Paul Eggert [Wed, 5 Oct 2005 06:39:08 +0000 (06:39 +0000)]
Undo the parts of the unlocked-I/O change that substituted
putc or puts for printf. This might hurt performance a bit,
but some people prefer the printf style.
* data/c.m4 (yysymprint): Prefer printf to puts and putc.
* data/glr.c (YYFPUTC, YYFPUTS, YYDPUTS): Remove.
All uses replaced by YYFPRINTF and YYDPRINTF.
* data/yacc.c: Likewise.
* lib/bitset.c (bitset_print): Likewise.
* lib/bitsetv.c (bitsetv_dump, debug-bitsetv): Prefer printf to
putc and puts.
* lib/lbitset.c (debug_lbitset): Likewise.
* src/closure.c (print_firsts, print_fderives): Likewise.
* src/gram.c (grammar_dump): Likewise.
* src/lalr.c (look_ahead_tokens_print): Likewise.
* src/output.c (escaped_output): Likewise.
(user_actions_output): Break apart two printfs.
* src/parse-gram.y (%printer): Prefer printf to putc and puts.
* src/reduce.c (reduce_print): Likewise.
* src/state.c (state_rule_look_ahead_tokens_print): Likewise.
* src/system.h: Include unlocked-io.h rathe than stdio.h.
Paul Eggert [Wed, 5 Oct 2005 06:05:22 +0000 (06:05 +0000)]
* data/glr.c (yyuserMerge, yyreportAmbiguity, yyreportSyntaxError):
Use assignments rather than casts-to-void to suppress
unused-variable warnings. This pacifies 'lint'.
* data/lalr1.cc (yysymprint_, yydestruct_): Use a call to suppress
unused-variable warnings.
Paul Eggert [Sun, 2 Oct 2005 21:24:12 +0000 (21:24 +0000)]
Use unlocked I/O for a minor performance improvement on hosts like
GNU/Linux and Solaris that support unlocked I/O. The basic idea
is to use the gnlib unlocked-io module, and to prefer putc and
puts to printf when either will work (since the latter doesn't
come in an unlocked flavor).
* bootstrap (gnulib_modules): Add unlocked-io.
* data/c.m4 (yysymprint): Prefer puts and putc to printf.
* data/glr.c (YYFPUTC, YYFPUTS, YYDPUTS): New macros.
Prefer them to YYFPRINTF and YYDPRINTF if either will do,
and similarly for puts and putc and printf.
* data/yacc.c: Likewise.
* lib/bitset.c (bitset_print): Likewise.
* lib/bitset.h [USE_UNLOCKED_IO]: Include unlocked-io.h.
* lib/bitsetv.c (bitsetv_dump, debug-bitsetv): Prefer putc and puts
to printf.
* lib/lbitset.c (debug_lbitset): Likewise.
* src/closure.c (print_firsts, print_fderives): Likewise.
* src/gram.c (grammar_dump): Likewise.
* src/lalr.c (look_ahead_tokens_print): Likewise.
* src/output.c (escaped_output): Likewise.
(user_actions_output): Coalesce two printfs.
* src/parse-gram.h (%printer): Prefer putc and puts to printf.
* src/reduce.c (reduce_print): Likewise.
* src/state.c (state_rule_look_ahead_tokens_print): Likewise.
* src/system.h: Include unlocked-io.h rathe than stdio.h.
Akim Demaille [Sun, 2 Oct 2005 16:34:02 +0000 (16:34 +0000)]
* data/location.cc: New, extract from...
* data/lalr1.cc: here.
(location.hh): Include it after the user prologue, in case the
filename type is defined by the user.
Forward declation location and position before the pre-prologue.
(yyresult_): Rename as...
(yyresult): this, it's a local variable, not an attribute.
* data/Makefile.am (dist_pkgdata_DATA): Adjust.
Alexandre Duret-Lutz <adl@gnu.org>
Move the token type and YYSTYPE in the parser class.
* data/lalr1.cc (stack.hh, location.hh): Include earlier.
