified that one of her tokens is numbered 256, then error
2293 will be mapped onto another number.
2294
402b123d 2295** Verbose error messages
e4ab1254 2296 They no longer report "..., expecting error or..." for states where
217598da
AD
2297 error recovery is possible.
2298
402b123d 2299** End token
e4ab1254 2300 Defaults to "$end" instead of "$".
217598da 2301
402b123d 2302** Error recovery now conforms to documentation and to POSIX
68cd8af3
PE
2303 When a Bison-generated parser encounters a syntax error, it now pops
2304 the stack until it finds a state that allows shifting the error
2305 token. Formerly, it popped the stack until it found a state that
2306 allowed some non-error action other than a default reduction on the
2307 error token. The new behavior has long been the documented behavior,
2308 and has long been required by POSIX. For more details, please see
337116ba
PE
2309 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
2310 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
68cd8af3 2311
402b123d 2312** Traces
5504898e
AD
2313 Popped tokens and nonterminals are now reported.
2314
402b123d 2315** Larger grammars
a861a339
PE
2316 Larger grammars are now supported (larger token numbers, larger grammar
2317 size (= sum of the LHS and RHS lengths), larger LALR tables).
2318 Formerly, many of these numbers ran afoul of 16-bit limits;
2319 now these limits are 32 bits on most hosts.
355e7c1c 2320
402b123d 2321** Explicit initial rule
643a5994
AD
2322 Bison used to play hacks with the initial rule, which the user does
2323 not write. It is now explicit, and visible in the reports and
2324 graphs as rule 0.
23c5a174 2325
402b123d 2326** Useless rules
643a5994 2327 Before, Bison reported the useless rules, but, although not used,
77714df2 2328 included them in the parsers. They are now actually removed.
23c5a174 2329
402b123d 2330** Useless rules, useless nonterminals
6b98e4b5
AD
2331 They are now reported, as a warning, with their locations.
2332
402b123d 2333** Rules never reduced
e8832397
AD
2334 Rules that can never be reduced because of conflicts are now
2335 reported.
2336
e4ab1254 2337** Incorrect "Token not used"
11652ab3
AD
2338 On a grammar such as
2339
e29f0771
AD
2340 %token useless useful
2341 %%
2342 exp: '0' %prec useful;
11652ab3
AD
2343
2344 where a token was used to set the precedence of the last rule,
e4ab1254 2345 bison reported both "useful" and "useless" as useless tokens.
11652ab3 2346
402b123d 2347** Revert the C++ namespace changes introduced in 1.31
77714df2 2348 as they caused too many portability hassles.
0179dd65 2349
402b123d 2350** Default locations
b2d52318
AD
2351 By an accident of design, the default computation of @$ was
2352 performed after another default computation was performed: @$ = @1.
2353 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
2354 the computation of @$.
adc8c848 2355
402b123d 2356** Token end-of-file
b7c49edf
AD
2357 The token end of file may be specified by the user, in which case,
2358 the user symbol is used in the reports, the graphs, and the verbose
e4ab1254 2359 error messages instead of "$end", which remains being the default.
b7c49edf 2360 For instance
e29f0771 2361 %token MYEOF 0
b7c49edf 2362 or
e29f0771 2363 %token MYEOF 0 "end of file"
fdbcd8e2 2364
402b123d 2365** Semantic parser
fdbcd8e2
AD
2366 This old option, which has been broken for ages, is removed.
2367
402b123d 2368** New translations
a861a339 2369 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
84614e13
AD
2370 Croatian, thanks to Denis Lackovic.
2371
402b123d 2372** Incorrect token definitions
e4ab1254
AD
2373 When given
2374 %token 'a' "A"
2375 bison used to output
2376 #define 'a' 65
b87f8b21 2377
402b123d 2378** Token definitions as enums
77714df2
AD
2379 Tokens are output both as the traditional #define's, and, provided
2380 the compiler supports ANSI C or is a C++ compiler, as enums.
e88dbdbf 2381 This lets debuggers display names instead of integers.
77714df2 2382
402b123d 2383** Reports
ec3bc396
AD
2384 In addition to --verbose, bison supports --report=THINGS, which
2385 produces additional information:
b408954b
AD
2386 - itemset
2387 complete the core item sets with their closure
e4ab1254 2388 - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back]
9e6e7ed2 2389 explicitly associate lookahead tokens to items
b408954b
AD
2390 - solved
2391 describe shift/reduce conflicts solving.
