]> git.saurik.com Git - bison.git/blame - NEWS
style: formatting and comment changes
[bison.git] / NEWS
CommitLineData
ed7658fe 1GNU Bison NEWS
3af4feb2 2
7ca7a3ab
AD
3* Noteworthy changes in release ?.? (????-??-??) [?]
4
5
c4e686a1 6* Noteworthy changes in release 3.0.4 (2015-01-23) [stable]
11d82017 7
3e82dfaa
AD
8** Bug fixes
9
10*** C++ with Variants (lalr1.cc)
11
12 Fix a compiler warning when no %destructor use $$.
11d82017 13
658b1893
AD
14*** Test suites
15
16 Several portability issues in tests were fixed.
17
da6f0746 18* Noteworthy changes in release 3.0.3 (2015-01-15) [stable]
0bd5ee5f 19
461f1516
AD
20** Bug fixes
21
ee028dce
AD
22*** C++ with Variants (lalr1.cc)
23
24 Problems with %destructor and '%define parse.assert' have been fixed.
25
26*** Named %union support (yacc.c, glr.c)
827bc59c
AD
27
28 Bison 3.0 introduced a regression on named %union such as
29
30 %union foo { int ival; };
31
ee028dce 32 The possibility to use a name was introduced "for Yacc compatibility".
827bc59c
AD
33 It is however not required by POSIX Yacc, and its usefulness is not clear.
34
ee028dce 35*** %define api.value.type union with %defines (yacc.c, glr.c)
8d4dc896 36
ee028dce
AD
37 The C parsers were broken when %defines was used together with "%define
38 api.value.type union".
8d4dc896 39
461f1516
AD
40*** Redeclarations are reported in proper order
41
42 On
43
44 %token FOO "foo"
45 %printer {} "foo"
46 %printer {} FOO
47
48 bison used to report:
49
50 /tmp/foo.yy:2.10-11: error: %printer redeclaration for FOO
51 %printer {} "foo"
52 ^^
53 /tmp/foo.yy:3.10-11: previous declaration
54 %printer {} FOO
55 ^^
56
57 Now, the "previous" declaration is always the first one.
58
0bd5ee5f 59
b167e7ba
AD
60** Documentation
61
62 Bison now installs various files in its docdir (which defaults to
63 '/usr/local/share/doc/bison'), including the three fully blown examples
64 extracted from the documentation:
65
66 - rpcalc
67 Reverse polish calculator, a simple introductory example.
68 - mfcalc
69 Multi-function Calc, a calculator with memory and functions and located
70 error messages.
71 - calc++
72 a calculator in C++ using variant support and token constructors.
73
a3efd13e 74* Noteworthy changes in release 3.0.2 (2013-12-05) [stable]
6d94eebb 75
7b0ca050
AD
76** Bug fixes
77
184b42c8
AD
78*** Generated source files when errors are reported
79
80 When warnings are issued and -Werror is set, bison would still generate
81 the source files (*.c, *.h...). As a consequence, some runs of "make"
82 could fail the first time, but not the second (as the files were generated
83 anyway).
84
85 This is fixed: bison no longer generates this source files, but, of
86 course, still produces the various reports (*.output, *.xml, etc.).
87
be29c71d
AD
88*** %empty is used in reports
89
90 Empty right-hand sides are denoted by '%empty' in all the reports (text,
91 dot, XML and formats derived from it).
92
7b0ca050
AD
93*** YYERROR and variants
94
95 When C++ variant support is enabled, an error triggered via YYERROR, but
96 not caught via error recovery, resulted in a double deletion.
6d94eebb 97
c2ecada3 98* Noteworthy changes in release 3.0.1 (2013-11-12) [stable]
e386b50f 99
0a244a22
AD
100** Bug fixes
101
39bace5d
AD
102*** Errors in caret diagnostics
103
104 On some platforms, some errors could result in endless diagnostics.
e4678430
AD
105
106*** Fixes of the -Werror option
107
108 Options such as "-Werror -Wno-error=foo" were still turning "foo"
109 diagnostics into errors instead of warnings. This is fixed.
110
111 Actually, for consistency with GCC, "-Wno-error=foo -Werror" now also
112 leaves "foo" diagnostics as warnings. Similarly, with "-Werror=foo
113 -Wno-error", "foo" diagnostics are now errors.
e386b50f 114
071863b3
AD
115*** GLR Predicates
116
117 As demonstrated in the documentation, one can now leave spaces between
118 "%?" and its "{".
119
265640d5
AD
120*** Installation
121
122 The yacc.1 man page is no longer installed if --disable-yacc was
123 specified.
124
39bace5d
AD
125*** Fixes in the test suite
126
127 Bugs and portability issues.
128
534497f5 129* Noteworthy changes in release 3.0 (2013-07-25) [stable]
facb910c 130
8458a411 131** WARNING: Future backward-incompatibilities!
597afd73 132
597afd73
AD
133 Like other GNU packages, Bison will start using some of the C99 features
134 for its own code, especially the definition of variables after statements.
135 The generated C parsers still aim at C90.
136
8458a411 137** Backward incompatible changes
47db7ed1
AD
138
139*** Obsolete features
140
40bb6f78
AD
141 Support for YYFAIL is removed (deprecated in Bison 2.4.2): use YYERROR.
142
143 Support for yystype and yyltype is removed (deprecated in Bison 1.875):
144 use YYSTYPE and YYLTYPE.
145
146 Support for YYLEX_PARAM and YYPARSE_PARAM is removed (deprecated in Bison
147 1.875): use %lex-param, %parse-param, or %param.
47db7ed1 148
05e25f23
AD
149 Missing semicolons at the end of actions are no longer added (as announced
150 in the release 2.5).
151
1fa19a76
AD
152*** Use of YACC='bison -y'
153
154 TL;DR: With Autoconf <= 2.69, pass -Wno-yacc to (AM_)YFLAGS if you use
155 Bison extensions.
156
157 Traditional Yacc generates 'y.tab.c' whatever the name of the input file.
158 Therefore Makefiles written for Yacc expect 'y.tab.c' (and possibly
159 'y.tab.h' and 'y.outout') to be generated from 'foo.y'.
160
161 To this end, for ages, AC_PROG_YACC, Autoconf's macro to look for an
162 implementation of Yacc, was using Bison as 'bison -y'. While it does
163 ensure compatible output file names, it also enables warnings for
164 incompatibilities with POSIX Yacc. In other words, 'bison -y' triggers
165 warnings for Bison extensions.
166
167 Autoconf 2.70+ fixes this incompatibility by using YACC='bison -o y.tab.c'
168 (which also generates 'y.tab.h' and 'y.output' when needed).
169 Alternatively, disable Yacc warnings by passing '-Wno-yacc' to your Yacc
170 flags (YFLAGS, or AM_YFLAGS with Automake).
171
597afd73
AD
172** Bug fixes
173
c21e515e 174*** The epilogue is no longer affected by internal #defines (glr.c)
597afd73
AD
175
176 The glr.c skeleton uses defines such as #define yylval (yystackp->yyval) in
177 generated code. These weren't properly undefined before the inclusion of
178 the user epilogue, so functions such as the following were butchered by the
179 preprocessor expansion:
180
181 int yylex (YYSTYPE *yylval);
182
6c7022f7 183 This is fixed: yylval, yynerrs, yychar, and yylloc are now valid
597afd73
AD
184 identifiers for user-provided variables.
185
f0f95a50
AD
186*** stdio.h is no longer needed when locations are enabled (yacc.c)
187
188 Changes in Bison 2.7 introduced a dependency on FILE and fprintf when
189 locations are enabled. This is fixed.
190
bb4b189b
AD
191*** Warnings about useless %pure-parser/%define api.pure are restored
192
597afd73
AD
193** Diagnostics reported by Bison
194
195 Most of these features were contributed by Théophile Ranquet and Victor
196 Santet.
73370a9d 197
016426c1
TR
198*** Carets
199
200 Version 2.7 introduced caret errors, for a prettier output. These are now
201 activated by default. The old format can still be used by invoking Bison
202 with -fno-caret (or -fnone).
203
fec5f3c0
AD
204 Some error messages that reproduced excerpts of the grammar are now using
205 the caret information only. For instance on:
206
207 %%
208 exp: 'a' | 'a';
209
210 Bison 2.7 reports:
211
212 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
213 in.y:2.12-14: warning: rule useless in parser due to conflicts: exp: 'a' [-Wother]
214
215 Now bison reports:
216
217 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
218 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
219 exp: 'a' | 'a';
220 ^^^
221
222 and "bison -fno-caret" reports:
223
224 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
225 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
226
1048a1c9 227*** Enhancements of the -Werror option
518e8830 228
1048a1c9 229 The -Werror=CATEGORY option is now recognized, and will treat specified
d949eefd
AD
230 warnings as errors. The warnings need not have been explicitly activated
231 using the -W option, this is similar to what GCC 4.7 does.
1048a1c9
AD
232
233 For example, given the following command line, Bison will treat both
d949eefd 234 warnings related to POSIX Yacc incompatibilities and S/R conflicts as
1048a1c9
AD
235 errors (and only those):
236
237 $ bison -Werror=yacc,error=conflicts-sr input.y
238
239 If no categories are specified, -Werror will make all active warnings into
240 errors. For example, the following line does the same the previous example:
241
242 $ bison -Werror -Wnone -Wyacc -Wconflicts-sr input.y
243
244 (By default -Wconflicts-sr,conflicts-rr,deprecated,other is enabled.)
245
246 Note that the categories in this -Werror option may not be prefixed with
247 "no-". However, -Wno-error[=CATEGORY] is valid.
248
249 Note that -y enables -Werror=yacc. Therefore it is now possible to require
250 Yacc-like behavior (e.g., always generate y.tab.c), but to report
251 incompatibilities as warnings: "-y -Wno-error=yacc".
518e8830 252
46bdb8ec 253*** The display of warnings is now richer
73370a9d 254
46bdb8ec 255 The option that controls a given warning is now displayed:
73370a9d 256
46bdb8ec 257 foo.y:4.6: warning: type clash on default action: <foo> != <bar> [-Wother]
73370a9d 258
46bdb8ec
TR
259 In the case of warnings treated as errors, the prefix is changed from
260 "warning: " to "error: ", and the suffix is displayed, in a manner similar
d949eefd 261 to GCC, as [-Werror=CATEGORY].
1048a1c9 262
46bdb8ec
TR
263 For instance, where the previous version of Bison would report (and exit
264 with failure):
1048a1c9
AD
265
266 bison: warnings being treated as errors
46bdb8ec 267 input.y:1.1: warning: stray ',' treated as white space
1048a1c9 268
46bdb8ec 269 it now reports:
1048a1c9 270
1048a1c9
AD
271 input.y:1.1: error: stray ',' treated as white space [-Werror=other]
272
273*** Deprecated constructs
274
275 The new 'deprecated' warning category flags obsolete constructs whose
276 support will be discontinued. It is enabled by default. These warnings
277 used to be reported as 'other' warnings.
278
73370a9d 279*** Useless semantic types
9641b918
VS
280
281 Bison now warns about useless (uninhabited) semantic types. Since
282 semantic types are not declared to Bison (they are defined in the opaque
283 %union structure), it is %printer/%destructor directives about useless
284 types that trigger the warning:
285
286 %token <type1> term
287 %type <type2> nterm
288 %printer {} <type1> <type3>
289 %destructor {} <type2> <type4>
290 %%
291 nterm: term { $$ = $1; };
292
293 3.28-34: warning: type <type3> is used, but is not associated to any symbol
294 4.28-34: warning: type <type4> is used, but is not associated to any symbol
295
31557b9e 296*** Undefined but unused symbols
b921d92f 297
31557b9e
AD
298 Bison used to raise an error for undefined symbols that are not used in
299 the grammar. This is now only a warning.
b921d92f
VS
300
301 %printer {} symbol1
302 %destructor {} symbol2
31557b9e 303 %type <type> symbol3
b921d92f
VS
304 %%
305 exp: "a";
306
73370a9d 307*** Useless destructors or printers
ea9a35c6
VS
308
309 Bison now warns about useless destructors or printers. In the following
310 example, the printer for <type1>, and the destructor for <type2> are
311 useless: all symbols of <type1> (token1) already have a printer, and all
312 symbols of type <type2> (token2) already have a destructor.
313
314 %token <type1> token1
315 <type2> token2
316 <type3> token3
317 <type4> token4
318 %printer {} token1 <type1> <type3>
319 %destructor {} token2 <type2> <type4>
320
d87ea54c
AD
321*** Conflicts
322
323 The warnings and error messages about shift/reduce and reduce/reduce
324 conflicts have been normalized. For instance on the following foo.y file:
325
326 %glr-parser
327 %%
328 exp: exp '+' exp | '0' | '0';
329
330 compare the previous version of bison:
331
332 $ bison foo.y
333 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
334 $ bison -Werror foo.y
335 bison: warnings being treated as errors
336 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
337
338 with the new behavior:
339
340 $ bison foo.y
341 foo.y: warning: 1 shift/reduce conflict [-Wconflicts-sr]
342 foo.y: warning: 2 reduce/reduce conflicts [-Wconflicts-rr]
343 $ bison -Werror foo.y
9503b0a4
TR
344 foo.y: error: 1 shift/reduce conflict [-Werror=conflicts-sr]
345 foo.y: error: 2 reduce/reduce conflicts [-Werror=conflicts-rr]
d87ea54c
AD
346
347 When %expect or %expect-rr is used, such as with bar.y:
348
349 %expect 0
350 %glr-parser
351 %%
352 exp: exp '+' exp | '0' | '0';
353
354 Former behavior:
355
356 $ bison bar.y
357 bar.y: conflicts: 1 shift/reduce, 2 reduce/reduce
358 bar.y: expected 0 shift/reduce conflicts
359 bar.y: expected 0 reduce/reduce conflicts
360
361 New one:
362
363 $ bison bar.y
9503b0a4
TR
364 bar.y: error: shift/reduce conflicts: 1 found, 0 expected
365 bar.y: error: reduce/reduce conflicts: 2 found, 0 expected
d87ea54c 366
f24695ef
AD
367** Incompatibilities with POSIX Yacc
368
369 The 'yacc' category is no longer part of '-Wall', enable it explicitly
370 with '-Wyacc'.
