Akim Demaille [Sat, 15 Jun 2002 18:24:08 +0000 (18:24 +0000)]
* src/conflicts.c (log_resolution): Accept the rule involved in
the sr conflicts instead of the lookahead number that points to
that rule.
(flush_reduce): Accept the current lookahead vector as argument,
instead of the index in LA.
(resolve_sr_conflict): Accept the current number of lookahead
bitset to consider for the STATE, instead of the index in LA.
(set_conflicts): Adjust.
* src/lalr.c, src/lalr.h, src/state.h: Comment changes.
Akim Demaille [Sat, 15 Jun 2002 18:23:12 +0000 (18:23 +0000)]
Copy BYacc's nice way to report the grammar.
* src/gram.h, src/gram.c (grammar_rhs_print, grammar_rules_print):
New.
Don't print the rules' location, it is confusing and useless.
(rule_print): Use grammar_rhs_print.
* src/print.c (print_grammar): Use grammar_rules_print.
Akim Demaille [Sat, 15 Jun 2002 18:21:46 +0000 (18:21 +0000)]
Let symbols have a location.
* src/symtab.h, src/symtab.c (symbol_t): Location is a new member.
(getsym): Adjust.
Adjust all callers.
* src/complain.h, src/complain.c (complain_at, fatal_at, warn_at):
Use location_t, not int.
* src/symtab.c (symbol_check_defined): Take advantage of the
location.
* tests/regression.at (Invalid inputs): Adjust.
Akim Demaille [Sat, 15 Jun 2002 18:21:11 +0000 (18:21 +0000)]
* src/parse-gram.y (YYLLOC_DEFAULT, current_lhs_location): New.
(input): Don't try to initialize yylloc here, do it in the
scanner.
* src/scan-gram.l (YY_USER_INIT): Initialize yylloc.
* src/gram.h (rule_t): Change line and action_line into location
and action_location, of location_t type.
Adjust all dependencies.
* src/location.h, src/location.c (empty_location): New.
* src/reader.h, src/reader.c (grammar_start_symbol_set)
(grammar_symbol_append, grammar_rule_begin, grammar_rule_end)
(grammar_current_rule_symbol_append)
(grammar_current_rule_action_append): Expect a location as argument.
* src/reader.c (grammar_midrule_action): Adjust to attach an
action's location as dummy symbol location.
* src/symtab.h, src/symtab.c (startsymbol_location): New.
* tests/regression.at (Web2c Report, Rule Line Numbers): Adjust
the line numbers.
Akim Demaille [Fri, 14 Jun 2002 17:38:14 +0000 (17:38 +0000)]
Grammar declarations may be found in the grammar section.
* src/parse-gram.y (rules_or_grammar_declaration): New.
(declarations): Each declaration may end with a semicolon, not
just...
(grammar_declaration): `"%union"'.
(grammar): Branch to rules_or_grammar_declaration.
Akim Demaille [Fri, 14 Jun 2002 17:36:24 +0000 (17:36 +0000)]
* src/files.c (action_obstack): Remove, unused.
(output_obstack): Remove it, and all its dependencies, as it is no
longer needed.
* src/reader.c (epilogue_set): Build the epilogue in the
muscle_obstack.
* src/output.h, src/output.c (muscle_obstack): Move to...
* src/muscle_tab.h, src/muscle_tab.h: here.
(muscle_init): Initialize muscle_obstack.
(muscle_free): New.
* src/main.c (main): Call it.
Akim Demaille [Fri, 14 Jun 2002 17:29:17 +0000 (17:29 +0000)]
* data/bison.simple, data/bison.c++: Be sure to restore the
current #line when returning to the skeleton contents after having
exposed the input file's #line.
Akim Demaille [Tue, 11 Jun 2002 21:46:16 +0000 (21:46 +0000)]
* src/scan-gram.l (YY_OBS_INIT): Remove, replace with...
[SC_COMMENT,SC_STRING,SC_CHARACTER,SC_BRACED_CODE,SC_PROLOGUE]
[SC_EPILOGUE]: Output the quadrigraphs only when not in a comment.
* src/reader.h, src/scan-gram.l (scanner_initialize): this.
* src/reader.c (reader): Use it.
Akim Demaille [Tue, 11 Jun 2002 21:43:18 +0000 (21:43 +0000)]
* src/scan-gram.l (YY_INIT, YY_GROW, YY_FINISH): Rename as...
(YY_OBS_INIT, YY_OBS_GROW, YY_OBS_FINISH): these.
