doc: create a new Tuning LR section in the manual.

[bison.git] / NEWS

diff --git a/NEWS b/NEWS
index 342b49fc473b03b48c793997b04615aac2d9b3d4..a657bcc85ed50955a8d44ae53341692e3c53b43d 100644 (file)
--- a/NEWS
+++ b/NEWS
@@ -3,7 +3,15 @@ Bison News
 
 * Changes in version 2.5 (????-??-??):
 
 
 * Changes in version 2.5 (????-??-??):
 

-** Named References Support
+** Grammar symbol names can now contain non-initial dashes:
+
+  Consistently with directives (such as %error-verbose) and with
+  %define variables (e.g. push-pull), grammar symbol names may contain
+  dashes in any position except the beginning.  This is a GNU
+  extension over POSIX Yacc.  Thus, use of this extension is reported
+  by -Wyacc and rejected in Yacc mode (--yacc).
+
+** Named references:

 
   Historically, Yacc and Bison have supported positional references
   ($n, $$) to allow access to symbol values from inside of semantic
 
   Historically, Yacc and Bison have supported positional references
   ($n, $$) to allow access to symbol values from inside of semantic


@@ -28,13 +36,13 @@ Bison News
   These features are experimental in this version.  More user feedback
   will help to stabilize them.
 
   These features are experimental in this version.  More user feedback
   will help to stabilize them.
 

-** IELR(1) and Canonical LR(1) Support
+** IELR(1) and canonical LR(1):

 
   IELR(1) is a minimal LR(1) parser table generation algorithm.  That
   is, given any context-free grammar, IELR(1) generates parser tables
   with the full language recognition power of canonical LR(1) but with
 
   IELR(1) is a minimal LR(1) parser table generation algorithm.  That
   is, given any context-free grammar, IELR(1) generates parser tables
   with the full language recognition power of canonical LR(1) but with

-  nearly the same number of parser states as LALR(1).  This reduction in
-  parser states is often an order of magnitude.  More importantly,
+  nearly the same number of parser states as LALR(1).  This reduction
+  in parser states is often an order of magnitude.  More importantly,

   because canonical LR(1)'s extra parser states may contain duplicate
   conflicts in the case of non-LR(1) grammars, the number of conflicts
   for IELR(1) is often an order of magnitude less as well.  This can
   because canonical LR(1)'s extra parser states may contain duplicate
   conflicts in the case of non-LR(1) grammars, the number of conflicts
   for IELR(1) is often an order of magnitude less as well.  This can


@@ -49,24 +57,58 @@ Bison News
     %define lr.type ielr
     %define lr.type canonical-lr
 
     %define lr.type ielr
     %define lr.type canonical-lr
 

-  The default reduction optimization in the parser tables can also be
-  adjusted using `%define lr.default-reductions'.  See the documentation
-  for `%define lr.type' and `%define lr.default-reductions' in the
-  section `Bison Declaration Summary' in the Bison manual for the
-  details.
+  The default-reduction optimization in the parser tables can also be
+  adjusted using `%define lr.default-reductions'.  For details on both
+  of these features, see the new section `Tuning LR' in the Bison
+  manual.

 
   These features are experimental.  More user feedback will help to
   stabilize them.
 
 
   These features are experimental.  More user feedback will help to
   stabilize them.
 

-** Unrecognized %code qualifiers are now an error not a warning.
+** LAC (Lookahead Correction) for syntax error handling:

 
 

-** %define improvements.
+  Canonical LR, IELR, and LALR can suffer from a couple of problems
+  upon encountering a syntax error.  First, the parser might perform
+  additional parser stack reductions before discovering the syntax
+  error.  Such reductions can perform user semantic actions that are
+  unexpected because they are based on an invalid token, and they
+  cause error recovery to begin in a different syntactic context than
+  the one in which the invalid token was encountered.  Second, when
+  verbose error messages are enabled (with %error-verbose or the
+  obsolete `#define YYERROR_VERBOSE'), the expected token list in the
+  syntax error message can both contain invalid tokens and omit valid
+  tokens.

 
 

-*** Unrecognized variables are now an error not a warning.
+  The culprits for the above problems are %nonassoc, default
+  reductions in inconsistent states, and parser state merging.  Thus,
+  IELR and LALR suffer the most.  Canonical LR can suffer only if
+  %nonassoc is used or if default reductions are enabled for
+  inconsistent states.

 
 

-*** Multiple invocations for any variable is now an error not a warning.
+  LAC is a new mechanism within the parsing algorithm that solves
+  these problems for canonical LR, IELR, and LALR without sacrificing
+  %nonassoc, default reductions, or state merging.  When LAC is in
+  use, canonical LR and IELR behave almost exactly the same for both
+  syntactically acceptable and syntactically unacceptable input.
+  While LALR still does not support the full language-recognition
+  power of canonical LR and IELR, LAC at least enables LALR's syntax
+  error handling to correctly reflect LALR's language-recognition
+  power.
+
+  Currently, LAC is only supported for deterministic parsers in C.
+  You can enable LAC with the following directive:

 
 

-*** Can now be invoked via the command line.
+    %define parse.lac full
+
+  See the new section `LAC' in the Bison manual for additional
+  details including a few caveats.
+
+  LAC is an experimental feature.  More user feedback will help to
+  stabilize it.
+
+** %define improvements:
+
+*** Can now be invoked via the command line:

 
   Each of these command-line options
 
 
   Each of these command-line options
 


@@ -85,7 +127,7 @@ Bison News
   quietly override %define, but -D and --define do not.  For further
   details, see the section "Bison Options" in the Bison manual.
 
   quietly override %define, but -D and --define do not.  For further
   details, see the section "Bison Options" in the Bison manual.
 

-*** Variables renamed.
+*** Variables renamed:

 
   The following %define variables
 
 
   The following %define variables
 


@@ -100,7 +142,7 @@ Bison News
   The old names are now deprecated but will be maintained indefinitely
   for backward compatibility.
 
   The old names are now deprecated but will be maintained indefinitely
   for backward compatibility.
 

-*** Values no longer need to be quoted in grammar file.
+*** Values no longer need to be quoted in grammar file:

 
   If a %define value is an identifier, it no longer needs to be placed
   within quotations marks.  For example,
 
   If a %define value is an identifier, it no longer needs to be placed
   within quotations marks.  For example,


@@ -111,15 +153,86 @@ Bison News
 
     %define api.push-pull push
 
 
     %define api.push-pull push
 

-** Symbol names.
+*** Unrecognized variables are now errors not warnings.

 
 

-  Consistently with directives (such as %error-verbose) and variables
-  (e.g. push-pull), symbol names may include dashes in any position,
-  similarly to periods and underscores.  This is GNU extension over
-  POSIX Yacc whose use is reported by -Wyacc, and rejected in Yacc
-  mode (--yacc).
+*** Multiple invocations for any variable is now an error not a warning.

 
 

-** Temporary hack for adding a semicolon to the user action.
+** Unrecognized %code qualifiers are now errors not warnings.
+
+** Character literals not of length one:
+
+  Previously, Bison quietly converted all character literals to length
+  one.  For example, without warning, Bison interpreted the operators in
+  the following grammar to be the same token:
+
+    exp: exp '++'
+       | exp '+' exp
+       ;
+
+  Bison now warns when a character literal is not of length one.  In
+  some future release, Bison will start reporting an error instead.
+
+** Destructor calls fixed for lookaheads altered in semantic actions:
+
+  Previously for deterministic parsers in C, if a user semantic action
+  altered yychar, the parser in some cases used the old yychar value to
+  determine which destructor to call for the lookahead upon a syntax
+  error or upon parser return.  This bug has been fixed.
+
+** C++ parsers use YYRHSLOC:
+
+  Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
+  macro and use it in the default YYLLOC_DEFAULT.  You are encouraged
+  to use it.  If, for instance, your location structure has "first"
+  and "last" members, instead of
+
+      # define YYLLOC_DEFAULT(Current, Rhs, N)                             \
+        do                                                                 \
+          if (N)                                                           \
+            {                                                              \
+              (Current).first = (Rhs)[1].location.first;                   \
+              (Current).last  = (Rhs)[N].location.last;                    \
+            }                                                              \
+          else                                                             \
+            {                                                              \
+              (Current).first = (Current).last = (Rhs)[0].location.last;   \
+            }                                                              \
+        while (false)
+
+  use:
+
+      # define YYLLOC_DEFAULT(Current, Rhs, N)                             \
+        do                                                                 \
+          if (N)                                                           \
+            {                                                              \
+              (Current).first = YYRHSLOC (Rhs, 1).first;                   \
+              (Current).last  = YYRHSLOC (Rhs, N).last;                    \
+            }                                                              \
+          else                                                             \
+            {                                                              \
+              (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last;   \
+            }                                                              \
+        while (false)
+
+** YYLLOC_DEFAULT in C++:
+
+  The default implementation of YYLLOC_DEFAULT used to be issued in
+  the header file.  It is now output in the implementation file, after
+  the user %code sections so that its #ifndef guard does not try to
+  override the user's YYLLOC_DEFAULT if provided.
+
+** YYFAIL now produces warnings and Java parsers no longer implement it:
+
+  YYFAIL has existed for many years as an undocumented feature of
+  deterministic parsers in C generated by Bison.  More recently, it was
+  a documented feature of Bison's experimental Java parsers.  As
+  promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
+  semantic action now produces a deprecation warning, and Java parsers
+  no longer implement YYFAIL at all.  For further details, including a
+  discussion of how to suppress C preprocessor warnings about YYFAIL
+  being unused, see the Bison 2.4.2 NEWS entry.
+
+** Temporary hack for adding a semicolon to the user action:

