]> git.saurik.com Git - bison.git/blob - NEWS
(po-check): Scan .l and .y files instead
[bison.git] / NEWS
1 Bison News
2 ----------
3
4 Changes in version 1.75b:
5
6 * %destructor
7 It is now possible to reclaim the memory associated to symbols
8 discarded during error recovery. This feature is still experimental.
9
10 * Bison now parses C99 lexical constructs like digraphs, UCNs, and
11 backslash-newline within C escape sequences, as POSIX now requires.
12
13 * --no-line works properly.
14
15 * %error-verbose
16 This new directive is preferred over YYERROR_VERBOSE.
17
18 * %lex-param, %parse-param
19 These new directives are preferred over PARSE_PARAM and LEX_PARAM.
20 In addition, they provide a means for yyerror to remain pure, and
21 to access to the current location.
22
23 #defining yyerror to steal internal variables is strongly
24 discouraged. It is not guaranteed to work for ever.
25
26 * #line
27 Bison now recognizes #line in its input, and forwards them.
28
29 * #line
30 File names are properly escaped. E.g. foo\bar.y give #line 123 "foo\\bar.y".
31
32 Changes in version 1.75a, 2002-10-24:
33
34 * Bison can no longer be built by a K&R C compiler; it requires C89 or
35 later to be built. This change originally took place a few versions
36 ago, but nobody noticed until we recently asked someone to try
37 building Bison with a K&R C compiler.
38
39 Changes in version 1.75, 2002-10-14:
40
41 * Bison should now work on 64-bit hosts.
42
43 * Indonesian translation thanks to Tedi Heriyanto.
44
45 * GLR parsers
46 Fix spurious parse errors.
47
48 * Pure parsers
49 Some people redefine yyerror to steal yyparse' private variables.
50 Reenable this trick until an official feature replaces it.
51
52 * Type Clashes
53 In agreement with POSIX and with other Yaccs, leaving a default
54 action is valid when $$ is untyped, and $1 typed:
55
56 untyped: ... typed;
57
58 but the converse remains an error:
59
60 typed: ... untyped;
61
62 * Values of mid-rule actions
63 The following code:
64
65 foo: { ... } { $$ = $1; } ...
66
67 was incorrectly rejected: $1 is defined in the second mid-rule
68 action, and is equal to the $$ of the first mid-rule action.
69
70 Changes in version 1.50, 2002-10-04:
71
72 * GLR parsing
73 The declaration
74 %glr-parser
75 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
76 almost any context-free grammar, ambiguous or not. The new declarations
77 %dprec and %merge on grammar rules allow parse-time resolution of
78 ambiguities. Contributed by Paul Hilfinger.
79
80 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
81 like the Alpha, so please stick to 32-bit hosts for now.
82
83 * Output Directory
84 When not in Yacc compatibility mode, when the output file was not
85 specified, running `bison foo/bar.y' created `foo/bar.c'. It
86 now creates `bar.c'.
87
88 * Undefined token
89 The undefined token was systematically mapped to 2 which prevented
90 the use of 2 by the user. This is no longer the case.
91
92 * Unknown token numbers
93 If yylex returned an out of range value, yyparse could die. This is
94 no longer the case.
95
96 * Error token
97 According to POSIX, the error token must be 256.
98 Bison extends this requirement by making it a preference: *if* the
99 user specified that one of her tokens is numbered 256, then error
100 will be mapped onto another number.
101
102 * Verbose error messages
103 They no longer report `..., expecting error or...' for states where
104 error recovery is possible.
105
106 * End token
107 Defaults to `$end' instead of `$'.
108
109 * Error recovery now conforms to documentation and to POSIX
110 When a Bison-generated parser encounters a syntax error, it now pops
111 the stack until it finds a state that allows shifting the error
112 token. Formerly, it popped the stack until it found a state that
113 allowed some non-error action other than a default reduction on the
114 error token. The new behavior has long been the documented behavior,
115 and has long been required by POSIX. For more details, please see
116 <http://mail.gnu.org/pipermail/bug-bison/2002-May/001452.html>.