(last_string, YY_OBS_FREE): New.
Use them when returning an ID.
Akim Demaille [Tue, 11 Jun 2002 20:16:05 +0000 (20:16 +0000)]
Have Bison grammars parsed by a Bison grammar.
* src/reader.c, src/reader.h (prologue_augment): New.
* src/reader.c (copy_definition): Remove.
* src/reader.h, src/reader.c (gram_start_symbol_set, prologue_augment)
(grammar_symbol_append, grammar_rule_begin, grammar_midrule_action)
(grammar_current_rule_prec_set, grammar_current_rule_check)
(grammar_current_rule_symbol_append)
(grammar_current_rule_action_append): Export.
* src/parse-gram.y (symbol_list_new, symbol_list_symbol_append_
(symbol_list_action_append): Remove.
Hook the routines from reader.
* src/scan-gram.l: In INITIAL, characters and strings are tokens.
* src/system.h (ATTRIBUTE_NORETURN, ATTRIBUTE_UNUSED): Now.
* src/reader.c (read_declarations): Remove, unused.
* src/parse-gram.y: Handle the epilogue.
* src/reader.h, src/reader.c (gram_start_symbol_set): Rename as...
(grammar_start_symbol_set): this.
* src/scan-gram.l: Be sure to ``use'' yycontrol to keep GCC quiet.
* src/reader.c (readgram): Remove, unused.
(reader): Adjust to insert eoftoken and axiom where appropriate.
* src/reader.c (copy_dollar): Replace with...
* src/scan-gram.h (handle_dollar): this.
* src/parse-gram.y: Remove `%thong'.
* src/reader.c (copy_at): Replace with...
* src/scan-gram.h (handle_at): this.
* src/complain.h, src/complain.c (warn_at, complain_at, fatal_at):
New.
* src/scan-gram.l (YY_LINES): Keep lineno synchronized for the
time being.
* src/reader.h, src/reader.c (grammar_rule_end): New.
* src/parse.y (current_type, current_class): New.
Implement `%nterm', `%token' support.
Merge `%term' into `%token'.
(string_as_id): New.
* src/symtab.h, src/symtab.c (symbol_make_alias): Don't pass the
type name.
* src/parse-gram.y: Be sure to handle properly the beginning of
rules.
* src/parse-gram.y: Handle %type.
* src/reader.c (grammar_rule_end): Call grammar_current_rule_check.
* src/parse-gram.y: More directives support.
* src/options.c: No longer handle source directives.
* src/parse-gram.y: Fix %output.
* src/parse-gram.y: Handle %union.
Use the prologue locations.
* src/reader.c (parse_union_decl): Remove.
* src/reader.h, src/reader.c (epilogue_set): New.
* src/parse-gram.y: Use it.
* data/bison.simple, data/bison.c++: b4_stype is now either not
defined, then default to int, or to the contents of %union,
without `union' itself.
Adjust.
* src/muscle_tab.c (muscle_init): Don't predefine `stype'.
* src/output.c (actions_output): Don't output braces, as they are
already handled by the scanner.
* src/scan-gram.l (SC_CHARACTER): Set the user_token_number of
characters to themselves.
* tests/reduce.at (Reduced Automaton): End the grammars with %% so
that the epilogue has a proper #line.
* src/parse-gram.y: Handle precedence/associativity.
* src/symtab.c (symbol_precedence_set): Requires the symbol to be
a terminal.
* src/scan-gram.l (SC_BRACED_CODE): Catch strings and characters.
* tests/calc.at: Do not use `%token "foo"' as it makes not sense
at all to define terminals that cannot be emitted.
* src/scan-gram.l: Escape M4 characters.
* src/scan-gram.l: Working properly with escapes in user
strings/characters.
* tests/torture.at (AT_DATA_TRIANGULAR_GRAMMAR)
(AT_DATA_HORIZONTAL_GRAMMAR): Respect the `%token ID NUM STRING'
grammar.
Use more modest sizes, as for the time being the parser does not
release memory, and therefore the process swallows a huge amount
of memory.
* tests/torture.at (AT_DATA_LOOKAHEADS_GRAMMAR): Adjust to the
stricter %token grammar.
* src/symtab.h (associativity): Add `undef_assoc'.
(symbol_precedence_set): Do nothing when passed an undef_assoc.
* src/symtab.c (symbol_check_alias_consistence): Adjust.
* tests/regression.at (Invalid %directive): Remove, as it is now
meaningless.
(Invalid inputs): Adjust to the new error messages.
