run parsers in all tests so that Valgrind is invoked during
maintainer-check-valgrind.
(Duplicate representation of merged trees): Free all semantic values.
(Duplicated user destructor for lookahead): Likewise.
<http://lists.gnu.org/archive/html/bug-bison/2006-10/msg00000.html>
and try to make such failures easier to catch in the future.
* data/glr.c (YYTRANSLATE): Don't check for nonpositive arg;
that's now the caller's responsibility.
(yyprocessOneStack, yyrecoverSyntaxError, yyparse):
Set yychar = YYEOF if it's negative.
* tests/actions.at (yylex): Abort if asked to read past EOF.
* tests/conflicts.at (yylex): Likewise.
* tests/cxx-type.at (yylex): Likewise.
* tests/glr-regression.at (yylex): Likewise.
* tests/input.at (yylex): Likewise.
* tests/regression.at (yylex): Likewise.
* tests/torture.at (yylex): Likewise.
declaration".
* src/symtab.c (redeclaration): Do that here.
* src/reader.c (record_merge_function_type): In the case of a result
type clash, report the previous declaration rather than the very first
one in the grammar file.
* tests/glr-regression.at (Missed %merge type warnings when LHS type is
declared later): Add a third declaration to check this behavior.
* tests/input.at (Incompatible Aliases): Update output.
declared after the %merge.
* src/reader.c (get_merge_function): Don't set the merger type yet.
(record_merge_function_type): New function for setting the merger type
and checking for clashes.
(grammar_current_rule_merge_set): Set the location of the %merge for
the current rule.
(packgram): Invoke record_merge_function_type for each rule now that
all symbol type declarations have been parsed.
* src/reader.h (merger_list.type_declaration_location): New member
storing the location of the first %merge from which the type for this
merging function was derived.
* src/symlist.h (symbol_list.merger_declaration_location): New member
storing the location of a rule's %merge, if any.
* tests/glr-regression.at (Missed %merge type warnings when LHS type is
declared later): New test to catch the error fixed by the above patch.
`look_ahead'. Discussed starting at
<http://lists.gnu.org/archive/html/bison-patches/2006-01/msg00049.html>
and then at
<http://lists.gnu.org/archive/html/bison-patches/2006-06/msg00017.html>.
* NEWS: For the next release, note the change to `--report'.
* TODO, doc/bison.1: Update English.
* doc/bison.texinfo: Update English.
(Understanding Your Parser, Bison Options): Document as
`--report=lookahead' rather than `--report=look-ahead'.
* src/conflicts.c: Update English in comments.
(lookahead_set): Rename from look_ahead_set.
(flush_reduce): Rename argument look_ahead_tokens to lookahead_tokens.
(resolve_sr_conflict): Rename local look_ahead_tokens to
lookahead_tokens, and update other uses.
(flush_shift, set_conflicts, conflicts_solve, count_sr_conflicts,
count_rr_conflicts, conflicts_free): Update uses.
* src/getargs.c (report_args): Move "lookahead" before alternate
spellings.
(report_types): Update uses.
(usage): For `--report' usage description, state `lookahead' spelling
rather than `look-ahead'.
* src/getargs.h (report.report_lookahead_tokens): Rename from
report_look_ahead_tokens.
* src/lalr.c: Update English in comments.
(compute_lookahead_tokens): Rename from compute_look_ahead_tokens.
(state_lookahead_tokens_count): Rename from
state_look_ahead_tokens_count.
Rename local n_look_ahead_tokens to n_lookahead_tokens.
(lookahead_tokens_print): Rename from look_ahead_tokens_print.
Rename local n_look_ahead_tokens to n_lookahead_tokens.
Update other uses.
Update English in output.
(add_lookback_edge, initialize_LA, lalr, lalr_free): Update uses.
* src/print.c: Update English in comments.
(lookahead_set): Rename from look_ahead_set.
(print_reduction): Rename argument lookahead_token from
look_ahead_token.
(print_core, state_default_rule, print_reductions, print_results):
Update uses.
* src/print_graph.c: Update English in comments.
