]> git.saurik.com Git - bison.git/blobdiff - ChangeLog
* src/scan-skel.l (<<EOF>>): Don't close standard output.
[bison.git] / ChangeLog
index 11f6835a8c5e2bbe39be49e90d4650f8b46c8cf5..5121a98de14c550e07a42007ea29e4d26e93f28f 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,184 @@
+2005-01-02  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * src/scan-skel.l (<<EOF>>): Don't close standard output.
+       Problem reported by Hans Aberg.
+
+2005-01-01  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * src/getargs.c (version): Happy new year; update overall
+       program copyright date from 2004 to 2005.
+
+       * src/scan-skel.l ("@output ".*\n): Don't close standard output.
+       Problem reported by Hans Aberg.
+       * tests/output.at (AT_CHECK_OUTPUT): New arg SHELLIO.
+       (Output file names.): Add a test for the case when standard output
+       is closed.
+
+2004-12-26  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * doc/bison.texinfo (@copying): Update FDL version number to 1.2,
+       to fix an oversight in the Bison 2.0 manual.
+
+2004-12-25  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * NEWS: Version 2.0.  Reformat the existing news items since
+       1.875, so that related items are grouped together.
+       * configure.ac (AC_INIT): Bump version to 2.0.
+       * src/parse-gram.c, src/parse-gram.h: Regenerate with 2.0.
+
+       * tests/torture.at (Exploding the Stack Size with Alloca): Set
+       YYSTACK_USE_ALLOCA to 1 if __GNUC__ or alloca are defined;
+       otherwise, we're not testing alloca.  Unfortunately there's no
+       simple way to consult HAVE_ALLOCA here.
+
+       * data/lalr1.cc (yydestruct_): Pacify unused variable warning
+       for yymsg, too.
+
+       * src/LR0.c (new_itemsets): Use memset rather than zeroing by
+       hand.  This avoids a warning about comparing int to size_t when
+       GCC warnings are enabled.
+
+2004-12-22  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * NEWS: Bison-generated parsers no longer default to using the
+       alloca function (when available) to extend the parser stack, due
+       to widespread problems in unchecked stack-overflow detection.
+       * data/glr.c (YYMAXDEPTH): Remove undef when zero.  It's the user's
+       responsibility to set it to a positive value.  This lets the user
+       specify a value that is not a preprocessor constant.
+       * data/yacc.c (YYMAXDEPTH): Likewise.
+       (YYSTACK_ALLOC): Define only if YYSTACK_USE_ALLOCA is nonzero.
+       * doc/bison.texinfo (Stack Overflow): YYMAXDEPTH no longer needs
+       to be a compile-time constant.  However, explain the constraints on it.
+       Also, explain the constraints on YYINITDEPTH.
+       (Table of Symbols): Explain that alloca is no longer the default.
+       Explain the user's responsibility if they define YYSTACK_USE_ALLOCA
+       to 1.
+
+       * doc/bison.texinfo (Location Default Action): Mention that n must
+       be zero when k is zero.  Suggested by Frank Heckenbach.
+
+2004-12-22  Akim Demaille  <akim@epita.fr>
+
+       * data/lalr1.cc (parser::token_number_type, parser::rhs_number_type)
+       (parser::state_type, parser::semantic_type, parser::location_type):
+       Private, not public.
+       (parser::parse): Return ints, not bool.
+       Returning a bool introduces a problem: 0 corresponds to false, and
+       it seems weird to return false on success.  Returning true changes
+       the conventions for yyparse.
+       Alternatively we could return void and send an exception.
+       There is no clear consensus (yet?).
+       (state_stack, semantic_stack, location_stack): Rename as...
+       (state_stack_type, semantic_stack_type, location_stack_type): these.
+       Private, not public.
+       * tests/c++.at: New.
+       * tests/testsuite.at, tests/Makefile.am: Adjust.
+
+2004-12-21  Akim Demaille  <akim@epita.fr>
+
+       * data/lalr1.cc (parser::parse): Return a bool instead of an int.
+
+2004-12-21  Akim Demaille  <akim@epita.fr>
+
+       Don't impose std::string for filenames.
+
+       * data/lalr1.cc (b4_filename_type): New.
+       (position::filename): Use it.
+       (parser.hh): Move the inclusion of stack.hh and location.hh below
+       the user code, so that needed headers for the filename type can be
+       included first.
+       Forward declare them before the user code.