117
118 * Traces
119 Popped tokens and nonterminals are now reported.
120
121 * Larger grammars
122 Larger grammars are now supported (larger token numbers, larger grammar
123 size (= sum of the LHS and RHS lengths), larger LALR tables).
124 Formerly, many of these numbers ran afoul of 16-bit limits;
125 now these limits are 32 bits on most hosts.
126
127 * Explicit initial rule
128 Bison used to play hacks with the initial rule, which the user does
129 not write. It is now explicit, and visible in the reports and
130 graphs as rule 0.
131
132 * Useless rules
133 Before, Bison reported the useless rules, but, although not used,
134 included them in the parsers. They are now actually removed.
135
136 * Useless rules, useless nonterminals
137 They are now reported, as a warning, with their locations.
138
139 * Rules never reduced
140 Rules that can never be reduced because of conflicts are now
141 reported.
142
143 * Incorrect `Token not used'
144 On a grammar such as
145
146 %token useless useful
147 %%
148 exp: '0' %prec useful;
149
150 where a token was used to set the precedence of the last rule,
151 bison reported both `useful' and `useless' as useless tokens.
152
153 * Revert the C++ namespace changes introduced in 1.31
154 as they caused too many portability hassles.
155
156 * Default locations
157 By an accident of design, the default computation of @$ was
158 performed after another default computation was performed: @$ = @1.
159 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
160 the computation of @$.
161
162 * Token end-of-file
163 The token end of file may be specified by the user, in which case,
164 the user symbol is used in the reports, the graphs, and the verbose
165 error messages instead of `$end', which remains being the default.
166 For instance
167 %token MYEOF 0
168 or
169 %token MYEOF 0 "end of file"
170
171 * Semantic parser
172 This old option, which has been broken for ages, is removed.
173
174 * New translations
175 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
176 Croatian, thanks to Denis Lackovic.
177
178 * Incorrect token definitions
179 When given `%token 'a' "A"', Bison used to output `#define 'a' 65'.
180
181 * Token definitions as enums
182 Tokens are output both as the traditional #define's, and, provided
183 the compiler supports ANSI C or is a C++ compiler, as enums.
184 This lets debuggers display names instead of integers.
185
186 * Reports
187 In addition to --verbose, bison supports --report=THINGS, which
188 produces additional information:
189 - itemset
190 complete the core item sets with their closure
191 - lookahead
192 explicitly associate lookaheads to items
193 - solved
194 describe shift/reduce conflicts solving.
195 Bison used to systematically output this information on top of
196 the report. Solved conflicts are now attached to their states.
197
198 * Type clashes
199 Previous versions don't complain when there is a type clash on
200 the default action if the rule has a mid-rule action, such as in:
201
202 %type <foo> bar
203 %%
204 bar: '0' {} '0';
205
206 This is fixed.
207
208 * GNU M4 is now required when using Bison.
209 \f
210 Changes in version 1.35, 2002-03-25:
211
212 * C Skeleton
213 Some projects use Bison's C parser with C++ compilers, and define
214 YYSTYPE as a class. The recent adjustment of C parsers for data
215 alignment and 64 bit architectures made this impossible.
216
217 Because for the time being no real solution for C++ parser
218 generation exists, kludges were implemented in the parser to
219 maintain this use. In the future, when Bison has C++ parsers, this
220 kludge will be disabled.
221
222 This kludge also addresses some C++ problems when the stack was
223 extended.
224 \f
225 Changes in version 1.34, 2002-03-12:
226
227 * File name clashes are detected
228 $ bison foo.y -d -o foo.x
229 fatal error: header and parser would both be named `foo.x'
230
231 * A missing `;' at the end of a rule triggers a warning
232 In accordance with POSIX, and in agreement with other
233 Yacc implementations, Bison will mandate this semicolon in the near
234 future. This eases the implementation of a Bison parser of Bison
235 grammars by making this grammar LALR(1) instead of LR(2). To
236 facilitate the transition, this release introduces a warning.
237
238 * Revert the C++ namespace changes introduced in 1.31, as they caused too
239 many portability hassles.
240
241 * DJGPP support added.
242
243 * Fix test suite portability problems.
244 \f
245 Changes in version 1.33, 2002-02-07:
246
247 * Fix C++ issues
248 Groff could not be compiled for the definition of size_t was lacking
249 under some conditions.
250
251 * Catch invalid @n
252 As is done with $n.
253 \f
254 Changes in version 1.32, 2002-01-23:
255
256 * Fix Yacc output file names
257
258 * Portability fixes
259
260 * Italian, Dutch translations
261 \f
262 Changes in version 1.31, 2002-01-14:
263
264 * Many Bug Fixes
265
266 * GNU Gettext and %expect
267 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
268 Bison dies on incorrect %expectations, we fear there will be
269 too many bug reports for Gettext, so _for the time being_, %expect
270 does not trigger an error when the input file is named `plural.y'.
271
272 * Use of alloca in parsers
273 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
274 malloc exclusively. Since 1.29, but was not NEWS'ed.
275
276 alloca is used only when compiled with GCC, to avoid portability
277 problems as on AIX.
278
279 * When the generated parser lacks debugging code, YYDEBUG is now 0
280 (as POSIX requires) instead of being undefined.
281
282 * User Actions
283 Bison has always permitted actions such as { $$ = $1 }: it adds the
284 ending semicolon. Now if in Yacc compatibility mode, the semicolon
285 is no longer output: one has to write { $$ = $1; }.
286
287 * Better C++ compliance
288 The output parsers try to respect C++ namespaces.
289 [This turned out to be a failed experiment, and it was reverted later.]
290
291 * Reduced Grammars
292 Fixed bugs when reporting useless nonterminals.
293
294 * 64 bit hosts
295 The parsers work properly on 64 bit hosts.
296
297 * Error messages
298 Some calls to strerror resulted in scrambled or missing error messages.
299
300 * %expect
301 When the number of shift/reduce conflicts is correct, don't issue
302 any warning.
303
304 * The verbose report includes the rule line numbers.
305
306 * Rule line numbers are fixed in traces.
307
308 * Swedish translation
309
310 * Parse errors
311 Verbose parse error messages from the parsers are better looking.
312 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
313 Now: parse error: unexpected '/', expecting "number" or '-' or '('
314
315 * Fixed parser memory leaks.
316 When the generated parser was using malloc to extend its stacks, the
317 previous allocations were not freed.
318
319 * Fixed verbose output file.
320 Some newlines were missing.
321 Some conflicts in state descriptions were missing.
322
323 * Fixed conflict report.
324 Option -v was needed to get the result.
325
326 * %expect
327 Was not used.
328 Mismatches are errors, not warnings.
329
330 * Fixed incorrect processing of some invalid input.
331
332 * Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
333
334 * Fixed some typos in the documentation.
335
336 * %token MY_EOF 0 is supported.
337 Before, MY_EOF was silently renumbered as 257.
338
339 * doc/refcard.tex is updated.
340
341 * %output, %file-prefix, %name-prefix.
342 New.
343
344 * --output
345 New, aliasing `--output-file'.
346 \f
347 Changes in version 1.30, 2001-10-26:
348
349 * `--defines' and `--graph' have now an optional argument which is the
350 output file name. `-d' and `-g' do not change; they do not take any
351 argument.
352
353 * `%source_extension' and `%header_extension' are removed, failed
354 experiment.
355
356 * Portability fixes.
357 \f
358 Changes in version 1.29, 2001-09-07:
359
360 * The output file does not define const, as this caused problems when used
361 with common autoconfiguration schemes. If you still use ancient compilers
362 that lack const, compile with the equivalent of the C compiler option
363 `-Dconst='. autoconf's AC_C_CONST macro provides one way to do this.