(print_core): Update uses.
* src/state.c: Update English in comments.
(reductions_new): Update uses.
(state_rule_lookahead_tokens_print): Rename from
state_rule_look_ahead_tokens_print, and update other uses.
* src/state.h: Update English in comments.
(reductions.lookahead_tokens): Rename from look_ahead_tokens.
(state_rule_lookahead_tokens_print): Rename from
state_rule_look_ahead_tokens_print.
* src/tables.c: Update English in comments.
(conflict_row, action_row): Update uses.
* tests/glr-regression.at
(Incorrect lookahead during deterministic GLR,
Incorrect lookahead during nondeterministic GLR): Rename
print_look_ahead to print_lookahead.
* tests/torture.at: Update English in comments.
(AT_DATA_LOOKAHEAD_TOKENS_GRAMMAR): Rename from
AT_DATA_LOOK_AHEAD_TOKENS_GRAMMAR.
(Many lookahead tokens): Update uses.
* data/glr.c: Update English in comments.
* lalr1.cc: Likewise.
* yacc.c: Likewise.
* src/conflicts.h: Likewise.
* src/lalr.h: Likewise.
* src/main.c: Likewise.
* src/output.c: Likewise.
* src/parse-gram.c: Likewise.
* src/tables.h: Likewise.
* tests/calc.at: Likewise.
state before an empty RHS is always resolved here. Only the location
of that state is guaranteed to be resolved, and that's enough. This
fixes the remaining bug reported by Derek M. Jones in
<http://lists.gnu.org/archive/html/bug-bison/2006-05/msg00027.html>.
* tests/glr-regression.at (Uninitialized location when reporting
ambiguity): Test the above case.
Also, the embedded comments in this test case claim it checks the case
of an empty RHS that has inherited the initial location. However, the
corresponding LHS was already resolved, so yyresolveLocations didn't
actually have reason to modify it. Fix this by forcing
nondeterministic operation at the beginning of the parse.
shadowing warnings. Use usual patter for iterating through RHS.
* tests/glr-regression.at
(Uninitialized location when reporting ambiguity):
Modify yylex so that it uses its argument, rather than trying
to rely on ARGSUSED (which doesn't work for gcc with warnings).
const char -> char const.
ambiguity): New test case.
* data/glr.c (yyresolveLocations): New function, which uses
YYLLOC_DEFAULT.
(yyresolveValue): Invoke yyresolveLocations before reporting an
ambiguity.
* doc/bison.texinfo (Default Action for Locations): Note
YYLLOC_DEFAULT's usage for ambiguity locations.
(GLR Semantic Actions): Cross-reference those notes.
ambiguity): Remove unnecessary union and type declarations.
(Leaked lookahead after nondeterministic parse syntax error): New test
case.
* data/glr.c (yyparse): Check for zero stacks remaining before
attempting to shift the lookahead so that you don't lose it.
* tests/glr-regression.at (Leaked semantic values when reporting
ambiguity): New test case.
* data/glr.c (yyreportAmbiguity): Invoke yyyerror directly and return
yyabort rather than invoking yyFail, which invokes longjmp. Remove the
now unnecessary yystackp parameter.
(yyresolveValue): Return yyreportAmbiguity's result. Now the necessary
destructors can be called.
* tests/glr-regression.at: Don't invoke bison with `-t' unnecessarily
in existing testcases.
parser, and clean up related code a bit.
* tests/glr-regression.at (Leaked merged semantic value if user action
cuts parse) Rename to...
(Leaked semantic values if user action cuts parse) ... this. Add check
for leaked parent RHS values.
* data/glr.c (yydestroyGLRState): In debugging output, distinguish
between an unresolved state (non-empty chain of semantic options) and
an incomplete one (signaled by an empty chain).
(yyresolveStates): Document the interface. Move all manipulation of an
successfully or unsuccessfully resolved yyGLRState to...
(yyresolveValue): ... here so that yyresolveValue always leaves a
yyGLRState with consistent data and thus is easier to understand.
