]>
Commit | Line | Data |
---|---|---|
1 | GNU Bison NEWS | |
2 | ||
3 | * Noteworthy changes in release ?.? (????-??-??) [?] | |
4 | ||
5 | ** Bug fixes | |
6 | ||
7 | *** Fix compiler attribute portability (yacc.c) | |
8 | ||
9 | With locations enabled, __attribute__ was used unprotected. | |
10 | ||
11 | *** Fix some compiler warnings (lalr1.cc) | |
12 | ||
13 | * Noteworthy changes in release 2.7 (2012-12-12) [stable] | |
14 | ||
15 | ** Bug fixes | |
16 | ||
17 | Warnings about uninitialized yylloc in yyparse have been fixed. | |
18 | ||
19 | Restored C90 compliance (yet no report was ever made). | |
20 | ||
21 | ** Diagnostics are improved | |
22 | ||
23 | *** Changes in the format of error messages | |
24 | ||
25 | This used to be the format of many error reports: | |
26 | ||
27 | input.y:2.7-12: %type redeclaration for exp | |
28 | input.y:1.7-12: previous declaration | |
29 | ||
30 | It is now: | |
31 | ||
32 | input.y:2.7-12: error: %type redeclaration for exp | |
33 | input.y:1.7-12: previous declaration | |
34 | ||
35 | *** New format for error reports: carets | |
36 | ||
37 | Caret errors have been added to Bison: | |
38 | ||
39 | input.y:2.7-12: error: %type redeclaration for exp | |
40 | %type <sval> exp | |
41 | ^^^^^^ | |
42 | input.y:1.7-12: previous declaration | |
43 | %type <ival> exp | |
44 | ^^^^^^ | |
45 | ||
46 | or | |
47 | ||
48 | input.y:3.20-23: error: ambiguous reference: '$exp' | |
49 | exp: exp '+' exp { $exp = $1 + $3; }; | |
50 | ^^^^ | |
51 | input.y:3.1-3: refers to: $exp at $$ | |
52 | exp: exp '+' exp { $exp = $1 + $3; }; | |
53 | ^^^ | |
54 | input.y:3.6-8: refers to: $exp at $1 | |
55 | exp: exp '+' exp { $exp = $1 + $3; }; | |
56 | ^^^ | |
57 | input.y:3.14-16: refers to: $exp at $3 | |
58 | exp: exp '+' exp { $exp = $1 + $3; }; | |
59 | ^^^ | |
60 | ||
61 | The default behaviour for now is still not to display these unless | |
62 | explictly asked with -fcaret (or -fall). However, in a later release, it | |
63 | will be made the default behavior (but may still be deactivated with | |
64 | -fno-caret). | |
65 | ||
66 | ** New value for %define variable: api.pure full | |
67 | ||
68 | The %define variable api.pure requests a pure (reentrant) parser. However, | |
69 | for historical reasons, using it in a location-tracking Yacc parser | |
70 | resulted in a yyerror function that did not take a location as a | |
71 | parameter. With this new value, the user may request a better pure parser, | |
72 | where yyerror does take a location as a parameter (in location-tracking | |
73 | parsers). | |
74 | ||
75 | The use of "%define api.pure true" is deprecated in favor of this new | |
76 | "%define api.pure full". | |
77 | ||
78 | ** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java) | |
79 | ||
80 | The %define variable api.location.type defines the name of the type to use | |
81 | for locations. When defined, Bison no longer generates the position.hh | |
82 | and location.hh files, nor does the parser will include them: the user is | |
83 | then responsible to define her type. | |
84 | ||
85 | This can be used in programs with several parsers to factor their location | |
86 | and position files: let one of them generate them, and the others just use | |
87 | them. | |
88 | ||
89 | This feature was actually introduced, but not documented, in Bison 2.5, | |
90 | under the name "location_type" (which is maintained for backward | |
91 | compatibility). | |
92 | ||
93 | For consistency, lalr1.java's %define variables location_type and | |
94 | position_type are deprecated in favor of api.location.type and | |
95 | api.position.type. | |
96 | ||
97 | ** Exception safety (lalr1.cc) | |
98 | ||
99 | The parse function now catches exceptions, uses the %destructors to | |
100 | release memory (the lookahead symbol and the symbols pushed on the stack) | |
101 | before re-throwing the exception. | |
102 | ||
103 | This feature is somewhat experimental. User feedback would be | |
104 | appreciated. | |
105 | ||
106 | ** Graph improvements in DOT and XSLT | |
107 | ||
108 | The graphical presentation of the states is more readable: their shape is | |
109 | now rectangular, the state number is clearly displayed, and the items are | |
110 | numbered and left-justified. | |
111 | ||
112 | The reductions are now explicitly represented as transitions to other | |
113 | diamond shaped nodes. | |
114 | ||
115 | These changes are present in both --graph output and xml2dot.xsl XSLT | |
116 | processing, with minor (documented) differences. | |
117 | ||
118 | ** %language is no longer an experimental feature. | |
119 | ||
120 | The introduction of this feature, in 2.4, was four years ago. The | |
121 | --language option and the %language directive are no longer experimental. | |
122 | ||
123 | ** Documentation | |
124 | ||
125 | The sections about shift/reduce and reduce/reduce conflicts resolution | |
126 | have been fixed and extended. | |
127 | ||
128 | Although introduced more than four years ago, XML and Graphviz reports | |
129 | were not properly documented. | |
130 | ||
131 | The translation of mid-rule actions is now described. | |
132 | ||
133 | * Noteworthy changes in release 2.6.5 (2012-11-07) [stable] | |
134 | ||
135 | We consider compiler warnings about Bison generated parsers to be bugs. | |
136 | Rather than working around them in your own project, please consider | |
137 | reporting them to us. | |
138 | ||
139 | ** Bug fixes | |
140 | ||
141 | Warnings about uninitialized yylval and/or yylloc for push parsers with a | |
142 | pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to | |
143 | 3.2. | |
144 | ||
145 | Other issues in the test suite have been addressed. | |
146 | ||
147 | Nul characters are correctly displayed in error messages. | |
148 | ||
149 | When possible, yylloc is correctly initialized before calling yylex. It | |
150 | is no longer necessary to initialize it in the %initial-action. | |
151 | ||
152 | * Noteworthy changes in release 2.6.4 (2012-10-23) [stable] | |
153 | ||
154 | Bison 2.6.3's --version was incorrect. This release fixes this issue. | |
155 | ||
156 | * Noteworthy changes in release 2.6.3 (2012-10-22) [stable] | |
157 | ||
158 | ** Bug fixes | |
159 | ||
160 | Bugs and portability issues in the test suite have been fixed. | |
161 | ||
162 | Some errors in translations have been addressed, and --help now directs | |
163 | users to the appropriate place to report them. | |
164 | ||
165 | Stray Info files shipped by accident are removed. | |
166 | ||
167 | Incorrect definitions of YY_, issued by yacc.c when no parser header is | |
168 | generated, are removed. | |
169 | ||
170 | All the generated headers are self-contained. | |
171 | ||
172 | ** Header guards (yacc.c, glr.c, glr.cc) | |
173 | ||
174 | In order to avoid collisions, the header guards are now | |
175 | YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>. | |
176 | For instance the header generated from | |
177 | ||
178 | %define api.prefix "calc" | |
179 | %defines "lib/parse.h" | |
180 | ||
181 | will use YY_CALC_LIB_PARSE_H_INCLUDED as guard. | |
182 | ||
183 | ** Fix compiler warnings in the generated parser (yacc.c, glr.c) | |
184 | ||
185 | The compilation of pure parsers (%define api.pure) can trigger GCC | |
186 | warnings such as: | |
187 | ||
188 | input.c: In function 'yyparse': | |
189 | input.c:1503:12: warning: 'yylval' may be used uninitialized in this | |
190 | function [-Wmaybe-uninitialized] | |
191 | *++yyvsp = yylval; | |
192 | ^ | |
193 | ||
194 | This is now fixed; pragmas to avoid these warnings are no longer needed. | |
195 | ||
196 | Warnings from clang ("equality comparison with extraneous parentheses" and | |
197 | "function declared 'noreturn' should not return") have also been | |
198 | addressed. | |
199 | ||
200 | * Noteworthy changes in release 2.6.2 (2012-08-03) [stable] | |
201 | ||
202 | ** Bug fixes | |
203 | ||
204 | Buffer overruns, complaints from Flex, and portability issues in the test | |
205 | suite have been fixed. | |
206 | ||
207 | ** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc) | |
208 | ||
209 | Trailing end-of-lines in %parse-param or %lex-param would result in | |
210 | invalid C++. This is fixed. | |
211 | ||
212 | ** Spurious spaces and end-of-lines | |
213 | ||
214 | The generated files no longer end (nor start) with empty lines. | |
215 | ||
216 | * Noteworthy changes in release 2.6.1 (2012-07-30) [stable] | |
217 | ||
218 | Bison no longer executes user-specified M4 code when processing a grammar. | |
219 | ||
220 | ** Future Changes | |
221 | ||
222 | In addition to the removal of the features announced in Bison 2.6, the | |
223 | next major release will remove the "Temporary hack for adding a semicolon | |
224 | to the user action", as announced in the release 2.5. Instead of: | |
225 | ||
226 | exp: exp "+" exp { $$ = $1 + $3 }; | |
227 | ||
228 | write: | |
229 | ||
230 | exp: exp "+" exp { $$ = $1 + $3; }; | |
231 | ||
232 | ** Bug fixes | |
233 | ||
234 | *** Type names are now properly escaped. | |
235 | ||
236 | *** glr.cc: set_debug_level and debug_level work as expected. | |
237 | ||
238 | *** Stray @ or $ in actions | |
239 | ||
240 | While Bison used to warn about stray $ or @ in action rules, it did not | |
241 | for other actions such as printers, destructors, or initial actions. It | |
242 | now does. | |
243 | ||
244 | ** Type names in actions | |
245 | ||
246 | For consistency with rule actions, it is now possible to qualify $$ by a | |
247 | type-name in destructors, printers, and initial actions. For instance: | |
248 | ||
249 | %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>; | |
250 | ||
251 | will display two values for each typed and untyped symbol (provided | |
252 | that YYSTYPE has both "ival" and "fval" fields). | |
253 | ||
254 | * Noteworthy changes in release 2.6 (2012-07-19) [stable] | |
255 | ||
256 | ** Future Changes | |
257 | ||
258 | The next major release of Bison will drop support for the following | |
259 | deprecated features. Please report disagreements to bug-bison@gnu.org. | |
260 | ||
261 | *** K&R C parsers | |
262 | ||
263 | Support for generating parsers in K&R C will be removed. Parsers | |
264 | generated for C support ISO C90, and are tested with ISO C99 and ISO C11 | |
265 | compilers. | |
266 | ||
267 | *** Features deprecated since Bison 1.875 | |
268 | ||
269 | The definitions of yystype and yyltype will be removed; use YYSTYPE and | |
270 | YYLTYPE. | |
271 | ||
272 | YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and | |
273 | %lex-param, will no longer be supported. | |
274 | ||
275 | Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use | |
276 | %error-verbose. | |
277 | ||
278 | *** The generated header will be included (yacc.c) | |
279 | ||
280 | Instead of duplicating the content of the generated header (definition of | |
281 | YYSTYPE, yyparse declaration etc.), the generated parser will include it, | |
282 | as is already the case for GLR or C++ parsers. This change is deferred | |
283 | because existing versions of ylwrap (e.g., Automake 1.12.1) do not support | |
284 | it. | |
285 | ||
286 | ** Generated Parser Headers | |
287 | ||
288 | *** Guards (yacc.c, glr.c, glr.cc) | |
289 | ||
290 | The generated headers are now guarded, as is already the case for C++ | |
291 | parsers (lalr1.cc). For instance, with --defines=foo.h: | |
292 | ||
293 | #ifndef YY_FOO_H | |
294 | # define YY_FOO_H | |