371
2055a44e
AD
372** Additional yylex/yyparse arguments
373
6dc4663d
AD
374 The new directive %param declares additional arguments to both yylex and
375 yyparse. The %lex-param, %parse-param, and %param directives support one
376 or more arguments. Instead of
2055a44e 377
e436fa67
AD
378 %lex-param {arg1_type *arg1}
379 %lex-param {arg2_type *arg2}
380 %parse-param {arg1_type *arg1}
381 %parse-param {arg2_type *arg2}
2055a44e
AD
382
383 one may now declare
384
e436fa67 385 %param {arg1_type *arg1} {arg2_type *arg2}
2055a44e 386
630a0218
AD
387** Types of values for %define variables
388
389 Bison used to make no difference between '%define foo bar' and '%define
390 foo "bar"'. The former is now called a 'keyword value', and the latter a
391 'string value'. A third kind was added: 'code values', such as '%define
392 foo {bar}'.
393
394 Keyword variables are used for fixed value sets, e.g.,
395
396 %define lr.type lalr
397
398 Code variables are used for value in the target language, e.g.,
399
400 %define api.value.type {struct semantic_type}
401
402 String variables are used remaining cases, e.g. file names.
403
2a6b66c5 404** Variable api.token.prefix
99c08fb6 405
2a6b66c5 406 The variable api.token.prefix changes the way tokens are identified in
99c08fb6
AD
407 the generated files. This is especially useful to avoid collisions
408 with identifiers in the target language. For instance
409
e436fa67 410 %token FILE for ERROR
630a0218 411 %define api.token.prefix {TOK_}
e436fa67
AD
412 %%
413 start: FILE for ERROR;
99c08fb6
AD
414
415 will generate the definition of the symbols TOK_FILE, TOK_for, and
416 TOK_ERROR in the generated sources. In particular, the scanner must
417 use these prefixed token names, although the grammar itself still
418 uses the short names (as in the sample rule given above).
419
6574576c
AD
420** Variable api.value.type
421
422 This new %define variable supersedes the #define macro YYSTYPE. The use
423 of YYSTYPE is discouraged. In particular, #defining YYSTYPE *and* either
424 using %union or %defining api.value.type results in undefined behavior.
425
426 Either define api.value.type, or use "%union":
427
428 %union
429 {
430 int ival;
431 char *sval;
432 }
433 %token <ival> INT "integer"
434 %token <sval> STRING "string"
435 %printer { fprintf (yyo, "%d", $$); } <ival>
436 %destructor { free ($$); } <sval>
437
438 /* In yylex(). */
439 yylval.ival = 42; return INT;
440 yylval.sval = "42"; return STRING;
441
630a0218
AD
442 The %define variable api.value.type supports both keyword and code values.
443
444 The keyword value 'union' means that the user provides genuine types, not
435575cb 445 union member names such as "ival" and "sval" above (WARNING: will fail if
1fa19a76 446 -y/--yacc/%yacc is enabled).
6574576c 447
435575cb 448 %define api.value.type union
6574576c
AD
449 %token <int> INT "integer"
450 %token <char *> STRING "string"
451 %printer { fprintf (yyo, "%d", $$); } <int>
452 %destructor { free ($$); } <char *>
453
454 /* In yylex(). */
455 yylval.INT = 42; return INT;
456 yylval.STRING = "42"; return STRING;
457
435575cb
AD
458 The keyword value variant is somewhat equivalent, but for C++ special
459 provision is made to allow classes to be used (more about this below).
6574576c 460
435575cb 461 %define api.value.type variant
6574576c
AD
462 %token <int> INT "integer"
463 %token <std::string> STRING "string"
464
630a0218 465 Code values (in braces) denote user defined types. This is where YYSTYPE
435575cb 466 used to be used.
6574576c
AD
467
468 %code requires
469 {
470 struct my_value
471 {
472 enum
473 {
474 is_int, is_string
475 } kind;
476 union
477 {
478 int ival;
479 char *sval;
480 } u;
481 };
482 }
435575cb 483 %define api.value.type {struct my_value}
6574576c
AD
484 %token <u.ival> INT "integer"
485 %token <u.sval> STRING "string"
486 %printer { fprintf (yyo, "%d", $$); } <u.ival>
487 %destructor { free ($$); } <u.sval>
488
489 /* In yylex(). */
490 yylval.u.ival = 42; return INT;
491 yylval.u.sval = "42"; return STRING;
492
31b850d2
AD
493** Variable parse.error
494
1f77b2e0
AD
495 This variable controls the verbosity of error messages. The use of the
496 %error-verbose directive is deprecated in favor of "%define parse.error
497 verbose".
31b850d2 498
c21e515e
AD
499** Renamed %define variables
500
501 The following variables have been renamed for consistency. Backward
502 compatibility is ensured, but upgrading is recommended.
503
504 lr.default-reductions -> lr.default-reduction
505 lr.keep-unreachable-states -> lr.keep-unreachable-state
506 namespace -> api.namespace
507 stype -> api.value.type
508
ca2a6d15
PH
509** Semantic predicates
510
597afd73
AD
511 Contributed by Paul Hilfinger.
512
1f77b2e0
AD
513 The new, experimental, semantic-predicate feature allows actions of the
514 form "%?{ BOOLEAN-EXPRESSION }", which cause syntax errors (as for
ca2a6d15 515 YYERROR) if the expression evaluates to 0, and are evaluated immediately
1f77b2e0
AD
516 in GLR parsers, rather than being deferred. The result is that they allow
517 the programmer to prune possible parses based on the values of run-time
518 expressions.
ca2a6d15 519
d1400569
AD
520** The directive %expect-rr is now an error in non GLR mode
521
522 It used to be an error only if used in non GLR mode, _and_ if there are
523 reduce/reduce conflicts.
524
5202b6ac 525** Tokens are numbered in their order of appearance
93561c21 526
5202b6ac
VT
527 Contributed by Valentin Tolmer.
528
529 With '%token A B', A had a number less than the one of B. However,
530 precedence declarations used to generate a reversed order. This is now
531 fixed, and introducing tokens with any of %token, %left, %right,
532 %precedence, or %nonassoc yields the same result.
533
534 When mixing declarations of tokens with a litteral character (e.g., 'a')
535 or with an identifier (e.g., B) in a precedence declaration, Bison
536 numbered the litteral characters first. For example
537
538 %right A B 'c' 'd'
539
540 would lead to the tokens declared in this order: 'c' 'd' A B. Again, the
541 input order is now preserved.
542
543 These changes were made so that one can remove useless precedence and
544 associativity declarations (i.e., map %nonassoc, %left or %right to
545 %precedence, or to %token) and get exactly the same output.
93561c21 546
cc2235ac
VT
547** Useless precedence and associativity
548
d2f9ae18
AD
549 Contributed by Valentin Tolmer.
550
1282c124 551 When developing and maintaining a grammar, useless associativity and
cc2235ac
VT
552 precedence directives are common. They can be a nuisance: new ambiguities
553 arising are sometimes masked because their conflicts are resolved due to
554 the extra precedence or associativity information. Furthermore, it can
555 hinder the comprehension of a new grammar: one will wonder about the role
556 of a precedence, where in fact it is useless. The following changes aim
557 at detecting and reporting these extra directives.
558
559*** Precedence warning category
560
561 A new category of warning, -Wprecedence, was introduced. It flags the
562 useless precedence and associativity directives.
563
564*** Useless associativity
565
566 Bison now warns about symbols with a declared associativity that is never
567 used to resolve conflicts. In that case, using %precedence is sufficient;
568 the parsing tables will remain unchanged. Solving these warnings may raise
569 useless precedence warnings, as the symbols no longer have associativity.
570 For example:
571
572 %left '+'
573 %left '*'
574 %%
575 exp:
1282c124
AD
576 "number"
577 | exp '+' "number"
cc2235ac
VT
578 | exp '*' exp
579 ;
580
581 will produce a
582
583 warning: useless associativity for '+', use %precedence [-Wprecedence]
584 %left '+'
585 ^^^
586
587*** Useless precedence
588
589 Bison now warns about symbols with a declared precedence and no declared
590 associativity (i.e., declared with %precedence), and whose precedence is
591 never used. In that case, the symbol can be safely declared with %token
592 instead, without modifying the parsing tables. For example:
593
594 %precedence '='
595 %%
1282c124 596 exp: "var" '=' "number";
cc2235ac
VT
597
598 will produce a
599
600 warning: useless precedence for '=' [-Wprecedence]
601 %precedence '='
602 ^^^
603
604*** Useless precedence and associativity
605
606 In case of both useless precedence and associativity, the issue is flagged
607 as follows:
608
609 %nonassoc '='
610 %%
1282c124 611 exp: "var" '=' "number";
cc2235ac
VT
612
613 The warning is:
614
615 warning: useless precedence and associativity for '=' [-Wprecedence]
616 %nonassoc '='
617 ^^^
09add9c2
AD
618
619** Empty rules
620
6240346a
AD
621 With help from Joel E. Denny and Gabriel Rassoul.
622
09add9c2
AD
623 Empty rules (i.e., with an empty right-hand side) can now be explicitly
624 marked by the new %empty directive. Using %empty on a non-empty rule is
625 an error. The new -Wempty-rule warning reports empty rules without
626 %empty. On the following grammar:
627
628 %%
629 s: a b c;
630 a: ;
631 b: %empty;
632 c: 'a' %empty;
633
634 bison reports:
635
636 3.4-5: warning: empty rule without %empty [-Wempty-rule]
637 a: {}
638 ^^
639 5.8-13: error: %empty on non-empty rule
640 c: 'a' %empty {};
641 ^^^^^^
cc2235ac 642
c21e515e
AD
643** Java skeleton improvements
644
c21e515e
AD
645 The constants for token names were moved to the Lexer interface. Also, it
646 is possible to add code to the parser's constructors using "%code init"
647 and "%define init_throws".
aa94def1
DH
648 Contributed by Paolo Bonzini.
649
de1a2f20
AD
650 The Java skeleton now supports push parsing.
651 Contributed by Dennis Heimbigner.
652
c21e515e
AD
653** C++ skeletons improvements
654
655*** The parser header is no longer mandatory (lalr1.cc, glr.cc)
656
657 Using %defines is now optional. Without it, the needed support classes
658 are defined in the generated parser, instead of additional files (such as
659 location.hh, position.hh and stack.hh).
660
661*** Locations are no longer mandatory (lalr1.cc, glr.cc)
662
663 Both lalr1.cc and glr.cc no longer require %location.
664
665*** syntax_error exception (lalr1.cc)
666
667 The C++ parser features a syntax_error exception, which can be
668 thrown from the scanner or from user rules to raise syntax errors.
669 This facilitates reporting errors caught in sub-functions (e.g.,
670 rejecting too large integral literals from a conversion function
671 used by the scanner, or rejecting invalid combinations from a
672 factory invoked by the user actions).
673
674*** %define api.value.type variant
675
676 This is based on a submission from Michiel De Wilde. With help
677 from Théophile Ranquet.
678
679 In this mode, complex C++ objects can be used as semantic values. For
680 instance:
681
682 %token <::std::string> TEXT;
683 %token <int> NUMBER;
684 %token SEMICOLON ";"
685 %type <::std::string> item;
686 %type <::std::list<std::string>> list;
687 %%
688 result:
689 list { std::cout << $1 << std::endl; }
690 ;
691
692 list:
6240346a 693 %empty { /* Generates an empty string list. */ }
c21e515e
AD
694 | list item ";" { std::swap ($$, $1); $$.push_back ($2); }
695 ;
696
697 item:
698 TEXT { std::swap ($$, $1); }
699 | NUMBER { $$ = string_cast ($1); }
700 ;
701
702*** %define api.token.constructor
703
704 When variants are enabled, Bison can generate functions to build the
705 tokens. This guarantees that the token type (e.g., NUMBER) is consistent
706 with the semantic value (e.g., int):
707
708 parser::symbol_type yylex ()
709 {
710 parser::location_type loc = ...;
711 ...
712 return parser::make_TEXT ("Hello, world!", loc);
713 ...
714 return parser::make_NUMBER (42, loc);
715 ...
716 return parser::make_SEMICOLON (loc);
717 ...
718 }
719
75ae8299
AD
720*** C++ locations
721
722 There are operator- and operator-= for 'location'. Negative line/column
723 increments can no longer underflow the resulting value.
724
1f5542fe 725* Noteworthy changes in release 2.7.1 (2013-04-15) [stable]
cc8962bd 726
80a2826e
AD
727** Bug fixes
728
729*** Fix compiler attribute portability (yacc.c)
730
731 With locations enabled, __attribute__ was used unprotected.
0a7b8559 732
e83be476
AD
733*** Fix some compiler warnings (lalr1.cc)
734
c13bb348 735* Noteworthy changes in release 2.7 (2012-12-12) [stable]
effd30c0 736
edf9a06f 737** Bug fixes
7bada535 738
edf9a06f 739 Warnings about uninitialized yylloc in yyparse have been fixed.
7bada535 740
1127a75a
AD
741 Restored C90 compliance (yet no report was ever made).
742
d4fe9e88 743** Diagnostics are improved
7bada535 744
8458a411
AD
745 Contributed by Théophile Ranquet.
746
d4fe9e88 747*** Changes in the format of error messages
7bada535 748
d4fe9e88 749 This used to be the format of many error reports:
1f1bd572 750
d4fe9e88
AD
751 input.y:2.7-12: %type redeclaration for exp
752 input.y:1.7-12: previous declaration
1f1bd572 753
d4fe9e88 754 It is now:
1f1bd572 755
d4fe9e88
AD
756 input.y:2.7-12: error: %type redeclaration for exp
757 input.y:1.7-12: previous declaration
cbaea010 758
d4fe9e88 759*** New format for error reports: carets
cbaea010 760
d4fe9e88 761 Caret errors have been added to Bison:
cbaea010 762
d4fe9e88
AD
763 input.y:2.7-12: error: %type redeclaration for exp
764 %type <sval> exp
765 ^^^^^^
766 input.y:1.7-12: previous declaration
767 %type <ival> exp
768 ^^^^^^
cbaea010 769
d4fe9e88 770 or
cbaea010 771
7bada535 772 input.y:3.20-23: error: ambiguous reference: '$exp'
fb6040f0 773 exp: exp '+' exp { $exp = $1 + $3; };
7bada535 774 ^^^^
fb6040f0
TR
775 input.y:3.1-3: refers to: $exp at $$
776 exp: exp '+' exp { $exp = $1 + $3; };
777 ^^^
778 input.y:3.6-8: refers to: $exp at $1
779 exp: exp '+' exp { $exp = $1 + $3; };
780 ^^^
781 input.y:3.14-16: refers to: $exp at $3
782 exp: exp '+' exp { $exp = $1 + $3; };
783 ^^^
7bada535 784
1282c124
AD
785 The default behavior for now is still not to display these unless
786 explicitly asked with -fcaret (or -fall). However, in a later release, it
d4fe9e88
AD
787 will be made the default behavior (but may still be deactivated with
788 -fno-caret).
d3e4409a 789
1f1bd572 790** New value for %define variable: api.pure full
d3e4409a 791
1f1bd572 792 The %define variable api.pure requests a pure (reentrant) parser. However,
d4fe9e88
AD
793 for historical reasons, using it in a location-tracking Yacc parser
794 resulted in a yyerror function that did not take a location as a
795 parameter. With this new value, the user may request a better pure parser,
796 where yyerror does take a location as a parameter (in location-tracking
797 parsers).