364
365 * Added `-g' and `--graph'.
366
367 * The Bison manual is now distributed under the terms of the GNU FDL.
368
369 * The input and the output files has automatically a similar extension.
370
371 * Russian translation added.
372
373 * NLS support updated; should hopefully be less troublesome.
374
375 * Added the old Bison reference card.
376
377 * Added `--locations' and `%locations'.
378
379 * Added `-S' and `--skeleton'.
380
381 * `%raw', `-r', `--raw' is disabled.
382
383 * Special characters are escaped when output. This solves the problems
384 of the #line lines with path names including backslashes.
385
386 * New directives.
387 `%yacc', `%fixed_output_files', `%defines', `%no_parser', `%verbose',
388 `%debug', `%source_extension' and `%header_extension'.
389
390 * @$
391 Automatic location tracking.
392 \f
393 Changes in version 1.28, 1999-07-06:
394
395 * Should compile better now with K&R compilers.
396
397 * Added NLS.
398
399 * Fixed a problem with escaping the double quote character.
400
401 * There is now a FAQ.
402 \f
403 Changes in version 1.27:
404
405 * The make rule which prevented bison.simple from being created on
406 some systems has been fixed.
407 \f
408 Changes in version 1.26:
409
410 * Bison now uses automake.
411
412 * New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
413
414 * Token numbers now start at 257 as previously documented, not 258.
415
416 * Bison honors the TMPDIR environment variable.
417
418 * A couple of buffer overruns have been fixed.
419
420 * Problems when closing files should now be reported.
421
422 * Generated parsers should now work even on operating systems which do
423 not provide alloca().
424 \f
425 Changes in version 1.25, 1995-10-16:
426
427 * Errors in the input grammar are not fatal; Bison keeps reading
428 the grammar file, and reports all the errors found in it.
429
430 * Tokens can now be specified as multiple-character strings: for
431 example, you could use "<=" for a token which looks like <=, instead
432 of chosing a name like LESSEQ.
433
434 * The %token_table declaration says to write a table of tokens (names
435 and numbers) into the parser file. The yylex function can use this
436 table to recognize multiple-character string tokens, or for other
437 purposes.
438
439 * The %no_lines declaration says not to generate any #line preprocessor
440 directives in the parser file.
441
442 * The %raw declaration says to use internal Bison token numbers, not
443 Yacc-compatible token numbers, when token names are defined as macros.
444
445 * The --no-parser option produces the parser tables without including
446 the parser engine; a project can now use its own parser engine.
447 The actions go into a separate file called NAME.act, in the form of
448 a switch statement body.
449 \f
450 Changes in version 1.23:
451
452 The user can define YYPARSE_PARAM as the name of an argument to be
453 passed into yyparse. The argument should have type void *. It should
454 actually point to an object. Grammar actions can access the variable
455 by casting it to the proper pointer type.
456
457 Line numbers in output file corrected.
458 \f
459 Changes in version 1.22:
460
461 --help option added.
462 \f
463 Changes in version 1.20:
464
465 Output file does not redefine const for C++.
466
467 Local Variables:
468 mode: outline
469 End:
470
471 -----
472
473 Copyright (C) 2001, 2002 Free Software Foundation, Inc.
474
475 This file is part of GNU Autoconf.
476
477 GNU Autoconf is free software; you can redistribute it and/or modify
478 it under the terms of the GNU General Public License as published by
479 the Free Software Foundation; either version 2, or (at your option)
480 any later version.
481
482 GNU Autoconf is distributed in the hope that it will be useful,
483 but WITHOUT ANY WARRANTY; without even the implied warranty of
484 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
485 GNU General Public License for more details.
486
487 You should have received a copy of the GNU General Public License
488 along with autoconf; see the file COPYING. If not, write to
489 the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
490 Boston, MA 02111-1307, USA.