295 | ... | |
296 | #endif /* !YY_FOO_H */ | |
297 | ||
298 | *** New declarations (yacc.c, glr.c) | |
299 | ||
300 | The generated header now declares yydebug and yyparse. Both honor | |
301 | --name-prefix=bar_, and yield | |
302 | ||
303 | int bar_parse (void); | |
304 | ||
305 | rather than | |
306 | ||
307 | #define yyparse bar_parse | |
308 | int yyparse (void); | |
309 | ||
310 | in order to facilitate the inclusion of several parser headers inside a | |
311 | single compilation unit. | |
312 | ||
313 | *** Exported symbols in C++ | |
314 | ||
315 | The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the | |
316 | header, are removed, as they prevent the possibility of including several | |
317 | generated headers from a single compilation unit. | |
318 | ||
319 | *** YYLSP_NEEDED | |
320 | ||
321 | For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no | |
322 | longer defined. | |
323 | ||
324 | ** New %define variable: api.prefix | |
325 | ||
326 | Now that the generated headers are more complete and properly protected | |
327 | against multiple inclusions, constant names, such as YYSTYPE are a | |
328 | problem. While yyparse and others are properly renamed by %name-prefix, | |
329 | YYSTYPE, YYDEBUG and others have never been affected by it. Because it | |
330 | would introduce backward compatibility issues in projects not expecting | |
331 | YYSTYPE to be renamed, instead of changing the behavior of %name-prefix, | |
332 | it is deprecated in favor of a new %define variable: api.prefix. | |
333 | ||
334 | The following examples compares both: | |
335 | ||
336 | %name-prefix "bar_" | %define api.prefix "bar_" | |
337 | %token <ival> FOO %token <ival> FOO | |
338 | %union { int ival; } %union { int ival; } | |
339 | %% %% | |
340 | exp: 'a'; exp: 'a'; | |
341 | ||
342 | bison generates: | |
343 | ||
344 | #ifndef BAR_FOO_H #ifndef BAR_FOO_H | |
345 | # define BAR_FOO_H # define BAR_FOO_H | |
346 | ||
347 | /* Enabling traces. */ /* Enabling traces. */ | |
348 | # ifndef YYDEBUG | # ifndef BAR_DEBUG | |
349 | > # if defined YYDEBUG | |
350 | > # if YYDEBUG | |
351 | > # define BAR_DEBUG 1 | |
352 | > # else | |
353 | > # define BAR_DEBUG 0 | |
354 | > # endif | |
355 | > # else | |
356 | # define YYDEBUG 0 | # define BAR_DEBUG 0 | |
357 | > # endif | |
358 | # endif | # endif | |
359 | ||
360 | # if YYDEBUG | # if BAR_DEBUG | |
361 | extern int bar_debug; extern int bar_debug; | |
362 | # endif # endif | |
363 | ||
364 | /* Tokens. */ /* Tokens. */ | |
365 | # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE | |
366 | # define YYTOKENTYPE | # define BAR_TOKENTYPE | |
367 | enum yytokentype { | enum bar_tokentype { | |
368 | FOO = 258 FOO = 258 | |
369 | }; }; | |
370 | # endif # endif | |
371 | ||
372 | #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \ | |
373 | && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED | |
374 | typedef union YYSTYPE | typedef union BAR_STYPE | |
375 | { { | |
376 | int ival; int ival; | |
377 | } YYSTYPE; | } BAR_STYPE; | |
378 | # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1 | |
379 | #endif #endif | |
380 | ||
381 | extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval; | |
382 | ||
383 | int bar_parse (void); int bar_parse (void); | |
384 | ||
385 | #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */ | |
386 | ||
387 | * Noteworthy changes in release 2.5.1 (2012-06-05) [stable] | |
388 | ||
389 | ** Future changes: | |
390 | ||
391 | The next major release will drop support for generating parsers in K&R C. | |
392 | ||
393 | ** yacc.c: YYBACKUP works as expected. | |
394 | ||
395 | ** glr.c improvements: | |
396 | ||
397 | *** Location support is eliminated when not requested: | |
398 | ||
399 | GLR parsers used to include location-related code even when locations were | |
400 | not requested, and therefore not even usable. | |
401 | ||
402 | *** __attribute__ is preserved: | |
403 | ||
404 | __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e., | |
405 | when -std is passed to GCC). | |
406 | ||
407 | ** lalr1.java: several fixes: | |
408 | ||
409 | The Java parser no longer throws ArrayIndexOutOfBoundsException if the | |
410 | first token leads to a syntax error. Some minor clean ups. | |
411 | ||
412 | ** Changes for C++: | |
413 | ||
414 | *** C++11 compatibility: | |
415 | ||
416 | C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L | |
417 | or higher. | |
418 | ||
419 | *** Header guards | |
420 | ||
421 | The header files such as "parser.hh", "location.hh", etc. used a constant | |
422 | name for preprocessor guards, for instance: | |
423 | ||
424 | #ifndef BISON_LOCATION_HH | |
425 | # define BISON_LOCATION_HH | |
426 | ... | |
427 | #endif // !BISON_LOCATION_HH | |
428 | ||
429 | The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower | |
430 | case characters are converted to upper case, and series of | |
431 | non-alphanumerical characters are converted to an underscore. | |
432 | ||
433 | With "bison -o lang++/parser.cc", "location.hh" would now include: | |
434 | ||
435 | #ifndef YY_LANG_LOCATION_HH | |
436 | # define YY_LANG_LOCATION_HH | |
437 | ... | |
438 | #endif // !YY_LANG_LOCATION_HH | |
439 | ||
440 | *** C++ locations: | |
441 | ||
442 | The position and location constructors (and their initialize methods) | |
443 | accept new arguments for line and column. Several issues in the | |
444 | documentation were fixed. | |
445 | ||
446 | ** liby is no longer asking for "rpl_fprintf" on some platforms. | |
447 | ||
448 | ** Changes in the manual: | |
449 | ||
450 | *** %printer is documented | |
451 | ||
452 | The "%printer" directive, supported since at least Bison 1.50, is finally | |
453 | documented. The "mfcalc" example is extended to demonstrate it. | |
454 | ||
455 | For consistency with the C skeletons, the C++ parsers now also support | |
456 | "yyoutput" (as an alias to "debug_stream ()"). | |
457 | ||
458 | *** Several improvements have been made: | |
459 | ||
460 | The layout for grammar excerpts was changed to a more compact scheme. | |
461 | Named references are motivated. The description of the automaton | |
462 | description file (*.output) is updated to the current format. Incorrect | |
463 | index entries were fixed. Some other errors were fixed. | |
464 | ||
465 | ** Building bison: | |
466 | ||
467 | *** Conflicting prototypes with recent/modified Flex. | |
468 | ||
469 | Fixed build problems with the current, unreleased, version of Flex, and | |
470 | some modified versions of 2.5.35, which have modified function prototypes. | |
471 | ||
472 | *** Warnings during the build procedure have been eliminated. | |
473 | ||
474 | *** Several portability problems in the test suite have been fixed: | |
475 | ||
476 | This includes warnings with some compilers, unexpected behavior of tools | |
477 | such as diff, warning messages from the test suite itself, etc. | |
478 | ||
479 | *** The install-pdf target works properly: | |
480 | ||
481 | Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer | |
482 | halts in the middle of its course. | |
483 | ||
484 | * Changes in version 2.5 (2011-05-14): | |
485 | ||
486 | ** Grammar symbol names can now contain non-initial dashes: | |
487 | ||
488 | Consistently with directives (such as %error-verbose) and with | |
489 | %define variables (e.g. push-pull), grammar symbol names may contain | |
490 | dashes in any position except the beginning. This is a GNU | |
491 | extension over POSIX Yacc. Thus, use of this extension is reported | |
492 | by -Wyacc and rejected in Yacc mode (--yacc). | |
493 | ||
494 | ** Named references: | |
495 | ||
496 | Historically, Yacc and Bison have supported positional references | |
497 | ($n, $$) to allow access to symbol values from inside of semantic | |
498 | actions code. | |
499 | ||
500 | Starting from this version, Bison can also accept named references. | |
501 | When no ambiguity is possible, original symbol names may be used | |
502 | as named references: | |
503 | ||
504 | if_stmt : "if" cond_expr "then" then_stmt ';' | |
505 | { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); } | |
506 | ||
507 | In the more common case, explicit names may be declared: | |
508 | ||
509 | stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';' | |
510 | { $res = mk_if_stmt($cond, $then, $else); } | |
511 | ||
512 | Location information is also accessible using @name syntax. When | |
513 | accessing symbol names containing dots or dashes, explicit bracketing | |
514 | ($[sym.1]) must be used. | |
515 | ||
516 | These features are experimental in this version. More user feedback | |
517 | will help to stabilize them. | |
518 | Contributed by Alex Rozenman. | |
519 | ||
520 | ** IELR(1) and canonical LR(1): | |
521 | ||
522 | IELR(1) is a minimal LR(1) parser table generation algorithm. That | |
523 | is, given any context-free grammar, IELR(1) generates parser tables | |
524 | with the full language-recognition power of canonical LR(1) but with | |
525 | nearly the same number of parser states as LALR(1). This reduction | |
526 | in parser states is often an order of magnitude. More importantly, | |
527 | because canonical LR(1)'s extra parser states may contain duplicate | |
528 | conflicts in the case of non-LR(1) grammars, the number of conflicts | |
529 | for IELR(1) is often an order of magnitude less as well. This can | |
530 | significantly reduce the complexity of developing of a grammar. | |
531 | ||
532 | Bison can now generate IELR(1) and canonical LR(1) parser tables in | |
533 | place of its traditional LALR(1) parser tables, which remain the | |
534 | default. You can specify the type of parser tables in the grammar | |
535 | file with these directives: | |
536 | ||
537 | %define lr.type lalr | |
538 | %define lr.type ielr | |
539 | %define lr.type canonical-lr | |
540 | ||
541 | The default-reduction optimization in the parser tables can also be | |
542 | adjusted using "%define lr.default-reductions". For details on both | |
543 | of these features, see the new section "Tuning LR" in the Bison | |
544 | manual. | |
545 | ||
546 | These features are experimental. More user feedback will help to | |
547 | stabilize them. | |
548 | ||
549 | ** LAC (Lookahead Correction) for syntax error handling: | |
550 | ||
551 | Canonical LR, IELR, and LALR can suffer from a couple of problems | |
552 | upon encountering a syntax error. First, the parser might perform | |
553 | additional parser stack reductions before discovering the syntax | |
554 | error. Such reductions can perform user semantic actions that are | |
555 | unexpected because they are based on an invalid token, and they | |
556 | cause error recovery to begin in a different syntactic context than | |
557 | the one in which the invalid token was encountered. Second, when | |
558 | verbose error messages are enabled (with %error-verbose or the | |
559 | obsolete "#define YYERROR_VERBOSE"), the expected token list in the | |
560 | syntax error message can both contain invalid tokens and omit valid | |
561 | tokens. | |
562 | ||
563 | The culprits for the above problems are %nonassoc, default | |
564 | reductions in inconsistent states, and parser state merging. Thus, | |
565 | IELR and LALR suffer the most. Canonical LR can suffer only if | |
566 | %nonassoc is used or if default reductions are enabled for | |
567 | inconsistent states. | |
568 | ||
569 | LAC is a new mechanism within the parsing algorithm that solves | |