2392 Bison used to systematically output this information on top of
2393 the report. Solved conflicts are now attached to their states.
ec3bc396 2394
402b123d 2395** Type clashes
9af3fbce
AD
2396 Previous versions don't complain when there is a type clash on
2397 the default action if the rule has a mid-rule action, such as in:
2398
e29f0771
AD
2399 %type <foo> bar
2400 %%
2401 bar: '0' {} '0';
9af3fbce
AD
2402
2403 This is fixed.
a861a339 2404
402b123d 2405** GNU M4 is now required when using Bison.
f987e9d2 2406\f
402b123d 2407* Changes in version 1.35, 2002-03-25:
76551463 2408
402b123d 2409** C Skeleton
76551463
AD
2410 Some projects use Bison's C parser with C++ compilers, and define
2411 YYSTYPE as a class. The recent adjustment of C parsers for data
2412 alignment and 64 bit architectures made this impossible.
2413
2414 Because for the time being no real solution for C++ parser
2415 generation exists, kludges were implemented in the parser to
2416 maintain this use. In the future, when Bison has C++ parsers, this
2417 kludge will be disabled.
2418
2419 This kludge also addresses some C++ problems when the stack was
2420 extended.
76551463 2421\f
402b123d 2422* Changes in version 1.34, 2002-03-12:
76551463 2423
402b123d 2424** File name clashes are detected
76551463 2425 $ bison foo.y -d -o foo.x
e4ab1254 2426 fatal error: header and parser would both be named "foo.x"
76551463 2427
e4ab1254 2428** A missing ";" at the end of a rule triggers a warning
76551463
AD
2429 In accordance with POSIX, and in agreement with other
2430 Yacc implementations, Bison will mandate this semicolon in the near
2431 future. This eases the implementation of a Bison parser of Bison
2432 grammars by making this grammar LALR(1) instead of LR(2). To
2433 facilitate the transition, this release introduces a warning.
2434
402b123d 2435** Revert the C++ namespace changes introduced in 1.31, as they caused too
76551463
AD
2436 many portability hassles.
2437
402b123d 2438** DJGPP support added.
76551463 2439
402b123d 2440** Fix test suite portability problems.
76551463 2441\f
402b123d 2442* Changes in version 1.33, 2002-02-07:
76551463 2443
402b123d 2444** Fix C++ issues
76551463
AD
2445 Groff could not be compiled for the definition of size_t was lacking
2446 under some conditions.
2447
402b123d 2448** Catch invalid @n
76551463
AD
2449 As is done with $n.
2450\f
402b123d 2451* Changes in version 1.32, 2002-01-23:
76551463 2452
402b123d 2453** Fix Yacc output file names
76551463 2454
402b123d 2455** Portability fixes
76551463 2456
402b123d 2457** Italian, Dutch translations
76551463 2458\f
402b123d 2459* Changes in version 1.31, 2002-01-14:
52d1aeee 2460
402b123d 2461** Many Bug Fixes
52d1aeee 2462
402b123d 2463** GNU Gettext and %expect
52d1aeee
MA
2464 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
2465 Bison dies on incorrect %expectations, we fear there will be
2466 too many bug reports for Gettext, so _for the time being_, %expect
e4ab1254 2467 does not trigger an error when the input file is named "plural.y".
52d1aeee 2468
402b123d 2469** Use of alloca in parsers
52d1aeee
MA
2470 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
2471 malloc exclusively. Since 1.29, but was not NEWS'ed.
2472
2473 alloca is used only when compiled with GCC, to avoid portability
2474 problems as on AIX.
2475
402b123d 2476** yyparse now returns 2 if memory is exhausted; formerly it dumped core.
b47dbebe 2477
402b123d 2478** When the generated parser lacks debugging code, YYDEBUG is now 0
52d1aeee
MA
2479 (as POSIX requires) instead of being undefined.
2480
402b123d 2481** User Actions
52d1aeee
MA
2482 Bison has always permitted actions such as { $$ = $1 }: it adds the
2483 ending semicolon. Now if in Yacc compatibility mode, the semicolon
2484 is no longer output: one has to write { $$ = $1; }.
2485
402b123d 2486** Better C++ compliance
52d1aeee 2487 The output parsers try to respect C++ namespaces.
76551463 2488 [This turned out to be a failed experiment, and it was reverted later.]
52d1aeee 2489
402b123d 2490** Reduced Grammars
52d1aeee
MA
2491 Fixed bugs when reporting useless nonterminals.
2492
402b123d 2493** 64 bit hosts
52d1aeee
MA
2494 The parsers work properly on 64 bit hosts.
2495
402b123d 2496** Error messages
52d1aeee
MA
2497 Some calls to strerror resulted in scrambled or missing error messages.
2498
402b123d 2499** %expect
52d1aeee
MA
2500 When the number of shift/reduce conflicts is correct, don't issue
2501 any warning.
