NEWS

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