Some consistency checks on symbols are performed after all the
symbols were read, by an iteration over the symbol table. This
traversal is nondeterministic, which can be a problem for test
cases.
Avoid this.
Addresses another form of nondeterminism reported by Joel E. Denny.
http://lists.gnu.org/archive/html/bison-patches/2009-05/msg00023.html
* tests/input.at (Numbered tokens): Split the hexadecimal/decimal
test in two.
Use different file names for the three tests to make the
maintenance easier.
* gnulib: Update to latest.
* lib/.cvsignore, lib/.gitignore, m4/.cvsignore,
* m4/.gitignore: Regen.
* src/symtab.c (symbol_from_uniqstr, semantic_type_from_uniqstr):
Call xalloc_die on hash_insert failures.
Requested by the new __warn_unused_result__ attribute of
hash_insert.
* data/bison.m4 (b4_check_user_names): In header comments, say
"user occurrence" instead of "grammar occurrence".
* src/muscle-tab.h (muscle_percent_define_insert): Likewise.
(muscle_percent_code_grow): Likewise just for consistency.
(cherry picked from commit 246c4efafe)
Conflicts:
src/muscle-tab.h
* data/c++.m4 (b4_namespace_close): Don't let it be printed
among the closing braces here. This fix might make the
generated code easier to debug, but otherwise it should be
insignificant because a trailing `:' is a C++ error already.
(cherry picked from commit 8c221795af)
* doc/bison.simple (Decl Summary): Document token.prefix.
(Calc++ Parser): Various fixes.
Formatting changes.
Use token.prefix.
Introduce a macro TOKEN to shorten the code and make it more
readable.
(Calc++ Scanner): Adjust.
* NEWS (Variable token.prefix): New.
For now, just api.push-pull and lr.keep-unreachable-states.
Maintain old names for backward compatibility.
* NEWS (2.5): Document.
* data/c.m4 (b4_identification): Update comment.
* data/yacc.c: Update access.
* doc/bison.texinfo: Update.
* etc/bench.pl.in (bench_push_parser): Update use.
* src/files.c (tr): Move to...
* src/getargs.c, src/getargs.h (tr): ... here because I can't
think of a better place to expose it. My logic is that, for all
uses of tr so far, command-line arguments can be involved, and
getargs.h is already included.
* src/main.c (main): Update access.
* src/muscle_tab.c (muscle_percent_define_insert): Convert old
variable names to new variable names before assigning value.
* src/reader.c (reader): Update setting default.
* tests/calc.at: Update uses.
* tests/conflicts.at (Unreachable States After Conflict
Resolution): Update use.
* tests/input.at (%define enum variables): Update use.
(%define backward compatibility): New test group.
* tests/push.at: Update uses.
* tests/reduce.at: Update uses.
* tests/torture.at: Update uses.
(cherry picked from commit 812775a039)
Conflicts:
data/c.m4
etc/bench.pl.in
src/parse-gram.c
src/parse-gram.h
tests/conflicts.at
* NEWS (2.5): Add an entry.
* src/getargs.c (usage): Mention IELR(1) and canonical LR(1)
besides just LALR(1) and GLR(1).
* doc/bison.texinfo (Introduction): Likewise.
(Language and Grammar): Bison is no longer limited to LALR(1)
restrictions.
(GLR parsing): Say deterministic or LR(1) rather than LALR(1)
when trying to distinguish from GLR. Talk about LR(1) grammars
rather than LALR(1) grammars.
(Decl Summary): In %define api.push_pull entry, say it applies
to deterministic parsers in C rather than LALR(1) parsers in C.
Add lr.default_rules entry.
Add lr.type entry.
(Mystery Conflicts): Bison is no longer limited to LALR(1)
restrictions.
(Generalized LR Parsing): Same changes as for the previous GLR
section.
(Memory Management): Say deterministic rather than LALR(1).
(Understanding): Correct some bison output.
Index discussion of "accepting state".
Say deterministic rather than LALR(1).
(Bison Options): In --yacc entry, say deterministic rather than
LALR(1).
In --report, --graph, and --xml entries, just don't mention
LALR(1).
(C++ Parsers): Say deterministic rather than LALR(1).
(Table of Symbols): Likewise in YYSTACK_USE_ALLOCA entry.
(Glossary): Add Accepting State, Consistent State, Default Rule,
and IELR(1) definitions.
In Generalized LR (GLR) definition, make same changes as in
previous GLR sections.
In LALR(1) definition, say Bison uses LALR(1) by default rather
than implying Bison is limited to LALR(1).