 
   Previously, Bison appended a semicolon to every user action for
   reductions when the output language defaulted to C (specifically, when
 
   Previously, Bison appended a semicolon to every user action for
   reductions when the output language defaulted to C (specifically, when


@@ -140,32 +253,92 @@ Bison News
   about a missing semicolon where it did not before.  Future releases of
   Bison will cease to append semicolons entirely.
 
   about a missing semicolon where it did not before.  Future releases of
   Bison will cease to append semicolons entirely.
 

-** Character literals not of length one.
-
-  Previously, Bison quietly converted all character literals to length
-  one.  For example, without warning, Bison interpreted the operators in
-  the following grammar to be the same token:
-
-    exp: exp '++'
-       | exp '+' exp
-       ;
-
-  Bison now warns when a character literal is not of length one.  In
-  some future release, Bison will report an error instead.
-
-** Verbose error messages fixed for nonassociative tokens.
-
-  When %error-verbose is specified, syntax error messages produced by
-  the generated parser include the unexpected token as well as a list of
-  expected tokens.  Previously, this list erroneously included tokens
-  that would actually induce a syntax error because conflicts for them
-  were resolved with %nonassoc.  Such tokens are now properly omitted
-  from the list.
-
-* Changes in version 2.4.2 (????-??-??):
+** Verbose syntax error message fixes:
+
+  When %error-verbose or the obsolete `#define YYERROR_VERBOSE' is
+  specified, syntax error messages produced by the generated parser
+  include the unexpected token as well as a list of expected tokens.
+  The effect of %nonassoc on these verbose messages has been corrected
+  in two ways, but a more complete fix requires LAC, described above:
+
+*** When %nonassoc is used, there can exist parser states that accept no
+    tokens, and so the parser does not always require a lookahead token
+    in order to detect a syntax error.  Because no unexpected token or
+    expected tokens can then be reported, the verbose syntax error
+    message described above is suppressed, and the parser instead
+    reports the simpler message, "syntax error".  Previously, this
+    suppression was sometimes erroneously triggered by %nonassoc when a
+    lookahead was actually required.  Now verbose messages are
+    suppressed only when all previous lookaheads have already been
+    shifted or discarded.
+
+*** Previously, the list of expected tokens erroneously included tokens
+    that would actually induce a syntax error because conflicts for them
+    were resolved with %nonassoc in the current parser state.  Such
+    tokens are now properly omitted from the list.
+
+*** Expected token lists are still often wrong due to state merging
+    (from LALR or IELR) and default reductions, which can both add
+    invalid tokens and subtract valid tokens.  Canonical LR almost
+    completely fixes this problem by eliminating state merging and
+    default reductions.  However, there is one minor problem left even
+    when using canonical LR and even after the fixes above.  That is,
+    if the resolution of a conflict with %nonassoc appears in a later
+    parser state than the one at which some syntax error is
+    discovered, the conflicted token is still erroneously included in
+    the expected token list.  Bison's new LAC implementation,
+    described above, eliminates this problem and the need for
+    canonical LR.  However, LAC is still experimental and is disabled
+    by default.
+
+** A location handling bug in the Java skeleton has been fixed.
+
+* Changes in version 2.4.3 (2010-08-05):
+
+** Bison now obeys -Werror and --warnings=error for warnings about
+   grammar rules that are useless in the parser due to conflicts.
+
+** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
+   been fixed.
+
+** Failures in the test suite for GCC 4.5 have been fixed.
+
+** Failures in the test suite for some versions of Sun Studio C++ have
+   been fixed.
+
+** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
+   warnings about undefined %prec identifiers will not be converted to
+   errors in Bison 2.5.  They will remain warnings, which should be
+   sufficient for POSIX while avoiding backward compatibility issues.
+
+** Minor documentation fixes.
+
+* Changes in version 2.4.2 (2010-03-20):
+
+** Some portability problems that resulted in failures and livelocks
+   in the test suite on some versions of at least Solaris, AIX, HP-UX,
+   RHEL4, and Tru64 have been addressed.  As a result, fatal Bison
+   errors should no longer cause M4 to report a broken pipe on the
+   affected platforms.
+
+** `%prec IDENTIFIER' requires IDENTIFIER to be defined separately.
+
+  POSIX specifies that an error be reported for any identifier that does
+  not appear on the LHS of a grammar rule and that is not defined by
+  %token, %left, %right, or %nonassoc.  Bison 2.3b and later lost this
+  error report for the case when an identifier appears only after a
+  %prec directive.  It is now restored.  However, for backward
+  compatibility with recent Bison releases, it is only a warning for
+  now.  In Bison 2.5 and later, it will return to being an error.
+  [Between the 2.4.2 and 2.4.3 releases, it was decided that this
+  warning will not be converted to an error in Bison 2.5.]