1f1bd572
TR
798
799 The use of "%define api.pure true" is deprecated in favor of this new
800 "%define api.pure full".
d3e4409a 801
7287be84 802** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java)
db8ab2be
AD
803
804 The %define variable api.location.type defines the name of the type to use
805 for locations. When defined, Bison no longer generates the position.hh
806 and location.hh files, nor does the parser will include them: the user is
807 then responsible to define her type.
808
809 This can be used in programs with several parsers to factor their location
7287be84
AD
810 and position files: let one of them generate them, and the others just use
811 them.
db8ab2be
AD
812
813 This feature was actually introduced, but not documented, in Bison 2.5,
814 under the name "location_type" (which is maintained for backward
815 compatibility).
816
7287be84
AD
817 For consistency, lalr1.java's %define variables location_type and
818 position_type are deprecated in favor of api.location.type and
819 api.position.type.
820
d4fe9e88
AD
821** Exception safety (lalr1.cc)
822
823 The parse function now catches exceptions, uses the %destructors to
824 release memory (the lookahead symbol and the symbols pushed on the stack)
825 before re-throwing the exception.
826
827 This feature is somewhat experimental. User feedback would be
828 appreciated.
829
9c16d399 830** Graph improvements in DOT and XSLT
fc4fdd62 831
8458a411
AD
832 Contributed by Théophile Ranquet.
833
fc4fdd62
TR
834 The graphical presentation of the states is more readable: their shape is
835 now rectangular, the state number is clearly displayed, and the items are
836 numbered and left-justified.
837
838 The reductions are now explicitly represented as transitions to other
839 diamond shaped nodes.
840
9c16d399
TR
841 These changes are present in both --graph output and xml2dot.xsl XSLT
842 processing, with minor (documented) differences.
843
d4fe9e88 844** %language is no longer an experimental feature.
fb4c8a7c 845
d4fe9e88
AD
846 The introduction of this feature, in 2.4, was four years ago. The
847 --language option and the %language directive are no longer experimental.
fb4c8a7c 848
53e2cd1e
AD
849** Documentation
850
851 The sections about shift/reduce and reduce/reduce conflicts resolution
852 have been fixed and extended.
9d2423f5 853
d4fe9e88
AD
854 Although introduced more than four years ago, XML and Graphviz reports
855 were not properly documented.
856
be22823e
AD
857 The translation of mid-rule actions is now described.
858
9d3f7eaf 859* Noteworthy changes in release 2.6.5 (2012-11-07) [stable]
6f1360bd 860
a68b1f23
AD
861 We consider compiler warnings about Bison generated parsers to be bugs.
862 Rather than working around them in your own project, please consider
863 reporting them to us.
864
865** Bug fixes
866
867 Warnings about uninitialized yylval and/or yylloc for push parsers with a
868 pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to
869 3.2.
870
871 Other issues in the test suite have been addressed.
6f1360bd 872
1282c124 873 Null characters are correctly displayed in error messages.
95066e92 874
a1d1ab50
AD
875 When possible, yylloc is correctly initialized before calling yylex. It
876 is no longer necessary to initialize it in the %initial-action.
877
0ac15849 878* Noteworthy changes in release 2.6.4 (2012-10-23) [stable]
a4eb820f 879
468455e1 880 Bison 2.6.3's --version was incorrect. This release fixes this issue.
a4eb820f 881
6eb8f74f 882* Noteworthy changes in release 2.6.3 (2012-10-22) [stable]
933ec544 883
6b4cb804
AD
884** Bug fixes
885
a1a77e1f 886 Bugs and portability issues in the test suite have been fixed.
6b4cb804
AD
887
888 Some errors in translations have been addressed, and --help now directs
889 users to the appropriate place to report them.
890
891 Stray Info files shipped by accident are removed.
892
893 Incorrect definitions of YY_, issued by yacc.c when no parser header is
894 generated, are removed.
9c26b8fc 895
a2b3f101
TR
896 All the generated headers are self-contained.
897
c9d5bcc9
AD
898** Header guards (yacc.c, glr.c, glr.cc)
899
900 In order to avoid collisions, the header guards are now
901 YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>.
902 For instance the header generated from
903
904 %define api.prefix "calc"
905 %defines "lib/parse.h"
906
907 will use YY_CALC_LIB_PARSE_H_INCLUDED as guard.
908
c12c4c50 909** Fix compiler warnings in the generated parser (yacc.c, glr.c)
321d3e35
AD
910
911 The compilation of pure parsers (%define api.pure) can trigger GCC
912 warnings such as:
913
914 input.c: In function 'yyparse':
915 input.c:1503:12: warning: 'yylval' may be used uninitialized in this
916 function [-Wmaybe-uninitialized]
917 *++yyvsp = yylval;
918 ^
919
920 This is now fixed; pragmas to avoid these warnings are no longer needed.
921
c12c4c50
AD
922 Warnings from clang ("equality comparison with extraneous parentheses" and
923 "function declared 'noreturn' should not return") have also been
924 addressed.
925
e1eeecd3 926* Noteworthy changes in release 2.6.2 (2012-08-03) [stable]
9c26b8fc 927
43ca8040
AD
928** Bug fixes
929
930 Buffer overruns, complaints from Flex, and portability issues in the test
931 suite have been fixed.
932
c9d546b2
AD
933** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc)
934
935 Trailing end-of-lines in %parse-param or %lex-param would result in
936 invalid C++. This is fixed.
9c26b8fc 937
dcd5344d
AD
938** Spurious spaces and end-of-lines
939
940 The generated files no longer end (nor start) with empty lines.
941
77b214ef 942* Noteworthy changes in release 2.6.1 (2012-07-30) [stable]
a4107f24 943
8617d87e
AD
944 Bison no longer executes user-specified M4 code when processing a grammar.
945
e20e6a50
AD
946** Future Changes
947
948 In addition to the removal of the features announced in Bison 2.6, the
949 next major release will remove the "Temporary hack for adding a semicolon
950 to the user action", as announced in the release 2.5. Instead of:
951
952 exp: exp "+" exp { $$ = $1 + $3 };
953
954 write:
955
956 exp: exp "+" exp { $$ = $1 + $3; };
957
8617d87e
AD
958** Bug fixes
959
0e164d43
AD
960*** Type names are now properly escaped.
961
962*** glr.cc: set_debug_level and debug_level work as expected.
a4107f24 963
26313726
AD
964*** Stray @ or $ in actions
965
966 While Bison used to warn about stray $ or @ in action rules, it did not
967 for other actions such as printers, destructors, or initial actions. It
968 now does.
969
cd735a8c 970** Type names in actions
4982f078
AD
971
972 For consistency with rule actions, it is now possible to qualify $$ by a
cd735a8c 973 type-name in destructors, printers, and initial actions. For instance:
4982f078
AD
974
975 %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>;
976
977 will display two values for each typed and untyped symbol (provided
cd735a8c 978 that YYSTYPE has both "ival" and "fval" fields).
60aa04a2 979
1505e8bb 980* Noteworthy changes in release 2.6 (2012-07-19) [stable]
0f11eec2 981
d0a30438 982** Future changes
9553083c 983
55d1006f
AD
984 The next major release of Bison will drop support for the following
985 deprecated features. Please report disagreements to bug-bison@gnu.org.
0f11eec2 986
aaf61036 987*** K&R C parsers
55d1006f
AD
988
989 Support for generating parsers in K&R C will be removed. Parsers
242cc08e 990 generated for C support ISO C90, and are tested with ISO C99 and ISO C11
55d1006f
AD
991 compilers.
992
258cddbc 993*** Features deprecated since Bison 1.875
0f11eec2 994
258cddbc
AD
995 The definitions of yystype and yyltype will be removed; use YYSTYPE and
996 YYLTYPE.
0f11eec2 997
258cddbc
AD
998 YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and
999 %lex-param, will no longer be supported.
1000
1001 Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use
1002 %error-verbose.
55d1006f
AD
1003
1004*** The generated header will be included (yacc.c)
0f11eec2
AD
1005
1006 Instead of duplicating the content of the generated header (definition of
55d1006f
AD
1007 YYSTYPE, yyparse declaration etc.), the generated parser will include it,
1008 as is already the case for GLR or C++ parsers. This change is deferred
1009 because existing versions of ylwrap (e.g., Automake 1.12.1) do not support
1010 it.
0f11eec2 1011
c2425191 1012** Generated Parser Headers
56ca3d8f 1013
258cddbc 1014*** Guards (yacc.c, glr.c, glr.cc)
c3e9f08f
AD
1015
1016 The generated headers are now guarded, as is already the case for C++
242cc08e 1017 parsers (lalr1.cc). For instance, with --defines=foo.h:
c3e9f08f 1018
e29f0771
AD
1019 #ifndef YY_FOO_H
1020 # define YY_FOO_H
1021 ...
1022 #endif /* !YY_FOO_H */
c3e9f08f 1023
258cddbc 1024*** New declarations (yacc.c, glr.c)
56ca3d8f
AD
1025
1026 The generated header now declares yydebug and yyparse. Both honor
1027 --name-prefix=bar_, and yield
1028
e29f0771 1029 int bar_parse (void);
56ca3d8f
AD
1030
1031 rather than
1032
e29f0771
AD
1033 #define yyparse bar_parse
1034 int yyparse (void);
56ca3d8f
AD
1035
1036 in order to facilitate the inclusion of several parser headers inside a
1037 single compilation unit.
c3e9f08f 1038
258cddbc
AD
1039*** Exported symbols in C++
1040
1041 The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the
1042 header, are removed, as they prevent the possibility of including several
1043 generated headers from a single compilation unit.
1044
694af10c
AD
1045*** YYLSP_NEEDED
1046
1047 For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no
1048 longer defined.
1049
4b3847c3
AD
1050** New %define variable: api.prefix
1051
1052 Now that the generated headers are more complete and properly protected
1053 against multiple inclusions, constant names, such as YYSTYPE are a
1054 problem. While yyparse and others are properly renamed by %name-prefix,
1055 YYSTYPE, YYDEBUG and others have never been affected by it. Because it
1056 would introduce backward compatibility issues in projects not expecting
1057 YYSTYPE to be renamed, instead of changing the behavior of %name-prefix,
1058 it is deprecated in favor of a new %define variable: api.prefix.
1059
1060 The following examples compares both:
1061
1062 %name-prefix "bar_" | %define api.prefix "bar_"
1063 %token <ival> FOO %token <ival> FOO
1064 %union { int ival; } %union { int ival; }
1065 %% %%
1066 exp: 'a'; exp: 'a';
1067
1068 bison generates:
1069
1070 #ifndef BAR_FOO_H #ifndef BAR_FOO_H
1071 # define BAR_FOO_H # define BAR_FOO_H
1072
1073 /* Enabling traces. */ /* Enabling traces. */
5f108727
AD
1074 # ifndef YYDEBUG | # ifndef BAR_DEBUG
1075 > # if defined YYDEBUG
1076 > # if YYDEBUG
1077 > # define BAR_DEBUG 1
1078 > # else
1079 > # define BAR_DEBUG 0
1080 > # endif
1081 > # else
1082 # define YYDEBUG 0 | # define BAR_DEBUG 0
1083 > # endif
1084 # endif | # endif
1085
1086 # if YYDEBUG | # if BAR_DEBUG
4b3847c3
AD
1087 extern int bar_debug; extern int bar_debug;
1088 # endif # endif
1089
1090 /* Tokens. */ /* Tokens. */
1091 # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE
1092 # define YYTOKENTYPE | # define BAR_TOKENTYPE
1093 enum yytokentype { | enum bar_tokentype {
1094 FOO = 258 FOO = 258
1095 }; };
1096 # endif # endif
1097
1098 #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \
1099 && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED
1100 typedef union YYSTYPE | typedef union BAR_STYPE
1101 { {
1102 int ival; int ival;
1103 } YYSTYPE; | } BAR_STYPE;
1104 # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1
1105 #endif #endif
1106
1107 extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval;
1108
1109 int bar_parse (void); int bar_parse (void);
1110
1111 #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */
1112
dfaac272 1113* Noteworthy changes in release 2.5.1 (2012-06-05) [stable]
df6e3db0 1114
debe2c03 1115** Future changes:
765e1bd4 1116
e4ab1254 1117 The next major release will drop support for generating parsers in K&R C.
041308d0 1118
466b4cf2 1119** yacc.c: YYBACKUP works as expected.
ef51bfa7 1120
d834eca0 1121** glr.c improvements:
041308d0 1122
d834eca0 1123*** Location support is eliminated when not requested:
041308d0 1124
e4ab1254
AD
1125 GLR parsers used to include location-related code even when locations were
1126 not requested, and therefore not even usable.
378e917c 1127
d834eca0 1128*** __attribute__ is preserved:
d115aad9 1129
e4ab1254
AD
1130 __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e.,
1131 when -std is passed to GCC).
041308d0 1132
466b4cf2 1133** lalr1.java: several fixes:
041308d0 1134
e4ab1254
AD
1135 The Java parser no longer throws ArrayIndexOutOfBoundsException if the
1136 first token leads to a syntax error. Some minor clean ups.
041308d0 1137
22172d47 1138** Changes for C++:
ef51bfa7 1139
22172d47 1140*** C++11 compatibility:
ef51bfa7 1141
e4ab1254
AD
1142 C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L
1143 or higher.
936c88d1 1144
22172d47
AD
1145*** Header guards
1146
1147 The header files such as "parser.hh", "location.hh", etc. used a constant
1148 name for preprocessor guards, for instance:
1149
e29f0771
AD
1150 #ifndef BISON_LOCATION_HH
1151 # define BISON_LOCATION_HH
1152 ...
1153 #endif // !BISON_LOCATION_HH
22172d47
AD
1154
1155 The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower
1156 case characters are converted to upper case, and series of
1157 non-alphanumerical characters are converted to an underscore.
1158
1159 With "bison -o lang++/parser.cc", "location.hh" would now include:
1160
e29f0771
AD
1161 #ifndef YY_LANG_LOCATION_HH
1162 # define YY_LANG_LOCATION_HH
1163 ...
1164 #endif // !YY_LANG_LOCATION_HH
22172d47
AD
1165
1166*** C++ locations:
936c88d1 1167
e4ab1254
AD
1168 The position and location constructors (and their initialize methods)
1169 accept new arguments for line and column. Several issues in the
1170 documentation were fixed.
936c88d1 1171
466b4cf2
AD
1172** liby is no longer asking for "rpl_fprintf" on some platforms.
1173
7e508a2b
AD
1174** Changes in the manual:
1175
1176*** %printer is documented
1177
e4ab1254
AD
1178 The "%printer" directive, supported since at least Bison 1.50, is finally
1179 documented. The "mfcalc" example is extended to demonstrate it.
7e508a2b 1180
e4ab1254
AD
1181 For consistency with the C skeletons, the C++ parsers now also support
1182 "yyoutput" (as an alias to "debug_stream ()").
7e508a2b
AD
1183
1184*** Several improvements have been made:
466b4cf2 1185
e4ab1254
AD
1186 The layout for grammar excerpts was changed to a more compact scheme.
1187 Named references are motivated. The description of the automaton
1188 description file (*.output) is updated to the current format. Incorrect
1189 index entries were fixed. Some other errors were fixed.
466b4cf2 1190
86b08b49
AD
1191** Building bison:
1192
1193*** Conflicting prototypes with recent/modified Flex.
466b4cf2 1194
e4ab1254
AD
1195 Fixed build problems with the current, unreleased, version of Flex, and
1196 some modified versions of 2.5.35, which have modified function prototypes.
466b4cf2 1197
8ef26c2a
AD
1198*** Warnings during the build procedure have been eliminated.
1199
1200*** Several portability problems in the test suite have been fixed:
466b4cf2 1201
e4ab1254
AD
1202 This includes warnings with some compilers, unexpected behavior of tools
1203 such as diff, warning messages from the test suite itself, etc.
466b4cf2 1204
91aadcc7 1205*** The install-pdf target works properly:
8ef26c2a 1206
e4ab1254
AD
1207 Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer
1208 halts in the middle of its course.
8ef26c2a 1209
28801043 1210* Changes in version 2.5 (2011-05-14):
50cca368 1211
82f3355e
JD
1212** Grammar symbol names can now contain non-initial dashes:
1213
1214 Consistently with directives (such as %error-verbose) and with
1215 %define variables (e.g. push-pull), grammar symbol names may contain
1216 dashes in any position except the beginning. This is a GNU
1217 extension over POSIX Yacc. Thus, use of this extension is reported
1218 by -Wyacc and rejected in Yacc mode (--yacc).
1219
f1b238df 1220** Named references:
66381412
AR
1221
1222 Historically, Yacc and Bison have supported positional references
1223 ($n, $$) to allow access to symbol values from inside of semantic
1224 actions code.
1225
1226 Starting from this version, Bison can also accept named references.
1227 When no ambiguity is possible, original symbol names may be used
1228 as named references:
1229
4b568fc0 1230 if_stmt : "if" cond_expr "then" then_stmt ';'
66381412
AR
1231 { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }
1232
1233 In the more common case, explicit names may be declared:
1234
4b568fc0 1235 stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';'
66381412
AR
1236 { $res = mk_if_stmt($cond, $then, $else); }
1237
5b1ff423 1238 Location information is also accessible using @name syntax. When
66381412
AR
1239 accessing symbol names containing dots or dashes, explicit bracketing
1240 ($[sym.1]) must be used.
1241
5b1ff423 1242 These features are experimental in this version. More user feedback
66381412 1243 will help to stabilize them.
2bd435c3 1244 Contributed by Alex Rozenman.
66381412 1245
f1b238df 1246** IELR(1) and canonical LR(1):
eb45ef3b
JD
1247
1248 IELR(1) is a minimal LR(1) parser table generation algorithm. That
1249 is, given any context-free grammar, IELR(1) generates parser tables
7262f54f 1250 with the full language-recognition power of canonical LR(1) but with
f1b238df
JD
1251 nearly the same number of parser states as LALR(1). This reduction
1252 in parser states is often an order of magnitude. More importantly,
eb45ef3b
JD
1253 because canonical LR(1)'s extra parser states may contain duplicate
1254 conflicts in the case of non-LR(1) grammars, the number of conflicts
1255 for IELR(1) is often an order of magnitude less as well. This can
1256 significantly reduce the complexity of developing of a grammar.
1257
1258 Bison can now generate IELR(1) and canonical LR(1) parser tables in
1259 place of its traditional LALR(1) parser tables, which remain the
1260 default. You can specify the type of parser tables in the grammar
1261 file with these directives:
1262
cf499cff
JD
1263 %define lr.type lalr
1264 %define lr.type ielr
1265 %define lr.type canonical-lr
eb45ef3b 1266
7fceb615 1267 The default-reduction optimization in the parser tables can also be
e4ab1254
AD
1268 adjusted using "%define lr.default-reductions". For details on both
1269 of these features, see the new section "Tuning LR" in the Bison
7fceb615 1270 manual.
eb45ef3b
JD
1271
1272 These features are experimental. More user feedback will help to
1273 stabilize them.
1274
8458a411
AD
1275** LAC (Lookahead Correction) for syntax error handling
1276
1277 Contributed by Joel E. Denny.
fcf834f9
JD
1278
1279 Canonical LR, IELR, and LALR can suffer from a couple of problems
1280 upon encountering a syntax error. First, the parser might perform
1281 additional parser stack reductions before discovering the syntax
7fceb615 1282 error. Such reductions can perform user semantic actions that are
fcf834f9
JD
1283 unexpected because they are based on an invalid token, and they
1284 cause error recovery to begin in a different syntactic context than
1285 the one in which the invalid token was encountered. Second, when
7fceb615 1286 verbose error messages are enabled (with %error-verbose or the
e4ab1254 1287 obsolete "#define YYERROR_VERBOSE"), the expected token list in the
7fceb615
JD
1288 syntax error message can both contain invalid tokens and omit valid
1289 tokens.
fcf834f9
JD
1290
1291 The culprits for the above problems are %nonassoc, default
1292 reductions in inconsistent states, and parser state merging. Thus,
1293 IELR and LALR suffer the most. Canonical LR can suffer only if
1294 %nonassoc is used or if default reductions are enabled for
1295 inconsistent states.
1296
7fceb615
JD
1297 LAC is a new mechanism within the parsing algorithm that solves
1298 these problems for canonical LR, IELR, and LALR without sacrificing
1299 %nonassoc, default reductions, or state merging. When LAC is in
1300 use, canonical LR and IELR behave almost exactly the same for both
1301 syntactically acceptable and syntactically unacceptable input.
fcf834f9
JD
1302 While LALR still does not support the full language-recognition
1303 power of canonical LR and IELR, LAC at least enables LALR's syntax
1304 error handling to correctly reflect LALR's language-recognition
1305 power.
1306
1307 Currently, LAC is only supported for deterministic parsers in C.
1308 You can enable LAC with the following directive:
1309
1310 %define parse.lac full
1311
e4ab1254 1312 See the new section "LAC" in the Bison manual for additional
7fceb615 1313 details including a few caveats.
fcf834f9
JD
1314
1315 LAC is an experimental feature. More user feedback will help to
1316 stabilize it.
1317
d397d9f0 1318** %define improvements:
cf499cff 1319
f1b238df 1320*** Can now be invoked via the command line:
50cca368 1321
de5ab940 1322 Each of these command-line options
50cca368 1323
de5ab940
JD
1324 -D NAME[=VALUE]
1325 --define=NAME[=VALUE]
1326
1327 -F NAME[=VALUE]
1328 --force-define=NAME[=VALUE]
50cca368
JD
1329
1330 is equivalent to this grammar file declaration
1331
de5ab940 1332 %define NAME ["VALUE"]
50cca368 1333
de5ab940
JD
1334 except that the manner in which Bison processes multiple definitions
1335 for the same NAME differs. Most importantly, -F and --force-define
1336 quietly override %define, but -D and --define do not. For further
e4ab1254 1337 details, see the section "Bison Options" in the Bison manual.
50cca368 1338
f1b238df 1339*** Variables renamed:
67212941
JD
1340
1341 The following %define variables
1342
1343 api.push_pull
1344 lr.keep_unreachable_states
1345
1346 have been renamed to
1347
1348 api.push-pull
1349 lr.keep-unreachable-states
1350
1351 The old names are now deprecated but will be maintained indefinitely
1352 for backward compatibility.
1353
7262f54f 1354*** Values no longer need to be quoted in the grammar file:
cf499cff
JD
1355
1356 If a %define value is an identifier, it no longer needs to be placed
1357 within quotations marks. For example,
1358
1359 %define api.push-pull "push"
1360
1361 can be rewritten as
1362
1363 %define api.push-pull push
1364
d397d9f0 1365*** Unrecognized variables are now errors not warnings.
cdf3f113 1366
d397d9f0
JD
1367*** Multiple invocations for any variable is now an error not a warning.
1368
1369** Unrecognized %code qualifiers are now errors not warnings.
1370
1371** Character literals not of length one:
1372
1373 Previously, Bison quietly converted all character literals to length
1374 one. For example, without warning, Bison interpreted the operators in
1375 the following grammar to be the same token:
1376
1377 exp: exp '++'
1378 | exp '+' exp
1379 ;
1380
1381 Bison now warns when a character literal is not of length one. In
1382 some future release, Bison will start reporting an error instead.
1383
1384** Destructor calls fixed for lookaheads altered in semantic actions:
1385
1386 Previously for deterministic parsers in C, if a user semantic action
1387 altered yychar, the parser in some cases used the old yychar value to
1388 determine which destructor to call for the lookahead upon a syntax
1389 error or upon parser return. This bug has been fixed.
1390
1391** C++ parsers use YYRHSLOC:
1392
1393 Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
1394 macro and use it in the default YYLLOC_DEFAULT. You are encouraged
e4ab1254
AD
1395 to use it. If, for instance, your location structure has "first"
1396 and "last" members, instead of
d397d9f0 1397
e29f0771
AD
1398 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1399 do \
1400 if (N) \
1401 { \
1402 (Current).first = (Rhs)[1].location.first; \
1403 (Current).last = (Rhs)[N].location.last; \
1404 } \
1405 else \
1406 { \
1407 (Current).first = (Current).last = (Rhs)[0].location.last; \
1408 } \
1409 while (false)
d397d9f0
JD
1410
1411 use:
1412
e29f0771
AD
1413 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1414 do \
1415 if (N) \
1416 { \
1417 (Current).first = YYRHSLOC (Rhs, 1).first; \
1418 (Current).last = YYRHSLOC (Rhs, N).last; \
1419 } \
1420 else \
1421 { \
1422 (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \
1423 } \
1424 while (false)
d397d9f0
JD
1425
1426** YYLLOC_DEFAULT in C++:
1427
1428 The default implementation of YYLLOC_DEFAULT used to be issued in
1429 the header file. It is now output in the implementation file, after
1430 the user %code sections so that its #ifndef guard does not try to
1431 override the user's YYLLOC_DEFAULT if provided.
cdf3f113 1432
f1b238df 1433** YYFAIL now produces warnings and Java parsers no longer implement it:
4395a9ff
JD
1434
1435 YYFAIL has existed for many years as an undocumented feature of
1436 deterministic parsers in C generated by Bison. More recently, it was
1437 a documented feature of Bison's experimental Java parsers. As
1438 promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
1439 semantic action now produces a deprecation warning, and Java parsers
1440 no longer implement YYFAIL at all. For further details, including a
1441 discussion of how to suppress C preprocessor warnings about YYFAIL
1442 being unused, see the Bison 2.4.2 NEWS entry.
1443
f1b238df 1444** Temporary hack for adding a semicolon to the user action:
197b82ba
JD
1445
1446 Previously, Bison appended a semicolon to every user action for
1447 reductions when the output language defaulted to C (specifically, when
1448 neither %yacc, %language, %skeleton, or equivalent command-line
1449 options were specified). This allowed actions such as
1450
1451 exp: exp "+" exp { $$ = $1 + $3 };
1452
1453 instead of
1454
1455 exp: exp "+" exp { $$ = $1 + $3; };
1456
1457 As a first step in removing this misfeature, Bison now issues a
1458 warning when it appends a semicolon. Moreover, in cases where Bison
1459 cannot easily determine whether a semicolon is needed (for example, an
1460 action ending with a cpp directive or a braced compound initializer),
1461 it no longer appends one. Thus, the C compiler might now complain
1462 about a missing semicolon where it did not before. Future releases of
1463 Bison will cease to append semicolons entirely.
1464
d2060f06
JD
1465** Verbose syntax error message fixes:
1466
e4ab1254 1467 When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is
7fceb615
JD
1468 specified, syntax error messages produced by the generated parser
1469 include the unexpected token as well as a list of expected tokens.
1470 The effect of %nonassoc on these verbose messages has been corrected
1471 in two ways, but a more complete fix requires LAC, described above:
d2060f06
JD
1472
1473*** When %nonassoc is used, there can exist parser states that accept no
1474 tokens, and so the parser does not always require a lookahead token
1475 in order to detect a syntax error. Because no unexpected token or
1476 expected tokens can then be reported, the verbose syntax error
1477 message described above is suppressed, and the parser instead
e4ab1254 1478 reports the simpler message, "syntax error". Previously, this
d2060f06
JD
1479 suppression was sometimes erroneously triggered by %nonassoc when a
1480 lookahead was actually required. Now verbose messages are
1481 suppressed only when all previous lookaheads have already been
1482 shifted or discarded.
1483
1484*** Previously, the list of expected tokens erroneously included tokens
1485 that would actually induce a syntax error because conflicts for them
1486 were resolved with %nonassoc in the current parser state. Such
1487 tokens are now properly omitted from the list.
1488
1489*** Expected token lists are still often wrong due to state merging
fcf834f9
JD
1490 (from LALR or IELR) and default reductions, which can both add
1491 invalid tokens and subtract valid tokens. Canonical LR almost
1492 completely fixes this problem by eliminating state merging and
1493 default reductions. However, there is one minor problem left even
1494 when using canonical LR and even after the fixes above. That is,
1495 if the resolution of a conflict with %nonassoc appears in a later
1496 parser state than the one at which some syntax error is
1497 discovered, the conflicted token is still erroneously included in
1498 the expected token list. Bison's new LAC implementation,
1499 described above, eliminates this problem and the need for
1500 canonical LR. However, LAC is still experimental and is disabled
1501 by default.
53f036ce 1502
1a33f4f6
JD
1503** Java skeleton fixes:
1504
1505*** A location handling bug has been fixed.
1506
1507*** The top element of each of the value stack and location stack is now
1508 cleared when popped so that it can be garbage collected.
6771a463 1509
02803d55
JD
1510*** Parser traces now print the top element of the stack.
1511
86408959
JD
1512** -W/--warnings fixes:
1513
e4ab1254 1514*** Bison now properly recognizes the "no-" versions of categories:
86408959
JD
1515
1516 For example, given the following command line, Bison now enables all
1517 warnings except warnings for incompatibilities with POSIX Yacc:
1518
1519 bison -Wall,no-yacc gram.y
1520
786743d5
JD
1521*** Bison now treats S/R and R/R conflicts like other warnings:
1522
1523 Previously, conflict reports were independent of Bison's normal
1524 warning system. Now, Bison recognizes the warning categories
e4ab1254 1525 "conflicts-sr" and "conflicts-rr". This change has important
786743d5
JD
1526 consequences for the -W and --warnings command-line options. For
1527 example:
1528
1529 bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
1530 bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
1531 bison -Wnone gram.y # no conflicts are reported
1532 bison -Werror gram.y # any conflict is an error
1533
1534 However, as before, if the %expect or %expect-rr directive is
1535 specified, an unexpected number of conflicts is an error, and an
1536 expected number of conflicts is not reported, so -W and --warning
1537 then have no effect on the conflict report.
1538
e4ab1254 1539*** The "none" category no longer disables a preceding "error":
bf0e44e8
JD
1540
1541 For example, for the following command line, Bison now reports
1542 errors instead of warnings for incompatibilities with POSIX Yacc:
1543
1544 bison -Werror,none,yacc gram.y
1545
e4ab1254 1546*** The "none" category now disables all Bison warnings:
c39014ae 1547
e4ab1254 1548 Previously, the "none" category disabled only Bison warnings for
c39014ae
JD
1549 which there existed a specific -W/--warning category. However,
1550 given the following command line, Bison is now guaranteed to
1551 suppress all warnings:
1552
1553 bison -Wnone gram.y
1554
1f36f544
JD
1555** Precedence directives can now assign token number 0:
1556
1557 Since Bison 2.3b, which restored the ability of precedence
1558 directives to assign token numbers, doing so for token number 0 has
1559 produced an assertion failure. For example:
1560
1561 %left END 0
1562
1563 This bug has been fixed.
1564
64877e5e 1565* Changes in version 2.4.3 (2010-08-05):
8b9e021f 1566
2bfcac9a
JD
1567** Bison now obeys -Werror and --warnings=error for warnings about
1568 grammar rules that are useless in the parser due to conflicts.
1569
8b9e021f
JD
1570** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
1571 been fixed.
1572
4ad3921d
JD
1573** Failures in the test suite for GCC 4.5 have been fixed.
1574
06cb07d5
JD
1575** Failures in the test suite for some versions of Sun Studio C++ have
1576 been fixed.
1577
9b5049bd
JD
1578** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
1579 warnings about undefined %prec identifiers will not be converted to
1580 errors in Bison 2.5. They will remain warnings, which should be
1581 sufficient for POSIX while avoiding backward compatibility issues.
1582
93d7dde9
JD
1583** Minor documentation fixes.
1584
e19a049c 1585* Changes in version 2.4.2 (2010-03-20):
74553c98 1586
f39ab286
JD
1587** Some portability problems that resulted in failures and livelocks
1588 in the test suite on some versions of at least Solaris, AIX, HP-UX,
e19a049c
JD
1589 RHEL4, and Tru64 have been addressed. As a result, fatal Bison
1590 errors should no longer cause M4 to report a broken pipe on the
f39ab286
JD
1591 affected platforms.
1592
e4ab1254 1593** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately.
8bb3a2e7
JD
1594
1595 POSIX specifies that an error be reported for any identifier that does
1596 not appear on the LHS of a grammar rule and that is not defined by
1597 %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this
1598 error report for the case when an identifier appears only after a
1599 %prec directive. It is now restored. However, for backward
1600 compatibility with recent Bison releases, it is only a warning for
1601 now. In Bison 2.5 and later, it will return to being an error.
9b5049bd
JD
1602 [Between the 2.4.2 and 2.4.3 releases, it was decided that this
1603 warning will not be converted to an error in Bison 2.5.]
8bb3a2e7 1604
d8911864
EB
1605** Detection of GNU M4 1.4.6 or newer during configure is improved.
1606
a603c6e0
JD
1607** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
1608 YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
1609 avoided.
c938d650 1610
98a345a2
JD
1611** %code is now a permanent feature.
1612
1613 A traditional Yacc prologue directive is written in the form:
1614
1615 %{CODE%}
1616
1617 To provide a more flexible alternative, Bison 2.3b introduced the
1618 %code directive with the following forms for C/C++:
1619
1620 %code {CODE}
1621 %code requires {CODE}
1622 %code provides {CODE}
1623 %code top {CODE}
1624
1625 These forms are now considered permanent features of Bison. See the
1626 %code entries in the section "Bison Declaration Summary" in the Bison
1627 manual for a summary of their functionality. See the section
1628 "Prologue Alternatives" for a detailed discussion including the
1629 advantages of %code over the traditional Yacc prologue directive.
1630
1631 Bison's Java feature as a whole including its current usage of %code
1632 is still considered experimental.
1633
1625df5b
JD
1634** YYFAIL is deprecated and will eventually be removed.
1635
1636 YYFAIL has existed for many years as an undocumented feature of
1637 deterministic parsers in C generated by Bison. Previously, it was
1638 documented for Bison's experimental Java parsers. YYFAIL is no longer
1639 documented for Java parsers and is formally deprecated in both cases.
1640 Users are strongly encouraged to migrate to YYERROR, which is
1641 specified by POSIX.
1642
1643 Like YYERROR, you can invoke YYFAIL from a semantic action in order to
1644 induce a syntax error. The most obvious difference from YYERROR is
1645 that YYFAIL will automatically invoke yyerror to report the syntax
1646 error so that you don't have to. However, there are several other
1647 subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
e4ab1254 1648 inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is
1625df5b
JD
1649 used. For a more detailed discussion, see:
1650
1651 http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
1652
1653 The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
1654 deterministic parsers in C will continue to implement it. However,
1655 because YYFAIL is already flawed, it seems futile to try to make new
1656 Bison features compatible with it. Thus, during parser generation,
1657 Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
1658 rule action. In a later release, YYFAIL will be disabled for
e4ab1254 1659 %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will
1625df5b
JD
1660 be removed altogether.
1661
1662 There exists at least one case where Bison 2.5's YYFAIL warning will
1663 be a false positive. Some projects add phony uses of YYFAIL and other
1664 Bison-defined macros for the sole purpose of suppressing C
1665 preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
1666 To avoid Bison's future warning, such YYFAIL uses can be moved to the
e4ab1254 1667 epilogue (that is, after the second "%%") in the Bison input file. In
1625df5b
JD
1668 this release (2.4.2), Bison already generates its own code to suppress
1669 C preprocessor warnings for YYFAIL, so projects can remove their own
1670 phony uses of YYFAIL if compatibility with Bison releases prior to
1671 2.4.2 is not necessary.
1672
2755de8f
AD
1673** Internationalization.
1674
1675 Fix a regression introduced in Bison 2.4: Under some circumstances,
1676 message translations were not installed although supported by the
1677 host system.
1678
74553c98 1679* Changes in version 2.4.1 (2008-12-11):
c9ba9e59 1680
0ea583d2
AD
1681** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
1682 declarations have been fixed.
1979121c 1683
0ea583d2
AD
1684** Temporary hack for adding a semicolon to the user action.
1685
1686 Bison used to prepend a trailing semicolon at the end of the user
1687 action for reductions. This allowed actions such as
1688
1689 exp: exp "+" exp { $$ = $1 + $3 };
1690
1691 instead of
1692
1693 exp: exp "+" exp { $$ = $1 + $3; };
1694
e4ab1254 1695 Some grammars still depend on this "feature". Bison 2.4.1 restores
0ea583d2
AD
1696 the previous behavior in the case of C output (specifically, when
1697 neither %language or %skeleton or equivalent command-line options
1698 are used) to leave more time for grammars depending on the old
1699 behavior to be adjusted. Future releases of Bison will disable this
1700 feature.
1701
1702** A few minor improvements to the Bison manual.
c9ba9e59 1703
402b123d 1704* Changes in version 2.4 (2008-11-02):
7bd1665a 1705
402b123d 1706** %language is an experimental feature.
ed4d67dc
JD
1707
1708 We first introduced this feature in test release 2.3b as a cleaner
1709 alternative to %skeleton. Since then, we have discussed the possibility of
1710 modifying its effect on Bison's output file names. Thus, in this release,
1711 we consider %language to be an experimental feature that will likely evolve
1712 in future releases.
7bd1665a 1713
402b123d 1714** Forward compatibility with GNU M4 has been improved.
241fda7a 1715
402b123d 1716** Several bugs in the C++ skeleton and the experimental Java skeleton have been
241fda7a
JD
1717 fixed.
1718
402b123d 1719* Changes in version 2.3b (2008-05-27):
35fe0834 1720
402b123d 1721** The quotes around NAME that used to be required in the following directive
d9df47b6
JD
1722 are now deprecated:
1723
1724 %define NAME "VALUE"
1725
e4ab1254 1726** The directive "%pure-parser" is now deprecated in favor of:
d9df47b6
JD
1727
1728 %define api.pure
1729
1730 which has the same effect except that Bison is more careful to warn about
1731 unreasonable usage in the latter case.
1732
402b123d 1733** Push Parsing
c373bf8b
JD
1734
1735 Bison can now generate an LALR(1) parser in C with a push interface. That
e4ab1254
AD
1736 is, instead of invoking "yyparse", which pulls tokens from "yylex", you can
1737 push one token at a time to the parser using "yypush_parse", which will
c373bf8b
JD
1738 return to the caller after processing each token. By default, the push
1739 interface is disabled. Either of the following directives will enable it:
1740
1741 %define api.push_pull "push" // Just push; does not require yylex.
1742 %define api.push_pull "both" // Push and pull; requires yylex.
1743
e4ab1254 1744 See the new section "A Push Parser" in the Bison manual for details.
c373bf8b 1745
59da312b
JD
1746 The current push parsing interface is experimental and may evolve. More user
1747 feedback will help to stabilize it.
1748
402b123d 1749** The -g and --graph options now output graphs in Graphviz DOT format,
8e55b3aa
JD
1750 not VCG format. Like --graph, -g now also takes an optional FILE argument
1751 and thus cannot be bundled with other short options.
c373bf8b 1752
402b123d 1753** Java
59da312b
JD
1754
1755 Bison can now generate an LALR(1) parser in Java. The skeleton is
e4ab1254 1756 "data/lalr1.java". Consider using the new %language directive instead of
59da312b
JD
1757 %skeleton to select it.
1758
e4ab1254 1759 See the new section "Java Parsers" in the Bison manual for details.
59da312b
JD
1760
1761 The current Java interface is experimental and may evolve. More user
1762 feedback will help to stabilize it.
2bd435c3 1763 Contributed by Paolo Bonzini.
59da312b 1764
402b123d 1765** %language
59da312b
JD
1766
1767 This new directive specifies the programming language of the generated
d43f77e7
PB
1768 parser, which can be C (the default), C++, or Java. Besides the skeleton
1769 that Bison uses, the directive affects the names of the generated files if
1770 the grammar file's name ends in ".y".
59da312b 1771
402b123d 1772** XML Automaton Report
59da312b
JD
1773
1774 Bison can now generate an XML report of the LALR(1) automaton using the new
e4ab1254 1775 "--xml" option. The current XML schema is experimental and may evolve. More
59da312b 1776 user feedback will help to stabilize it.
2bd435c3 1777 Contributed by Wojciech Polak.
c373bf8b 1778
402b123d 1779** The grammar file may now specify the name of the parser header file using
c373bf8b
JD
1780 %defines. For example:
1781
1782 %defines "parser.h"
1783
402b123d 1784** When reporting useless rules, useless nonterminals, and unused terminals,
d80fb37a
JD
1785 Bison now employs the terms "useless in grammar" instead of "useless",
1786 "useless in parser" instead of "never reduced", and "unused in grammar"
1787 instead of "unused".
cff03fb2 1788
402b123d 1789** Unreachable State Removal
c373bf8b
JD
1790
1791 Previously, Bison sometimes generated parser tables containing unreachable
31984206
JD
1792 states. A state can become unreachable during conflict resolution if Bison
1793 disables a shift action leading to it from a predecessor state. Bison now:
75ad86ee
JD
1794
1795 1. Removes unreachable states.
1796
1797 2. Does not report any conflicts that appeared in unreachable states.
1798 WARNING: As a result, you may need to update %expect and %expect-rr
1799 directives in existing grammar files.
1800
1801 3. For any rule used only in such states, Bison now reports the rule as
cff03fb2 1802 "useless in parser due to conflicts".
75ad86ee 1803
31984206
JD
1804 This feature can be disabled with the following directive:
1805
1806 %define lr.keep_unreachable_states
1807
e4ab1254 1808 See the %define entry in the "Bison Declaration Summary" in the Bison manual
31984206
JD
1809 for further discussion.
1810
e4ab1254 1811** Lookahead Set Correction in the ".output" Report
b1cc23c4 1812
e4ab1254
AD
1813 When instructed to generate a ".output" file including lookahead sets
1814 (using "--report=lookahead", for example), Bison now prints each reduction's
88c78747
JD
1815 lookahead set only next to the associated state's one item that (1) is
1816 associated with the same rule as the reduction and (2) has its dot at the end
1817 of its RHS. Previously, Bison also erroneously printed the lookahead set
1818 next to all of the state's other items associated with the same rule. This
e4ab1254 1819 bug affected only the ".output" file and not the generated parser source
88c78747
JD
1820 code.
1821
e4ab1254 1822** --report-file=FILE is a new option to override the default ".output" file
59da312b 1823 name.
1bb2bd75 1824
e4ab1254 1825** The "=" that used to be required in the following directives is now
02975b9a
JD
1826 deprecated:
1827
1828 %file-prefix "parser"
1829 %name-prefix "c_"
1830 %output "parser.c"
1831
e4ab1254 1832** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}"
c373bf8b
JD
1833
1834 Bison 2.3a provided a new set of directives as a more flexible alternative to
8e0a5e9e
JD
1835 the traditional Yacc prologue blocks. Those have now been consolidated into
1836 a single %code directive with an optional qualifier field, which identifies
1837 the purpose of the code and thus the location(s) where Bison should generate
1838 it:
1839
e4ab1254
AD
1840 1. "%code {CODE}" replaces "%after-header {CODE}"
1841 2. "%code requires {CODE}" replaces "%start-header {CODE}"
1842 3. "%code provides {CODE}" replaces "%end-header {CODE}"
1843 4. "%code top {CODE}" replaces "%before-header {CODE}"
8e0a5e9e 1844
e4ab1254
AD
1845 See the %code entries in section "Bison Declaration Summary" in the Bison
1846 manual for a summary of the new functionality. See the new section "Prologue
1847 Alternatives" for a detailed discussion including the advantages of %code
8e0a5e9e
JD
1848 over the traditional Yacc prologues.
1849
1850 The prologue alternatives are experimental. More user feedback will help to
1851 determine whether they should become permanent features.
1852
402b123d 1853** Revised warning: unset or unused mid-rule values
17bd8a73
JD
1854
1855 Since Bison 2.2, Bison has warned about mid-rule values that are set but not
1856 used within any of the actions of the parent rule. For example, Bison warns
1857 about unused $2 in:
1858
1859 exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };
1860
1861 Now, Bison also warns about mid-rule values that are used but not set. For
1862 example, Bison warns about unset $$ in the mid-rule action in:
1863
1864 exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };
1865
1866 However, Bison now disables both of these warnings by default since they
1867 sometimes prove to be false alarms in existing grammars employing the Yacc
1868 constructs $0 or $-N (where N is some positive integer).
1869
e4ab1254
AD
1870 To enable these warnings, specify the option "--warnings=midrule-values" or
1871 "-W", which is a synonym for "--warnings=all".
17bd8a73 1872
e4ab1254 1873** Default %destructor or %printer with "<*>" or "<>"
c373bf8b
JD
1874
1875 Bison now recognizes two separate kinds of default %destructor's and
12e35840
JD
1876 %printer's:
1877
e4ab1254 1878 1. Place "<*>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1879 %destructor/%printer for all grammar symbols for which you have formally
1880 declared semantic type tags.
1881
e4ab1254 1882 2. Place "<>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1883 %destructor/%printer for all grammar symbols without declared semantic
1884 type tags.
1885
e4ab1254
AD
1886 Bison no longer supports the "%symbol-default" notation from Bison 2.3a.
1887 "<*>" and "<>" combined achieve the same effect with one exception: Bison no
12e35840
JD
1888 longer applies any %destructor to a mid-rule value if that mid-rule value is
1889 not actually ever referenced using either $$ or $n in a semantic action.
1890
85894313
JD
1891 The default %destructor's and %printer's are experimental. More user
1892 feedback will help to determine whether they should become permanent
1893 features.
1894
e4ab1254 1895 See the section "Freeing Discarded Symbols" in the Bison manual for further
12e35840
JD
1896 details.
1897
402b123d 1898** %left, %right, and %nonassoc can now declare token numbers. This is required
e4ab1254 1899 by POSIX. However, see the end of section "Operator Precedence" in the Bison
ab7f29f8
JD
1900 manual for a caveat concerning the treatment of literal strings.
1901
402b123d 1902** The nonfunctional --no-parser, -n, and %no-parser options have been
b1cc23c4
JD
1903 completely removed from Bison.
1904
402b123d 1905* Changes in version 2.3a, 2006-09-13:
742e4900 1906
402b123d 1907** Instead of %union, you can define and use your own union type
ddc8ede1
PE
1908 YYSTYPE if your grammar contains at least one <type> tag.
1909 Your YYSTYPE need not be a macro; it can be a typedef.
1910 This change is for compatibility with other Yacc implementations,
1911 and is required by POSIX.
1912
402b123d 1913** Locations columns and lines start at 1.
cd48d21d
AD
1914 In accordance with the GNU Coding Standards and Emacs.
1915
402b123d 1916** You may now declare per-type and default %destructor's and %printer's:
ec5479ce
JD
1917
1918 For example:
1919
b2a0b7ca
JD
1920 %union { char *string; }
1921 %token <string> STRING1
1922 %token <string> STRING2
1923 %type <string> string1
1924 %type <string> string2
1925 %union { char character; }
1926 %token <character> CHR
1927 %type <character> chr
1928 %destructor { free ($$); } %symbol-default
1929 %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
1930 %destructor { } <character>
1931
1932 guarantees that, when the parser discards any user-defined symbol that has a
e4ab1254
AD
1933 semantic type tag other than "<character>", it passes its semantic value to
1934 "free". However, when the parser discards a "STRING1" or a "string1", it
1935 also prints its line number to "stdout". It performs only the second
1936 "%destructor" in this case, so it invokes "free" only once.
ec5479ce 1937
85894313
JD
1938 [Although we failed to mention this here in the 2.3a release, the default
1939 %destructor's and %printer's were experimental, and they were rewritten in
1940 future versions.]
1941
e4ab1254
AD
1942** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y",
1943 "--yacc", or "%yacc"), Bison no longer generates #define statements for
b931235e
JD
1944 associating token numbers with token names. Removing the #define statements
1945 helps to sanitize the global namespace during preprocessing, but POSIX Yacc
1946 requires them. Bison still generates an enum for token names in all cases.
1947
402b123d 1948** Handling of traditional Yacc prologue blocks is now more consistent but
34f98f46 1949 potentially incompatible with previous releases of Bison.
9bc0dd67
JD
1950
1951 As before, you declare prologue blocks in your grammar file with the
e4ab1254 1952 "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all
34f98f46
JD
1953 prologue blocks that you've declared before the first %union. To generate
1954 the post-prologue, Bison concatenates all prologue blocks that you've
ddc8ede1 1955 declared after the first %union.
9bc0dd67 1956
34f98f46 1957 Previous releases of Bison inserted the pre-prologue into both the header
9bc0dd67
JD
1958 file and the code file in all cases except for LALR(1) parsers in C. In the
1959 latter case, Bison inserted it only into the code file. For parsers in C++,
1960 the point of insertion was before any token definitions (which associate
1961 token numbers with names). For parsers in C, the point of insertion was
1962 after the token definitions.
1963
1964 Now, Bison never inserts the pre-prologue into the header file. In the code
1965 file, it always inserts it before the token definitions.
1966
402b123d 1967** Bison now provides a more flexible alternative to the traditional Yacc
34f98f46
JD
1968 prologue blocks: %before-header, %start-header, %end-header, and
1969 %after-header.
1970
1971 For example, the following declaration order in the grammar file reflects the
1972 order in which Bison will output these code blocks. However, you are free to
1973 declare these code blocks in your grammar file in whatever order is most
1974 convenient for you:
1975
1976 %before-header {
1977 /* Bison treats this block like a pre-prologue block: it inserts it into
1978 * the code file before the contents of the header file. It does *not*
1979 * insert it into the header file. This is a good place to put
1980 * #include's that you want at the top of your code file. A common
e4ab1254 1981 * example is '#include "system.h"'. */
34f98f46
JD
1982 }
1983 %start-header {
1984 /* Bison inserts this block into both the header file and the code file.
1985 * In both files, the point of insertion is before any Bison-generated
1986 * token, semantic type, location type, and class definitions. This is a
1987 * good place to define %union dependencies, for example. */
9bc0dd67
JD
1988 }
1989 %union {
34f98f46
JD
1990 /* Unlike the traditional Yacc prologue blocks, the output order for the
1991 * new %*-header blocks is not affected by their declaration position
1992 * relative to any %union in the grammar file. */
9bc0dd67 1993 }
34f98f46
JD
1994 %end-header {
1995 /* Bison inserts this block into both the header file and the code file.
1996 * In both files, the point of insertion is after the Bison-generated
1997 * definitions. This is a good place to declare or define public
1998 * functions or data structures that depend on the Bison-generated
1999 * definitions. */
9bc0dd67 2000 }
34f98f46
JD
2001 %after-header {
2002 /* Bison treats this block like a post-prologue block: it inserts it into
2003 * the code file after the contents of the header file. It does *not*
2004 * insert it into the header file. This is a good place to declare or
2005 * define internal functions or data structures that depend on the
2006 * Bison-generated definitions. */
2007 }
2008
2009 If you have multiple occurrences of any one of the above declarations, Bison
2010 will concatenate the contents in declaration order.
9bc0dd67 2011
85894313
JD
2012 [Although we failed to mention this here in the 2.3a release, the prologue
2013 alternatives were experimental, and they were rewritten in future versions.]
2014
e4ab1254 2015** The option "--report=look-ahead" has been changed to "--report=lookahead".
9e6e7ed2
PE
2016 The old spelling still works, but is not documented and may be removed
2017 in a future release.
742e4900 2018
402b123d 2019* Changes in version 2.3, 2006-06-05:
4ad3ed84 2020
e4ab1254 2021** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING",
4ad3ed84
PE
2022 for compatibility with LALR(1) grammars.
2023
402b123d 2024** It is now documented that any definition of YYSTYPE or YYLTYPE should
4ad3ed84
PE
2025 be to a type name that does not contain parentheses or brackets.
2026
402b123d 2027* Changes in version 2.2, 2006-05-19:
193d7c70 2028
402b123d 2029** The distribution terms for all Bison-generated parsers now permit
193d7c70
PE
2030 using the parsers in nonfree programs. Previously, this permission
2031 was granted only for Bison-generated LALR(1) parsers in C.
5f4236a0 2032
402b123d 2033** %name-prefix changes the namespace name in C++ outputs.
aa08666d 2034
402b123d 2035** The C++ parsers export their token_type.
5f4236a0 2036
402b123d 2037** Bison now allows multiple %union declarations, and concatenates
d6ca7905
PE
2038 their contents together.
2039
402b123d 2040** New warning: unused values
4d7bc38c
PE
2041 Right-hand side symbols whose values are not used are reported,
2042 if the symbols have destructors. For instance:
affac613 2043
8f3596a6 2044 exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
e9690142
JD
2045 | exp "+" exp
2046 ;
affac613 2047
8f3596a6
AD
2048 will trigger a warning about $$ and $5 in the first rule, and $3 in
2049 the second ($1 is copied to $$ by the default rule). This example
4e26c69e 2050 most likely contains three errors, and could be rewritten as:
affac613 2051
4e26c69e 2052 exp: exp "?" exp ":" exp
e9690142
JD
2053 { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
2054 | exp "+" exp
2055 { $$ = $1 ? $1 : $3; if ($1) free ($3); }
2056 ;
affac613 2057
4e26c69e
PE
2058 However, if the original actions were really intended, memory leaks
2059 and all, the warnings can be suppressed by letting Bison believe the
2060 values are used, e.g.:
721be13c 2061
8f3596a6 2062 exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
e9690142
JD
2063 | exp "+" exp { $$ = $1; (void) $3; }
2064 ;
721be13c 2065
84866159
AD
2066 If there are mid-rule actions, the warning is issued if no action
2067 uses it. The following triggers no warning: $1 and $3 are used.
2068
2069 exp: exp { push ($1); } '+' exp { push ($3); sum (); };
2070
721be13c
PE
2071 The warning is intended to help catching lost values and memory leaks.
2072 If a value is ignored, its associated memory typically is not reclaimed.
affac613 2073
402b123d 2074** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
9d9b8b70
PE
2075 Destructors are now called when user code invokes YYABORT, YYACCEPT,
2076 and YYERROR, for all objects on the stack, other than objects
2077 corresponding to the right-hand side of the current rule.
a85284cf 2078
402b123d 2079** %expect, %expect-rr
035aa4a0
PE
2080 Incorrect numbers of expected conflicts are now actual errors,
2081 instead of warnings.
2082
402b123d 2083** GLR, YACC parsers.
4e26c69e
PE
2084 The %parse-params are available in the destructors (and the
2085 experimental printers) as per the documentation.
4b367315 2086
e4ab1254 2087** Bison now warns if it finds a stray "$" or "@" in an action.
ad6a9b97 2088
402b123d 2089** %require "VERSION"
4e26c69e
PE
2090 This specifies that the grammar file depends on features implemented
2091 in Bison version VERSION or higher.
b50d2359 2092
402b123d 2093** lalr1.cc: The token and value types are now class members.
e14d0ab6
AD
2094 The tokens were defined as free form enums and cpp macros. YYSTYPE
2095 was defined as a free form union. They are now class members:
e4ab1254
AD
2096 tokens are enumerations of the "yy::parser::token" struct, and the
2097 semantic values have the "yy::parser::semantic_type" type.
fb9712a9
AD
2098
2099 If you do not want or can update to this scheme, the directive
e4ab1254 2100 '%define "global_tokens_and_yystype" "1"' triggers the global
b50d2359
AD
2101 definition of tokens and YYSTYPE. This change is suitable both
2102 for previous releases of Bison, and this one.
fb9712a9 2103
b50d2359 2104 If you wish to update, then make sure older version of Bison will
e4ab1254 2105 fail using '%require "2.2"'.
fb9712a9 2106
402b123d 2107** DJGPP support added.
193d7c70 2108\f
402b123d 2109* Changes in version 2.1, 2005-09-16:
1ce59070 2110
402b123d 2111** The C++ lalr1.cc skeleton supports %lex-param.
e14d0ab6 2112
402b123d 2113** Bison-generated parsers now support the translation of diagnostics like
baf785db
PE
2114 "syntax error" into languages other than English. The default
2115 language is still English. For details, please see the new
0410a6e0
PE
2116 Internationalization section of the Bison manual. Software
2117 distributors should also see the new PACKAGING file. Thanks to
2118 Bruno Haible for this new feature.
1ce59070 2119
402b123d 2120** Wording in the Bison-generated parsers has been changed slightly to
1a059451
PE
2121 simplify translation. In particular, the message "memory exhausted"
2122 has replaced "parser stack overflow", as the old message was not
2123 always accurate for modern Bison-generated parsers.
2124
402b123d 2125** Destructors are now called when the parser aborts, for all symbols left
258b75ca
PE
2126 behind on the stack. Also, the start symbol is now destroyed after a
2127 successful parse. In both cases, the behavior was formerly inconsistent.
2128
402b123d 2129** When generating verbose diagnostics, Bison-generated parsers no longer
72f000b0
PE
2130 quote the literal strings associated with tokens. For example, for
2131 a syntax error associated with '%token NUM "number"' they might
2132 print 'syntax error, unexpected number' instead of 'syntax error,
2133 unexpected "number"'.
193d7c70 2134\f
402b123d 2135* Changes in version 2.0, 2004-12-25:
efeed023 2136
402b123d 2137** Possibly-incompatible changes
d7e14fc0 2138
82de6b0d
PE
2139 - Bison-generated parsers no longer default to using the alloca function
2140 (when available) to extend the parser stack, due to widespread
2141 problems in unchecked stack-overflow detection. You can "#define
2142 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
2143 the manual to determine safe values for YYMAXDEPTH in that case.
8dd162d3 2144
82de6b0d
PE
2145 - Error token location.
2146 During error recovery, the location of the syntax error is updated
2147 to cover the whole sequence covered by the error token: it includes
2148 the shifted symbols thrown away during the first part of the error
2149 recovery, and the lookahead rejected during the second part.
18d192f0 2150
82de6b0d
PE
2151 - Semicolon changes:
2152 . Stray semicolons are no longer allowed at the start of a grammar.
2153 . Semicolons are now required after in-grammar declarations.
e342c3be 2154
82de6b0d
PE
2155 - Unescaped newlines are no longer allowed in character constants or
2156 string literals. They were never portable, and GCC 3.4.0 has
2157 dropped support for them. Better diagnostics are now generated if
2158 forget a closing quote.
8dd162d3 2159
82de6b0d 2160 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
f74b6f91 2161
402b123d 2162** New features
1452af69 2163
82de6b0d 2164 - GLR grammars now support locations.
4febdd96 2165
82de6b0d
PE
2166 - New directive: %initial-action.
2167 This directive allows the user to run arbitrary code (including
2168 initializing @$) from yyparse before parsing starts.
1452af69 2169
82de6b0d
PE
2170 - A new directive "%expect-rr N" specifies the expected number of
2171 reduce/reduce conflicts in GLR parsers.
1452af69 2172
e4ab1254 2173 - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d".
82de6b0d 2174 This is a GNU extension.
4febdd96 2175
e4ab1254 2176 - The option "--report=lookahead" was changed to "--report=look-ahead".
9e6e7ed2 2177 [However, this was changed back after 2.3.]
1452af69 2178
82de6b0d 2179 - Experimental %destructor support has been added to lalr1.cc.
1452af69 2180
82de6b0d
PE
2181 - New configure option --disable-yacc, to disable installation of the
2182 yacc command and -ly library introduced in 1.875 for POSIX conformance.
6040d338 2183
402b123d 2184** Bug fixes
d5a3fe37 2185
82de6b0d
PE
2186 - For now, %expect-count violations are now just warnings, not errors.
2187 This is for compatibility with Bison 1.75 and earlier (when there are
2188 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
2189 are too many or too few shift/reduce conflicts). However, in future
2190 versions of Bison we plan to improve the %expect machinery so that
2191 these violations will become errors again.
3473d0f8 2192
82de6b0d
PE
2193 - Within Bison itself, numbers (e.g., goto numbers) are no longer
2194 arbitrarily limited to 16-bit counts.
d600ee67 2195
82de6b0d 2196 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
d600ee67 2197\f
402b123d 2198* Changes in version 1.875, 2003-01-01:
963fcc17 2199
402b123d 2200** The documentation license has been upgraded to version 1.2
dc546b0f 2201 of the GNU Free Documentation License.
75eb3bc4 2202
402b123d 2203** syntax error processing
75eb3bc4 2204
dc546b0f
PE
2205 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
2206 locations too. This fixes bugs in error-location computation.
75eb3bc4 2207
dc546b0f
PE
2208 - %destructor
2209 It is now possible to reclaim the memory associated to symbols
2210 discarded during error recovery. This feature is still experimental.
20daca06 2211
dc546b0f
PE
2212 - %error-verbose
2213 This new directive is preferred over YYERROR_VERBOSE.
74724a70 2214
dc546b0f
PE
2215 - #defining yyerror to steal internal variables is discouraged.
2216 It is not guaranteed to work forever.
d1de5372 2217
402b123d 2218** POSIX conformance
d1de5372 2219
dc546b0f
PE
2220 - Semicolons are once again optional at the end of grammar rules.
2221 This reverts to the behavior of Bison 1.33 and earlier, and improves
2222 compatibility with Yacc.
74724a70 2223
e4ab1254
AD
2224 - "parse error" -> "syntax error"
2225 Bison now uniformly uses the term "syntax error"; formerly, the code
2226 and manual sometimes used the term "parse error" instead. POSIX
2227 requires "syntax error" in diagnostics, and it was thought better to
dc546b0f 2228 be consistent.
74724a70 2229
dc546b0f
PE
2230 - The documentation now emphasizes that yylex and yyerror must be
2231 declared before use. C99 requires this.
d1de5372 2232
dc546b0f
PE
2233 - Bison now parses C99 lexical constructs like UCNs and
2234 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
d1de5372 2235
dc546b0f
PE
2236 - File names are properly escaped in C output. E.g., foo\bar.y is
2237 output as "foo\\bar.y".
6780ca7a 2238
dc546b0f 2239 - Yacc command and library now available
e4ab1254 2240 The Bison distribution now installs a "yacc" command, as POSIX requires.
dc546b0f
PE
2241 Also, Bison now installs a small library liby.a containing
2242 implementations of Yacc-compatible yyerror and main functions.
2243 This library is normally not useful, but POSIX requires it.
6e649e65 2244
dc546b0f 2245 - Type clashes now generate warnings, not errors.
6e649e65 2246
dc546b0f
PE
2247 - If the user does not define YYSTYPE as a macro, Bison now declares it
2248 using typedef instead of defining it as a macro.
2249 For consistency, YYLTYPE is also declared instead of defined.
9501dc6e 2250
402b123d 2251** Other compatibility issues
886a425c 2252
e4ab1254
AD
2253 - %union directives can now have a tag before the "{", e.g., the
2254 directive "%union foo {...}" now generates the C code
2255 "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility.
2256 The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc.
2257 For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype".
dc546b0f 2258 This is for compatibility with both Yacc and Bison 1.35.
72f889cc 2259
e4ab1254 2260 - ";" is output before the terminating "}" of an action, for
dc546b0f 2261 compatibility with Bison 1.35.
886a425c 2262
dc546b0f 2263 - Bison now uses a Yacc-style format for conflict reports, e.g.,
e4ab1254 2264 "conflicts: 2 shift/reduce, 1 reduce/reduce".
437c2d80 2265
e4ab1254 2266 - "yystype" and "yyltype" are now obsolescent macros instead of being
dc546b0f
PE
2267 typedefs or tags; they are no longer documented and are planned to be
2268 withdrawn in a future release.
2a8d363a 2269
402b123d 2270** GLR parser notes
2a8d363a 2271
dc546b0f
PE
2272 - GLR and inline
2273 Users of Bison have to decide how they handle the portability of the
e4ab1254 2274 C keyword "inline".
959e5f51 2275
e4ab1254
AD
2276 - "parsing stack overflow..." -> "parser stack overflow"
2277 GLR parsers now report "parser stack overflow" as per the Bison manual.
900c5db5 2278
18ad57b3
AD
2279** %parse-param and %lex-param
2280 The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass
2281 additional context to yyparse and yylex. They suffer from several
2282 shortcomings:
2283
2284 - a single argument only can be added,
2285 - their types are weak (void *),
242cc08e 2286 - this context is not passed to ancillary functions such as yyerror,
18ad57b3
AD
2287 - only yacc.c parsers support them.
2288
2289 The new %parse-param/%lex-param directives provide a more precise control.
2290 For instance:
2291
2292 %parse-param {int *nastiness}
2293 %lex-param {int *nastiness}
2294 %parse-param {int *randomness}
2295
2296 results in the following signatures:
2297
2298 int yylex (int *nastiness);
2299 int yyparse (int *nastiness, int *randomness);
2300
2301 or, if both %pure-parser and %locations are used:
2302
2303 int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness);
2304 int yyparse (int *nastiness, int *randomness);
2305
402b123d 2306** Bison now warns if it detects conflicting outputs to the same file,
e4ab1254 2307 e.g., it generates a warning for "bison -d -o foo.h foo.y" since
dc546b0f 2308 that command outputs both code and header to foo.h.
6e40b4eb 2309
402b123d 2310** #line in output files
dc546b0f 2311 - --no-line works properly.
6e40b4eb 2312
402b123d 2313** Bison can no longer be built by a K&R C compiler; it requires C89 or
6e40b4eb
AD
2314 later to be built. This change originally took place a few versions
2315 ago, but nobody noticed until we recently asked someone to try
2316 building Bison with a K&R C compiler.
d600ee67 2317\f
402b123d 2318* Changes in version 1.75, 2002-10-14:
7933f2b5 2319
402b123d 2320** Bison should now work on 64-bit hosts.
7933f2b5 2321
402b123d 2322** Indonesian translation thanks to Tedi Heriyanto.
7933f2b5 2323
402b123d 2324** GLR parsers
f50adbbd
AD
2325 Fix spurious parse errors.
2326
402b123d 2327** Pure parsers
f50adbbd
AD
2328 Some people redefine yyerror to steal yyparse' private variables.
2329 Reenable this trick until an official feature replaces it.
2330
402b123d 2331** Type Clashes
d90c934c
AD
2332 In agreement with POSIX and with other Yaccs, leaving a default
2333 action is valid when $$ is untyped, and $1 typed:
2334
e9690142 2335 untyped: ... typed;
d90c934c
AD
2336
2337 but the converse remains an error:
2338
e9690142 2339 typed: ... untyped;
d90c934c 2340
402b123d 2341** Values of mid-rule actions
d90c934c
AD
2342 The following code:
2343
e9690142 2344 foo: { ... } { $$ = $1; } ...
d90c934c
AD
2345
2346 was incorrectly rejected: $1 is defined in the second mid-rule
2347 action, and is equal to the $$ of the first mid-rule action.
d600ee67 2348\f
402b123d 2349* Changes in version 1.50, 2002-10-04:
adc8c848 2350
402b123d 2351** GLR parsing
676385e2
PH
2352 The declaration
2353 %glr-parser
2354 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
2355 almost any context-free grammar, ambiguous or not. The new declarations
e8832397 2356 %dprec and %merge on grammar rules allow parse-time resolution of
676385e2
PH
2357 ambiguities. Contributed by Paul Hilfinger.
2358
7933f2b5 2359 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
420f93c8
PE
2360 like the Alpha, so please stick to 32-bit hosts for now.
2361
402b123d 2362** Output Directory
8c165d89 2363 When not in Yacc compatibility mode, when the output file was not
e4ab1254
AD
2364 specified, running "bison foo/bar.y" created "foo/bar.c". It
2365 now creates "bar.c".
8c165d89 2366
402b123d 2367** Undefined token
007a50a4 2368 The undefined token was systematically mapped to 2 which prevented
e88dbdbf 2369 the use of 2 by the user. This is no longer the case.
007a50a4 2370
402b123d 2371** Unknown token numbers
e88dbdbf 2372 If yylex returned an out of range value, yyparse could die. This is
007a50a4
AD
2373 no longer the case.
2374
402b123d 2375** Error token
e88dbdbf 2376 According to POSIX, the error token must be 256.
23c5a174
AD
2377 Bison extends this requirement by making it a preference: *if* the
2378 user specified that one of her tokens is numbered 256, then error
2379 will be mapped onto another number.
2380
402b123d 2381** Verbose error messages
e4ab1254 2382 They no longer report "..., expecting error or..." for states where
217598da
AD
2383 error recovery is possible.
2384
402b123d 2385** End token
e4ab1254 2386 Defaults to "$end" instead of "$".
217598da 2387
402b123d 2388** Error recovery now conforms to documentation and to POSIX
68cd8af3
PE
2389 When a Bison-generated parser encounters a syntax error, it now pops
2390 the stack until it finds a state that allows shifting the error
2391 token. Formerly, it popped the stack until it found a state that
2392 allowed some non-error action other than a default reduction on the
2393 error token. The new behavior has long been the documented behavior,
2394 and has long been required by POSIX. For more details, please see
337116ba
PE
2395 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
2396 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
68cd8af3 2397
402b123d 2398** Traces
5504898e
AD
2399 Popped tokens and nonterminals are now reported.
2400
402b123d 2401** Larger grammars
a861a339
PE
2402 Larger grammars are now supported (larger token numbers, larger grammar
2403 size (= sum of the LHS and RHS lengths), larger LALR tables).
2404 Formerly, many of these numbers ran afoul of 16-bit limits;
2405 now these limits are 32 bits on most hosts.
355e7c1c 2406
402b123d 2407** Explicit initial rule
643a5994
AD
2408 Bison used to play hacks with the initial rule, which the user does
2409 not write. It is now explicit, and visible in the reports and
2410 graphs as rule 0.
23c5a174 2411
402b123d 2412** Useless rules
643a5994 2413 Before, Bison reported the useless rules, but, although not used,
77714df2 2414 included them in the parsers. They are now actually removed.
23c5a174 2415
402b123d 2416** Useless rules, useless nonterminals
6b98e4b5
AD
2417 They are now reported, as a warning, with their locations.
2418
402b123d 2419** Rules never reduced
e8832397
AD
2420 Rules that can never be reduced because of conflicts are now
2421 reported.
2422
e4ab1254 2423** Incorrect "Token not used"
11652ab3
AD
2424 On a grammar such as
2425
e29f0771
AD
2426 %token useless useful
2427 %%
2428 exp: '0' %prec useful;
11652ab3
AD
2429
2430 where a token was used to set the precedence of the last rule,
e4ab1254 2431 bison reported both "useful" and "useless" as useless tokens.
11652ab3 2432
402b123d 2433** Revert the C++ namespace changes introduced in 1.31
77714df2 2434 as they caused too many portability hassles.
0179dd65 2435
402b123d 2436** Default locations
b2d52318
AD
2437 By an accident of design, the default computation of @$ was
2438 performed after another default computation was performed: @$ = @1.
2439 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
2440 the computation of @$.
adc8c848 2441
402b123d 2442** Token end-of-file
b7c49edf
AD
2443 The token end of file may be specified by the user, in which case,
2444 the user symbol is used in the reports, the graphs, and the verbose
e4ab1254 2445 error messages instead of "$end", which remains being the default.
b7c49edf 2446 For instance
e29f0771 2447 %token MYEOF 0
b7c49edf 2448 or
e29f0771 2449 %token MYEOF 0 "end of file"
fdbcd8e2 2450
402b123d 2451** Semantic parser
fdbcd8e2
AD
2452 This old option, which has been broken for ages, is removed.
2453
402b123d 2454** New translations
a861a339 2455 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
84614e13
AD
2456 Croatian, thanks to Denis Lackovic.
2457
402b123d 2458** Incorrect token definitions
e4ab1254
AD
2459 When given
2460 %token 'a' "A"
2461 bison used to output
2462 #define 'a' 65
b87f8b21 2463
402b123d 2464** Token definitions as enums
77714df2
AD
2465 Tokens are output both as the traditional #define's, and, provided
2466 the compiler supports ANSI C or is a C++ compiler, as enums.
e88dbdbf 2467 This lets debuggers display names instead of integers.
77714df2 2468
402b123d 2469** Reports
ec3bc396
AD
2470 In addition to --verbose, bison supports --report=THINGS, which
2471 produces additional information:
b408954b
AD
2472 - itemset
2473 complete the core item sets with their closure
e4ab1254 2474 - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back]
9e6e7ed2 2475 explicitly associate lookahead tokens to items
b408954b
AD
2476 - solved
2477 describe shift/reduce conflicts solving.
2478 Bison used to systematically output this information on top of
2479 the report. Solved conflicts are now attached to their states.
ec3bc396 2480
402b123d 2481** Type clashes
9af3fbce
AD
2482 Previous versions don't complain when there is a type clash on
2483 the default action if the rule has a mid-rule action, such as in:
2484
e29f0771
AD
2485 %type <foo> bar
2486 %%
2487 bar: '0' {} '0';
9af3fbce
AD
2488
2489 This is fixed.
a861a339 2490
402b123d 2491** GNU M4 is now required when using Bison.
f987e9d2 2492\f
402b123d 2493* Changes in version 1.35, 2002-03-25:
76551463 2494
402b123d 2495** C Skeleton
76551463
AD
2496 Some projects use Bison's C parser with C++ compilers, and define
2497 YYSTYPE as a class. The recent adjustment of C parsers for data
2498 alignment and 64 bit architectures made this impossible.
2499
2500 Because for the time being no real solution for C++ parser
2501 generation exists, kludges were implemented in the parser to
2502 maintain this use. In the future, when Bison has C++ parsers, this
2503 kludge will be disabled.
2504
2505 This kludge also addresses some C++ problems when the stack was
2506 extended.
76551463 2507\f
402b123d 2508* Changes in version 1.34, 2002-03-12:
76551463 2509
402b123d 2510** File name clashes are detected
76551463 2511 $ bison foo.y -d -o foo.x
e4ab1254 2512 fatal error: header and parser would both be named "foo.x"
76551463 2513
e4ab1254 2514** A missing ";" at the end of a rule triggers a warning
76551463
AD
2515 In accordance with POSIX, and in agreement with other
2516 Yacc implementations, Bison will mandate this semicolon in the near
2517 future. This eases the implementation of a Bison parser of Bison
2518 grammars by making this grammar LALR(1) instead of LR(2). To
2519 facilitate the transition, this release introduces a warning.
2520
402b123d 2521** Revert the C++ namespace changes introduced in 1.31, as they caused too
76551463
AD
2522 many portability hassles.
2523
402b123d 2524** DJGPP support added.
76551463 2525
402b123d 2526** Fix test suite portability problems.
76551463 2527\f
402b123d 2528* Changes in version 1.33, 2002-02-07:
76551463 2529
402b123d 2530** Fix C++ issues
76551463
AD
2531 Groff could not be compiled for the definition of size_t was lacking
2532 under some conditions.
2533
402b123d 2534** Catch invalid @n
76551463
AD
2535 As is done with $n.
2536\f
402b123d 2537* Changes in version 1.32, 2002-01-23:
76551463 2538
402b123d 2539** Fix Yacc output file names
76551463 2540
402b123d 2541** Portability fixes
76551463 2542
402b123d 2543** Italian, Dutch translations
76551463 2544\f
402b123d 2545* Changes in version 1.31, 2002-01-14:
52d1aeee 2546
402b123d 2547** Many Bug Fixes
52d1aeee 2548
402b123d 2549** GNU Gettext and %expect
52d1aeee
MA
2550 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
2551 Bison dies on incorrect %expectations, we fear there will be
2552 too many bug reports for Gettext, so _for the time being_, %expect
e4ab1254 2553 does not trigger an error when the input file is named "plural.y".
52d1aeee 2554
402b123d 2555** Use of alloca in parsers
52d1aeee
MA
2556 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
2557 malloc exclusively. Since 1.29, but was not NEWS'ed.
2558
2559 alloca is used only when compiled with GCC, to avoid portability
2560 problems as on AIX.
2561
402b123d 2562** yyparse now returns 2 if memory is exhausted; formerly it dumped core.
b47dbebe 2563
402b123d 2564** When the generated parser lacks debugging code, YYDEBUG is now 0
52d1aeee
MA
2565 (as POSIX requires) instead of being undefined.
2566
402b123d 2567** User Actions
52d1aeee
MA
2568 Bison has always permitted actions such as { $$ = $1 }: it adds the
2569 ending semicolon. Now if in Yacc compatibility mode, the semicolon
2570 is no longer output: one has to write { $$ = $1; }.
2571
402b123d 2572** Better C++ compliance
52d1aeee 2573 The output parsers try to respect C++ namespaces.
76551463 2574 [This turned out to be a failed experiment, and it was reverted later.]
52d1aeee 2575
402b123d 2576** Reduced Grammars
52d1aeee
MA
2577 Fixed bugs when reporting useless nonterminals.
2578
402b123d 2579** 64 bit hosts
52d1aeee
MA
2580 The parsers work properly on 64 bit hosts.
2581
402b123d 2582** Error messages
52d1aeee
MA
2583 Some calls to strerror resulted in scrambled or missing error messages.
2584
402b123d 2585** %expect
52d1aeee
MA
2586 When the number of shift/reduce conflicts is correct, don't issue
2587 any warning.
2588
402b123d 2589** The verbose report includes the rule line numbers.
52d1aeee 2590
402b123d 2591** Rule line numbers are fixed in traces.
52d1aeee 2592
402b123d 2593** Swedish translation
52d1aeee 2594
402b123d 2595** Parse errors
52d1aeee
MA
2596 Verbose parse error messages from the parsers are better looking.
2597 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
2598 Now: parse error: unexpected '/', expecting "number" or '-' or '('
2599
402b123d 2600** Fixed parser memory leaks.
52d1aeee
MA
2601 When the generated parser was using malloc to extend its stacks, the
2602 previous allocations were not freed.
2603
402b123d 2604** Fixed verbose output file.
52d1aeee
MA
2605 Some newlines were missing.
2606 Some conflicts in state descriptions were missing.
2607
402b123d 2608** Fixed conflict report.
52d1aeee
MA
2609 Option -v was needed to get the result.
2610
402b123d 2611** %expect
52d1aeee
MA
2612 Was not used.
2613 Mismatches are errors, not warnings.
2614
402b123d 2615** Fixed incorrect processing of some invalid input.
52d1aeee 2616
402b123d 2617** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
52d1aeee 2618
402b123d 2619** Fixed some typos in the documentation.
52d1aeee 2620
402b123d 2621** %token MY_EOF 0 is supported.
52d1aeee
MA
2622 Before, MY_EOF was silently renumbered as 257.
2623
402b123d 2624** doc/refcard.tex is updated.
52d1aeee 2625
402b123d 2626** %output, %file-prefix, %name-prefix.
52d1aeee
MA
2627 New.
2628
402b123d 2629** --output
e4ab1254 2630 New, aliasing "--output-file".
52d1aeee 2631\f
402b123d 2632* Changes in version 1.30, 2001-10-26:
342b8b6e 2633
e4ab1254
AD
2634** "--defines" and "--graph" have now an optional argument which is the
2635 output file name. "-d" and "-g" do not change; they do not take any
342b8b6e
AD
2636 argument.
2637
e4ab1254 2638** "%source_extension" and "%header_extension" are removed, failed
342b8b6e
AD
2639 experiment.
2640
402b123d 2641** Portability fixes.
f987e9d2 2642\f
402b123d 2643* Changes in version 1.29, 2001-09-07:
342b8b6e 2644
402b123d 2645** The output file does not define const, as this caused problems when used
342b8b6e
AD
2646 with common autoconfiguration schemes. If you still use ancient compilers
2647 that lack const, compile with the equivalent of the C compiler option
e4ab1254 2648 "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this.
342b8b6e 2649
e4ab1254 2650** Added "-g" and "--graph".
f87a2205 2651
402b123d 2652** The Bison manual is now distributed under the terms of the GNU FDL.
f2b5126e 2653
402b123d 2654** The input and the output files has automatically a similar extension.
234a3be3 2655
402b123d 2656** Russian translation added.
f87a2205 2657
402b123d 2658** NLS support updated; should hopefully be less troublesome.
f87a2205 2659
402b123d 2660** Added the old Bison reference card.
c33638bb 2661
e4ab1254 2662** Added "--locations" and "%locations".
6deb4447 2663
e4ab1254 2664** Added "-S" and "--skeleton".
cd5bd6ac 2665
e4ab1254 2666** "%raw", "-r", "--raw" is disabled.
62ab6972 2667
402b123d 2668** Special characters are escaped when output. This solves the problems
cd5bd6ac
AD
2669 of the #line lines with path names including backslashes.
2670
402b123d 2671** New directives.
e4ab1254
AD
2672 "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose",
2673 "%debug", "%source_extension" and "%header_extension".
f987e9d2 2674
402b123d 2675** @$
f987e9d2 2676 Automatic location tracking.
f87a2205 2677\f
402b123d 2678* Changes in version 1.28, 1999-07-06:
d2e00347 2679
402b123d 2680** Should compile better now with K&R compilers.
d2e00347 2681
402b123d 2682** Added NLS.
d2e00347 2683
402b123d 2684** Fixed a problem with escaping the double quote character.
d2e00347 2685
402b123d 2686** There is now a FAQ.
d2e00347 2687\f
402b123d 2688* Changes in version 1.27:
5c31c3c2 2689
402b123d 2690** The make rule which prevented bison.simple from being created on
5c31c3c2
JT
2691 some systems has been fixed.
2692\f
402b123d 2693* Changes in version 1.26:
4be07551 2694
7e508a2b 2695** Bison now uses Automake.
4be07551 2696
402b123d 2697** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
4be07551 2698
402b123d 2699** Token numbers now start at 257 as previously documented, not 258.
4be07551 2700
402b123d 2701** Bison honors the TMPDIR environment variable.
4be07551 2702
402b123d 2703** A couple of buffer overruns have been fixed.
f51dbca1 2704
402b123d 2705** Problems when closing files should now be reported.
f51dbca1 2706
402b123d 2707** Generated parsers should now work even on operating systems which do
f51dbca1 2708 not provide alloca().
4be07551 2709\f
402b123d 2710* Changes in version 1.25, 1995-10-16:
df8878c5 2711
402b123d 2712** Errors in the input grammar are not fatal; Bison keeps reading
df8878c5 2713the grammar file, and reports all the errors found in it.
8c44d3ec 2714
402b123d 2715** Tokens can now be specified as multiple-character strings: for
df8878c5 2716example, you could use "<=" for a token which looks like <=, instead
7e508a2b 2717of choosing a name like LESSEQ.
df8878c5 2718
402b123d 2719** The %token_table declaration says to write a table of tokens (names
df8878c5
RS
2720and numbers) into the parser file. The yylex function can use this
2721table to recognize multiple-character string tokens, or for other
2722purposes.
2723
402b123d 2724** The %no_lines declaration says not to generate any #line preprocessor
df8878c5
RS
2725directives in the parser file.
2726
402b123d 2727** The %raw declaration says to use internal Bison token numbers, not
df8878c5
RS
2728Yacc-compatible token numbers, when token names are defined as macros.
2729
402b123d 2730** The --no-parser option produces the parser tables without including
df8878c5
RS
2731the parser engine; a project can now use its own parser engine.
2732The actions go into a separate file called NAME.act, in the form of
2733a switch statement body.
2734\f
402b123d 2735* Changes in version 1.23:
6780ca7a 2736
4d019228
DM
2737The user can define YYPARSE_PARAM as the name of an argument to be
2738passed into yyparse. The argument should have type void *. It should
2739actually point to an object. Grammar actions can access the variable
2740by casting it to the proper pointer type.
6780ca7a 2741
6780ca7a 2742Line numbers in output file corrected.
6780ca7a 2743\f
402b123d 2744* Changes in version 1.22:
6780ca7a
DM
2745
2746--help option added.
6780ca7a 2747\f
402b123d 2748* Changes in version 1.20:
6780ca7a
DM
2749
2750Output file does not redefine const for C++.
9f4503d6 2751
76551463
AD
2752-----
2753
3209eb1c 2754Copyright (C) 1995-2015 Free Software Foundation, Inc.
76551463 2755
74553c98 2756This file is part of Bison, the GNU Parser Generator.
76551463 2757
f16b0819 2758This program is free software: you can redistribute it and/or modify
76551463 2759it under the terms of the GNU General Public License as published by
f16b0819
PE
2760the Free Software Foundation, either version 3 of the License, or
2761(at your option) any later version.
76551463 2762
f16b0819 2763This program is distributed in the hope that it will be useful,
76551463
AD
2764but WITHOUT ANY WARRANTY; without even the implied warranty of
2765MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2766GNU General Public License for more details.
2767
2768You should have received a copy of the GNU General Public License
f16b0819 2769along with this program. If not, see <http://www.gnu.org/licenses/>.
7e508a2b
AD
2770
2771 LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr
2772 LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr
2773 LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar
2774 LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr
2775 LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks
2776 LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG
2777 LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush
2778 LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly
2779 LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's
2780 LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de
2781 LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto
2782 LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs
2783 LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF
242cc08e 2784 LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY
7e508a2b 2785 LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL
242cc08e
AD
2786 LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh
2787 LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf
dcb366b1 2788 LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp Wother nterm arg init
1282c124
AD
2789 LocalWords: TOK calc yyo fval Wconflicts parsers yystackp yyval yynerrs
2790 LocalWords: Théophile Ranquet Santet fno fnone stype associativity Tolmer
2791 LocalWords: Wprecedence Rassoul Wempty Paolo Bonzini parser's Michiel loc
2792 LocalWords: redeclaration sval fcaret reentrant XSLT xsl Wmaybe yyvsp Tedi
2793 LocalWords: pragmas noreturn untyped Rozenman unexpanded Wojciech Polak
6574576c 2794 LocalWords: Alexandre MERCHANTABILITY yytype
7e508a2b
AD
2795
2796Local Variables:
2797mode: outline
e4ab1254 2798fill-column: 76
7e508a2b 2799End: