newsBison - News: Bison 3.5 released [stable]

 
 
Latest News
Bison 3.5 released [stable] posted by akim, Wed 01 Jan 2020 09:37:26 AM UTC - 0 replies
Bison 3.4.2 released [stable] posted by akim, Sat 14 Sep 2019 07:01:33 AM UTC - 0 replies
Bison 3.4 released [stable] posted by akim, Sun 19 May 2019 10:01:36 AM UTC - 0 replies
Bison 3.3 released [stable] posted by akim, Sat 26 Jan 2019 02:15:08 PM UTC - 0 replies
Bison 3.2.2 released [stable] posted by akim, Wed 21 Nov 2018 08:32:46 PM UTC - 0 replies
[Submit News]
[18 news in archive]

Bison 3.5 released [stable]

Item posted by Akim Demaille <akim> on Wed 01 Jan 2020 09:37:26 AM UTC.

We are very happy to announce the release of Bison 3.5, the best release
ever of Bison!  Better than 3.4, although it was a big improvement over 3.3,
which was huge upgrade compared to 3.2, itself way ahead Bison 3.1.  Ethic
demands that we don't mention 3.0.  Rumor has it that Bison 3.5 is not as
good as 3.6 will be though...

Paul Eggert revised the use of integral types in both the generator and the
generated parsers.  As a consequence small parsers have a smaller footprint,
and very large automata are now possible with the default back-end (yacc.c).
If you are interested in smaller parsers, also have a look at api.token.raw.

Adrian Vogelsgesang contributed lookahead correction for C++.

The purpose of string literals has been clarified.  Indeed, they are used
for two different purposes: freeing from having to implement the keyword
matching in the scanner, and improving error messages.  Most of the time
both can be achieved at the same time, but on occasions, it does not work so
well.  We promote their use for error messages.  We still support the former
case (at least for historical skeletons), but it is not a recommended
practice.  The documentation now warns against this use.  A new warning,
-Wdangling-alias, should help users who want to enforce the use of aliases
only for error messages.

An experimental back-end for the D programming language was added thanks to
Oliver Mangold and H. S. Teoh.  It is looking for active support from the D
community.

Happy parsing!

==================================================================

Bison is a general-purpose parser generator that converts an annotated
context-free grammar into a deterministic LR or generalized LR (GLR) parser
employing LALR(1) parser tables.  Bison can also generate IELR(1) or
canonical LR(1) parser tables.  Once you are proficient with Bison, you can
use it to develop a wide range of language parsers, from those used in
simple desk calculators to complex programming languages.

Bison is upward compatible with Yacc: all properly-written Yacc grammars
work with Bison with no change.  Anyone familiar with Yacc should be able to
use Bison with little trouble.  You need to be fluent in C, C++ or Java
programming in order to use Bison.

Here is the GNU Bison home page:
   https://gnu.org/software/bison/

==================================================================

Here are the compressed sources:
  https://ftp.gnu.org/gnu/bison/bison-3.5.tar.gz   (5.1MB)
  https://ftp.gnu.org/gnu/bison/bison-3.5.tar.xz   (3.1MB)

Here are the GPG detached signatures[*]:
  https://ftp.gnu.org/gnu/bison/bison-3.5.tar.gz.sig
  https://ftp.gnu.org/gnu/bison/bison-3.5.tar.xz.sig

Use a mirror for higher download bandwidth:
  https://www.gnu.org/order/ftp.html

[*] Use a .sig file to verify that the corresponding file (without the
.sig suffix) is intact.  First, be sure to download both the .sig file
and the corresponding tarball.  Then, run a command like this:

  gpg --verify bison-3.5.tar.gz.sig

If that command fails because you don't have the required public key,
then run this command to import it:

  gpg --keyserver keys.gnupg.net --recv-keys 0DDCAA3278D5264E

and rerun the 'gpg --verify' command.

This release was bootstrapped with the following tools:
  Autoconf 2.69
  Automake 1.16.1
  Flex 2.6.4
  Gettext 0.19.8.1
  Gnulib v0.1-2971-gb943dd664

==================================================================

* Noteworthy changes in release 3.5 (2019-12-11) [stable]

** Backward incompatible changes

  Lone carriage-return characters (aka \r or ^M) in the grammar files are no
  longer treated as end-of-lines.  This changes the diagnostics, and in
  particular their locations.

  In C++, line numbers and columns are now represented as 'int' not
  'unsigned', so that integer overflow on positions is easily checkable via
  'gcc -fsanitize=undefined' and the like.  This affects the API for
  positions.  The default position and location classes now expose
  'counter_type' (int), used to define line and column numbers.

** Deprecated features

  The YYPRINT macro, which works only with yacc.c and only for tokens, was
  obsoleted long ago by %printer, introduced in Bison 1.50 (November 2002).
  It is deprecated and its support will be removed eventually.

** New features

*** Lookahead correction in C++

  Contributed by Adrian Vogelsgesang.

  The C++ deterministic skeleton (lalr1.cc) now supports LAC, via the
  %define variable parse.lac.

*** Variable api.token.raw: Optimized token numbers (all skeletons)

  In the generated parsers, tokens have two numbers: the "external" token
  number as returned by yylex (which starts at 257), and the "internal"
  symbol number (which starts at 3).  Each time yylex is called, a table
  lookup maps the external token number to the internal symbol number.

  When the %define variable api.token.raw is set, tokens are assigned their
  internal number, which saves one table lookup per token, and also saves
  the generation of the mapping table.