 
 ** Detection of GNU M4 1.4.6 or newer during configure is improved.
 
 
 ** Detection of GNU M4 1.4.6 or newer during configure is improved.
 

+** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
+   YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
+   avoided.
+

 ** %code is now a permanent feature.
 
   A traditional Yacc prologue directive is written in the form:
 ** %code is now a permanent feature.
 
   A traditional Yacc prologue directive is written in the form:


@@ -189,6 +362,45 @@ Bison News
   Bison's Java feature as a whole including its current usage of %code
   is still considered experimental.
 
   Bison's Java feature as a whole including its current usage of %code
   is still considered experimental.
 

+** YYFAIL is deprecated and will eventually be removed.
+
+  YYFAIL has existed for many years as an undocumented feature of
+  deterministic parsers in C generated by Bison.  Previously, it was
+  documented for Bison's experimental Java parsers.  YYFAIL is no longer
+  documented for Java parsers and is formally deprecated in both cases.
+  Users are strongly encouraged to migrate to YYERROR, which is
+  specified by POSIX.
+
+  Like YYERROR, you can invoke YYFAIL from a semantic action in order to
+  induce a syntax error.  The most obvious difference from YYERROR is
+  that YYFAIL will automatically invoke yyerror to report the syntax
+  error so that you don't have to.  However, there are several other
+  subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
+  inherent flaws when %error-verbose or `#define YYERROR_VERBOSE' is
+  used.  For a more detailed discussion, see:
+
+    http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
+
+  The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
+  deterministic parsers in C will continue to implement it.  However,
+  because YYFAIL is already flawed, it seems futile to try to make new
+  Bison features compatible with it.  Thus, during parser generation,
+  Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
+  rule action.  In a later release, YYFAIL will be disabled for
+  %error-verbose and `#define YYERROR_VERBOSE'.  Eventually, YYFAIL will
+  be removed altogether.
+
+  There exists at least one case where Bison 2.5's YYFAIL warning will
+  be a false positive.  Some projects add phony uses of YYFAIL and other
+  Bison-defined macros for the sole purpose of suppressing C
+  preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
+  To avoid Bison's future warning, such YYFAIL uses can be moved to the
+  epilogue (that is, after the second `%%') in the Bison input file.  In
+  this release (2.4.2), Bison already generates its own code to suppress
+  C preprocessor warnings for YYFAIL, so projects can remove their own
+  phony uses of YYFAIL if compatibility with Bison releases prior to
+  2.4.2 is not necessary.
+

 ** Internationalization.
 
   Fix a regression introduced in Bison 2.4: Under some circumstances,
 ** Internationalization.
 
   Fix a regression introduced in Bison 2.4: Under some circumstances,


@@ -1242,8 +1454,7 @@ End:
 
 -----
 
 
 -----
 

-Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
-2004, 2005, 2006, 2007, 2008, 2009 Free Software Foundation, Inc.
+Copyright (C) 1995-2011 Free Software Foundation, Inc.

 
 This file is part of Bison, the GNU Parser Generator.
 
 
 This file is part of Bison, the GNU Parser Generator.