(Token definitions): The new grammar doesn't allow too many
eccentricities.
* src/lex.h, src/lex.c: Remove.
* src/reader.c (lastprec, skip_to_char, read_signed_integer)
(copy_character, copy_string2, copy_string, copy_identifier)
(copy_comment, parse_token_decl, parse_type_decl, parse_assoc_decl)
(parse_muscle_decl, parse_dquoted_param, parse_skel_decl)
(parse_action): Remove.
* po/POTFILES.in: Adjust.
Akim Demaille [Tue, 11 Jun 2002 08:08:36 +0000 (08:08 +0000)]
* src/reader.c (parse_action): Don't store directly into the
rule's action member: return the action as a string.
Don't require `rule_length' as an argument: compute it.
(grammar_current_rule_symbol_append)
(grammar_current_rule_action_append): New, eved out from
(readgram): here.
Remove `action_flag', `rulelength', unused now.
Akim Demaille [Tue, 11 Jun 2002 08:01:41 +0000 (08:01 +0000)]
GCJ's parser requires the tokens to be defined before the prologue.
* data/bison.simple: Output the token definition before the user's
prologue.
* tests/regression.at (Braces parsing, Duplicate string)
(Mixing %token styles): Check the output from bison.
(Early token definitions): New.
Akim Demaille [Mon, 10 Jun 2002 08:38:10 +0000 (08:38 +0000)]
* src/symtab.c (symbol_user_token_number_set): Don't complain when
assigning twice the same user number to a token, so that we can
use it in...
* src/lex.c (lex): here.
Also use `symbol_class_set' instead of hand written code.
* src/reader.c (parse_assoc_decl): Likewise.
Akim Demaille [Mon, 10 Jun 2002 08:37:55 +0000 (08:37 +0000)]
* src/symtab.c, src/symtab.c (symbol_class_set)
(symbol_user_token_number_set): New.
* src/reader.c (parse_token_decl): Use them.
Use a switch instead of ifs.
Use a single argument.
Akim Demaille [Mon, 10 Jun 2002 08:37:30 +0000 (08:37 +0000)]
Remove `%thong' support as it is undocumented, unused, duplicates
`%token's job, and creates useless e-mail traffic with people who
want to know what it is, why it is undocumented, unused, and
duplicates `%token's job.
* src/reader.c (parse_thong_decl): Remove.
* src/options.c (option_table): Remove "thong".
* src/lex.h (tok_thong): Remove.
Akim Demaille [Tue, 28 May 2002 09:53:35 +0000 (09:53 +0000)]
* data/bison.c++: Catch up with bison.simple:
2002-05-24 Paul Hilfinger <Hilfinger@CS.Berkeley.EDU>
and Paul Eggert <eggert@twinsun.com>: `error' handing.
2002-05-26 Akim Demaille <akim@epita.fr>: stos_, token_number_,
and popping traces.
Paul Hilfinger [Mon, 27 May 2002 22:33:54 +0000 (22:33 +0000)]
* src/output.c (output_skeleton): Put an explicit path in front of
the skeleton file name, rather than relying on the -I directory,
to partially alleviate effects of having a skeleton file lying around
in the current directory.
Akim Demaille [Sun, 26 May 2002 20:25:52 +0000 (20:25 +0000)]
* src/state.h (state_t): `solved_conflicts' is a new member.
* src/LR0.c (new_state): Set it to 0.
* src/conflicts.h, src/conflicts.c (print_conflicts)
(free_conflicts, solve_conflicts): Rename as...
(conflicts_print, conflicts_free, conflicts_solve): these.
Adjust callers.
* src/conflicts.c (enum conflict_resolution_e)
(solved_conflicts_obstack): New, used by...
(log_resolution): this.
Adjust to attach the conflict resolution to each state.
Complete the description with the precedence/associativity
information.
(resolve_sr_conflict): Adjust.
* src/print.c (print_state): Output its solved_conflicts.
* tests/conflicts.at (Unresolved SR Conflicts)
(Solved SR Conflicts): Exercise --report=all.
Akim Demaille [Sun, 26 May 2002 18:38:41 +0000 (18:38 +0000)]
* tests/calc.at (AT_CHECK_CALC): Adjust: there are now additional
`Error:' lines.
* data/bison.simple (yystos) [YYDEBUG]: New.
(yyparse) [YYDEBUG]: Display the symbols which are popped during
error recovery.
* tests/regression.at (Web2c Actions): Adjust: yystos is output now.
