Fix some memory leaks, and fix a bug: state 0 was examined twice.
* src/LR0.c (new_state): Merge into...
(state_list_append): this.
(new_states): Merge into...
(generate_states): here.
(set_states): Don't ensure a proper `errs' state member here, do it...
* src/conflicts.c (conflicts_solve): here.
* src/state.h, src/state.c: Comment changes.
(state_t): Rename member `shifts' as `transitions'.
Adjust all dependencies.
(errs_new): For consistency, also take the values as argument.
(errs_dup): Remove.
(state_errs_set): New.
(state_reductions_set, state_transitions_set): Assert that no
previous value was assigned.
(state_free): New.
(states_free): Use it.
* src/conflicts.c (resolve_sr_conflict): Don't use an `errs_t' as
temporary storage: use `errs' and `nerrs' as elsewhere.
(set_conflicts): Allocate and free this `errs'.

* lib/libiberty.h: New.
* lib: Update the bitset implementation from upstream.
* src/closure.c, src/lalr.c, src/output.c, src/print_graph.c,
* src/state.c: Use BITSET_FOR_EACH, not BITSET_EXECUTE.
* src/main.c: Adjust bitset stats calls.

Regen.

* src/scan-gram.l (<SC_ESCAPED_CHARACTER>): Convert to unsigned
char, so that negative chars don't collide with $.

(<SC_ESCAPED_CHARACTER>): Convert to unsigned char, so that negative
chars don't collide with $.

Typo.

Regen.

Have the GLR tests be `warning' checked, and fix the warnings.
* data/glr.c (YYFPRINTF): Always define it, not only when YYDEBUG
(yyuserAction, yyreportAmbiguity): `Use' all the arguments.
(yyremoveDeletes): `yyi' and `yyj' are size_t.
Use YYFPRINTF when under if (YYDEBUG) to avoid empty `if' bodies.
(yyaddDeferredAction): static.
(yyglrReduce): yyi, yyk, amd yyposn are size_t.
(yyreportParseError): yyprefix is const.
yytokenp is used only when verbose.
(yy__GNUC__): Replace with __GNUC__.
(yypdumpstack): yyi is size_t.
(yypreference): Un-yy local variables and arguments, to avoid
clashes with `yyr1'.  Anyway, we are not in the user name space.
(yytname_size): be an int, as is compared with ints.
* tests/testsuite.at (AT_COMPILE, AT_PARSER_CHECK): New.
Use them.
* tests/cxx-gram.at: Use quotation to protect $1.
Use AT_COMPILE to enable warnings hunts.
Prototype yylex and yyerror.
`Use' argc.
Include `string.h', not `strings.h'.
Produce and prototype stmtMerge only when used.
yylex takes a location.

We spend a lot of time in quotearg, in particular when --verbose.
* src/symtab.c (symbol_get): Store a quoted version of the key.
(symbol_tag_get, symbol_tag_get_n, symbol_tag_print): Remove.
Adjust all callers.

* src/state.h (reductions_t): Rename member `nreds' as num.
(errs_t): Rename members `nerrs' and `errs' as `num' and `symbols'.
* src/state.c (ERRS_ALLOC, REDUCTIONS_ALLOC): Use the correct types.

* src/state.h, src/state.c (shift_t, SHIFT_SYMBOL, SHIFT_IS_SHIFT)
(SHIFT_IS_GOTO, SHIFT_IS_ERROR, SHIFT_DISABLE, SHIFT_IS_DISABLED)
(shifts_to): Rename as...
(transition_t, TRANSITION_SYMBOL, TRANSITION_IS_TRANSITION)
(TRANSITION_IS_GOTO, TRANSITION_IS_ERROR, TRANSITION_DISABLE)
(TRANSITION_IS_DISABLED, transitions_to): these.

* src/print.c (print_shifts, print_gotos): Merge into...
(print_transitions): this.
(print_transitions, print_errs, print_reductions): Align the
lookaheads columns.
(print_core, print_transitions, print_errs, print_state,
print_grammar): Output empty lines separator before, not after.
(state_default_rule_compute): Rename as...
(state_default_rule): this.
* tests/conflicts.at (Defaulted Conflicted Reduction),
(Unresolved SR Conflicts, Resolved SR Conflicts): Adjust.
* tests/regression.at (Rule Line Numbers, Web2c Report): Adjust.

Display items as we display rules.
* src/gram.h, src/gram.c (rule_lhs_print): New.
* src/gram.c (grammar_rules_partial_print): Use it.
* src/print.c (print_core): Likewise.
* tests/conflicts.at (Defaulted Conflicted Reduction),
(Unresolved SR Conflicts): Adjust.
(Unresolved SR Conflicts): Adjust and rename as...
(Resolved SR Conflicts): this, as was meant.
* tests/regression.at (Web2c Report): Adjust.

* src/print.c (state_default_rule_compute): New, extracted from...
(print_reductions): here.
Pessimize, but clarify the code.
* tests/conflicts.at (Defaulted Conflicted Reduction): New.

* src/output.c (action_row): Let default_rule be always a rule
number.

* src/closure.c (print_firsts, print_fderives, closure):
Use BITSET_EXECUTE.
* src/lalr.c (lookaheads_print): Likewise.
* src/state.c (state_rule_lookaheads_print): Likewise.
* src/print_graph.c (print_core): Likewise.
* src/print.c (print_reductions): Likewise.
* src/output.c (action_row): Likewise.
Use SHIFT_IS_DISABLED, SHIFT_IS_SHIFT and SHIFT_SYMBOL.

* src/print_graph.c: Use report_flag.

* src/lalr.c (traverse, digraph, matrix_print, transpose): Move
to...
* src/relation.h, src/relation.c (traverse, relation_digraph)
(relation_print, relation_transpose): New.

* src/state.h, src/state.c (shifts_to): New.
* src/lalr.c (build_relations): Use it.

* src/gram.h (rule_number_t, RULE_NUMBER_MAX, int_of_rule_number)
(item_number_of_rule_number, rule_number_of_item_number): New.
* src/LR0.c, src/closure.c, src/derives.c, src/derives.h,
* src/gram.c, src/lalr.c, src/nullable.c, src/output.c, src/print.c,
* src/print_graph.c, src/reader.c, src/reduce.c, src/reduce.h:
Propagate their use.
Much remains to be done, in particular wrt `shorts' from types.h.

* src/symtab.c (symbol_new): Initialize the `printer' member.

* src/LR0.c (save_reductions): Remove, replaced by...
* src/state.h, src/state.c (state_reductions_set): New.
(reductions, errs): Rename as...
(reductions_t, errs_t): these.
Adjust all dependencies.

* src/LR0.c (state_list_t, state_list_append): New.
(first_state, last_state): Now symbol_list_t.
(this_state): Remove.
(new_itemsets, append_states, save_reductions): Take a state_t as
argument.
(set_states, generate_states): Adjust.
(save_shifts): Remove, replaced by...
* src/state.h, src/state.c (state_shifts_set): New.
(shifts): Rename as...
(shifts_t): this.
Adjust all dependencies.
* src/state.h (state_t): Remove the `next' member.

* src/vcg.c (quote): Use slot 2, since we often pass symbol tag
escaped in slot 0.

Use hash.h for the state hash table.
* src/LR0.c (STATE_HASH_SIZE, state_hash): Remove.
(allocate_storage): Use state_hash_new.
(free_storage): Use state_hash_free.
(new_state, get_state): Adjust.
* src/lalr.h, src/lalr.c (states): Move to...
* src/states.h (state_t): Remove the `link' member, no longer
used.
* src/states.h, src/states.c: here.
(state_hash_new, state_hash_free, state_hash_lookup)
(state_hash_insert, states_free): New.
* src/states.c (state_table, state_compare, state_hash): New.
* src/output.c (output_actions): Do not free states now, since we
still need to know the final_state number in `prepare', called
afterwards.  Do it...
* src/main.c (main): here: call states_free after `output'.

* src/state.h, src/state.c (state_new): New, extracted from...
* src/LR0.c (new_state): here.
* src/state.h (STATE_ALLOC): Move to...
* src/state.c: here.
* src/LR0.h, src/LR0.c (nstates, final_state): Move to...
* src/state.h, src/state.c: here.

* src/reader.c (gensym): Rename as...
* src/symtab.h, src/symtab.c (dummy_symbol_get): this.
(getsym): Rename as...
(symbol_get): this.

* src/state.h (state_number_t, STATE_NUMBER_MAX): New.
* src/LR0.c, src/LR0.h, src/conflicts.c, src/lalr.c, src/lalr.h,
* src/output.c, src/print.c, src/print_graph.c: Propagate.
* src/LR0.h, src/LR0.h (final_state): Is a state_t*.

Make the test suite pass with warnings checked.
* tests/actions.at (Printers and Destructors): Improve.
Avoid unsigned vs. signed issues.
* tests/calc.at: Don't exercise the scanner here, do it...
* tests/input.at (Torturing the Scanner): here.

* data/glr.c: Correct typo in Emacs-mode directive.  Slightly
  reorganize first lines parallel to yacc.c.

* data/c.m4 (b4_copyright, b4_sint_type, b4_uint_type, b4_token_define)
(b4_token_enum, b4_token_defines): New, factored from...
* data/lalr1.cc, data/yacc.c, glr.c: here.

regen.

* data/yacc.c (yydestruct, yysymprint): Pacify GCC warnings for
unused variables.
* src/output.c (merger_output): static.

* src/reader.h: s/grammer_current_rule_merge_set/grammar_.../.
* src/conflicts.c (conflicts_total_count): `i' is unsigned, to
pacify GCC.
* src/output.c (save_row): Initialize all the variables to pacify GCC.
(merger_output): static.

Fix small formatting glitch.

Initial check-in introducing experimental GLR parsing.  See entry in
ChangeLog dated 2002-06-27 from Paul Hilfinger for details.

Renamed file from bison.glr.

Renaming file to glr.c.

* src/options.h, src/options.c: Remove.
* src/getargs.c (short_options, long_options): New.

Update.

* data/bison.simple, data/bison.c++: Rename as...
* data/yacc.c, data/lalr1.cc: these.
* doc/bison.texinfo (Environment Variables): Remove.

Update.

New file: Tests for GLR parsing.

New file: skeleton for GLR (Generalized LR) parser.

* src/getargs.c (report_argmatch): Initialize strtok().

Update.

Update.

* data/bison.simple (b4_symbol_actions): New, replaces...
(b4_symbol_destructor, b4_symbol_printer): these.
(yysymprint): Be sure to call YYPRINT only for tokens, and using
user token numbers.

* data/bison.simple (yydestructor): Rename as...
(yydestruct): this.

* src/symtab.h, src/symtab.c (symbol_type_set)
(symbol_destructor_set, symbol_precedence_set): The location is
the last argument.
Adjust all callers.

* src/parse-gram.y (YYPRINT, yyprint): Don't mess with the parser
internals.
* src/reader.h, src/reader.c (grammar_current_rule_prec_set):
Takes a location.
* src/symtab.h, src/symtab.c (symbol_class_set)
(symbol_user_token_number_set): Likewise.
Adjust all callers.
Promote complain_at.
* tests/input.at (Type Clashes): Adjust.

* data/bison.simple (YYLEX): Fix the declaration when
%pure-parser.

* data/bison.simple (yysymprint): Don't print the token number,
just its name.
* tests/actions.at (Destructors): Rename as...
(Printers and Destructors): this.
Also exercise %printer.

* data/bison.simple (YYDSYMPRINT): New.
Use it to remove many of the #if YYDEBUG/if (yydebug).

* src/symtab.h, src/symtab.c (symbol_t): printer and
printer_location are new members.
(symbol_printer_set): New.
* src/parse-gram.y (PERCENT_PRINTER): New token.
Handle its associated rule.
* src/scan-gram.l: Adjust.
(handle_destructor_at, handle_destructor_dollar): Rename as...
(handle_symbol_code_at, handle_symbol_code_dollar): these.
* src/output.c (symbol_printers_output): New.
(output_skeleton): Call it.
* data/bison.simple (yysymprint): New.  Cannot be named yyprint
since there are already many grammar files with a user `yyprint'.
Replace the calls to YYPRINT to calls to yysymprint.
* tests/calc.at: Adjust.
* tests/torture.at (AT_DATA_STACK_TORTURE): Remove YYPRINT: it was
taking advantage of parser very internal details (stack size!).

Regen.

* src/scan-gram.l: Complete the scanner with the missing patterns
to pacify Flex.
Use `quote' and `symbol_tag_get' where appropriate.

* tests/actions.at (Destructors): Augment to test locations.
* data/bison.simple (yydestructor): Pass it the current location
if locations are enabled.
Prototype only when __STDC__ or C++.
Change the argument names to move into the yy name space: there is
user code here.

* data/bison.simple (b4_pure_if): New.
Use it instead of #ifdef YYPURE.

* data/bison.simple (m4_location_if): New.
Use it instead of #ifdef YYLSP_NEEDED.

Regen.

Prepare @$ in %destructor, but currently don't bind it in the
skeleton, as %location use is not cleaned up yet.
* src/scan-gram.l (handle_dollar, handle_destructor_at)
(handle_action_at): New.
(handle_at, handle_action_dollar, handle_destructor_dollar): Take
a braced_code_t and a location as additional arguments.
(handle_destructor_dollar): Instead of requiring `b4_eval', just
unquote one when outputting `b4_dollar_dollar'.
Adjust callers.
* data/bison.simple (b4_eval): Remove.
(b4_symbol_destructor): Adjust.
* tests/input.at (Invalid @n): Adjust.

* doc/bison.texinfo: Document ability to have multiple
prologue sections.

* src/files.c (compute_base_names): When computing the output file
names from the input file name, strip the directory part.

* data/bison.simple.new: Comment changes.
Reported by Andreas Schwab.

* data/bison.simple (yyoverflowlab): #ifndef yyoverflow, so that
there are no `label `yyoverflowlab' defined but not used' warnings
when yyoverflow is defined.

