This changes the traces from
Reading a token:
Now at end of input.
to
Reading a token:
Next token is token $end (7FFEE56E6474)
which is ok. Actually it is even better, as it gives the location
when locations are enabled, and is clearer when rules explicitly use
the EOF token.
* data/skeletons/lalr1.d (yytranslate_): Handle eof here, as is done
in lalr1.cc.
Bison used to feature %raw, documented as follows:
@item %raw
The output file @file{@var{name}.h} normally defines the tokens with
Yacc-compatible token numbers. If this option is specified, the
internal Bison numbers are used instead. (Yacc-compatible numbers start
at 257 except for single character tokens; Bison assigns token numbers
sequentially for all tokens starting at 3.)
Unfortunately, as far as I can tell, it never worked: token numbers
are indeed changed in the generated tables (from external token number
to internal), yet the code was still applying the mapping from
external token numbers to internal token numbers.
This commit reintroduces the feature as it was expected to be.
* data/skeletons/bison.m4 (b4_token_format): When api.token.raw is
enabled, use the internal token number.
* data/skeletons/yacc.c (yytranslate): Don't emit if api.token.raw is
enabled.
(YYTRANSLATE): Adjust.
The CI, with CC='gcc-7 -fsanitize=undefined,address
-fno-omit-frame-pointer', reports:
calc.cc:1652:50: runtime error: load of value 190, which is not a valid value for type 'bool'
../../tests/calc.at:867: cat stderr
--- expout 2019-09-05 20:30:37.887257545 +0000
+++ /home/travis/build/bison-3.4.1.72-79a1-dirty/_build/tests/testsuite.dir/at-groups/438/stdout 2019-09-05 20:30:37.887257545 +0000
@@ -1 +1,2 @@
syntax error
+calc.cc:1652:50: runtime error: load of value 190, which is not a valid value for type 'bool'
438. calc.at:867: 438. Calculator glr.cc (calc.at:867): FAILED (calc.at:867)
The problem is that yylookaheadNeeds is not initialized in
yyinitStateSet, and when it is copied, the value is not 0 or 1.
* data/skeletons/glr.c (yylookaheadNeeds): Initialize yylookaheadNeeds.
Implement lookahead correction (LAC) for the C++ skeleton. LAC is a
mechanism to make sure that we report the correct list of expected
tokens if a syntax error occurs. So far, LAC was only supported for
the C skeleton "yacc.c".
* data/skeletons/lalr1.cc: Add LAC support.
* doc/bison.texi: Update.
* data/skeletons/yacc.c (yysyntax_error): Change the nesting of `m4`
conditions slightly to make it more readable.
The generated C code stays unchanged.
* configure.ac (DCFLAGS): Pass -g.
* data/skeletons/d.m4 (b4_locations_if): Remove, let bison.m4's one do
its job.
* data/skeletons/lalr1.d (position): Leave filename empty by default.
(position::toString): Don't print empty file names.
(location::this): New ctor.
(location::toString): Match the implementations of C/C++.
(yy_semantic_null): Leave undefined, the previous implementation does
not compile.
* tests/calc.at: Improve the implementation for D.
Enable more checks, in particular using locations.
* tests/local.at (AT_YYERROR_DEFINE(d)): Fix its implementation.
* data/skeletons/lalr1.java: Use more conventional function names for
Java.
Prefer < and <= to => and >.
Use the same approach for m4 quotation as in the other skeletons.
Fix indentation issues.
* tests/calc.at, tests/java.at, tests/javapush.at: Fix quotation style.
(main): Use 'args', not 'argv', the former seems more conventional and
is used elsewhere in Bison.
Prefer character literals to integers to denote characters.
* examples/java/Calc.y: Likewise.
This makes reading the trace slightly easier. It would be very nice
to highlight the "big steps", especially reductions. But this is a
private experiment: do not use it.
* data/diagnostics.css (value): New.
* src/parse-gram.y: Use no delimiters and no c quotation for strings
to facilitate debugging.
(tron, troff, TRACE): New.
Not very elegant, but until there is support for printf-formats in
libtextstyle, it shall be enough.
Reported by Balázs Scheidler.
* data/skeletons/c.m4 (b4_location_type_define): Use api.location.type
if defined.
* doc/bison.texi: Document it.
* tests/local.at (AT_C_IF, AT_LANG_CASE): New.
Support Span in C.
* tests/calc.at (Span): Convert it to be usable in C and C++.
Check api.location.type with yacc.c and glr.c.
Currently, with --no-lines, instead of "#line file line\n", we emit
"\n". Let's emit nothing.
* data/skeletons/bison.m4 (b4_syncline): Emit at end-of-line when enabled.
* data/skeletons/bison.m4, data/skeletons/c.m4, data/skeletons/glr.cc,
* data/skeletons/lalr1.cc, src/output.c: Use dnl after b4_syncline to
avoid spurious empty lines.
* tests/synclines.at (Sync Lines): Make sure that --no-lines is like
grep -v #line.
* tests/calc.at: Make sure that a rich grammar file behaves properly
with %no-lines.
Currently we use the syncline to report errors about a symbol's
destructor/printer. This is not accurate (only file and line), and
this is incorrect: the file name is double quotes (a recent change,
needed to make sure we escape properly double quotes in it). And
worst of all: with --no-line, b4_syncline expands to nothing.
Rather, push the locations into the backend, and use them.
* src/muscle-tab.h, src/muscle-tab.c (muscle_location_grow): Make it
public.
* src/output.c (prepare_symbol_definitions): Use it to pubish the
location of the printer and destructor.