(LocalWords): Add IELR.
(cherry picked from commit 34a6c2d19b)
Its value can be "LALR", "IELR", or "canonical LR".
* lib/timevar.def (TV_IELR_PHASE1): New var.
(TV_IELR_PHASE2): New var.
(TV_IELR_PHASE3): New var.
(TV_IELR_PHASE4): New var.
* src/local.mk (src_bison_SOURCES): Add AnnotationList.c,
AnnotationList.h, InadequacyList.c, InadequacyList.h, Sbitset.c,
Sbitset.h, ielr.c, and ielr.h.
* src/getargs.h, src/getargs.c (enum trace, trace_args,
trace_types): Add trace_ielr.
* src/lalr.h, src/lalr.c (ngotos): Export it.
(F): Rename to...
(goto_follows): ... this, update all uses, and export it.
(set_goto_map): Export it.
(map_goto): Export it.
(compute_lookahead_tokens): Don't free goto_follows yet. Now
handled in ielr.
(initialize_LA): Export it. Move lookback allocation to...
(lalr): ... here because, for canonical LR, initialize_LA must
be invoked but lookback and much of the rest of LALR isn't
needed.
* main.c (main): Instead of lalr, invoke ielr, which invokes
lalr.
* src/reader.c (reader): Default lr.type to "LALR".
Default lr.default_rules to "accepting" if lr.type is "canonical
LR". Leave the default as "all" otherwise.
Check for a valid lr.type value.
* src/state.h, src/state.c (struct state_list): Add state_list
member.
(state_new): Initialize state_list member to NULL.
(state_new_isocore): New function, exported.
* tests/existing.at (AT_TEST_EXISTING_GRAMMAR): New macro that
exercises all values of lr.type.
(GNU AWK Grammar): Rename test group to...
(GNU AWK 3.1.0 Grammar): ... this, and extend to use
AT_TEST_EXISTING_GRAMMAR.
(GNU Cim Grammar): Extend to use AT_TEST_EXISTING_GRAMMAR.
(GNU pic Grammar): Rename test group to...
(GNU pic (Groff 1.18.1) Grammar): ... this, and extend to use
AT_TEST_EXISTING_GRAMMAR.
* tests/reduce.at (AT_TEST_LR_TYPE): New macro that exercises
all values of lr.type.
(Single State Split): New test groups using AT_TEST_LR_TYPE.
(Lane Split): Likewise.
(Complex Lane Split): Likewise.
(Split During Added Lookahead Propagation): Likewise.
Its value describes the states that are permitted to contain
default rules: "all", "consistent", or "accepting".
* src/reader.c (reader): Default lr.default_rules to "all".
Check for a valid lr.default_rules value.
* src/lalr.c (state_lookahead_tokens_count): If lr.default_rules
is "accepting", then only mark the accepting state as
consistent.
(initialize_LA): Tell state_lookahead_tokens_count whether
lr.default_rules is "accepting".
* src/tables.c (action_row): If lr.default_rules is not "all",
then disable default rules in inconsistent states.
* src/print.c (print_reductions): Use this opportunity to
perform some assertions about whether lr.default_rules was
obeyed correctly.
* tests/local.at (AT_TEST_TABLES_AND_PARSE): New macro that
helps with checking the parser tables for a grammar.
* tests/input.at (%define lr.default_rules invalid values): New
test group.
* tests/reduce.at (AT_TEST_LR_DEFAULT_RULES): New macro using
AT_TEST_TABLES_AND_PARSE.
(`no %define lr.default_rules'): New test group generated by
AT_TEST_LR_DEFAULT_RULES.
(`%define lr.default_rules "all"'): Likewise.
(`%define lr.default_rules "consistent"'): Likewise.
(`%define lr.default_rules "accepting"'): Likewise.
* data/bison.m4 (b4_percent_define_if_define_): Also map dashes to
underscores.
* src/scan-gram.l ({id}): Also accept dashes after the initial
letter.
({directive}): Use {id}.
* src/parse-gram.y: Comment and formatting changes.
* doc/bison.texinfo (Symbols): Adjust the lexical definitions of
symbols.
* src/complain.h, src/complain.c (yacc_at): New.
* src/symtab.c (symbol_new): Use yacc_at to report inappropriate
symbol names.
* src/output.c (token_definitions_output): Do not #define token
names with dashes.
* tests/output.at (AT_CHECK_OUTPUT_FILE_NAME): Avoid tripping
autoconf bug related to # in test.
Signed-off-by: Eric Blake <ebb9@byu.net>
(cherry picked from commit ab3a683f05)
