]> git.saurik.com Git - bison.git/blame_incremental - NEWS
Normalize conflict related messages.
[bison.git] / NEWS
... / ...
CommitLineData
1Bison News
2----------
3
4Changes in version 1.49b:
5
6* GLR parsing
7 The declaration
8 %glr-parser
9 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
10 almost any context-free grammar, ambiguous or not. The new declarations
11 %dprec and %merge on grammar rules allow parse-time resolution of
12 ambiguities. Contributed by Paul Hilfinger.
13
14* Output Directory
15 When not in Yacc compatibility mode, when the output file was not
16 specified, runnning `bison foo/bar.y' created `foo/bar.c'. It
17 now creates `bar.c'.
18
19* Undefined token
20 The undefined token was systematically mapped to 2 which prevented
21 the use of 2 from the user. This is no longer the case.
22
23* Unknown token numbers
24 If yylex returned a code out of range, yyparse could die. This is
25 no longer the case.
26
27* Error token
28 According to POSIX, the error token should be numbered as 256.
29 Bison extends this requirement by making it a preference: *if* the
30 user specified that one of her tokens is numbered 256, then error
31 will be mapped onto another number.
32
33* Error recovery now conforms to documentation and to POSIX
34 When a Bison-generated parser encounters a syntax error, it now pops
35 the stack until it finds a state that allows shifting the error
36 token. Formerly, it popped the stack until it found a state that
37 allowed some non-error action other than a default reduction on the
38 error token. The new behavior has long been the documented behavior,
39 and has long been required by POSIX. For more details, please see
40 <http://mail.gnu.org/pipermail/bug-bison/2002-May/001452.html>.
41
42* Traces
43 Popped tokens and nonterminals are now reported.
44
45* Large grammars
46 Large grammars are now supported (large token numbers, large grammar
47 size (= sum of the LHS and RHS lengths), large LALR tables).
48
49* Explicit initial rule
50 Bison used to play hacks with the initial rule, which the user does
51 not write. It is now explicit, and visible in the reports and
52 graphs as rule 0.
53
54* Useless rules
55 Before, Bison reported the useless rules, but, although not used,
56 included them in the parsers. They are now actually removed.
57
58* Useless rules, useless nonterminals
59 They are now reported, as a warning, with their locations.
60
61* Rules never reduced
62 Rules that can never be reduced because of conflicts are now
63 reported.
64
65* Incorrect `Token not used'
66 On a grammar such as
67
68 %token useless useful
69 %%
70 exp: '0' %prec useful;
71
72 where a token was used to set the precedence of the last rule,
73 bison reported both `useful' and `useless' as useless tokens.
74
75* Revert the C++ namespace changes introduced in 1.31
76 as they caused too many portability hassles.
77
78* Default locations
79 By an accident of design, the default computation of @$ was
80 performed after another default computation was performed: @$ = @1.
81 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
82 the computation of @$.
83
84* Token end-of-file
85 The token end of file may be specified by the user, in which case,
86 the user symbol is used in the reports, the graphs, and the verbose
87 error messages instead of `$', which remains being the defaults.
88 For instance
89 %token YYEOF 0
90 or
91 %token YYEOF 0 "end of file"
92
93* Semantic parser
94 This old option, which has been broken for ages, is removed.
95
96* New translation
97 Croatian, thanks to Denis Lackovic.
98
99* Incorrect token definitions
100 When fed with `%token 'a' "A"', Bison used to output `#define 'a' 65'.
101
102* Token definitions as enums
103 Tokens are output both as the traditional #define's, and, provided
104 the compiler supports ANSI C or is a C++ compiler, as enums.
105 This helps debuggers producing symbols instead of values.
106
107* Reports
108 In addition to --verbose, bison supports --report=THINGS, which
109 produces additional information:
110 - itemset
111 complete the core item sets with their closure
112 - lookahead
113 explicitly associate lookaheads to items
114 - solved
115 describe shift/reduce conflicts solving.
116 Bison used to systematically output this information on top of
117 the report. Solved conflicts are now attached to their states.