Remove the yyvalp and yylocp parameters since they are always just
taken from the yys parameter. When reporting a discarded merged value
in debugging output, note that it is incompletely merged. Document the
interface.
(yyresolveAction): If resolving any of the RHS states fails, destroy
them all rather than leaking them. Thus, as long as user actions are
written to clean up the RHS correctly, yyresolveAction always cleans up
the RHS of a semantic option. Document the interface.
influence the parse by changing yychar. To make this easier to fix and
to make glr.c easier to evolve in general, don't maintain yytoken in
parallel with yychar; just compute yytoken when needed.
* tests/glr-regression.at (Incorrect lookahead during deterministic
GLR): Check that setting yychar in a user action has the intended
effect.
* data/glr.c (yyGLRStack): Remove yytokenp member.
(yyclearin): Don't set *yytokenp.
(yyprocessOneStack, yyreportSyntaxError, yyrecoverSyntaxError): Examine
yychar rather than *yytokenp to determine the current lookahead.
Compute yytoken locally when needed.
(yyparse): Likewise. Remove the local yytoken that yytokenp used to
point to.
* doc/bison.texinfo (Bison Options): Remove stray sentence fragment
after `--report' documentation.
* configure.ac: Don't use -Wnested-externs, as it's incompatible
with the new verify.h implementation.
* data/c.m4 (b4_yy_symbol_print_generate): YYUSE (yyoutput).
* data/glr.c (YYUSE): Depend on __GNUC__ as well.
* data/yacc.c (YYUSE): Likewise.
* data/lalr1.cc (yysyntax_error_): YYUSE (yystate).
* lib/subpipe.c (end_of_output_subpipe): The args are unused.
* src/parse-gram.y (declaration): Don't pass a string constant
to a function that expects char *, since GCC might complain
about the constant value.
* src/reader.c (symbol_typed_p): Add parens to pacify GCC.
* tests/cxx-type.at (_AT_TEST_GLR_CXXTYPES): #undef yylloc and yylval
before #defining them.
* tests/glr-regression.at
(Incorrectly initialized location for empty right-hand side in GLR):
In yyerror, use the msg arg.
(Corrupted semantic options if user action cuts parse):
(Incorrect lookahead during deterministic GLR):
(Incorrect lookahead during nondeterministic GLR):
Don't name a local var 'index'; it shadows string.h's 'index'.
yylookaheadNeeds. All uses updated.
(yysplitStack): Rename local yynewLookaheadStatuses to
yynewLookaheadNeeds.
* data/glr-regression.at (Incorrect lookahead during nondeterministic
GLR): In comments, change `lookahead status' to `lookahead need'.
use during nondeterministic operation to track which stacks have
actually needed the current lookahead.
(yyinitStateSet, yyfreeStateSet, yyremoveDeletes, yysplitStack):
Allocate, deallocate, resize, and otherwise shuffle space for
yylookaheadStatuses in parallel with yystates member of yyGLRStateSet.
(yysplitStack, yyprocessOneStack, yyparse): Set lookahead status
appropriately during nondeterministic operation.
(yySemanticOption): Add int yyrawchar, YYSTYPE yyval, and YYLTYPE yyloc
members to store the current lookahead to be used by the deferred
user action.
(yyaddDeferredAction): Add size_t yyk parameter specifying the stack
from which the RHS is taken. Set the lookahead members of the new
yySemanticOption according to the lookahead status for stack yyk.
(yyglrShiftDefer, yyglrReduce): Pass yyk parameter on to
yyaddDeferredAction.
(yyresolveAction): Set yychar, yylval, and yylloc to the lookahead
members of yySemanticOption before invoking yyuserAction, and then set
them back to their current values afterward.
(yyparse): Set yychar = YYEMPTY where yytoken = YYEMPTY.
(yyreportAmbiguity): Add /*ARGSUSED*/ to pacify lint.
* tests/glr-regression.at: Remove `.' from the ends of recent test case
titles for consistency.
(Leaked merged semantic value if user action cuts parse): In order to
suppress lint warnings, use arguments in merge function, and assign
char value < 128 in main.
(Incorrect lookahead during deterministic GLR): New test case.