(parser::token): New, from the moved free definition of tokens.
(parser::semantic_value): Now a full definition instead of an
indirection to YYSTYPE.
(b4_post_prologue): No longer included in the header file, but
in the implementation file.
* doc/bison.texi (C+ Language Interface): Update.
* src/parse-gram.y: Support unary %define.
* tests/actions.at: Define global_tokens_and_yystype for backward
compatibility until we update the tests.
* tests/calc.at: Idem.
(first_line, first_column, last_line, last_column): Define for lalr1.cc
to simplify the code.
* data/c.m4 (b4_error_verbose_if): New.
* data/lalr1.cc: Use it.
(YYERROR_VERBOSE_IF): Remove.
(yyn_, yylen_, yystate_, yynerrs_, yyerrstatus_): Remove as
parser members, replaced by...
(yyn, yylen, yystate, yynerss, yyerrstatus): these parser::parse
local variables.
(yysyntax_error_): Takes the state number as argument.
(yyreduce_print_): Use the argument yyrule, not the former
attribute yyn_.
Paul Eggert [Tue, 27 Sep 2005 06:08:27 +0000 (06:08 +0000)]
* bootstrap (gnulib_modules): Add verify.
* lib/.cvsignore: Add verify.h.
* src/getargs.c: Use ARGMATCH_VERIFY rather than verify.
* src/system.h (verify): Remove.
Include verify.h instead.
* src/tables.c (tables_generate): Use new API for 'verify'.
Paul Eggert [Wed, 21 Sep 2005 19:54:48 +0000 (19:54 +0000)]
More alloca cleanups.
* bootstrap (gnulib_modules): Remove alloca. Bison doesn't need
it itself.
* src/main.c (main) [C_ALLOCA]: Don't flush alloca'ed memory; we
don't use alloca any more.
* data/yacc.c [YYSTACK_USE_ALLOCA && !defined __GNUC__ && ! defined
__BUILTIN_VA_ARG_INCR && ! defined _AIX && ! defined _MSC_VER &&
(defined alloca || defined _ALLOCA_H)]: Don't include <stdlib.h>;
not needed in this case.
* tests/torture.at (Exploding the Stack Size with Alloca): Adjust
to match yacc.c, to test more hosts.
Paul Eggert [Tue, 20 Sep 2005 23:06:10 +0000 (23:06 +0000)]
* data/yacc.c (YYSIZE_T): Reindent to make it clearer. This
doesn't affect behavior.
(YYSTACK_ALLOC) [YYSTACK_USE_ALLOCA]: Improve support for
Solaris, AIX, MSC.
(_STDLIB_H): Renamed from YYINCLUDED_STDLIB_H. All uses changed.
This works a bit better with glibc, if user code has already included
stdlib.h.
* doc/bison.texinfo (Bison Parser): Document that users can't
arbitrarily use malloc and free for other purposes. Document
that <alloca.h> and <malloc.h> might be included.
(Table of Symbols): Under YYSTACK_USE_ALLOCA, Don't claim that the
user must declare alloca.
Paul Eggert [Tue, 20 Sep 2005 06:06:42 +0000 (06:06 +0000)]
(YYSIZE_MAXIMUM): New macro.
(YYSTACK_ALLOC_MAXIMUM): Use it.
(yysyntax_error): New function.
(yyparse) [YYERROR_VERBOSE]: Don't leak memory indefinitely if
multiple syntax errors are reported, and alloca is being used.
Instead, reallocate buffers twice as big each time, so that
we waste at most half the allocated memory. Start with a small
(128-byte) buffer that will suffice in most cases anyway.
Use yysyntax_error to do most of the work.
Paul Eggert [Mon, 19 Sep 2005 21:08:21 +0000 (21:08 +0000)]
* tests/glr-regression.at (Duplicated user destructor for lookahead):
Mark it as expected to fail.
Cast result of malloc; problem reported by twlevo@xs4all.nl.