118
119* Type clashes
120 Previous versions don't complain when there is a type clash on
121 the default action if the rule has a mid-rule action, such as in:
122
123 %type <foo> bar
124 %%
125 bar: '0' {} '0';
126
127 This is fixed.
128\f
129Changes in version 1.35, 2002-03-25:
130
131* C Skeleton
132 Some projects use Bison's C parser with C++ compilers, and define
133 YYSTYPE as a class. The recent adjustment of C parsers for data
134 alignment and 64 bit architectures made this impossible.
135
136 Because for the time being no real solution for C++ parser
137 generation exists, kludges were implemented in the parser to
138 maintain this use. In the future, when Bison has C++ parsers, this
139 kludge will be disabled.
140
141 This kludge also addresses some C++ problems when the stack was
142 extended.
143
144\f
145Changes in version 1.34, 2002-03-12:
146
147* File name clashes are detected
148 $ bison foo.y -d -o foo.x
149 fatal error: header and parser would both be named `foo.x'
150
151* A missing `;' at the end of a rule triggers a warning
152 In accordance with POSIX, and in agreement with other
153 Yacc implementations, Bison will mandate this semicolon in the near
154 future. This eases the implementation of a Bison parser of Bison
155 grammars by making this grammar LALR(1) instead of LR(2). To
156 facilitate the transition, this release introduces a warning.
157
158* Revert the C++ namespace changes introduced in 1.31, as they caused too
159 many portability hassles.
160
161* DJGPP support added.
162
163* Fix test suite portability problems.
164\f
165Changes in version 1.33, 2002-02-07:
166
167* Fix C++ issues
168 Groff could not be compiled for the definition of size_t was lacking
169 under some conditions.
170
171* Catch invalid @n
172 As is done with $n.
173\f
174Changes in version 1.32, 2002-01-23:
175
176* Fix Yacc output file names
177
178* Portability fixes
179
180* Italian, Dutch translations
181\f
182Changes in version 1.31, 2002-01-14:
183
184* Many Bug Fixes
185
186* GNU Gettext and %expect
187 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
188 Bison dies on incorrect %expectations, we fear there will be
189 too many bug reports for Gettext, so _for the time being_, %expect
190 does not trigger an error when the input file is named `plural.y'.
191
192* Use of alloca in parsers
193 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
194 malloc exclusively. Since 1.29, but was not NEWS'ed.
195
196 alloca is used only when compiled with GCC, to avoid portability
197 problems as on AIX.
198
199* When the generated parser lacks debugging code, YYDEBUG is now 0
200 (as POSIX requires) instead of being undefined.
201
202* User Actions
203 Bison has always permitted actions such as { $$ = $1 }: it adds the
204 ending semicolon. Now if in Yacc compatibility mode, the semicolon
205 is no longer output: one has to write { $$ = $1; }.
206
207* Better C++ compliance
208 The output parsers try to respect C++ namespaces.
209 [This turned out to be a failed experiment, and it was reverted later.]
210
211* Reduced Grammars
212 Fixed bugs when reporting useless nonterminals.
213
214* 64 bit hosts
215 The parsers work properly on 64 bit hosts.
216
217* Error messages
218 Some calls to strerror resulted in scrambled or missing error messages.
219
220* %expect
221 When the number of shift/reduce conflicts is correct, don't issue
222 any warning.
223
224* The verbose report includes the rule line numbers.
225
226* Rule line numbers are fixed in traces.
227
228* Swedish translation
229
230* Parse errors
231 Verbose parse error messages from the parsers are better looking.
232 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
233 Now: parse error: unexpected '/', expecting "number" or '-' or '('
234
235* Fixed parser memory leaks.
236 When the generated parser was using malloc to extend its stacks, the
237 previous allocations were not freed.
238
239* Fixed verbose output file.
240 Some newlines were missing.
241 Some conflicts in state descriptions were missing.
242
243* Fixed conflict report.
244 Option -v was needed to get the result.
245
246* %expect
247 Was not used.
248 Mismatches are errors, not warnings.
249
250* Fixed incorrect processing of some invalid input.
251
252* Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
253
254* Fixed some typos in the documentation.
255
256* %token MY_EOF 0 is supported.
257 Before, MY_EOF was silently renumbered as 257.
258
259* doc/refcard.tex is updated.
260
261* %output, %file-prefix, %name-prefix.
262 New.
263
264* --output
265 New, aliasing `--output-file'.
266\f
267Changes in version 1.30, 2001-10-26:
268
269* `--defines' and `--graph' have now an optionnal argument which is the
270 output file name. `-d' and `-g' do not change, they do not take any
271 argument.
272
273* `%source_extension' and `%header_extension' are removed, failed
274 experiment.
275
276* Portability fixes.
277\f
278Changes in version 1.29, 2001-09-07:
279
280* The output file does not define const, as this caused problems when used
281 with common autoconfiguration schemes. If you still use ancient compilers
282 that lack const, compile with the equivalent of the C compiler option
283 `-Dconst='. autoconf's AC_C_CONST macro provides one way to do this.
284
285* Added `-g' and `--graph'.
286
287* The Bison manual is now distributed under the terms of the GNU FDL.
288
289* The input and the output files has automatically a similar extension.
290
291* Russian translation added.
292
293* NLS support updated; should hopefully be less troublesome.
294
295* Added the old Bison reference card.
296
297* Added `--locations' and `%locations'.
298
299* Added `-S' and `--skeleton'.
300
301* `%raw', `-r', `--raw' is disabled.
302
303* Special characters are escaped when output. This solves the problems
304 of the #line lines with path names including backslashes.
305
306* New directives.
307 `%yacc', `%fixed_output_files', `%defines', `%no_parser', `%verbose',
308 `%debug', `%source_extension' and `%header_extension'.
309
310* @$
311 Automatic location tracking.
312\f
313Changes in version 1.28, 1999-07-06:
314
315* Should compile better now with K&R compilers.
316
317* Added NLS.
318
319* Fixed a problem with escaping the double quote character.
320
321* There is now a FAQ.
322\f
323Changes in version 1.27:
324
325* The make rule which prevented bison.simple from being created on
326 some systems has been fixed.
327\f
328Changes in version 1.26:
329
330* Bison now uses automake.
331
332* New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
333
334* Token numbers now start at 257 as previously documented, not 258.
335
336* Bison honors the TMPDIR environment variable.
337
338* A couple of buffer overruns have been fixed.
339
340* Problems when closing files should now be reported.
341
342* Generated parsers should now work even on operating systems which do
343 not provide alloca().
344\f
345Changes in version 1.25, 1995-10-16:
346
347* Errors in the input grammar are not fatal; Bison keeps reading
348the grammar file, and reports all the errors found in it.
349
350* Tokens can now be specified as multiple-character strings: for
351example, you could use "<=" for a token which looks like <=, instead
352of chosing a name like LESSEQ.
353
354* The %token_table declaration says to write a table of tokens (names
355and numbers) into the parser file. The yylex function can use this
356table to recognize multiple-character string tokens, or for other
357purposes.
358
359* The %no_lines declaration says not to generate any #line preprocessor
360directives in the parser file.
361
362* The %raw declaration says to use internal Bison token numbers, not
363Yacc-compatible token numbers, when token names are defined as macros.
364
365* The --no-parser option produces the parser tables without including
366the parser engine; a project can now use its own parser engine.
367The actions go into a separate file called NAME.act, in the form of
368a switch statement body.
369\f
370Changes in version 1.23:
371
372The user can define YYPARSE_PARAM as the name of an argument to be
373passed into yyparse. The argument should have type void *. It should
374actually point to an object. Grammar actions can access the variable
375by casting it to the proper pointer type.
376
377Line numbers in output file corrected.
378\f
379Changes in version 1.22:
380
381--help option added.
382\f
383Changes in version 1.20:
384
385Output file does not redefine const for C++.
386
387Local Variables:
388mode: outline
389End:
390
391-----
392
393Copyright (C) 2001, 2002 Free Software Foundation, Inc.
394
395This file is part of GNU Autoconf.
396
397GNU Autoconf is free software; you can redistribute it and/or modify
398it under the terms of the GNU General Public License as published by
399the Free Software Foundation; either version 2, or (at your option)
400any later version.
401
402GNU Autoconf is distributed in the hope that it will be useful,
403but WITHOUT ANY WARRANTY; without even the implied warranty of
404MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
405GNU General Public License for more details.
406
407You should have received a copy of the GNU General Public License
408along with autoconf; see the file COPYING. If not, write to
409the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
410Boston, MA 02111-1307, USA.