Akim Demaille [Sat, 25 May 2002 16:12:40 +0000 (16:12 +0000)]
* doc/bison.texinfo (Debugging): Split into...
(Tracing): this new section, its former contents, and...
(Understanding): this new section.
* src/getargs.h, src/getargs.c (verbose_flag): Remove, replaced
by...
(report_flag): this.
Adjust all dependencies.
(report_args, report_types, report_argmatch): New.
(usage, getargs): Report/support -r, --report.
* src/options.h
(struct option_table_struct): Rename as..,
(struct option_table_s): this.
Rename the `set_flag' member to `flag' to match with getopt_long's
struct.
* src/options.c (option_table): Split verbose into an entry for
%verbose, and another for --verbose.
Support --report/-r, so remove -r from the obsolete --raw.
* src/print.c: Attach full item sets and lookaheads reports to
report_flag instead of trace_flag.
* lib/argmatch.h, lib/argmatch.c: New, from Fileutils 4.1.
Paul Eggert [Fri, 24 May 2002 12:45:23 +0000 (12:45 +0000)]
* data/bison.simple (yyparse): Correct error handling to conform to
POSIX and yacc. Specifically, after syntax error is discovered,
do not reduce further before shifting the error token.
Clean up the code a bit by removing the labels yyerrdefault,
yyerrhandle, yyerrpop.
* NEWS: Document the above.
Paul Eggert [Fri, 24 May 2002 12:37:12 +0000 (12:37 +0000)]
(yyparse): Correct error handling to conform to
POSIX and yacc. Specifically, after syntax error is discovered,
do not reduce further before shifting the error token.
Clean up the code a bit by removing the labels yyerrdefault,
yyerrhandle, yyerrpop.
Paul Hilfinger [Mon, 20 May 2002 22:56:53 +0000 (22:56 +0000)]
* data/bison.simple (yyr1): Don't use yy_token_number_type as element
type; it isn't always big enough, since it doesn't necessarily
include non-terminals.
(yytranslate): Expand definition of yy_token_number_type, so that
the latter can be removed.
(yy_token_number_type): Remove, only one use.
* data/bison.c++ (r1_): Parallel change to yyr1 in bison.simple---
don't use TokenNumberType as element type.
* tests/regression.at: Modify expected output to agree with change
to yyr1 and yytranslate.
Akim Demaille [Tue, 7 May 2002 08:07:32 +0000 (08:07 +0000)]
Kill GCC warnings.
* src/reduce.c (nonterminals_reduce): Don't loop over RITEM: loop
over the RHS of each rule.
* src/gram.h, src/gram.c (nritems): Is `unsigned int', not int.
* src/state.h (state_t): Member `nitems' is unsigned short.
* src/LR0.c (get_state): Adjust.
* src/reader.c (packgram): Likewise.
* src/output.c (GENERATE_MUSCLE_INSERT_TABLE): `max' is of type
`Type'.
(muscle_insert_int_table): Remove, unused.
(prepare_rules): Remove `max'.
Akim Demaille [Mon, 6 May 2002 08:23:41 +0000 (08:23 +0000)]
* src/LR0.c (new_state, get_state): Instead of using the global
`kernel_size' and `kernel_base', have two new arguments:
`core_size' and `core'.
Adjust callers.
Akim Demaille [Sun, 5 May 2002 11:56:49 +0000 (11:56 +0000)]
New experimental feature: display the lookaheads in the report and
graph.
* src/print (print_core): When --trace-flag, display the rules
lookaheads.
* src/print_graph.c (print_core): Likewise.
Swap the arguments.
Adjust caller.
Akim Demaille [Sun, 5 May 2002 11:56:06 +0000 (11:56 +0000)]
* src/output.c (output_unsigned_int_table): New.
(prepare_rules): `i' is unsigned.
`prhs', `rline', `r2' are unsigned int.
Rename muscle `rhs_number_max' as `rhs_max'.
Output muscles `prhs_max', `rline_max', and `r2_max'.
Free rline and r1.
* data/bison.simple, data/bison.c++: Adjust to use these muscles
to compute types instead of constant types.
* tests/regression.at (Web2c Actions): Adjust.
Akim Demaille [Fri, 3 May 2002 08:42:48 +0000 (08:42 +0000)]
* data/bison.simple (b4_token_enum): New.
(b4_token_defines): Use it to output tokens both as #define and
enums.
Suggested by Paul Eggert.
* src/output.c (token_definitions_output): Don't output spurious
white spaces.