2502
402b123d 2503** The verbose report includes the rule line numbers.
52d1aeee 2504
402b123d 2505** Rule line numbers are fixed in traces.
52d1aeee 2506
402b123d 2507** Swedish translation
52d1aeee 2508
402b123d 2509** Parse errors
52d1aeee
MA
2510 Verbose parse error messages from the parsers are better looking.
2511 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
2512 Now: parse error: unexpected '/', expecting "number" or '-' or '('
2513
402b123d 2514** Fixed parser memory leaks.
52d1aeee
MA
2515 When the generated parser was using malloc to extend its stacks, the
2516 previous allocations were not freed.
2517
402b123d 2518** Fixed verbose output file.
52d1aeee
MA
2519 Some newlines were missing.
2520 Some conflicts in state descriptions were missing.
2521
402b123d 2522** Fixed conflict report.
52d1aeee
MA
2523 Option -v was needed to get the result.
2524
402b123d 2525** %expect
52d1aeee
MA
2526 Was not used.
2527 Mismatches are errors, not warnings.
2528
402b123d 2529** Fixed incorrect processing of some invalid input.
52d1aeee 2530
402b123d 2531** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
52d1aeee 2532
402b123d 2533** Fixed some typos in the documentation.
52d1aeee 2534
402b123d 2535** %token MY_EOF 0 is supported.
52d1aeee
MA
2536 Before, MY_EOF was silently renumbered as 257.
2537
402b123d 2538** doc/refcard.tex is updated.
52d1aeee 2539
402b123d 2540** %output, %file-prefix, %name-prefix.
52d1aeee
MA
2541 New.
2542
402b123d 2543** --output
e4ab1254 2544 New, aliasing "--output-file".
52d1aeee 2545\f
402b123d 2546* Changes in version 1.30, 2001-10-26:
342b8b6e 2547
e4ab1254
AD
2548** "--defines" and "--graph" have now an optional argument which is the
2549 output file name. "-d" and "-g" do not change; they do not take any
342b8b6e
AD
2550 argument.
2551
e4ab1254 2552** "%source_extension" and "%header_extension" are removed, failed
342b8b6e
AD
2553 experiment.
2554
402b123d 2555** Portability fixes.
f987e9d2 2556\f
402b123d 2557* Changes in version 1.29, 2001-09-07:
342b8b6e 2558
402b123d 2559** The output file does not define const, as this caused problems when used
342b8b6e
AD
2560 with common autoconfiguration schemes. If you still use ancient compilers
2561 that lack const, compile with the equivalent of the C compiler option
e4ab1254 2562 "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this.
342b8b6e 2563
e4ab1254 2564** Added "-g" and "--graph".
f87a2205 2565
402b123d 2566** The Bison manual is now distributed under the terms of the GNU FDL.
f2b5126e 2567
402b123d 2568** The input and the output files has automatically a similar extension.
234a3be3 2569
402b123d 2570** Russian translation added.
f87a2205 2571
402b123d 2572** NLS support updated; should hopefully be less troublesome.
f87a2205 2573
402b123d 2574** Added the old Bison reference card.
c33638bb 2575
e4ab1254 2576** Added "--locations" and "%locations".
6deb4447 2577
e4ab1254 2578** Added "-S" and "--skeleton".
cd5bd6ac 2579
e4ab1254 2580** "%raw", "-r", "--raw" is disabled.
62ab6972 2581
402b123d 2582** Special characters are escaped when output. This solves the problems
cd5bd6ac
AD
2583 of the #line lines with path names including backslashes.
2584
402b123d 2585** New directives.
e4ab1254
AD
2586 "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose",
2587 "%debug", "%source_extension" and "%header_extension".
f987e9d2 2588
402b123d 2589** @$
f987e9d2 2590 Automatic location tracking.
f87a2205 2591\f
402b123d 2592* Changes in version 1.28, 1999-07-06:
d2e00347 2593
402b123d 2594** Should compile better now with K&R compilers.
d2e00347 2595
402b123d 2596** Added NLS.
d2e00347 2597
402b123d 2598** Fixed a problem with escaping the double quote character.
d2e00347 2599
402b123d 2600** There is now a FAQ.
d2e00347 2601\f
402b123d 2602* Changes in version 1.27:
5c31c3c2 2603
402b123d 2604** The make rule which prevented bison.simple from being created on
5c31c3c2
JT
2605 some systems has been fixed.
2606\f
402b123d 2607* Changes in version 1.26:
4be07551 2608
7e508a2b 2609** Bison now uses Automake.