(Incorrect lookahead during nondeterministic GLR): New test case.
least one user action succeeds but a later one cuts the parse, then
destroy the semantic value before returning rather than leaking it.
(yyresolveStates): If a user action cuts the parse and thus
yyresolveValue fails, ignore the (unset) semantic value rather than
corrupting the yyGLRState, and empty the semantic options list since
the user actions should have called all necessary destructors.
Simplify code with YYCHK.
* tests/glr-regression.at (Corrupted semantic options if user action
cuts parse): New test case.
(Undesirable destructors if user action cuts parse): New test case.
Before applying any of this patch, this test case never actually failed
for me... but only because the corrupted semantic options usually
masked this bug.
(Leaked merged semantic value if user action cuts parse): New test
case.
stacks, and iterate another stack in order to call user
destructors.
* tests/glr-regression.at (No users destructors if stack 0 deleted):
New test case.
(Duplicated user destructor for lookahead): This test now is expected
to succeed.
(yyresolveAction): Initialize default location properly for empty right-hand
sides.
(yydoAction): Ditto. Add comment explaining apparently dead code.
* tests/glr-regression.at
(Incorrectly initialized location for empty right-hand side in GLR): New test.
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.
* 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.
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
* tests/cxx-type.at (_AT_TEST_GLR_CXXTYPES): Have yyerror return
void, not int.
* tests/glr-regression.at (Badly Collapsed GLR States):
Likewise.
(Improper handling of embedded actions and dollar(-N) in GLR parsers):
yylex should return 0 at EOF rather than aborting.
Problem reported by twlevo@xs4all.nl.
* data/glr.c (struct yyGLRStack): Remove yyerrflag member.
All uses removed.
(yyStackOverflow): Just longjmp, but with value 2 so that caller
can handle the problem.
(YYCHK1): Use goto (a la yacc.c) rather than setting a flag.
(yyparse): New local variable yyresult to record the result.
Use result of setjmp to set it, rather than storing itinto
struct.
(yyDone): Remove label.
(yyacceptlab, yyabortlab, yyoverflowlab, yyreturn): New labels,
to mimic yacc.c. Do not discard lookahead if it's EOF (possible
if YYABORT is used).
* tests/actions.at (_AT_CHECK_PRINTER_AND_DESTRUCTOR): Exit with
yyparse status; put status > 1 into diagnostic.
Check that status==2 works.
* tests/calc.at, tests/cxx-type.at, tests/glr-regression.at:
Use exit status 3 for failure to open (which shouldn't happen).
1 on syntax error; just let yyparse do its thing.
* tests/glr-regression.at (Badly Collapsed GLR States): Likewise.
* tests/torture.at (AT_DATA_STACK_TORTURE): Likewise.
(Exploding the Stack Size with Alloca):
(Exploding the Stack Size with Malloc):
Expect exit status 2, not 1, since the parser is supposed to blow
its stack. Problem reported by twlevo@xs4all.nl.
(yyuserMerge): Provide a default case if b4_mergers is empty.
* tests/cxx-type.at (_AT_TEST_GLR_CXXTYPES): Define YYSTACKEXPANDABLE.
* tests/glr-regression.at
(Improper handling of embedded actions and dollar(-N) in GLR parsers):
YYSTYPE is char *, not char const *, so that strcpy ($$, ...) works.
fix a small glitch in debugging output.
(yyprocessOneStack, yyrecoverSyntaxError, yyparse): Print newline
after YY_SYMBOL_PRINT where needed.
(struct yyGLRState): Add some comments.
(struct yySemanticOption): Add some comments.
(union yyGLRStackItem): Add comment.
(yymergeOptionSets): Correct this to properly perform the union,
avoiding infinite reported by Michael Rosien.
Update comment.
* tests/glr-regression.at: Add test for GLR merging error reported
by M. Rosien.
tests/glr-regression.at: Add new regression test for the problems
reported by S. Eken with semantic values for embedded rules.
Update copyright notice.
tests/testsuite.at, tests/Makefile.am: Rename glr-regr1.at to
glr-regression.at.