* src/symtab.h, src/symtab.c (symbol_t): destructor_location is a
new member.
(symbol_destructor_set): Adjust.
* src/output.c (symbol_destructors_output): Output the destructor
locations.
Output the symbol name.
* data/bison.simple (b4_symbol_destructor): Adjust.

and Akim Demaille  <akim@epita.fr>
* data/bison.simple.new (yyerrlab1): Be sure to pop and destroy
what's left on the stack when the error recovery hits EOF.
* tests/actions.at (Destructors): Complete to exercise this case.

Update.

Regen.

* data/m4sugar/m4sugar.m4 (m4_map): Recognize when the list of
arguments is really empty, not only equal to `[]'.
* src/symtab.h, src/symtab.c (symbol_t): `destructor' is a new
member.
(symbol_destructor_set): New.
* src/output.c (symbol_destructors_output): New.
* src/reader.h (brace_code_t, current_braced_code): New.
* src/scan-gram.l (BRACED_CODE): Use it to branch on...
(handle_dollar): Rename as...
(handle_action_dollar): this.
(handle_destructor_dollar): New.
* src/parse-gram.y (PERCENT_DESTRUCTOR): New.
(grammar_declaration): Use it.
* data/bison.simple (yystos): Is always defined.
(yydestructor): New.
* tests/actions.at (Destructors): New.
* tests/calc.at (_AT_CHECK_CALC_ERROR): Don't rely on egrep.

* src/symlist.h, src/symlist.c (symbol_list_length): New.
* src/scan-gram.l (handle_dollar, handle_at): Compute the
rule_length only when needed.
* src/output.c (actions_output, token_definitions_output): Output
the full M4 block.
* src/symtab.c: Don't access directly to the symbol tag, use
symbol_tag_get.
* src/parse-gram.y: Use symbol_list_free.

* src/reader.h, src/reader.c (symbol_list, symbol_list_new)
(symbol_list_prepend, get_type_name): Move to...
* src/symlist.h, src/symlist.c (symbol_list_t, symbol_list_new)
(symbol_list_prepend, symbol_list_n_type_name_get): here.
Adjust all callers.
(symbol_list_free): New.
* src/scan-gram.l (handle_dollar): Takes a location.
* tests/input.at (Invalid $n): Adjust.

* src/reader.h, src/reader.c (symbol_list_new): Export it.
(symbol_list_prepend): New.
* src/parse-gram.y (%union): `list' is a new member.
(symbols.1): New, replaces...
(terms_to_prec.1, nterms_to_type.1): these.
* src/symtab.h, src/symtab.c (symbol_type_set, symbol_precedence_set)
Take a location as additional argument.
Adjust all callers.

Regen.

* src/parse-gram.y: Move %token in the declaration section so that
we don't depend upon CVS Bison.

* src/state.h, src/state.c (state_rule_lookaheads_print): New.
* src/print.c (print_core): Use it.

* 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.

* src/state.h (state_t): Replace the `lookaheadsp' member, a
short, with `lookaheads' (bitsetv), `lookaheads_rule' (rule_t**).
Adjust all dependencies.
* src/lalr.c (initialize_lookaheads): Split into...
(states_lookaheads_count, states_lookaheads_initialize): these.
(lalr): Adjust.

* src/gram.h, src/gram.c (grammar_rules_partial_print): New, eved
out of...
(grammar_rules_print): here.
* src/reduce.c (reduce_output): Use it.
* tests/reduce.at (Useless Rules, Reduced Automaton)
(Underivable Rules): Adjust.

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.

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.

* 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.

Formatting changes.

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.

* src/main.c (main): Invoke scanner_free.

* src/output.c (m4_invoke): Extracted from...
(output_skeleton): here.
Free tempfile.

* src/parse-gram.y (directives, directive, gram)
(grammar_directives, precedence_directives, precedence_directive):
Rename as...
(declarations, declaration, grammar, grammar_declaration)
(precedence_declaration, precedence_declarator): these.
(symbol_declaration): New.

* 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.

* src/location.h: New, extracted from...
* src/reader.h: here.
* src/Makefile.am (noinst_HEADERS): Merge into
(bison_SOURCES): this.
Add location.h.
* src/parse-gram.y: Use location_t instead of Bison's.
* src/reader.h, src/reader.c (prologue_augment, epilogue_set):
Use location_t instead of ints.

* 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.

* src/scan-gram.l (SC_BRACED_CODE): Don't use `<.*>', it is too
eager.
* tests/actions.at (Exotic Dollars): New.

* src/scan-gram.l (SC_PROLOGUE): Don't eat characters amongst
['"/] too eagerly.
* tests/input.at (Torturing the Scanner): New.

White space changes.

* 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.

* src/scan-gram.l (YY_OBS_FINISH): Don't set yylval.
Adjust all callers.
(scanner_last_string_free): New.

* 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.

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.

Regen.

Update.