4be07551 2610
402b123d 2611** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
4be07551 2612
402b123d 2613** Token numbers now start at 257 as previously documented, not 258.
4be07551 2614
402b123d 2615** Bison honors the TMPDIR environment variable.
4be07551 2616
402b123d 2617** A couple of buffer overruns have been fixed.
f51dbca1 2618
402b123d 2619** Problems when closing files should now be reported.
f51dbca1 2620
402b123d 2621** Generated parsers should now work even on operating systems which do
f51dbca1 2622 not provide alloca().
4be07551 2623\f
402b123d 2624* Changes in version 1.25, 1995-10-16:
df8878c5 2625
402b123d 2626** Errors in the input grammar are not fatal; Bison keeps reading
df8878c5 2627the grammar file, and reports all the errors found in it.
8c44d3ec 2628
402b123d 2629** Tokens can now be specified as multiple-character strings: for
df8878c5 2630example, you could use "<=" for a token which looks like <=, instead
7e508a2b 2631of choosing a name like LESSEQ.
df8878c5 2632
402b123d 2633** The %token_table declaration says to write a table of tokens (names
df8878c5
RS
2634and numbers) into the parser file. The yylex function can use this
2635table to recognize multiple-character string tokens, or for other
2636purposes.
2637
402b123d 2638** The %no_lines declaration says not to generate any #line preprocessor
df8878c5
RS
2639directives in the parser file.
2640
402b123d 2641** The %raw declaration says to use internal Bison token numbers, not
df8878c5
RS
2642Yacc-compatible token numbers, when token names are defined as macros.
2643
402b123d 2644** The --no-parser option produces the parser tables without including
df8878c5
RS
2645the parser engine; a project can now use its own parser engine.
2646The actions go into a separate file called NAME.act, in the form of
2647a switch statement body.
2648\f
402b123d 2649* Changes in version 1.23:
6780ca7a 2650
4d019228
DM
2651The user can define YYPARSE_PARAM as the name of an argument to be
2652passed into yyparse. The argument should have type void *. It should
2653actually point to an object. Grammar actions can access the variable
2654by casting it to the proper pointer type.
6780ca7a 2655
6780ca7a 2656Line numbers in output file corrected.
6780ca7a 2657\f
402b123d 2658* Changes in version 1.22:
6780ca7a
DM
2659
2660--help option added.
6780ca7a 2661\f
402b123d 2662* Changes in version 1.20:
6780ca7a
DM
2663
2664Output file does not redefine const for C++.
9f4503d6 2665
76551463
AD
2666-----
2667
7d6bad19 2668Copyright (C) 1995-2013 Free Software Foundation, Inc.
76551463 2669
74553c98 2670This file is part of Bison, the GNU Parser Generator.
76551463 2671
f16b0819 2672This program is free software: you can redistribute it and/or modify
76551463 2673it under the terms of the GNU General Public License as published by
f16b0819
PE
2674the Free Software Foundation, either version 3 of the License, or
2675(at your option) any later version.
76551463 2676
f16b0819 2677This program is distributed in the hope that it will be useful,
76551463
AD
2678but WITHOUT ANY WARRANTY; without even the implied warranty of
2679MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2680GNU General Public License for more details.
2681
2682You should have received a copy of the GNU General Public License
f16b0819 2683along with this program. If not, see <http://www.gnu.org/licenses/>.
7e508a2b
AD
2684
2685 LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr
2686 LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr
2687 LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar
2688 LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr
2689 LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks
2690 LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG
2691 LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush
2692 LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly
2693 LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's
2694 LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de
2695 LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto
2696 LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs
2697 LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF
242cc08e 2698 LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY
7e508a2b 2699 LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL
242cc08e
AD
2700 LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh
2701 LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf
dcb366b1 2702 LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp Wother nterm arg init
1282c124
AD
2703 LocalWords: TOK calc yyo fval Wconflicts parsers yystackp yyval yynerrs
2704 LocalWords: Théophile Ranquet Santet fno fnone stype associativity Tolmer
2705 LocalWords: Wprecedence Rassoul Wempty Paolo Bonzini parser's Michiel loc
2706 LocalWords: redeclaration sval fcaret reentrant XSLT xsl Wmaybe yyvsp Tedi
2707 LocalWords: pragmas noreturn untyped Rozenman unexpanded Wojciech Polak
6574576c 2708 LocalWords: Alexandre MERCHANTABILITY yytype
7e508a2b
AD
2709
2710Local Variables:
2711mode: outline
e4ab1254 2712fill-column: 76
7e508a2b 2713End: