]> git.saurik.com Git - bison.git/blob - NEWS
* data/glr.c: Reorder typedef declarations for structs to match order
[bison.git] / NEWS
1 Bison News
2 ----------
3
4 Changes in version 2.1a:
5
6 * GLR, C++ LALR(1) parsers.
7 These parser skeletons are now distributed with the same special
8 license exception that the C LALR(1) parser skeleton has had since
9 Bison 1.24. Hence the FSF imposes no copyright restriction on the
10 output of Bison when it is generating such parsers.
11
12 * %expect, %expect-rr
13 Incorrect numbers of expected conflicts are now actual errors,
14 instead of warnings.
15
16 * GLR, YACC parsers.
17 The %parse-params are available in the %destructor's (and the
18 experimental %printer's) as per the documentation.
19
20 * Bison now warns if it finds a stray `$' or `@' in an action.
21
22 * %require "VERSION"
23 To specify that the grammar file depends on features implemented in
24 Bison version VERSION or higher.
25
26 * lalr1.cc: The token and value types are now class members.
27 The tokens where defined as free form enums and cpp macros. YYSTYPE
28 was defined as a free form union. Both are now class members:
29 tokens are enumerations of the `yy::parser::token' struct, and the
30 semantic values have the `yy::parser::semantic_type' type.
31
32 If you do not want or can update to this scheme, the directive
33 `%define "global_tokens_and_yystype" "1"' triggers the global
34 definition of tokens and YYSTYPE. This change is suitable both
35 for previous releases of Bison, and this one.
36
37 If you wish to update, then make sure older version of Bison will
38 fail using `%require "2.1a"'.
39
40 * DJGPP support added.
41
42 Changes in version 2.1, 2005-09-16:
43
44 * Bison-generated parsers now support the translation of diagnostics like
45 "syntax error" into languages other than English. The default
46 language is still English. For details, please see the new
47 Internationalization section of the Bison manual. Software
48 distributors should also see the new PACKAGING file. Thanks to
49 Bruno Haible for this new feature.
50
51 * Wording in the Bison-generated parsers has been changed slightly to
52 simplify translation. In particular, the message "memory exhausted"
53 has replaced "parser stack overflow", as the old message was not
54 always accurate for modern Bison-generated parsers.
55
56 * Destructors are now called when the parser aborts, for all symbols left
57 behind on the stack. Also, the start symbol is now destroyed after a
58 successful parse. In both cases, the behavior was formerly inconsistent.
59
60 * When generating verbose diagnostics, Bison-generated parsers no longer
61 quote the literal strings associated with tokens. For example, for
62 a syntax error associated with '%token NUM "number"' they might
63 print 'syntax error, unexpected number' instead of 'syntax error,
64 unexpected "number"'.
65
66 Changes in version 2.0, 2004-12-25:
67
68 * Possibly-incompatible changes
69
70 - Bison-generated parsers no longer default to using the alloca function
71 (when available) to extend the parser stack, due to widespread
72 problems in unchecked stack-overflow detection. You can "#define
73 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
74 the manual to determine safe values for YYMAXDEPTH in that case.
75
76 - Error token location.
77 During error recovery, the location of the syntax error is updated
78 to cover the whole sequence covered by the error token: it includes
79 the shifted symbols thrown away during the first part of the error
80 recovery, and the lookahead rejected during the second part.
81
82 - Semicolon changes:
83 . Stray semicolons are no longer allowed at the start of a grammar.
84 . Semicolons are now required after in-grammar declarations.
85
86 - Unescaped newlines are no longer allowed in character constants or
87 string literals. They were never portable, and GCC 3.4.0 has
88 dropped support for them. Better diagnostics are now generated if
89 forget a closing quote.
90
91 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
92
93 * New features
94
95 - GLR grammars now support locations.
96
97 - New directive: %initial-action.
98 This directive allows the user to run arbitrary code (including
99 initializing @$) from yyparse before parsing starts.
100
101 - A new directive "%expect-rr N" specifies the expected number of
102 reduce/reduce conflicts in GLR parsers.
103
104 - %token numbers can now be hexadecimal integers, e.g., `%token FOO 0x12d'.
105 This is a GNU extension.
106
107 - The option `--report=lookahead' was changed to `--report=look-ahead'.
108 The old spelling still works, but is not documented and will be
109 removed.
110
111 - Experimental %destructor support has been added to lalr1.cc.
112
113 - New configure option --disable-yacc, to disable installation of the
114 yacc command and -ly library introduced in 1.875 for POSIX conformance.
115
116 * Bug fixes
117
118 - For now, %expect-count violations are now just warnings, not errors.
119 This is for compatibility with Bison 1.75 and earlier (when there are
120 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
121 are too many or too few shift/reduce conflicts). However, in future
122 versions of Bison we plan to improve the %expect machinery so that
123 these violations will become errors again.
124
125 - Within Bison itself, numbers (e.g., goto numbers) are no longer
126 arbitrarily limited to 16-bit counts.
127
128 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
129 \f
130 Changes in version 1.875, 2003-01-01:
131
132 * The documentation license has been upgraded to version 1.2
133 of the GNU Free Documentation License.
134
135 * syntax error processing
136
137 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
138 locations too. This fixes bugs in error-location computation.
139
140 - %destructor
141 It is now possible to reclaim the memory associated to symbols
142 discarded during error recovery. This feature is still experimental.
143
144 - %error-verbose
145 This new directive is preferred over YYERROR_VERBOSE.
146
147 - #defining yyerror to steal internal variables is discouraged.
148 It is not guaranteed to work forever.
149
150 * POSIX conformance
151
152 - Semicolons are once again optional at the end of grammar rules.
153 This reverts to the behavior of Bison 1.33 and earlier, and improves
154 compatibility with Yacc.
155
156 - `parse error' -> `syntax error'
157 Bison now uniformly uses the term `syntax error'; formerly, the code
158 and manual sometimes used the term `parse error' instead. POSIX
159 requires `syntax error' in diagnostics, and it was thought better to
160 be consistent.
161
162 - The documentation now emphasizes that yylex and yyerror must be
163 declared before use. C99 requires this.
164
165 - Bison now parses C99 lexical constructs like UCNs and
166 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
167
168 - File names are properly escaped in C output. E.g., foo\bar.y is
169 output as "foo\\bar.y".
170
171 - Yacc command and library now available
172 The Bison distribution now installs a `yacc' command, as POSIX requires.
173 Also, Bison now installs a small library liby.a containing
174 implementations of Yacc-compatible yyerror and main functions.
175 This library is normally not useful, but POSIX requires it.
176
177 - Type clashes now generate warnings, not errors.
178
179 - If the user does not define YYSTYPE as a macro, Bison now declares it
180 using typedef instead of defining it as a macro.
181 For consistency, YYLTYPE is also declared instead of defined.
182
183 * Other compatibility issues
184
185 - %union directives can now have a tag before the `{', e.g., the
186 directive `%union foo {...}' now generates the C code
187 `typedef union foo { ... } YYSTYPE;'; this is for Yacc compatibility.
188 The default union tag is `YYSTYPE', for compatibility with Solaris 9 Yacc.
189 For consistency, YYLTYPE's struct tag is now `YYLTYPE' not `yyltype'.
190 This is for compatibility with both Yacc and Bison 1.35.
191
192 - `;' is output before the terminating `}' of an action, for
193 compatibility with Bison 1.35.
194
195 - Bison now uses a Yacc-style format for conflict reports, e.g.,
196 `conflicts: 2 shift/reduce, 1 reduce/reduce'.
197
198 - `yystype' and `yyltype' are now obsolescent macros instead of being
199 typedefs or tags; they are no longer documented and are planned to be
200 withdrawn in a future release.
201
202 * GLR parser notes
203
204 - GLR and inline
205 Users of Bison have to decide how they handle the portability of the
206 C keyword `inline'.
207
208 - `parsing stack overflow...' -> `parser stack overflow'
209 GLR parsers now report `parser stack overflow' as per the Bison manual.
210
211 * Bison now warns if it detects conflicting outputs to the same file,
212 e.g., it generates a warning for `bison -d -o foo.h foo.y' since
213 that command outputs both code and header to foo.h.
214
215 * #line in output files
216 - --no-line works properly.
217
218 * Bison can no longer be built by a K&R C compiler; it requires C89 or
219 later to be built. This change originally took place a few versions
220 ago, but nobody noticed until we recently asked someone to try
221 building Bison with a K&R C compiler.
222 \f
223 Changes in version 1.75, 2002-10-14:
224
225 * Bison should now work on 64-bit hosts.
226
227 * Indonesian translation thanks to Tedi Heriyanto.
228
229 * GLR parsers
230 Fix spurious parse errors.
231
232 * Pure parsers
233 Some people redefine yyerror to steal yyparse' private variables.
234 Reenable this trick until an official feature replaces it.
235
236 * Type Clashes
237 In agreement with POSIX and with other Yaccs, leaving a default
238 action is valid when $$ is untyped, and $1 typed:
239
240 untyped: ... typed;
241
242 but the converse remains an error:
243
244 typed: ... untyped;
245
246 * Values of mid-rule actions
247 The following code:
248
249 foo: { ... } { $$ = $1; } ...
250
251 was incorrectly rejected: $1 is defined in the second mid-rule
252 action, and is equal to the $$ of the first mid-rule action.
253 \f
254 Changes in version 1.50, 2002-10-04:
255
256 * GLR parsing
257 The declaration
258 %glr-parser
259 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
260 almost any context-free grammar, ambiguous or not. The new declarations
261 %dprec and %merge on grammar rules allow parse-time resolution of
262 ambiguities. Contributed by Paul Hilfinger.
263
264 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
265 like the Alpha, so please stick to 32-bit hosts for now.
266
267 * Output Directory
268 When not in Yacc compatibility mode, when the output file was not
269 specified, running `bison foo/bar.y' created `foo/bar.c'. It
270 now creates `bar.c'.
271
272 * Undefined token
273 The undefined token was systematically mapped to 2 which prevented
274 the use of 2 by the user. This is no longer the case.
275
276 * Unknown token numbers
277 If yylex returned an out of range value, yyparse could die. This is
278 no longer the case.
279
280 * Error token
281 According to POSIX, the error token must be 256.
282 Bison extends this requirement by making it a preference: *if* the
283 user specified that one of her tokens is numbered 256, then error
284 will be mapped onto another number.
285
286 * Verbose error messages
287 They no longer report `..., expecting error or...' for states where
288 error recovery is possible.
289
290 * End token
291 Defaults to `$end' instead of `$'.
292
293 * Error recovery now conforms to documentation and to POSIX
294 When a Bison-generated parser encounters a syntax error, it now pops
295 the stack until it finds a state that allows shifting the error
296 token. Formerly, it popped the stack until it found a state that
297 allowed some non-error action other than a default reduction on the
298 error token. The new behavior has long been the documented behavior,
299 and has long been required by POSIX. For more details, please see
300 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
301 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
302
303 * Traces
304 Popped tokens and nonterminals are now reported.
305
306 * Larger grammars
307 Larger grammars are now supported (larger token numbers, larger grammar
308 size (= sum of the LHS and RHS lengths), larger LALR tables).
309 Formerly, many of these numbers ran afoul of 16-bit limits;
310 now these limits are 32 bits on most hosts.
311
312 * Explicit initial rule
313 Bison used to play hacks with the initial rule, which the user does
314 not write. It is now explicit, and visible in the reports and
315 graphs as rule 0.
316
317 * Useless rules
318 Before, Bison reported the useless rules, but, although not used,
319 included them in the parsers. They are now actually removed.
320
321 * Useless rules, useless nonterminals
322 They are now reported, as a warning, with their locations.
323
324 * Rules never reduced
325 Rules that can never be reduced because of conflicts are now
326 reported.
327
328 * Incorrect `Token not used'
329 On a grammar such as
330
331 %token useless useful
332 %%
333 exp: '0' %prec useful;
334
335 where a token was used to set the precedence of the last rule,
336 bison reported both `useful' and `useless' as useless tokens.
337
338 * Revert the C++ namespace changes introduced in 1.31
339 as they caused too many portability hassles.
340
341 * Default locations
342 By an accident of design, the default computation of @$ was
343 performed after another default computation was performed: @$ = @1.
344 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
345 the computation of @$.
346
347 * Token end-of-file
348 The token end of file may be specified by the user, in which case,
349 the user symbol is used in the reports, the graphs, and the verbose
350 error messages instead of `$end', which remains being the default.
351 For instance
352 %token MYEOF 0
353 or
354 %token MYEOF 0 "end of file"
355
356 * Semantic parser
357 This old option, which has been broken for ages, is removed.
358
359 * New translations
360 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
361 Croatian, thanks to Denis Lackovic.
362
363 * Incorrect token definitions
364 When given `%token 'a' "A"', Bison used to output `#define 'a' 65'.
365
366 * Token definitions as enums
367 Tokens are output both as the traditional #define's, and, provided
368 the compiler supports ANSI C or is a C++ compiler, as enums.
369 This lets debuggers display names instead of integers.
370
371 * Reports
372 In addition to --verbose, bison supports --report=THINGS, which
373 produces additional information:
374 - itemset
375 complete the core item sets with their closure
376 - lookahead [changed to `look-ahead' in 1.875e and later]
377 explicitly associate look-ahead tokens to items
378 - solved
379 describe shift/reduce conflicts solving.
380 Bison used to systematically output this information on top of
381 the report. Solved conflicts are now attached to their states.
382
383 * Type clashes
384 Previous versions don't complain when there is a type clash on
385 the default action if the rule has a mid-rule action, such as in:
386
387 %type <foo> bar
388 %%
389 bar: '0' {} '0';
390
391 This is fixed.
392
393 * GNU M4 is now required when using Bison.
394 \f
395 Changes in version 1.35, 2002-03-25:
396
397 * C Skeleton
398 Some projects use Bison's C parser with C++ compilers, and define
399 YYSTYPE as a class. The recent adjustment of C parsers for data
400 alignment and 64 bit architectures made this impossible.
401
402 Because for the time being no real solution for C++ parser
403 generation exists, kludges were implemented in the parser to
404 maintain this use. In the future, when Bison has C++ parsers, this
405 kludge will be disabled.
406
407 This kludge also addresses some C++ problems when the stack was
408 extended.
409 \f
410 Changes in version 1.34, 2002-03-12:
411
412 * File name clashes are detected
413 $ bison foo.y -d -o foo.x
414 fatal error: header and parser would both be named `foo.x'
415
416 * A missing `;' at the end of a rule triggers a warning
417 In accordance with POSIX, and in agreement with other
418 Yacc implementations, Bison will mandate this semicolon in the near
419 future. This eases the implementation of a Bison parser of Bison
420 grammars by making this grammar LALR(1) instead of LR(2). To
421 facilitate the transition, this release introduces a warning.
422
423 * Revert the C++ namespace changes introduced in 1.31, as they caused too
424 many portability hassles.
425
426 * DJGPP support added.
427
428 * Fix test suite portability problems.
429 \f
430 Changes in version 1.33, 2002-02-07:
431
432 * Fix C++ issues
433 Groff could not be compiled for the definition of size_t was lacking
434 under some conditions.
435
436 * Catch invalid @n
437 As is done with $n.
438 \f
439 Changes in version 1.32, 2002-01-23:
440
441 * Fix Yacc output file names
442
443 * Portability fixes
444
445 * Italian, Dutch translations
446 \f
447 Changes in version 1.31, 2002-01-14:
448
449 * Many Bug Fixes
450
451 * GNU Gettext and %expect
452 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
453 Bison dies on incorrect %expectations, we fear there will be
454 too many bug reports for Gettext, so _for the time being_, %expect
455 does not trigger an error when the input file is named `plural.y'.
456
457 * Use of alloca in parsers
458 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
459 malloc exclusively. Since 1.29, but was not NEWS'ed.
460
461 alloca is used only when compiled with GCC, to avoid portability
462 problems as on AIX.
463
464 * yyparse now returns 2 if memory is exhausted; formerly it dumped core.
465
466 * When the generated parser lacks debugging code, YYDEBUG is now 0
467 (as POSIX requires) instead of being undefined.
468
469 * User Actions
470 Bison has always permitted actions such as { $$ = $1 }: it adds the
471 ending semicolon. Now if in Yacc compatibility mode, the semicolon
472 is no longer output: one has to write { $$ = $1; }.
473
474 * Better C++ compliance
475 The output parsers try to respect C++ namespaces.
476 [This turned out to be a failed experiment, and it was reverted later.]
477
478 * Reduced Grammars
479 Fixed bugs when reporting useless nonterminals.
480
481 * 64 bit hosts
482 The parsers work properly on 64 bit hosts.
483
484 * Error messages
485 Some calls to strerror resulted in scrambled or missing error messages.
486
487 * %expect
488 When the number of shift/reduce conflicts is correct, don't issue
489 any warning.
490
491 * The verbose report includes the rule line numbers.
492
493 * Rule line numbers are fixed in traces.
494
495 * Swedish translation
496
497 * Parse errors
498 Verbose parse error messages from the parsers are better looking.
499 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
500 Now: parse error: unexpected '/', expecting "number" or '-' or '('
501
502 * Fixed parser memory leaks.
503 When the generated parser was using malloc to extend its stacks, the
504 previous allocations were not freed.
505
506 * Fixed verbose output file.
507 Some newlines were missing.
508 Some conflicts in state descriptions were missing.
509
510 * Fixed conflict report.
511 Option -v was needed to get the result.
512
513 * %expect
514 Was not used.
515 Mismatches are errors, not warnings.
516
517 * Fixed incorrect processing of some invalid input.
518
519 * Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
520
521 * Fixed some typos in the documentation.
522
523 * %token MY_EOF 0 is supported.
524 Before, MY_EOF was silently renumbered as 257.
525
526 * doc/refcard.tex is updated.
527
528 * %output, %file-prefix, %name-prefix.
529 New.
530
531 * --output
532 New, aliasing `--output-file'.
533 \f
534 Changes in version 1.30, 2001-10-26:
535
536 * `--defines' and `--graph' have now an optional argument which is the
537 output file name. `-d' and `-g' do not change; they do not take any
538 argument.
539
540 * `%source_extension' and `%header_extension' are removed, failed
541 experiment.
542
543 * Portability fixes.
544 \f
545 Changes in version 1.29, 2001-09-07:
546
547 * The output file does not define const, as this caused problems when used
548 with common autoconfiguration schemes. If you still use ancient compilers
549 that lack const, compile with the equivalent of the C compiler option
550 `-Dconst='. autoconf's AC_C_CONST macro provides one way to do this.
551
552 * Added `-g' and `--graph'.
553
554 * The Bison manual is now distributed under the terms of the GNU FDL.
555
556 * The input and the output files has automatically a similar extension.
557
558 * Russian translation added.
559
560 * NLS support updated; should hopefully be less troublesome.
561
562 * Added the old Bison reference card.
563
564 * Added `--locations' and `%locations'.
565
566 * Added `-S' and `--skeleton'.
567
568 * `%raw', `-r', `--raw' is disabled.
569
570 * Special characters are escaped when output. This solves the problems
571 of the #line lines with path names including backslashes.
572
573 * New directives.
574 `%yacc', `%fixed_output_files', `%defines', `%no_parser', `%verbose',
575 `%debug', `%source_extension' and `%header_extension'.
576
577 * @$
578 Automatic location tracking.
579 \f
580 Changes in version 1.28, 1999-07-06:
581
582 * Should compile better now with K&R compilers.
583
584 * Added NLS.
585
586 * Fixed a problem with escaping the double quote character.
587
588 * There is now a FAQ.
589 \f
590 Changes in version 1.27:
591
592 * The make rule which prevented bison.simple from being created on
593 some systems has been fixed.
594 \f
595 Changes in version 1.26:
596
597 * Bison now uses automake.
598
599 * New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
600
601 * Token numbers now start at 257 as previously documented, not 258.
602
603 * Bison honors the TMPDIR environment variable.
604
605 * A couple of buffer overruns have been fixed.
606
607 * Problems when closing files should now be reported.
608
609 * Generated parsers should now work even on operating systems which do
610 not provide alloca().
611 \f
612 Changes in version 1.25, 1995-10-16:
613
614 * Errors in the input grammar are not fatal; Bison keeps reading
615 the grammar file, and reports all the errors found in it.
616
617 * Tokens can now be specified as multiple-character strings: for
618 example, you could use "<=" for a token which looks like <=, instead
619 of chosing a name like LESSEQ.
620
621 * The %token_table declaration says to write a table of tokens (names
622 and numbers) into the parser file. The yylex function can use this
623 table to recognize multiple-character string tokens, or for other
624 purposes.
625
626 * The %no_lines declaration says not to generate any #line preprocessor
627 directives in the parser file.
628
629 * The %raw declaration says to use internal Bison token numbers, not
630 Yacc-compatible token numbers, when token names are defined as macros.
631
632 * The --no-parser option produces the parser tables without including
633 the parser engine; a project can now use its own parser engine.
634 The actions go into a separate file called NAME.act, in the form of
635 a switch statement body.
636 \f
637 Changes in version 1.23:
638
639 The user can define YYPARSE_PARAM as the name of an argument to be
640 passed into yyparse. The argument should have type void *. It should
641 actually point to an object. Grammar actions can access the variable
642 by casting it to the proper pointer type.
643
644 Line numbers in output file corrected.
645 \f
646 Changes in version 1.22:
647
648 --help option added.
649 \f
650 Changes in version 1.20:
651
652 Output file does not redefine const for C++.
653
654 Local Variables:
655 mode: outline
656 End:
657
658 -----
659
660 Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005
661 Free Software Foundation, Inc.
662
663 This file is part of Bison, the GNU Compiler Compiler.
664
665 Bison is free software; you can redistribute it and/or modify
666 it under the terms of the GNU General Public License as published by
667 the Free Software Foundation; either version 2, or (at your option)
668 any later version.
669
670 Bison is distributed in the hope that it will be useful,
671 but WITHOUT ANY WARRANTY; without even the implied warranty of
672 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
673 GNU General Public License for more details.
674
675 You should have received a copy of the GNU General Public License
676 along with autoconf; see the file COPYING. If not, write to
677 the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
678 Boston, MA 02110-1301, USA.