  The gain is typically moderate, but in extreme cases (very simple user
  actions), a 10% improvement can be observed.

*** Generated parsers use better types for states

  Stacks now use the best integral type for state numbers, instead of always
  using 15 bits.  As a result "small" parsers now have a smaller memory
  footprint (they use 8 bits), and there is support for large automata (16
  bits), and extra large (using int, i.e., typically 31 bits).

*** Generated parsers prefer signed integer types

  Bison skeletons now prefer signed to unsigned integer types when either
  will do, as the signed types are less error-prone and allow for better
  checking with 'gcc -fsanitize=undefined'.  Also, the types chosen are now
  portable to unusual machines where char, short and int are all the same
  width.  On non-GNU platforms this may entail including <limits.h> and (if
  available) <stdint.h> to define integer types and constants.

*** A skeleton for the D programming language

  For the last few releases, Bison has shipped a stealth experimental
  skeleton: lalr1.d.  It was first contributed by Oliver Mangold, based on
  Paolo Bonzini's lalr1.java, and was cleaned and improved thanks to
  H. S. Teoh.

  However, because nobody has committed to improving, testing, and
  documenting this skeleton, it is not clear that it will be supported in
  the future.

  The lalr1.d skeleton *is functional*, and works well, as demonstrated in
  examples/d/calc.d.  Please try it, enjoy it, and... commit to support it.

*** Debug traces in Java

  The Java backend no longer emits code and data for parser tracing if the
  %define variable parse.trace is not defined.

** Diagnostics

*** New diagnostic: -Wdangling-alias

  String literals, which allow for better error messages, are (too)
  liberally accepted by Bison, which might result in silent errors.  For
  instance

    %type <exVal> cond "condition"

  does not define "condition" as a string alias to 'cond' (nonterminal
  symbols do not have string aliases).  It is rather equivalent to

    %nterm <exVal> cond
    %token <exVal> "condition"

  i.e., it gives the type 'exVal' to the "condition" token, which was
  clearly not the intention.

  Also, because string aliases need not be defined, typos such as "baz"
  instead of "bar" will be not reported.

  The option -Wdangling-alias catches these situations.  On

    %token BAR "bar"
    %type <ival> foo "foo"
    %%
    foo: "baz" {}

  bison -Wdangling-alias reports

    warning: string literal not attached to a symbol
          | %type <ival> foo "foo"
          |                  ^~~~~
    warning: string literal not attached to a symbol
          | foo: "baz" {}
          |      ^~~~~

   The -Wall option does not (yet?) include -Wdangling-alias.

*** Better POSIX Yacc compatibility diagnostics

  POSIX Yacc restricts %type to nonterminals.  This is now diagnosed by
  -Wyacc.

    %token TOKEN1
    %type  <ival> TOKEN1 TOKEN2 't'
    %token TOKEN2
    %%
    expr:

  gives with -Wyacc

    input.y:2.15-20: warning: POSIX yacc reserves %type to nonterminals [-Wyacc]
        2 | %type  <ival> TOKEN1 TOKEN2 't'
          |               ^~~~~~
    input.y:2.29-31: warning: POSIX yacc reserves %type to nonterminals [-Wyacc]
        2 | %type  <ival> TOKEN1 TOKEN2 't'
          |                             ^~~
    input.y:2.22-27: warning: POSIX yacc reserves %type to nonterminals [-Wyacc]
        2 | %type  <ival> TOKEN1 TOKEN2 't'
          |                      ^~~~~~

*** Diagnostics with insertion

  The diagnostics now display the suggestion below the underlined source.
  Replacement for undeclared symbols are now also suggested.

    $ cat /tmp/foo.y
    %%
    list: lis '.' |

    $ bison -Wall foo.y
    foo.y:2.7-9: error: symbol 'lis' is used, but is not defined as a token and has no rules; did you mean 'list'?
        2 | list: lis '.' |
          |       ^~~
          |       list
    foo.y:2.16: warning: empty rule without %empty [-Wempty-rule]
        2 | list: lis '.' |
          |                ^
          |                %empty
    foo.y: warning: fix-its can be applied.  Rerun with option '--update'. [-Wother]

*** Diagnostics about long lines

  Quoted sources may now be truncated to fit the screen.  For instance, on a
  30-column wide terminal:

    $ cat foo.y
    %token FOO                       FOO                         FOO
    %%
    exp: FOO
    $ bison foo.y
    foo.y:1.34-36: warning: symbol FOO redeclared [-Wother]
        1 | …         FOO                  …
          |           ^~~
    foo.y:1.8-10:      previous declaration
        1 | %token FOO                     …
          |        ^~~
    foo.y:1.62-64: warning: symbol FOO redeclared [-Wother]
        1 | …         FOO
          |           ^~~
    foo.y:1.8-10:      previous declaration
        1 | %token FOO                     …
          |        ^~~

** Changes

*** Debugging glr.c and glr.cc

  The glr.c skeleton always had asserts to check its own behavior (not the
  user's).  These assertions are now under the control of the parse.assert
  %define variable (disabled by default).

*** Clean up

  Several new compiler warnings in the generated output have been avoided.
  Some unused features are no longer emitted.  Cleaner generated code in
  general.

** Bug Fixes

  Portability issues in the test suite.

  In theory, parsers using %nonassoc could crash when reporting verbose
  error messages. This unlikely bug has been fixed.

  In Java, %define api.prefix was ignored.  It now behaves as expected.

No messages in Bison 3.5 released [stable]

 

Back to the top


Powered by Savane 3.5