+       * tests/Makefile.am (check-local, installcheck-local): Pass
+       TESTSUITEFLAGS to the TESTSUITE.
+
+2004-12-20  Akim Demaille  <akim@epita.fr>
+
+       Use more STL like names: my_class instead of MyClass.
+
+       * data/lalr1.cc (LocationStack, LocationType, RhsNumberType)
+       (SemanticStack, SemanticType, StateStack, StateType)
+       (TokenNumberType, Stack, Slice, Traits, Parser::location)
+       (Parser::value): Rename as...
+       (location_stack, location_type, rhs_number_type, semantic_stack)
+       (semantic_type, state_stack, state_type, token_number_type, stack)
+       (slice, traits, parser::yylloc, parser::yylval): these.
+
+       * tests/calc.at, tests/regression.at, tests/actions.at: Adjust.
+
+2004-12-19  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * data/glr.c (YYLLOC_DEFAULT): Use GNU spacing conventions.
+       * data/yacc.c (YYLLOC_DEFAULT): Likewise.
+
+2004-12-17  Paul Eggert  <eggert@cs.ucla.edu>
+
+       Remove uses of 'short int' and 'unsigned short int'.  This raises
+       some arbitrary limits.  It uses more memory but nowadays that's
+       not much of an issue.
+
+       This change does not affect the generated parsers; that's a different
+       task, as some users will want to conserve memory there.
+
+       Ideally we should use size_t to represent all object counts, and
+       something like ptrdiff_t to represent signed differences of object
+       counts; but that will require more code-cleanup than I have the
+       time to do right now.
+
+       * src/LR0.c (allocate_itemsets, new_itemsets, save_reductions):
+       Use size_t, not int or short int, to count objects.
+       * src/closure.c (nritemset, closure): Likewise.
+       * src/closure.h (nritemset, closure): Likewise.
+       * src/nullable.c (nullable_compute): Likewise.
+       * src/print.c (print_core): Likewise.
+       * src/print_graph.c (print_core): Likewise.
+       * src/state.c (state_compare, state_hash): Likewise.
+       * src/state.h (struct state): Likewise.
+       * src/tables.c (default_goto, goto_actions): Likewise.
+
+       * src/gram.h (rule_number, rule): Use int, not short int.
+       * src/output.c (prepare_rules): Likewise.
+       * src/state.h (state_number, STATE_NUMBER_MAXIMUM, transitions,
+       errs, reductions): Likewise.
+       * src/symtab.h (symbol_number, SYMBOL_NUMBER_MAXIMUM, struct symbol):
+       Likewise.
+       * src/tables.c (vector_number, tally, action_number,
+       ACTION_NUMBER_MINIMUM): Likewise.
+       * src/output.c (muscle_insert_short_int_table): Remove.
+
+2004-12-17  Akim Demaille  <akim@epita.fr>
+
+       * data/lalr1.cc: Extensive Doxygenation.
+       (error_): Rename as...
+       (error): this, since it is visible to the user.
+       Adjust callers.
+       (Parser::message): Now an automatic variable from...
+       (Parser::yyreport_syntax_error_): here.
+       * tests/actions.at, tests/calc.at, tests/regression.at: Adjust to
+       Parser::error.
+       * tests/input.at: Escape $.
+
+2004-12-16  Paul Eggert  <eggert@cs.ucla.edu>
+
+       * data/glr.c (b4_lhs_value, b4_rhs_value, b4_rhs-location):
+       Parenthesize rhs to avoid obscure problems with mistakes like
+       "foo$$bar = foo$1bar;".  Problem reported by twlevo at xs4all.
+       * data/lalr1.cc (b4_lhs_value, b4_rhs_value, b4_lhs_location,
+       b4_rhs_location): Likewise.
+       * data/yacc.c (b4_lhs_value, b4_rhs_value, b4_lhs_location,
+       b4_rhs_location): Likewise.
+
+2004-12-16  Akim Demaille  <akim@epita.fr>
+
+       * data/lalr1.cc (yyreport_syntax_error_): Catch up with glr.c and
+       yacc.c: be sure to stay within yycheck_.
+       * tests/actions.at: Re-enable C++ tests.
+
+2004-12-16  Akim Demaille  <akim@epita.fr>
+
+       * src/print_graph.c (print_graph): Remove layoutalgorithm uses for
+       real.
+
 2004-12-16  Akim Demaille  <akim@epita.fr>
 
        Use #define to handle the %name-prefix.