* tests/actions.at, tests/calc.at, tests/glr-regression.at:
Don't start user-code symbols with "yy", to avoid name space problems.
Remove the traits, failed experiment.
It never proved useful, and anyway because of the current
definition, it was not possible to have several specialization of
this traits, making it useless.
* data/lalr1.cc (yy:traits): Remove.
Inline its definitions in the parser class.
Paul Eggert [Fri, 16 Sep 2005 22:54:21 +0000 (22:54 +0000)]
* NEWS: Version 2.1.
* NEWS: Remove notice of yytname change, since it was never in an
official release.
* data/glr.c (yydestroyGLRState): Rename local var to avoid shadowing
diagnostic.
* src/output.c (prepare): Likewise.
* data/lalr1.cc (YYERROR_VERBOSE_IF): New macro.
(yysyntax_error_): Use it to avoid GCC warning when YYERROR_VERBOSE
is not defined. This is an awful hack, but it's enough for now.
All callers changed.
* tests/glr-regression-at (make_value): Args are const pointers now,
to avoid GCC warning.
(Duplicated user destructor for lookahead): New test. Currently
skipped. It fails on my host but I'm not sure it'll always fail.
* src/symtab.h (struct symbol): Declare the printer and destructor
as const, to avoid accidental calls to free.
(symbol_destructor_set, symbol_printer_set): Adjust.
* src/symtab.c: Adjust.
* data/lalr1.cc (yylex_): Remove, inline its code.
(yyreport_syntax_error_): Remove, replaced by...
(yysyntax_error_): this which returns a string and leaves to the
caller the call to the users' error function.
(yylooka_, yyilooka_, yylval, yylloc, yyerror_range_, yyval, yyloc):
Move from members of the parser object...
(yylooka, yyilooka, yylval, yylloc, yyerror_range, yyval, yyloc):
to local variables of the parse function.
Paul Eggert [Thu, 15 Sep 2005 14:06:52 +0000 (14:06 +0000)]
* NEWS: Document when yyparse started to return 2.
* doc/bison.texinfo (Parser Function): Document when yyparse
returns 2.
* data/lalr1.cc: Revert part of previous change, as it's incompatible.
(b4_filename_type): Renamed back from b4_file_name_type. All uses
changed.
(class position): file_name -> filename (reverting). All uses changed.
Paul Eggert [Mon, 12 Sep 2005 05:50:44 +0000 (05:50 +0000)]
[defined YYSETJMP]: Don't include <setjmp.h>.
(YYJMP_BUF, YYSETJMP, YYLONGJMP) [!defined YYSETJMP]: New macros.
All uses of jmp_buf, setjmp, longjmp changed to use these instead.
(yyparse): Abort if user code uses longjmp to throw an unexpected
value.
Paul Eggert [Fri, 9 Sep 2005 22:14:28 +0000 (22:14 +0000)]
Don't use "filename", as per GNU coding standards. Use "file name" or
"file" or "name", depending on the context.
(Invocation): The output of "bison hack/foo.y" goes to foo.tab.c,
not to hack/foo.tab.c.
(Calc++ Top Level): 2nd arg of main is not const.
Paul Eggert [Fri, 26 Aug 2005 20:16:16 +0000 (20:16 +0000)]
* data/glr.c (yydestroyGLRState): Renamed from yydestroyStackItem.
All uses changed. Invoke user destructor after an error during a
split parse (trivial change from Joel E. Denny).
* tests/glr-regression.at
(User destructor after an error during a split parse): New test case.
Problem reported by Joel E. Denny in:
http://lists.gnu.org/archive/html/bison-patches/2005-08/msg00029.html
Paul Eggert [Thu, 25 Aug 2005 19:06:24 +0000 (19:06 +0000)]
* README-cvs: Give URLs for recommended tools.
Mention Gzip version problem, and bootstrapping issues.
Remove troubleshooting section, as it's somewhat obsolete.
Paul Eggert [Thu, 25 Aug 2005 17:57:23 +0000 (17:57 +0000)]
(yydestroyStackItem): New function.
(yyrecoverSyntaxError, yyreturn): Use it to improve quality of
debugging information. Problem reported by Joel E. Denny.
Paul Eggert [Thu, 25 Aug 2005 06:11:35 +0000 (06:11 +0000)]
* data/glr.c (yyrecoverSyntaxError, yyreturn):
Don't invoke destructor on unresolved entries.
* tests/glr-regression.at
(User destructor for unresolved GLR semantic value): New test case.
Problem reported by Joel E. Denny in:
http://lists.gnu.org/archive/html/bison-patches/2005-08/msg00016.html
Paul Eggert [Mon, 22 Aug 2005 02:31:10 +0000 (02:31 +0000)]
* data/glr.c (yyreturn): Use "Cleanup:" rather than "Error:"
in yydestruct diagnostic, since it might not be an error.
Problem reported by Joel Denny near end of
<http://lists.gnu.org/archive/html/help-bison/2005-07/msg00040.html>.
* data/lalr1cc (yyerturn): Likewise.
* data/yacc.c (yyreturn): Likewise.
* tests/calc.at (_AT_CHECK_CALC_ERROR): Adjust to the above change.
Paul Eggert [Mon, 22 Aug 2005 01:05:52 +0000 (01:05 +0000)]
(YY_SYMBOL_PRINT): Append a newline, for consistency
with the other templates, and to fix bogus run-on messages such
as the one reported at the end of
<http://lists.gnu.org/archive/html/help-bison/2005-07/msg00040.html>.
All callers changed to avoid the newline.
(yyprocessOneStack): Output two lines rather than one, to accommodate
the above change. This changes the debug output format slightly.
Paul Eggert [Sun, 21 Aug 2005 23:43:56 +0000 (23:43 +0000)]
* data/glr.c (yyresolveValue): Fix redundant parse tree problem
reported by Joel E. Denny in
<http://lists.gnu.org/archive/html/bison-patches/2005-08/msg00004.html>
(trivial change).
* tests/glr-regression.at (Duplicate representation of merged trees):
New test, from Joel E. Denny in:
<http://lists.gnu.org/archive/html/help-bison/2005-07/msg00013.html>.
* THANKS: Add Joel E. Denny.
Paul Eggert [Mon, 25 Jul 2005 06:36:13 +0000 (06:36 +0000)]
* tests/regression.at (Token definitions): Don't rely on
AT_PARSER_CHECK for data that contains backslashes. It currently
uses 'echo', and 'echo' isn't portable if its argument contains
backslashes. Problem found on OpenBSD 3.4. Also, do not assume
that the byte '\0xff' is not printable in the C locale; it is,
under OpenBSD 3.4 (!). Luckily, '\0x80' through '\0x9e' are
not printable, so use '\0x81' to test.
Paul Eggert [Mon, 25 Jul 2005 03:38:42 +0000 (03:38 +0000)]
Fix core dump reported by Pablo De Napoli in
<http://lists.gnu.org/archive/html/bug-bison/2005-07/msg00053.html>.
* tests/regression.at (Invalid inputs with {}): New test.
* src/parse-gram.y (token_name): Translate type before using
it as an index.
Paul Eggert [Sun, 24 Jul 2005 07:24:22 +0000 (07:24 +0000)]
* NEWS: Version 2.0b.
* data/glr.c (ATTRIBUTE_UNUSED): Remove, since it infringes on
the user's name space. All uses changed to __attribute__
((__unused__)).
(yyFail, yyMemoryExhausted, yyreportAmbiguity):
Add __attribute__ ((__noreturn__)).
* etc/clcommit: Remove. We weren't using it, and it failed
"make maintainer-distcheck".
* Makefile.maint: Merge from coreutils.
(CVS_LIST, CVS_LIST_EXCEPT): New macros.
(syntax-check-rules): Change list of rules as described below.
(sc_cast_of_alloca_return_value, sc_dd_max_sym_length):
(sc_file_system, sc_obsolete_symbols, sc_prohibit_atoi_atof):
(sc_prohibit_jm_in_m4, sc_root_tests, sc_tight_scope):
(sc_trailing_space): New rules.