570 | these problems for canonical LR, IELR, and LALR without sacrificing | |
571 | %nonassoc, default reductions, or state merging. When LAC is in | |
572 | use, canonical LR and IELR behave almost exactly the same for both | |
573 | syntactically acceptable and syntactically unacceptable input. | |
574 | While LALR still does not support the full language-recognition | |
575 | power of canonical LR and IELR, LAC at least enables LALR's syntax | |
576 | error handling to correctly reflect LALR's language-recognition | |
577 | power. | |
578 | ||
579 | Currently, LAC is only supported for deterministic parsers in C. | |
580 | You can enable LAC with the following directive: | |
581 | ||
582 | %define parse.lac full | |
583 | ||
584 | See the new section "LAC" in the Bison manual for additional | |
585 | details including a few caveats. | |
586 | ||
587 | LAC is an experimental feature. More user feedback will help to | |
588 | stabilize it. | |
589 | ||
590 | ** %define improvements: | |
591 | ||
592 | *** Can now be invoked via the command line: | |
593 | ||
594 | Each of these command-line options | |
595 | ||
596 | -D NAME[=VALUE] | |
597 | --define=NAME[=VALUE] | |
598 | ||
599 | -F NAME[=VALUE] | |
600 | --force-define=NAME[=VALUE] | |
601 | ||
602 | is equivalent to this grammar file declaration | |
603 | ||
604 | %define NAME ["VALUE"] | |
605 | ||
606 | except that the manner in which Bison processes multiple definitions | |
607 | for the same NAME differs. Most importantly, -F and --force-define | |
608 | quietly override %define, but -D and --define do not. For further | |
609 | details, see the section "Bison Options" in the Bison manual. | |
610 | ||
611 | *** Variables renamed: | |
612 | ||
613 | The following %define variables | |
614 | ||
615 | api.push_pull | |
616 | lr.keep_unreachable_states | |
617 | ||
618 | have been renamed to | |
619 | ||
620 | api.push-pull | |
621 | lr.keep-unreachable-states | |
622 | ||
623 | The old names are now deprecated but will be maintained indefinitely | |
624 | for backward compatibility. | |
625 | ||
626 | *** Values no longer need to be quoted in the grammar file: | |
627 | ||
628 | If a %define value is an identifier, it no longer needs to be placed | |
629 | within quotations marks. For example, | |
630 | ||
631 | %define api.push-pull "push" | |
632 | ||
633 | can be rewritten as | |
634 | ||
635 | %define api.push-pull push | |
636 | ||
637 | *** Unrecognized variables are now errors not warnings. | |
638 | ||
639 | *** Multiple invocations for any variable is now an error not a warning. | |
640 | ||
641 | ** Unrecognized %code qualifiers are now errors not warnings. | |
642 | ||
643 | ** Character literals not of length one: | |
644 | ||
645 | Previously, Bison quietly converted all character literals to length | |
646 | one. For example, without warning, Bison interpreted the operators in | |
647 | the following grammar to be the same token: | |
648 | ||
649 | exp: exp '++' | |
650 | | exp '+' exp | |
651 | ; | |
652 | ||
653 | Bison now warns when a character literal is not of length one. In | |
654 | some future release, Bison will start reporting an error instead. | |
655 | ||
656 | ** Destructor calls fixed for lookaheads altered in semantic actions: | |
657 | ||
658 | Previously for deterministic parsers in C, if a user semantic action | |
659 | altered yychar, the parser in some cases used the old yychar value to | |
660 | determine which destructor to call for the lookahead upon a syntax | |
661 | error or upon parser return. This bug has been fixed. | |
662 | ||
663 | ** C++ parsers use YYRHSLOC: | |
664 | ||
665 | Similarly to the C parsers, the C++ parsers now define the YYRHSLOC | |
666 | macro and use it in the default YYLLOC_DEFAULT. You are encouraged | |
667 | to use it. If, for instance, your location structure has "first" | |
668 | and "last" members, instead of | |
669 | ||
670 | # define YYLLOC_DEFAULT(Current, Rhs, N) \ | |
671 | do \ | |
672 | if (N) \ | |
673 | { \ | |
674 | (Current).first = (Rhs)[1].location.first; \ | |
675 | (Current).last = (Rhs)[N].location.last; \ | |
676 | } \ | |
677 | else \ | |
678 | { \ | |
679 | (Current).first = (Current).last = (Rhs)[0].location.last; \ | |
680 | } \ | |
681 | while (false) | |
682 | ||
683 | use: | |
684 | ||
685 | # define YYLLOC_DEFAULT(Current, Rhs, N) \ | |
686 | do \ | |
687 | if (N) \ | |
688 | { \ | |
689 | (Current).first = YYRHSLOC (Rhs, 1).first; \ | |
690 | (Current).last = YYRHSLOC (Rhs, N).last; \ | |
691 | } \ | |
692 | else \ | |
693 | { \ | |
694 | (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \ | |
695 | } \ | |
696 | while (false) | |
697 | ||
698 | ** YYLLOC_DEFAULT in C++: | |
699 | ||
700 | The default implementation of YYLLOC_DEFAULT used to be issued in | |
701 | the header file. It is now output in the implementation file, after | |
702 | the user %code sections so that its #ifndef guard does not try to | |
703 | override the user's YYLLOC_DEFAULT if provided. | |
704 | ||
705 | ** YYFAIL now produces warnings and Java parsers no longer implement it: | |
706 | ||
707 | YYFAIL has existed for many years as an undocumented feature of | |
708 | deterministic parsers in C generated by Bison. More recently, it was | |
709 | a documented feature of Bison's experimental Java parsers. As | |
710 | promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a | |
711 | semantic action now produces a deprecation warning, and Java parsers | |
712 | no longer implement YYFAIL at all. For further details, including a | |
713 | discussion of how to suppress C preprocessor warnings about YYFAIL | |
714 | being unused, see the Bison 2.4.2 NEWS entry. | |
715 | ||
716 | ** Temporary hack for adding a semicolon to the user action: | |
717 | ||
718 | Previously, Bison appended a semicolon to every user action for | |
719 | reductions when the output language defaulted to C (specifically, when | |
720 | neither %yacc, %language, %skeleton, or equivalent command-line | |
721 | options were specified). This allowed actions such as | |
722 | ||
723 | exp: exp "+" exp { $$ = $1 + $3 }; | |
724 | ||
725 | instead of | |
726 | ||
727 | exp: exp "+" exp { $$ = $1 + $3; }; | |
728 | ||
729 | As a first step in removing this misfeature, Bison now issues a | |
730 | warning when it appends a semicolon. Moreover, in cases where Bison | |
731 | cannot easily determine whether a semicolon is needed (for example, an | |
732 | action ending with a cpp directive or a braced compound initializer), | |
733 | it no longer appends one. Thus, the C compiler might now complain | |
734 | about a missing semicolon where it did not before. Future releases of | |
735 | Bison will cease to append semicolons entirely. | |
736 | ||
737 | ** Verbose syntax error message fixes: | |
738 | ||
739 | When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is | |
740 | specified, syntax error messages produced by the generated parser | |
741 | include the unexpected token as well as a list of expected tokens. | |
742 | The effect of %nonassoc on these verbose messages has been corrected | |
743 | in two ways, but a more complete fix requires LAC, described above: | |
744 | ||
745 | *** When %nonassoc is used, there can exist parser states that accept no | |
746 | tokens, and so the parser does not always require a lookahead token | |
747 | in order to detect a syntax error. Because no unexpected token or | |
748 | expected tokens can then be reported, the verbose syntax error | |
749 | message described above is suppressed, and the parser instead | |
750 | reports the simpler message, "syntax error". Previously, this | |
751 | suppression was sometimes erroneously triggered by %nonassoc when a | |
752 | lookahead was actually required. Now verbose messages are | |
753 | suppressed only when all previous lookaheads have already been | |
754 | shifted or discarded. | |
755 | ||
756 | *** Previously, the list of expected tokens erroneously included tokens | |
757 | that would actually induce a syntax error because conflicts for them | |
758 | were resolved with %nonassoc in the current parser state. Such | |
759 | tokens are now properly omitted from the list. | |
760 | ||
761 | *** Expected token lists are still often wrong due to state merging | |
762 | (from LALR or IELR) and default reductions, which can both add | |
763 | invalid tokens and subtract valid tokens. Canonical LR almost | |
764 | completely fixes this problem by eliminating state merging and | |
765 | default reductions. However, there is one minor problem left even | |
766 | when using canonical LR and even after the fixes above. That is, | |
767 | if the resolution of a conflict with %nonassoc appears in a later | |
768 | parser state than the one at which some syntax error is | |
769 | discovered, the conflicted token is still erroneously included in | |
770 | the expected token list. Bison's new LAC implementation, | |
771 | described above, eliminates this problem and the need for | |
772 | canonical LR. However, LAC is still experimental and is disabled | |
773 | by default. | |
774 | ||
775 | ** Java skeleton fixes: | |
776 | ||
777 | *** A location handling bug has been fixed. | |
778 | ||
779 | *** The top element of each of the value stack and location stack is now | |
780 | cleared when popped so that it can be garbage collected. | |
781 | ||
782 | *** Parser traces now print the top element of the stack. | |
783 | ||
784 | ** -W/--warnings fixes: | |
785 | ||
786 | *** Bison now properly recognizes the "no-" versions of categories: | |
787 | ||
788 | For example, given the following command line, Bison now enables all | |
789 | warnings except warnings for incompatibilities with POSIX Yacc: | |
790 | ||
791 | bison -Wall,no-yacc gram.y | |
792 | ||
793 | *** Bison now treats S/R and R/R conflicts like other warnings: | |
794 | ||
795 | Previously, conflict reports were independent of Bison's normal | |
796 | warning system. Now, Bison recognizes the warning categories | |
797 | "conflicts-sr" and "conflicts-rr". This change has important | |
798 | consequences for the -W and --warnings command-line options. For | |
799 | example: | |
800 | ||
801 | bison -Wno-conflicts-sr gram.y # S/R conflicts not reported | |
802 | bison -Wno-conflicts-rr gram.y # R/R conflicts not reported | |
803 | bison -Wnone gram.y # no conflicts are reported | |
804 | bison -Werror gram.y # any conflict is an error | |
805 | ||
806 | However, as before, if the %expect or %expect-rr directive is | |
807 | specified, an unexpected number of conflicts is an error, and an | |
808 | expected number of conflicts is not reported, so -W and --warning | |
809 | then have no effect on the conflict report. | |
810 | ||
811 | *** The "none" category no longer disables a preceding "error": | |
812 | ||
813 | For example, for the following command line, Bison now reports | |
814 | errors instead of warnings for incompatibilities with POSIX Yacc: | |
815 | ||
816 | bison -Werror,none,yacc gram.y | |
817 | ||
818 | *** The "none" category now disables all Bison warnings: | |
819 | ||
820 | Previously, the "none" category disabled only Bison warnings for | |
821 | which there existed a specific -W/--warning category. However, | |
822 | given the following command line, Bison is now guaranteed to | |
823 | suppress all warnings: | |
824 | ||
825 | bison -Wnone gram.y | |
826 | ||
827 | ** Precedence directives can now assign token number 0: | |
828 | ||
829 | Since Bison 2.3b, which restored the ability of precedence | |
830 | directives to assign token numbers, doing so for token number 0 has | |
831 | produced an assertion failure. For example: | |
832 | ||
833 | %left END 0 | |
834 | ||
835 | This bug has been fixed. | |
836 | ||
837 | * Changes in version 2.4.3 (2010-08-05): | |
838 | ||
839 | ** Bison now obeys -Werror and --warnings=error for warnings about | |
840 | grammar rules that are useless in the parser due to conflicts. | |
841 | ||
842 | ** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have | |
843 | been fixed. | |
844 | ||
845 | ** Failures in the test suite for GCC 4.5 have been fixed. | |
846 | ||
847 | ** Failures in the test suite for some versions of Sun Studio C++ have | |
848 | been fixed. | |
849 | ||
850 | ** Contrary to Bison 2.4.2's NEWS entry, it has been decided that | |
851 | warnings about undefined %prec identifiers will not be converted to | |
852 | errors in Bison 2.5. They will remain warnings, which should be | |
853 | sufficient for POSIX while avoiding backward compatibility issues. | |
854 | ||
855 | ** Minor documentation fixes. | |
856 | ||
857 | * Changes in version 2.4.2 (2010-03-20): | |
858 | ||
859 | ** Some portability problems that resulted in failures and livelocks | |
860 | in the test suite on some versions of at least Solaris, AIX, HP-UX, | |
861 | RHEL4, and Tru64 have been addressed. As a result, fatal Bison | |
862 | errors should no longer cause M4 to report a broken pipe on the | |
863 | affected platforms. | |
864 | ||
865 | ** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately. | |
866 | ||
867 | POSIX specifies that an error be reported for any identifier that does | |
868 | not appear on the LHS of a grammar rule and that is not defined by | |
869 | %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this | |
870 | error report for the case when an identifier appears only after a | |
871 | %prec directive. It is now restored. However, for backward | |
872 | compatibility with recent Bison releases, it is only a warning for | |
873 | now. In Bison 2.5 and later, it will return to being an error. | |
874 | [Between the 2.4.2 and 2.4.3 releases, it was decided that this | |
875 | warning will not be converted to an error in Bison 2.5.] | |
876 | ||
877 | ** Detection of GNU M4 1.4.6 or newer during configure is improved. | |
878 | ||
879 | ** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS, | |
880 | YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now | |
881 | avoided. | |
882 | ||
883 | ** %code is now a permanent feature. | |
884 | ||
885 | A traditional Yacc prologue directive is written in the form: | |
886 | ||
887 | %{CODE%} | |
888 | ||
889 | To provide a more flexible alternative, Bison 2.3b introduced the | |
890 | %code directive with the following forms for C/C++: | |
891 | ||
892 | %code {CODE} | |
893 | %code requires {CODE} | |
894 | %code provides {CODE} | |
895 | %code top {CODE} | |
896 | ||
897 | These forms are now considered permanent features of Bison. See the | |
898 | %code entries in the section "Bison Declaration Summary" in the Bison | |
899 | manual for a summary of their functionality. See the section | |
900 | "Prologue Alternatives" for a detailed discussion including the | |
901 | advantages of %code over the traditional Yacc prologue directive. | |
902 | ||
903 | Bison's Java feature as a whole including its current usage of %code | |
904 | is still considered experimental. | |
905 | ||
906 | ** YYFAIL is deprecated and will eventually be removed. | |
907 | ||
908 | YYFAIL has existed for many years as an undocumented feature of | |
909 | deterministic parsers in C generated by Bison. Previously, it was | |
910 | documented for Bison's experimental Java parsers. YYFAIL is no longer | |
911 | documented for Java parsers and is formally deprecated in both cases. | |
912 | Users are strongly encouraged to migrate to YYERROR, which is | |
913 | specified by POSIX. | |
914 | ||
915 | Like YYERROR, you can invoke YYFAIL from a semantic action in order to | |
916 | induce a syntax error. The most obvious difference from YYERROR is | |
917 | that YYFAIL will automatically invoke yyerror to report the syntax | |
918 | error so that you don't have to. However, there are several other | |
919 | subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from | |
920 | inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is | |
921 | used. For a more detailed discussion, see: | |
922 | ||
923 | http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html | |
924 | ||
925 | The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but | |
926 | deterministic parsers in C will continue to implement it. However, | |
927 | because YYFAIL is already flawed, it seems futile to try to make new | |
928 | Bison features compatible with it. Thus, during parser generation, | |
929 | Bison 2.5 will produce a warning whenever it discovers YYFAIL in a | |
930 | rule action. In a later release, YYFAIL will be disabled for | |
931 | %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will | |
932 | be removed altogether. | |
933 | ||
934 | There exists at least one case where Bison 2.5's YYFAIL warning will | |
935 | be a false positive. Some projects add phony uses of YYFAIL and other | |
936 | Bison-defined macros for the sole purpose of suppressing C | |
937 | preprocessor warnings (from GCC cpp's -Wunused-macros, for example). | |
938 | To avoid Bison's future warning, such YYFAIL uses can be moved to the | |
939 | epilogue (that is, after the second "%%") in the Bison input file. In | |
940 | this release (2.4.2), Bison already generates its own code to suppress | |
941 | C preprocessor warnings for YYFAIL, so projects can remove their own | |
942 | phony uses of YYFAIL if compatibility with Bison releases prior to | |
943 | 2.4.2 is not necessary. | |
944 | ||
945 | ** Internationalization. | |
946 | ||
947 | Fix a regression introduced in Bison 2.4: Under some circumstances, | |
948 | message translations were not installed although supported by the | |
949 | host system. | |
950 | ||
951 | * Changes in version 2.4.1 (2008-12-11): | |
952 | ||
953 | ** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc | |
954 | declarations have been fixed. | |
955 | ||
956 | ** Temporary hack for adding a semicolon to the user action. | |
957 | ||
958 | Bison used to prepend a trailing semicolon at the end of the user | |
959 | action for reductions. This allowed actions such as | |
960 | ||
961 | exp: exp "+" exp { $$ = $1 + $3 }; | |
962 | ||
963 | instead of | |
964 | ||
965 | exp: exp "+" exp { $$ = $1 + $3; }; | |
966 | ||
967 | Some grammars still depend on this "feature". Bison 2.4.1 restores | |
968 | the previous behavior in the case of C output (specifically, when | |
969 | neither %language or %skeleton or equivalent command-line options | |
970 | are used) to leave more time for grammars depending on the old | |
971 | behavior to be adjusted. Future releases of Bison will disable this | |
972 | feature. | |
973 | ||
974 | ** A few minor improvements to the Bison manual. | |
975 | ||
976 | * Changes in version 2.4 (2008-11-02): | |
977 | ||
978 | ** %language is an experimental feature. | |
979 | ||
980 | We first introduced this feature in test release 2.3b as a cleaner | |
981 | alternative to %skeleton. Since then, we have discussed the possibility of | |
982 | modifying its effect on Bison's output file names. Thus, in this release, | |
983 | we consider %language to be an experimental feature that will likely evolve | |
984 | in future releases. | |
985 | ||
986 | ** Forward compatibility with GNU M4 has been improved. | |
987 | ||
988 | ** Several bugs in the C++ skeleton and the experimental Java skeleton have been | |
989 | fixed. | |
990 | ||
991 | * Changes in version 2.3b (2008-05-27): | |
992 | ||
993 | ** The quotes around NAME that used to be required in the following directive | |
994 | are now deprecated: | |
995 | ||
996 | %define NAME "VALUE" | |
997 | ||
998 | ** The directive "%pure-parser" is now deprecated in favor of: | |
999 | ||
1000 | %define api.pure | |
1001 | ||
1002 | which has the same effect except that Bison is more careful to warn about | |
1003 | unreasonable usage in the latter case. | |
1004 | ||
1005 | ** Push Parsing | |
1006 | ||
1007 | Bison can now generate an LALR(1) parser in C with a push interface. That | |
1008 | is, instead of invoking "yyparse", which pulls tokens from "yylex", you can | |
1009 | push one token at a time to the parser using "yypush_parse", which will | |
1010 | return to the caller after processing each token. By default, the push | |
1011 | interface is disabled. Either of the following directives will enable it: | |
1012 | ||
1013 | %define api.push_pull "push" // Just push; does not require yylex. | |
1014 | %define api.push_pull "both" // Push and pull; requires yylex. | |
1015 | ||
1016 | See the new section "A Push Parser" in the Bison manual for details. | |
1017 | ||
1018 | The current push parsing interface is experimental and may evolve. More user | |
1019 | feedback will help to stabilize it. | |
1020 | ||
1021 | ** The -g and --graph options now output graphs in Graphviz DOT format, | |
1022 | not VCG format. Like --graph, -g now also takes an optional FILE argument | |
1023 | and thus cannot be bundled with other short options. | |
1024 | ||
1025 | ** Java | |
1026 | ||
1027 | Bison can now generate an LALR(1) parser in Java. The skeleton is | |
1028 | "data/lalr1.java". Consider using the new %language directive instead of | |
1029 | %skeleton to select it. | |
1030 | ||
1031 | See the new section "Java Parsers" in the Bison manual for details. | |
1032 | ||
1033 | The current Java interface is experimental and may evolve. More user | |
1034 | feedback will help to stabilize it. | |
1035 | Contributed by Paolo Bonzini. | |
1036 | ||
1037 | ** %language | |
1038 | ||
1039 | This new directive specifies the programming language of the generated | |
1040 | parser, which can be C (the default), C++, or Java. Besides the skeleton | |
1041 | that Bison uses, the directive affects the names of the generated files if | |
1042 | the grammar file's name ends in ".y". | |
1043 | ||
1044 | ** XML Automaton Report | |
1045 | ||
1046 | Bison can now generate an XML report of the LALR(1) automaton using the new | |
1047 | "--xml" option. The current XML schema is experimental and may evolve. More | |
1048 | user feedback will help to stabilize it. | |
1049 | Contributed by Wojciech Polak. | |
1050 | ||
1051 | ** The grammar file may now specify the name of the parser header file using | |
1052 | %defines. For example: | |
1053 | ||
1054 | %defines "parser.h" | |
1055 | ||
1056 | ** When reporting useless rules, useless nonterminals, and unused terminals, | |
1057 | Bison now employs the terms "useless in grammar" instead of "useless", | |
1058 | "useless in parser" instead of "never reduced", and "unused in grammar" | |
1059 | instead of "unused". | |
1060 | ||
1061 | ** Unreachable State Removal | |
1062 | ||
1063 | Previously, Bison sometimes generated parser tables containing unreachable | |
1064 | states. A state can become unreachable during conflict resolution if Bison | |
1065 | disables a shift action leading to it from a predecessor state. Bison now: | |
1066 | ||
1067 | 1. Removes unreachable states. | |
1068 | ||
1069 | 2. Does not report any conflicts that appeared in unreachable states. | |
1070 | WARNING: As a result, you may need to update %expect and %expect-rr | |
1071 | directives in existing grammar files. | |
1072 | ||
1073 | 3. For any rule used only in such states, Bison now reports the rule as | |
1074 | "useless in parser due to conflicts". | |
1075 | ||
1076 | This feature can be disabled with the following directive: | |
1077 | ||
1078 | %define lr.keep_unreachable_states | |
1079 | ||
1080 | See the %define entry in the "Bison Declaration Summary" in the Bison manual | |
1081 | for further discussion. | |
1082 | ||
1083 | ** Lookahead Set Correction in the ".output" Report | |
1084 | ||
1085 | When instructed to generate a ".output" file including lookahead sets | |
1086 | (using "--report=lookahead", for example), Bison now prints each reduction's | |
1087 | lookahead set only next to the associated state's one item that (1) is | |
1088 | associated with the same rule as the reduction and (2) has its dot at the end | |
1089 | of its RHS. Previously, Bison also erroneously printed the lookahead set | |
1090 | next to all of the state's other items associated with the same rule. This | |
1091 | bug affected only the ".output" file and not the generated parser source | |
1092 | code. | |
1093 | ||
1094 | ** --report-file=FILE is a new option to override the default ".output" file | |
1095 | name. | |
1096 | ||
1097 | ** The "=" that used to be required in the following directives is now | |
1098 | deprecated: | |
1099 | ||
1100 | %file-prefix "parser" | |
1101 | %name-prefix "c_" | |
1102 | %output "parser.c" | |
1103 | ||
1104 | ** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}" | |
1105 | ||
1106 | Bison 2.3a provided a new set of directives as a more flexible alternative to | |
1107 | the traditional Yacc prologue blocks. Those have now been consolidated into | |
1108 | a single %code directive with an optional qualifier field, which identifies | |
1109 | the purpose of the code and thus the location(s) where Bison should generate | |
1110 | it: | |
1111 | ||
1112 | 1. "%code {CODE}" replaces "%after-header {CODE}" | |
1113 | 2. "%code requires {CODE}" replaces "%start-header {CODE}" | |
1114 | 3. "%code provides {CODE}" replaces "%end-header {CODE}" | |
1115 | 4. "%code top {CODE}" replaces "%before-header {CODE}" | |
1116 | ||
1117 | See the %code entries in section "Bison Declaration Summary" in the Bison | |
1118 | manual for a summary of the new functionality. See the new section "Prologue | |
1119 | Alternatives" for a detailed discussion including the advantages of %code | |
1120 | over the traditional Yacc prologues. | |
1121 | ||
1122 | The prologue alternatives are experimental. More user feedback will help to | |
1123 | determine whether they should become permanent features. | |
1124 | ||
1125 | ** Revised warning: unset or unused mid-rule values | |
1126 | ||
1127 | Since Bison 2.2, Bison has warned about mid-rule values that are set but not | |
1128 | used within any of the actions of the parent rule. For example, Bison warns | |
1129 | about unused $2 in: | |
1130 | ||
1131 | exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; }; | |
1132 | ||
1133 | Now, Bison also warns about mid-rule values that are used but not set. For | |
1134 | example, Bison warns about unset $$ in the mid-rule action in: | |
1135 | ||
1136 | exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; }; | |
1137 | ||
1138 | However, Bison now disables both of these warnings by default since they | |
1139 | sometimes prove to be false alarms in existing grammars employing the Yacc | |
1140 | constructs $0 or $-N (where N is some positive integer). | |
1141 | ||
1142 | To enable these warnings, specify the option "--warnings=midrule-values" or | |
1143 | "-W", which is a synonym for "--warnings=all". | |
1144 | ||
1145 | ** Default %destructor or %printer with "<*>" or "<>" | |
1146 | ||
1147 | Bison now recognizes two separate kinds of default %destructor's and | |
1148 | %printer's: | |
1149 | ||
1150 | 1. Place "<*>" in a %destructor/%printer symbol list to define a default | |
1151 | %destructor/%printer for all grammar symbols for which you have formally | |
1152 | declared semantic type tags. | |
1153 | ||
1154 | 2. Place "<>" in a %destructor/%printer symbol list to define a default | |
1155 | %destructor/%printer for all grammar symbols without declared semantic | |
1156 | type tags. | |
1157 | ||
1158 | Bison no longer supports the "%symbol-default" notation from Bison 2.3a. | |
1159 | "<*>" and "<>" combined achieve the same effect with one exception: Bison no | |
1160 | longer applies any %destructor to a mid-rule value if that mid-rule value is | |
1161 | not actually ever referenced using either $$ or $n in a semantic action. | |
1162 | ||
1163 | The default %destructor's and %printer's are experimental. More user | |
1164 | feedback will help to determine whether they should become permanent | |
1165 | features. | |
1166 | ||
1167 | See the section "Freeing Discarded Symbols" in the Bison manual for further | |
1168 | details. | |
1169 | ||
1170 | ** %left, %right, and %nonassoc can now declare token numbers. This is required | |
1171 | by POSIX. However, see the end of section "Operator Precedence" in the Bison | |
1172 | manual for a caveat concerning the treatment of literal strings. | |
1173 | ||
1174 | ** The nonfunctional --no-parser, -n, and %no-parser options have been | |
1175 | completely removed from Bison. | |
1176 | ||
1177 | * Changes in version 2.3a, 2006-09-13: | |
1178 | ||
1179 | ** Instead of %union, you can define and use your own union type | |
1180 | YYSTYPE if your grammar contains at least one <type> tag. | |
1181 | Your YYSTYPE need not be a macro; it can be a typedef. | |
1182 | This change is for compatibility with other Yacc implementations, | |
1183 | and is required by POSIX. | |
1184 | ||
1185 | ** Locations columns and lines start at 1. | |
1186 | In accordance with the GNU Coding Standards and Emacs. | |
1187 | ||
1188 | ** You may now declare per-type and default %destructor's and %printer's: | |
1189 | ||
1190 | For example: | |
1191 | ||
1192 | %union { char *string; } | |
1193 | %token <string> STRING1 | |
1194 | %token <string> STRING2 | |
1195 | %type <string> string1 | |
1196 | %type <string> string2 | |
1197 | %union { char character; } | |
1198 | %token <character> CHR | |
1199 | %type <character> chr | |
1200 | %destructor { free ($$); } %symbol-default | |
1201 | %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1 | |
1202 | %destructor { } <character> | |
1203 | ||
1204 | guarantees that, when the parser discards any user-defined symbol that has a | |
1205 | semantic type tag other than "<character>", it passes its semantic value to | |
1206 | "free". However, when the parser discards a "STRING1" or a "string1", it | |
1207 | also prints its line number to "stdout". It performs only the second | |
1208 | "%destructor" in this case, so it invokes "free" only once. | |
1209 | ||
1210 | [Although we failed to mention this here in the 2.3a release, the default | |
1211 | %destructor's and %printer's were experimental, and they were rewritten in | |
1212 | future versions.] | |
1213 | ||
1214 | ** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y", | |
1215 | "--yacc", or "%yacc"), Bison no longer generates #define statements for | |
1216 | associating token numbers with token names. Removing the #define statements | |
1217 | helps to sanitize the global namespace during preprocessing, but POSIX Yacc | |
1218 | requires them. Bison still generates an enum for token names in all cases. | |
1219 | ||
1220 | ** Handling of traditional Yacc prologue blocks is now more consistent but | |
1221 | potentially incompatible with previous releases of Bison. | |
1222 | ||
1223 | As before, you declare prologue blocks in your grammar file with the | |
1224 | "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all | |
1225 | prologue blocks that you've declared before the first %union. To generate | |
1226 | the post-prologue, Bison concatenates all prologue blocks that you've | |
1227 | declared after the first %union. | |
1228 | ||
1229 | Previous releases of Bison inserted the pre-prologue into both the header | |
1230 | file and the code file in all cases except for LALR(1) parsers in C. In the | |
1231 | latter case, Bison inserted it only into the code file. For parsers in C++, | |
1232 | the point of insertion was before any token definitions (which associate | |
1233 | token numbers with names). For parsers in C, the point of insertion was | |
1234 | after the token definitions. | |
1235 | ||
1236 | Now, Bison never inserts the pre-prologue into the header file. In the code | |
1237 | file, it always inserts it before the token definitions. | |
1238 | ||
1239 | ** Bison now provides a more flexible alternative to the traditional Yacc | |
1240 | prologue blocks: %before-header, %start-header, %end-header, and | |
1241 | %after-header. | |
1242 | ||
1243 | For example, the following declaration order in the grammar file reflects the | |
1244 | order in which Bison will output these code blocks. However, you are free to | |
1245 | declare these code blocks in your grammar file in whatever order is most | |
1246 | convenient for you: | |
1247 | ||
1248 | %before-header { | |
1249 | /* Bison treats this block like a pre-prologue block: it inserts it into | |
1250 | * the code file before the contents of the header file. It does *not* | |
1251 | * insert it into the header file. This is a good place to put | |
1252 | * #include's that you want at the top of your code file. A common | |
1253 | * example is '#include "system.h"'. */ | |
1254 | } | |
1255 | %start-header { | |
1256 | /* Bison inserts this block into both the header file and the code file. | |
1257 | * In both files, the point of insertion is before any Bison-generated | |
1258 | * token, semantic type, location type, and class definitions. This is a | |
1259 | * good place to define %union dependencies, for example. */ | |
1260 | } | |
1261 | %union { | |
1262 | /* Unlike the traditional Yacc prologue blocks, the output order for the | |
1263 | * new %*-header blocks is not affected by their declaration position | |
1264 | * relative to any %union in the grammar file. */ | |
1265 | } | |
1266 | %end-header { | |
1267 | /* Bison inserts this block into both the header file and the code file. | |
1268 | * In both files, the point of insertion is after the Bison-generated | |
1269 | * definitions. This is a good place to declare or define public | |
1270 | * functions or data structures that depend on the Bison-generated | |
1271 | * definitions. */ | |
1272 | } | |
1273 | %after-header { | |
1274 | /* Bison treats this block like a post-prologue block: it inserts it into | |
1275 | * the code file after the contents of the header file. It does *not* | |
1276 | * insert it into the header file. This is a good place to declare or | |
1277 | * define internal functions or data structures that depend on the | |
1278 | * Bison-generated definitions. */ | |
1279 | } | |
1280 | ||
1281 | If you have multiple occurrences of any one of the above declarations, Bison | |
1282 | will concatenate the contents in declaration order. | |
1283 | ||
1284 | [Although we failed to mention this here in the 2.3a release, the prologue | |
1285 | alternatives were experimental, and they were rewritten in future versions.] | |
1286 | ||
1287 | ** The option "--report=look-ahead" has been changed to "--report=lookahead". | |
1288 | The old spelling still works, but is not documented and may be removed | |
1289 | in a future release. | |
1290 | ||
1291 | * Changes in version 2.3, 2006-06-05: | |
1292 | ||
1293 | ** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING", | |
1294 | for compatibility with LALR(1) grammars. | |
1295 | ||
1296 | ** It is now documented that any definition of YYSTYPE or YYLTYPE should | |
1297 | be to a type name that does not contain parentheses or brackets. | |
1298 | ||
1299 | * Changes in version 2.2, 2006-05-19: | |
1300 | ||
1301 | ** The distribution terms for all Bison-generated parsers now permit | |
1302 | using the parsers in nonfree programs. Previously, this permission | |
1303 | was granted only for Bison-generated LALR(1) parsers in C. | |
1304 | ||
1305 | ** %name-prefix changes the namespace name in C++ outputs. | |
1306 | ||
1307 | ** The C++ parsers export their token_type. | |
1308 | ||
1309 | ** Bison now allows multiple %union declarations, and concatenates | |
1310 | their contents together. | |
1311 | ||
1312 | ** New warning: unused values | |
1313 | Right-hand side symbols whose values are not used are reported, | |
1314 | if the symbols have destructors. For instance: | |
1315 | ||
1316 | exp: exp "?" exp ":" exp { $1 ? $1 : $3; } | |
1317 | | exp "+" exp | |
1318 | ; | |
1319 | ||
1320 | will trigger a warning about $$ and $5 in the first rule, and $3 in | |
1321 | the second ($1 is copied to $$ by the default rule). This example | |
1322 | most likely contains three errors, and could be rewritten as: | |
1323 | ||
1324 | exp: exp "?" exp ":" exp | |
1325 | { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); } | |
1326 | | exp "+" exp | |
1327 | { $$ = $1 ? $1 : $3; if ($1) free ($3); } | |
1328 | ; | |
1329 | ||
1330 | However, if the original actions were really intended, memory leaks | |
1331 | and all, the warnings can be suppressed by letting Bison believe the | |
1332 | values are used, e.g.: | |
1333 | ||
1334 | exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); } | |
1335 | | exp "+" exp { $$ = $1; (void) $3; } | |
1336 | ; | |
1337 | ||
1338 | If there are mid-rule actions, the warning is issued if no action | |
1339 | uses it. The following triggers no warning: $1 and $3 are used. | |
1340 | ||
1341 | exp: exp { push ($1); } '+' exp { push ($3); sum (); }; | |
1342 | ||
1343 | The warning is intended to help catching lost values and memory leaks. | |
1344 | If a value is ignored, its associated memory typically is not reclaimed. | |
1345 | ||
1346 | ** %destructor vs. YYABORT, YYACCEPT, and YYERROR. | |
1347 | Destructors are now called when user code invokes YYABORT, YYACCEPT, | |
1348 | and YYERROR, for all objects on the stack, other than objects | |
1349 | corresponding to the right-hand side of the current rule. | |
1350 | ||
1351 | ** %expect, %expect-rr | |
1352 | Incorrect numbers of expected conflicts are now actual errors, | |
1353 | instead of warnings. | |
1354 | ||
1355 | ** GLR, YACC parsers. | |
1356 | The %parse-params are available in the destructors (and the | |
1357 | experimental printers) as per the documentation. | |
1358 | ||
1359 | ** Bison now warns if it finds a stray "$" or "@" in an action. | |
1360 | ||
1361 | ** %require "VERSION" | |
1362 | This specifies that the grammar file depends on features implemented | |
1363 | in Bison version VERSION or higher. | |
1364 | ||
1365 | ** lalr1.cc: The token and value types are now class members. | |
1366 | The tokens were defined as free form enums and cpp macros. YYSTYPE | |
1367 | was defined as a free form union. They are now class members: | |
1368 | tokens are enumerations of the "yy::parser::token" struct, and the | |
1369 | semantic values have the "yy::parser::semantic_type" type. | |
1370 | ||
1371 | If you do not want or can update to this scheme, the directive | |
1372 | '%define "global_tokens_and_yystype" "1"' triggers the global | |
1373 | definition of tokens and YYSTYPE. This change is suitable both | |
1374 | for previous releases of Bison, and this one. | |
1375 | ||
1376 | If you wish to update, then make sure older version of Bison will | |
1377 | fail using '%require "2.2"'. | |
1378 | ||
1379 | ** DJGPP support added. | |
1380 | \f | |
1381 | * Changes in version 2.1, 2005-09-16: | |
1382 | ||
1383 | ** The C++ lalr1.cc skeleton supports %lex-param. | |
1384 | ||
1385 | ** Bison-generated parsers now support the translation of diagnostics like | |
1386 | "syntax error" into languages other than English. The default | |
1387 | language is still English. For details, please see the new | |
1388 | Internationalization section of the Bison manual. Software | |
1389 | distributors should also see the new PACKAGING file. Thanks to | |
1390 | Bruno Haible for this new feature. | |
1391 | ||
1392 | ** Wording in the Bison-generated parsers has been changed slightly to | |
1393 | simplify translation. In particular, the message "memory exhausted" | |
1394 | has replaced "parser stack overflow", as the old message was not | |
1395 | always accurate for modern Bison-generated parsers. | |
1396 | ||
1397 | ** Destructors are now called when the parser aborts, for all symbols left | |
1398 | behind on the stack. Also, the start symbol is now destroyed after a | |
1399 | successful parse. In both cases, the behavior was formerly inconsistent. | |
1400 | ||
1401 | ** When generating verbose diagnostics, Bison-generated parsers no longer | |
1402 | quote the literal strings associated with tokens. For example, for | |
1403 | a syntax error associated with '%token NUM "number"' they might | |
1404 | print 'syntax error, unexpected number' instead of 'syntax error, | |
1405 | unexpected "number"'. | |
1406 | \f | |
1407 | * Changes in version 2.0, 2004-12-25: | |
1408 | ||
1409 | ** Possibly-incompatible changes | |
1410 | ||
1411 | - Bison-generated parsers no longer default to using the alloca function | |
1412 | (when available) to extend the parser stack, due to widespread | |
1413 | problems in unchecked stack-overflow detection. You can "#define | |
1414 | YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read | |
1415 | the manual to determine safe values for YYMAXDEPTH in that case. | |
1416 | ||
1417 | - Error token location. | |
1418 | During error recovery, the location of the syntax error is updated | |
1419 | to cover the whole sequence covered by the error token: it includes | |
1420 | the shifted symbols thrown away during the first part of the error | |
1421 | recovery, and the lookahead rejected during the second part. | |
1422 | ||
1423 | - Semicolon changes: | |
1424 | . Stray semicolons are no longer allowed at the start of a grammar. | |
1425 | . Semicolons are now required after in-grammar declarations. | |
1426 | ||
1427 | - Unescaped newlines are no longer allowed in character constants or | |
1428 | string literals. They were never portable, and GCC 3.4.0 has | |
1429 | dropped support for them. Better diagnostics are now generated if | |
1430 | forget a closing quote. | |
1431 | ||
1432 | - NUL bytes are no longer allowed in Bison string literals, unfortunately. | |
1433 | ||
1434 | ** New features | |
1435 | ||
1436 | - GLR grammars now support locations. | |
1437 | ||
1438 | - New directive: %initial-action. | |
1439 | This directive allows the user to run arbitrary code (including | |
1440 | initializing @$) from yyparse before parsing starts. | |
1441 | ||
1442 | - A new directive "%expect-rr N" specifies the expected number of | |
1443 | reduce/reduce conflicts in GLR parsers. | |
1444 | ||
1445 | - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d". | |
1446 | This is a GNU extension. | |
1447 | ||
1448 | - The option "--report=lookahead" was changed to "--report=look-ahead". | |
1449 | [However, this was changed back after 2.3.] | |
1450 | ||
1451 | - Experimental %destructor support has been added to lalr1.cc. | |
1452 | ||
1453 | - New configure option --disable-yacc, to disable installation of the | |
1454 | yacc command and -ly library introduced in 1.875 for POSIX conformance. | |
1455 | ||
1456 | ** Bug fixes | |
1457 | ||
1458 | - For now, %expect-count violations are now just warnings, not errors. | |
1459 | This is for compatibility with Bison 1.75 and earlier (when there are | |
1460 | reduce/reduce conflicts) and with Bison 1.30 and earlier (when there | |
1461 | are too many or too few shift/reduce conflicts). However, in future | |
1462 | versions of Bison we plan to improve the %expect machinery so that | |
1463 | these violations will become errors again. | |
1464 | ||
1465 | - Within Bison itself, numbers (e.g., goto numbers) are no longer | |
1466 | arbitrarily limited to 16-bit counts. | |
1467 | ||
1468 | - Semicolons are now allowed before "|" in grammar rules, as POSIX requires. | |
1469 | \f | |
1470 | * Changes in version 1.875, 2003-01-01: | |
1471 | ||
1472 | ** The documentation license has been upgraded to version 1.2 | |
1473 | of the GNU Free Documentation License. | |
1474 | ||
1475 | ** syntax error processing | |
1476 | ||
1477 | - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error | |
1478 | locations too. This fixes bugs in error-location computation. | |
1479 | ||
1480 | - %destructor | |
1481 | It is now possible to reclaim the memory associated to symbols | |
1482 | discarded during error recovery. This feature is still experimental. | |
1483 | ||
1484 | - %error-verbose | |
1485 | This new directive is preferred over YYERROR_VERBOSE. | |
1486 | ||
1487 | - #defining yyerror to steal internal variables is discouraged. | |
1488 | It is not guaranteed to work forever. | |
1489 | ||
1490 | ** POSIX conformance | |
1491 | ||
1492 | - Semicolons are once again optional at the end of grammar rules. | |
1493 | This reverts to the behavior of Bison 1.33 and earlier, and improves | |
1494 | compatibility with Yacc. | |
1495 | ||
1496 | - "parse error" -> "syntax error" | |
1497 | Bison now uniformly uses the term "syntax error"; formerly, the code | |
1498 | and manual sometimes used the term "parse error" instead. POSIX | |
1499 | requires "syntax error" in diagnostics, and it was thought better to | |
1500 | be consistent. | |
1501 | ||
1502 | - The documentation now emphasizes that yylex and yyerror must be | |
1503 | declared before use. C99 requires this. | |
1504 | ||
1505 | - Bison now parses C99 lexical constructs like UCNs and | |
1506 | backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires. | |
1507 | ||
1508 | - File names are properly escaped in C output. E.g., foo\bar.y is | |
1509 | output as "foo\\bar.y". | |
1510 | ||
1511 | - Yacc command and library now available | |
1512 | The Bison distribution now installs a "yacc" command, as POSIX requires. | |
1513 | Also, Bison now installs a small library liby.a containing | |
1514 | implementations of Yacc-compatible yyerror and main functions. | |
1515 | This library is normally not useful, but POSIX requires it. | |
1516 | ||
1517 | - Type clashes now generate warnings, not errors. | |
1518 | ||
1519 | - If the user does not define YYSTYPE as a macro, Bison now declares it | |
1520 | using typedef instead of defining it as a macro. | |
1521 | For consistency, YYLTYPE is also declared instead of defined. | |
1522 | ||
1523 | ** Other compatibility issues | |
1524 | ||
1525 | - %union directives can now have a tag before the "{", e.g., the | |
1526 | directive "%union foo {...}" now generates the C code | |
1527 | "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility. | |
1528 | The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc. | |
1529 | For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype". | |
1530 | This is for compatibility with both Yacc and Bison 1.35. | |
1531 | ||
1532 | - ";" is output before the terminating "}" of an action, for | |
1533 | compatibility with Bison 1.35. | |
1534 | ||
1535 | - Bison now uses a Yacc-style format for conflict reports, e.g., | |
1536 | "conflicts: 2 shift/reduce, 1 reduce/reduce". | |
1537 | ||
1538 | - "yystype" and "yyltype" are now obsolescent macros instead of being | |
1539 | typedefs or tags; they are no longer documented and are planned to be | |
1540 | withdrawn in a future release. | |
1541 | ||
1542 | ** GLR parser notes | |
1543 | ||
1544 | - GLR and inline | |
1545 | Users of Bison have to decide how they handle the portability of the | |
1546 | C keyword "inline". | |
1547 | ||
1548 | - "parsing stack overflow..." -> "parser stack overflow" | |
1549 | GLR parsers now report "parser stack overflow" as per the Bison manual. | |
1550 | ||
1551 | ** %parse-param and %lex-param | |
1552 | The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass | |
1553 | additional context to yyparse and yylex. They suffer from several | |
1554 | shortcomings: | |
1555 | ||
1556 | - a single argument only can be added, | |
1557 | - their types are weak (void *), | |
1558 | - this context is not passed to ancillary functions such as yyerror, | |
1559 | - only yacc.c parsers support them. | |
1560 | ||
1561 | The new %parse-param/%lex-param directives provide a more precise control. | |
1562 | For instance: | |
1563 | ||
1564 | %parse-param {int *nastiness} | |
1565 | %lex-param {int *nastiness} | |
1566 | %parse-param {int *randomness} | |
1567 | ||
1568 | results in the following signatures: | |
1569 | ||
1570 | int yylex (int *nastiness); | |
1571 | int yyparse (int *nastiness, int *randomness); | |
1572 | ||
1573 | or, if both %pure-parser and %locations are used: | |
1574 | ||
1575 | int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness); | |
1576 | int yyparse (int *nastiness, int *randomness); | |
1577 | ||
1578 | ** Bison now warns if it detects conflicting outputs to the same file, | |
1579 | e.g., it generates a warning for "bison -d -o foo.h foo.y" since | |
1580 | that command outputs both code and header to foo.h. | |
1581 | ||
1582 | ** #line in output files | |
1583 | - --no-line works properly. | |
1584 | ||
1585 | ** Bison can no longer be built by a K&R C compiler; it requires C89 or | |
1586 | later to be built. This change originally took place a few versions | |
1587 | ago, but nobody noticed until we recently asked someone to try | |
1588 | building Bison with a K&R C compiler. | |
1589 | \f | |
1590 | * Changes in version 1.75, 2002-10-14: | |
1591 | ||
1592 | ** Bison should now work on 64-bit hosts. | |
1593 | ||
1594 | ** Indonesian translation thanks to Tedi Heriyanto. | |
1595 | ||
1596 | ** GLR parsers | |
1597 | Fix spurious parse errors. | |
1598 | ||
1599 | ** Pure parsers | |
1600 | Some people redefine yyerror to steal yyparse' private variables. | |
1601 | Reenable this trick until an official feature replaces it. | |
1602 | ||
1603 | ** Type Clashes | |
1604 | In agreement with POSIX and with other Yaccs, leaving a default | |
1605 | action is valid when $$ is untyped, and $1 typed: | |
1606 | ||
1607 | untyped: ... typed; | |
1608 | ||
1609 | but the converse remains an error: | |
1610 | ||
1611 | typed: ... untyped; | |
1612 | ||
1613 | ** Values of mid-rule actions | |
1614 | The following code: | |
1615 | ||
1616 | foo: { ... } { $$ = $1; } ... | |
1617 | ||
1618 | was incorrectly rejected: $1 is defined in the second mid-rule | |
1619 | action, and is equal to the $$ of the first mid-rule action. | |
1620 | \f | |
1621 | * Changes in version 1.50, 2002-10-04: | |
1622 | ||
1623 | ** GLR parsing | |
1624 | The declaration | |
1625 | %glr-parser | |
1626 | causes Bison to produce a Generalized LR (GLR) parser, capable of handling | |
1627 | almost any context-free grammar, ambiguous or not. The new declarations | |
1628 | %dprec and %merge on grammar rules allow parse-time resolution of | |
1629 | ambiguities. Contributed by Paul Hilfinger. | |
1630 | ||
1631 | Unfortunately Bison 1.50 does not work properly on 64-bit hosts | |
1632 | like the Alpha, so please stick to 32-bit hosts for now. | |
1633 | ||
1634 | ** Output Directory | |
1635 | When not in Yacc compatibility mode, when the output file was not | |
1636 | specified, running "bison foo/bar.y" created "foo/bar.c". It | |
1637 | now creates "bar.c". | |
1638 | ||
1639 | ** Undefined token | |
1640 | The undefined token was systematically mapped to 2 which prevented | |
1641 | the use of 2 by the user. This is no longer the case. | |
1642 | ||
1643 | ** Unknown token numbers | |
1644 | If yylex returned an out of range value, yyparse could die. This is | |
1645 | no longer the case. | |
1646 | ||
1647 | ** Error token | |
1648 | According to POSIX, the error token must be 256. | |
1649 | Bison extends this requirement by making it a preference: *if* the | |
1650 | user specified that one of her tokens is numbered 256, then error | |
1651 | will be mapped onto another number. | |
1652 | ||
1653 | ** Verbose error messages | |
1654 | They no longer report "..., expecting error or..." for states where | |
1655 | error recovery is possible. | |
1656 | ||
1657 | ** End token | |
1658 | Defaults to "$end" instead of "$". | |
1659 | ||
1660 | ** Error recovery now conforms to documentation and to POSIX | |
1661 | When a Bison-generated parser encounters a syntax error, it now pops | |
1662 | the stack until it finds a state that allows shifting the error | |
1663 | token. Formerly, it popped the stack until it found a state that | |
1664 | allowed some non-error action other than a default reduction on the | |
1665 | error token. The new behavior has long been the documented behavior, | |
1666 | and has long been required by POSIX. For more details, please see | |
1667 | Paul Eggert, "Reductions during Bison error handling" (2002-05-20) | |
1668 | <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>. | |
1669 | ||
1670 | ** Traces | |
1671 | Popped tokens and nonterminals are now reported. | |
1672 | ||
1673 | ** Larger grammars | |
1674 | Larger grammars are now supported (larger token numbers, larger grammar | |
1675 | size (= sum of the LHS and RHS lengths), larger LALR tables). | |
1676 | Formerly, many of these numbers ran afoul of 16-bit limits; | |
1677 | now these limits are 32 bits on most hosts. | |
1678 | ||
1679 | ** Explicit initial rule | |
1680 | Bison used to play hacks with the initial rule, which the user does | |
1681 | not write. It is now explicit, and visible in the reports and | |
1682 | graphs as rule 0. | |
1683 | ||
1684 | ** Useless rules | |
1685 | Before, Bison reported the useless rules, but, although not used, | |
1686 | included them in the parsers. They are now actually removed. | |
1687 | ||
1688 | ** Useless rules, useless nonterminals | |
1689 | They are now reported, as a warning, with their locations. | |
1690 | ||
1691 | ** Rules never reduced | |
1692 | Rules that can never be reduced because of conflicts are now | |
1693 | reported. | |
1694 | ||
1695 | ** Incorrect "Token not used" | |
1696 | On a grammar such as | |
1697 | ||
1698 | %token useless useful | |
1699 | %% | |
1700 | exp: '0' %prec useful; | |
1701 | ||
1702 | where a token was used to set the precedence of the last rule, | |
1703 | bison reported both "useful" and "useless" as useless tokens. | |
1704 | ||
1705 | ** Revert the C++ namespace changes introduced in 1.31 | |
1706 | as they caused too many portability hassles. | |
1707 | ||
1708 | ** Default locations | |
1709 | By an accident of design, the default computation of @$ was | |
1710 | performed after another default computation was performed: @$ = @1. | |
1711 | The latter is now removed: YYLLOC_DEFAULT is fully responsible of | |
1712 | the computation of @$. | |
1713 | ||
1714 | ** Token end-of-file | |
1715 | The token end of file may be specified by the user, in which case, | |
1716 | the user symbol is used in the reports, the graphs, and the verbose | |
1717 | error messages instead of "$end", which remains being the default. | |
1718 | For instance | |
1719 | %token MYEOF 0 | |
1720 | or | |
1721 | %token MYEOF 0 "end of file" | |
1722 | ||
1723 | ** Semantic parser | |
1724 | This old option, which has been broken for ages, is removed. | |
1725 | ||
1726 | ** New translations | |
1727 | Brazilian Portuguese, thanks to Alexandre Folle de Menezes. | |
1728 | Croatian, thanks to Denis Lackovic. | |
1729 | ||
1730 | ** Incorrect token definitions | |
1731 | When given | |
1732 | %token 'a' "A" | |
1733 | bison used to output | |
1734 | #define 'a' 65 | |
1735 | ||
1736 | ** Token definitions as enums | |
1737 | Tokens are output both as the traditional #define's, and, provided | |
1738 | the compiler supports ANSI C or is a C++ compiler, as enums. | |
1739 | This lets debuggers display names instead of integers. | |
1740 | ||
1741 | ** Reports | |
1742 | In addition to --verbose, bison supports --report=THINGS, which | |
1743 | produces additional information: | |
1744 | - itemset | |
1745 | complete the core item sets with their closure | |
1746 | - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back] | |
1747 | explicitly associate lookahead tokens to items | |
1748 | - solved | |
1749 | describe shift/reduce conflicts solving. | |
1750 | Bison used to systematically output this information on top of | |
1751 | the report. Solved conflicts are now attached to their states. | |
1752 | ||
1753 | ** Type clashes | |
1754 | Previous versions don't complain when there is a type clash on | |
1755 | the default action if the rule has a mid-rule action, such as in: | |
1756 | ||
1757 | %type <foo> bar | |
1758 | %% | |
1759 | bar: '0' {} '0'; | |
1760 | ||
1761 | This is fixed. | |
1762 | ||
1763 | ** GNU M4 is now required when using Bison. | |
1764 | \f | |
1765 | * Changes in version 1.35, 2002-03-25: | |
1766 | ||
1767 | ** C Skeleton | |
1768 | Some projects use Bison's C parser with C++ compilers, and define | |
1769 | YYSTYPE as a class. The recent adjustment of C parsers for data | |
1770 | alignment and 64 bit architectures made this impossible. | |
1771 | ||
1772 | Because for the time being no real solution for C++ parser | |
1773 | generation exists, kludges were implemented in the parser to | |
1774 | maintain this use. In the future, when Bison has C++ parsers, this | |
1775 | kludge will be disabled. | |
1776 | ||
1777 | This kludge also addresses some C++ problems when the stack was | |
1778 | extended. | |
1779 | \f | |
1780 | * Changes in version 1.34, 2002-03-12: | |
1781 | ||
1782 | ** File name clashes are detected | |
1783 | $ bison foo.y -d -o foo.x | |
1784 | fatal error: header and parser would both be named "foo.x" | |
1785 | ||
1786 | ** A missing ";" at the end of a rule triggers a warning | |
1787 | In accordance with POSIX, and in agreement with other | |
1788 | Yacc implementations, Bison will mandate this semicolon in the near | |
1789 | future. This eases the implementation of a Bison parser of Bison | |
1790 | grammars by making this grammar LALR(1) instead of LR(2). To | |
1791 | facilitate the transition, this release introduces a warning. | |
1792 | ||
1793 | ** Revert the C++ namespace changes introduced in 1.31, as they caused too | |
1794 | many portability hassles. | |
1795 | ||
1796 | ** DJGPP support added. | |
1797 | ||
1798 | ** Fix test suite portability problems. | |
1799 | \f | |
1800 | * Changes in version 1.33, 2002-02-07: | |
1801 | ||
1802 | ** Fix C++ issues | |
1803 | Groff could not be compiled for the definition of size_t was lacking | |
1804 | under some conditions. | |
1805 | ||
1806 | ** Catch invalid @n | |
1807 | As is done with $n. | |
1808 | \f | |
1809 | * Changes in version 1.32, 2002-01-23: | |
1810 | ||
1811 | ** Fix Yacc output file names | |
1812 | ||
1813 | ** Portability fixes | |
1814 | ||
1815 | ** Italian, Dutch translations | |
1816 | \f | |
1817 | * Changes in version 1.31, 2002-01-14: | |
1818 | ||
1819 | ** Many Bug Fixes | |
1820 | ||
1821 | ** GNU Gettext and %expect | |
1822 | GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that | |
1823 | Bison dies on incorrect %expectations, we fear there will be | |
1824 | too many bug reports for Gettext, so _for the time being_, %expect | |
1825 | does not trigger an error when the input file is named "plural.y". | |
1826 | ||
1827 | ** Use of alloca in parsers | |
1828 | If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use | |
1829 | malloc exclusively. Since 1.29, but was not NEWS'ed. | |
1830 | ||
1831 | alloca is used only when compiled with GCC, to avoid portability | |
1832 | problems as on AIX. | |
1833 | ||
1834 | ** yyparse now returns 2 if memory is exhausted; formerly it dumped core. | |
1835 | ||
1836 | ** When the generated parser lacks debugging code, YYDEBUG is now 0 | |
1837 | (as POSIX requires) instead of being undefined. | |
1838 | ||
1839 | ** User Actions | |
1840 | Bison has always permitted actions such as { $$ = $1 }: it adds the | |
1841 | ending semicolon. Now if in Yacc compatibility mode, the semicolon | |
1842 | is no longer output: one has to write { $$ = $1; }. | |
1843 | ||
1844 | ** Better C++ compliance | |
1845 | The output parsers try to respect C++ namespaces. | |
1846 | [This turned out to be a failed experiment, and it was reverted later.] | |
1847 | ||
1848 | ** Reduced Grammars | |
1849 | Fixed bugs when reporting useless nonterminals. | |
1850 | ||
1851 | ** 64 bit hosts | |
1852 | The parsers work properly on 64 bit hosts. | |
1853 | ||
1854 | ** Error messages | |
1855 | Some calls to strerror resulted in scrambled or missing error messages. | |
1856 | ||
1857 | ** %expect | |
1858 | When the number of shift/reduce conflicts is correct, don't issue | |
1859 | any warning. | |
1860 | ||
1861 | ** The verbose report includes the rule line numbers. | |
1862 | ||
1863 | ** Rule line numbers are fixed in traces. | |
1864 | ||
1865 | ** Swedish translation | |
1866 | ||
1867 | ** Parse errors | |
1868 | Verbose parse error messages from the parsers are better looking. | |
1869 | Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'('' | |
1870 | Now: parse error: unexpected '/', expecting "number" or '-' or '(' | |
1871 | ||
1872 | ** Fixed parser memory leaks. | |
1873 | When the generated parser was using malloc to extend its stacks, the | |
1874 | previous allocations were not freed. | |
1875 | ||
1876 | ** Fixed verbose output file. | |
1877 | Some newlines were missing. | |
1878 | Some conflicts in state descriptions were missing. | |
1879 | ||
1880 | ** Fixed conflict report. | |
1881 | Option -v was needed to get the result. | |
1882 | ||
1883 | ** %expect | |
1884 | Was not used. | |
1885 | Mismatches are errors, not warnings. | |
1886 | ||
1887 | ** Fixed incorrect processing of some invalid input. | |
1888 | ||
1889 | ** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H. | |
1890 | ||
1891 | ** Fixed some typos in the documentation. | |
1892 | ||
1893 | ** %token MY_EOF 0 is supported. | |
1894 | Before, MY_EOF was silently renumbered as 257. | |
1895 | ||
1896 | ** doc/refcard.tex is updated. | |
1897 | ||
1898 | ** %output, %file-prefix, %name-prefix. | |
1899 | New. | |
1900 | ||
1901 | ** --output | |
1902 | New, aliasing "--output-file". | |
1903 | \f | |
1904 | * Changes in version 1.30, 2001-10-26: | |
1905 | ||
1906 | ** "--defines" and "--graph" have now an optional argument which is the | |
1907 | output file name. "-d" and "-g" do not change; they do not take any | |
1908 | argument. | |
1909 | ||
1910 | ** "%source_extension" and "%header_extension" are removed, failed | |
1911 | experiment. | |
1912 | ||
1913 | ** Portability fixes. | |
1914 | \f | |
1915 | * Changes in version 1.29, 2001-09-07: | |
1916 | ||
1917 | ** The output file does not define const, as this caused problems when used | |
1918 | with common autoconfiguration schemes. If you still use ancient compilers | |
1919 | that lack const, compile with the equivalent of the C compiler option | |
1920 | "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this. | |
1921 | ||
1922 | ** Added "-g" and "--graph". | |
1923 | ||
1924 | ** The Bison manual is now distributed under the terms of the GNU FDL. | |
1925 | ||
1926 | ** The input and the output files has automatically a similar extension. | |
1927 | ||
1928 | ** Russian translation added. | |
1929 | ||
1930 | ** NLS support updated; should hopefully be less troublesome. | |
1931 | ||
1932 | ** Added the old Bison reference card. | |
1933 | ||
1934 | ** Added "--locations" and "%locations". | |
1935 | ||
1936 | ** Added "-S" and "--skeleton". | |
1937 | ||
1938 | ** "%raw", "-r", "--raw" is disabled. | |
1939 | ||
1940 | ** Special characters are escaped when output. This solves the problems | |
1941 | of the #line lines with path names including backslashes. | |
1942 | ||
1943 | ** New directives. | |
1944 | "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose", | |
1945 | "%debug", "%source_extension" and "%header_extension". | |
1946 | ||
1947 | ** @$ | |
1948 | Automatic location tracking. | |
1949 | \f | |
1950 | * Changes in version 1.28, 1999-07-06: | |
1951 | ||
1952 | ** Should compile better now with K&R compilers. | |
1953 | ||
1954 | ** Added NLS. | |
1955 | ||
1956 | ** Fixed a problem with escaping the double quote character. | |
1957 | ||
1958 | ** There is now a FAQ. | |
1959 | \f | |
1960 | * Changes in version 1.27: | |
1961 | ||
1962 | ** The make rule which prevented bison.simple from being created on | |
1963 | some systems has been fixed. | |
1964 | \f | |
1965 | * Changes in version 1.26: | |
1966 | ||
1967 | ** Bison now uses Automake. | |
1968 | ||
1969 | ** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>. | |
1970 | ||
1971 | ** Token numbers now start at 257 as previously documented, not 258. | |
1972 | ||
1973 | ** Bison honors the TMPDIR environment variable. | |
1974 | ||
1975 | ** A couple of buffer overruns have been fixed. | |
1976 | ||
1977 | ** Problems when closing files should now be reported. | |
1978 | ||
1979 | ** Generated parsers should now work even on operating systems which do | |
1980 | not provide alloca(). | |
1981 | \f | |
1982 | * Changes in version 1.25, 1995-10-16: | |
1983 | ||
1984 | ** Errors in the input grammar are not fatal; Bison keeps reading | |
1985 | the grammar file, and reports all the errors found in it. | |
1986 | ||
1987 | ** Tokens can now be specified as multiple-character strings: for | |
1988 | example, you could use "<=" for a token which looks like <=, instead | |
1989 | of choosing a name like LESSEQ. | |
1990 | ||
1991 | ** The %token_table declaration says to write a table of tokens (names | |
1992 | and numbers) into the parser file. The yylex function can use this | |
1993 | table to recognize multiple-character string tokens, or for other | |
1994 | purposes. | |
1995 | ||
1996 | ** The %no_lines declaration says not to generate any #line preprocessor | |
1997 | directives in the parser file. | |
1998 | ||
1999 | ** The %raw declaration says to use internal Bison token numbers, not | |
2000 | Yacc-compatible token numbers, when token names are defined as macros. | |
2001 | ||
2002 | ** The --no-parser option produces the parser tables without including | |
2003 | the parser engine; a project can now use its own parser engine. | |
2004 | The actions go into a separate file called NAME.act, in the form of | |
2005 | a switch statement body. | |
2006 | \f | |
2007 | * Changes in version 1.23: | |
2008 | ||
2009 | The user can define YYPARSE_PARAM as the name of an argument to be | |
2010 | passed into yyparse. The argument should have type void *. It should | |
2011 | actually point to an object. Grammar actions can access the variable | |
2012 | by casting it to the proper pointer type. | |
2013 | ||
2014 | Line numbers in output file corrected. | |
2015 | \f | |
2016 | * Changes in version 1.22: | |
2017 | ||
2018 | --help option added. | |
2019 | \f | |
2020 | * Changes in version 1.20: | |
2021 | ||
2022 | Output file does not redefine const for C++. | |
2023 | ||
2024 | ----- | |
2025 | ||
2026 | Copyright (C) 1995-2012 Free Software Foundation, Inc. | |
2027 | ||
2028 | This file is part of Bison, the GNU Parser Generator. | |
2029 | ||
2030 | This program is free software: you can redistribute it and/or modify | |
2031 | it under the terms of the GNU General Public License as published by | |
2032 | the Free Software Foundation, either version 3 of the License, or | |
2033 | (at your option) any later version. | |
2034 | ||
2035 | This program is distributed in the hope that it will be useful, | |
2036 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
2037 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
2038 | GNU General Public License for more details. | |
2039 | ||
2040 | You should have received a copy of the GNU General Public License | |
2041 | along with this program. If not, see <http://www.gnu.org/licenses/>. | |
2042 | ||
2043 | LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr | |
2044 | LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr | |
2045 | LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar | |
2046 | LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr | |
2047 | LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks | |
2048 | LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG | |
2049 | LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush | |
2050 | LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly | |
2051 | LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's | |
2052 | LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de | |
2053 | LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto | |
2054 | LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs | |
2055 | LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF | |
2056 | LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY | |
2057 | LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL | |
2058 | LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh | |
2059 | LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf | |
2060 | LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp calc yyo fval Wmaybe | |
2061 | LocalWords: yyvsp pragmas noreturn java's | |
2062 | ||
2063 | Local Variables: | |
2064 | mode: outline | |
2065 | fill-column: 76 | |
2066 | End: |