(sc_xalloc_h_in_src): Remove.
(sc_cast_of_argument_to_free, sc_cast_of_x_alloc_return_value):
(sc_space_tab, sc_error_exit_success, sc_changelog):
(sc_system_h_headers, sc_sun_os_names, sc_unmarked_diagnostics):
(makefile-check, po-check, author_mark_check):
(makefile_path_separator_check, copyright-check):
Use grep -n, to make it easier to find violations.
Use CVS_LIST and CVS_LIST_EXCEPT.
(header_regexp, h_re): Remove.
(dd_c): New macro.
(sc_dd_max_sym_length, .re-list, news-date-check): New rules.
(my-distcheck): Use more-modern GCC flags.
(signatures, %.asc): Remove.
(rel-files, announcement): Remove signatures.
Restore old updating code, even though we don't use it, so
that we're the same as coreutils.
(alpha, beta, major): Depend on news-date-check.
Make the upload commands.
* data/m4sugar/m4sugar.m4: Sync from Autoconf, except that
we comment out the Autoconf version number.
* doc/bison.texinfo (Calc++ Scanner): Don't use atoi, as
it's error-prone and "make maintainer-distcheck" rejects it.
* lib/subpipe.c: Include <fcntl.h> without checking for HAVE_FCNTL_H.
Indent calls to "error" to pacify "make maintainer-distcheck",
when the calls are not intended to be translated.
* m4/subpipe.m4 (BISON_PREREQ_SUBPIPE): Don't check for fcntl.h.
* src/Makefile.am (DEFS): Use +=, to pacify
"make maintainer-distcheck".
(bison_SOURCES): Add scan-skel.h.
(sc_tight_scope): New rule, from coreutils.
* src/files.c (src_extension, header_extension):
Now static, not extern.
* src/getargs.c (short_options): Likewise.
* src/muscle_tab.c (muscle_table): Likewise.
* src/parse-gram.y (current_class, current_type, current_prec):
Likewise.
* src/reader.c (grammar_end, previous_rule_end): Likewise.
* src/getargs.h: Redo comments to pacify "make maintainer-distcheck".
* src/main.c (main): Cast bindtextdomain and textdomain calls to
void, to avoid warning when NLS is disabled.
* src/output.c: Include scan-skel.h.
(scan_skel): Remove decl, since scan-skel.h does this.
(output_skeleton):
Indent calls to "error" to pacify "make maintainer-distcheck".
* src/print_graph.c: Don't include <obstack.h>, as system.h does this.
* src/reader.h (gram_end, gram_lineno): New decls to pacify
"make maintainer-distcheck".
* src/scan-skel.l (skel_lex, skel_get_lineno, skel_get_in):
(skel_get_out, skel_get_leng, skel_get_text, skel_set_lineno):
(skel_set_in, skel_set_out, skel_get_debug, skel_set_debug):
(skel_lex_destroy, scan_skel): Move these decls to...
* src/scan-skel.h: New file.
* src/uniqstr.c (uniqstr_assert):
Indent calls to "error" to pacify "make maintainer-distcheck".
* tests/Makefile.am ($(srcdir)/package.m4): Use $(VAR),
not @VAR@.
* tests/torture.at: Revamp to avoid misuse of atoi that
"make maintainer-distcheck" complained about.
Paul Eggert [Sun, 24 Jul 2005 07:17:06 +0000 (07:17 +0000)]
(ATTRIBUTE_UNUSED): Remove, since it infringes on
the user's name space. All uses changed to __attribute__
((__unused__)).
(yyFail, yyMemoryExhausted, yyreportAmbiguity):
Add __attribute__ ((__noreturn__)).
Paul Eggert [Sat, 23 Jul 2005 17:06:41 +0000 (17:06 +0000)]
(message): Don't print a message more than once,
and omit line-number decoration that makes Emacs compile think
that informative messages are worth worrying about.