* data/skeletons/lalr1.java: Use complain_at instead of complain.
* tests/java.at (Java invalid directives): Adjust expectations.
* data/skeletons/bison.m4 (b4_symbol_action_location): Remove.
We should not use b4_syncline this way.
Fatal errors are inconvenient, and should be reserved to cases where
we cannot continue. Here, it could even be warnings actually: these
directives will simply be ignored.
* data/skeletons/lalr1.java: Prefer error (b4_complain) to fatal
errors (b4_fatal).
* tests/java.at (Java invalid directives): New.
The timevar and bitset modules now use the c99 module which causes
$CXX to now include -std=gnu++11 when possible. Unfortunately, G++
4.7 does not implement [[noreturn]] in C++11 mode, so our tests of
glr.cc (which uses _Noreturn) fail with
input.cc:954:1: error: expected unqualified-id before '[' token
right before [[noreturn]]. 4.8 works fine.
* data/skeletons/c.m4 (b4_attribute_define): Do not use [[noreturn]]
with GCC 4.7.
The variable spec_defines_file denotes the name of the generated
header. Its name is derived from --defines/%defines, whose name in
turn is derived from the fact that the header, in Yacc, contained the
Not only does the header now contain a lot more than just the token
definitions, but we no longer even generate macros, but an enum...
Let's modernize our vocabulary.
* src/files.h, src/files.c (spec_defines_file): Rename as...
(spec_header_file): this.
Currently when --defines is used, we generate a header, and paste an
exact copy of it into the generated parser implementation file. Let's
provide a means to #include it instead.
We don't do it by default because of the Autotools' ylwrap. This
program wraps invocations of yacc (that uses a fixed output name:
y.tab.c, y.tab.h, y.output) to support a more modern naming
scheme (dir/foo.y -> dir/foo.tab.c, dir/foo.tab.h, etc.). It does
that by renaming the generated files, and then by running sed to
propagate these renamings inside the files themselves.
Unfortunately Automake's Makefiles uses Bison as if it were Yacc (with
--yacc or with -o y.tab.c) and invoke bison via ylwrap. As a
consequence, as far as Bison is concerned, the output files are
y.tab.c and y.tab.h, so it emits '#include "y.tab.h"'. So far, so
good. But now ylwrap processes this '#include "y.tab.h"' into
'#include "dir/foo.tab.h"', which is not guaranteed to always work.
So, let's do the Right Thing when the output file is not y.tab.c, in
which case the user should %define api.header.include. Binding this
behavior to --yacc is tempting, but we recently told people to stop
using --yacc (as it also enables the Yacc warnings), but rather to use
-o y.tab.c.
Yacc.c is the only skeleton concerned: all the others do include their
header.
* data/skeletons/yacc.c (b4_header_include_if): New.
(api.header.include): Provide a default value when the output is not
y.tab.c.
* src/parse-gram.y (api.header.include): Define.
* data/skeletons/yacc.c: here.
This is more logical for the time stamps, but it's also required by
following patches: the shared declarations are also in charge of
handling api.value.type=union. So far, they are run in the
implementation file in both cases (with or without header). But if we
run them only in the header, then the implementation file is emited
with incorrect support for api.value.type=union.
Arguably we should not have such dependencies. This is because we
have side-effects in our backend (redefining the symbols' type and
type_tag). In the future we should find a better solution for this,
without sacrificing the independence of the backend from bison
itself (i.e., I don't think we should handle api.value.type=union in
bison, leave it to m4).
Currently we generate things like:
#line 683 "src/parse-gram.y" /* yacc.c:316 */
The first part is of course very important: compilers point the users
to their grammar file rather than into the generated parser. The
second part points to the place in the skeletons that generated this
piece of code.
This dependency on the Bison skeletons generates lots of useless 'git
diff'. This location is useless for the regular user (who does not
care about the skeletons) and is actually not useful for Bison
developpers too (I never used this to locate the code in skeletons
that generated output). So disable it completely. If someone thinks
this was actually useful, a %define variable should be provided to
control the level of verbosity of '#line', in replacement of
--no-lines.
So now, generate:
#line 683 "src/parse-gram.y"
* data/skeletons/bison.m4 (b4_sync_end): Emit nothing.
* configure.ac (DCFLAGS): Define.
* tests/atlocal.in: Receive it.
* data/skeletons/d.m4 (api.parser.class): Remove spurious YY.
* data/skeletons/lalr1.d (yylex): Return an int instead of a
YYTokenType, so that we can use characters as tokens.
* examples/d/calc.y: Adjust.
* tests/local.at: Initial support for D.
(AT_D_IF, AT_DATA_GRAMMAR(D), AT_YYERROR_DECLARE(d))
(AT_YYERROR_DECLARE_EXTERN(d), AT_YYERROR_DEFINE(d))
(AT_MAIN_DEFINE(d), AT_COMPILE_D, AT_LANG_COMPILE(d), AT_LANG_EXT(d)):
New.
* tests/calc.at: Initial support for D.
* tests/headers.at
While hacking on the computation of the automaton, I had yystate being
equal to -1, and the parser loops. Let's catch this when
parser.assert is enabled.
* data/skeletons/yacc.c (YY_ASSERT): New.
Use it.
Not using the name YYASSERT, to make it clear that this is private.
glr.c should probably move to YY_ASSERT too.
Also, while at it, report 'Entering state...' even before growing the
stacks.
Since Bison 3.3, semantic values in rule actions (i.e., '$...') are
passed to the m4 backend as the symbol number. Unfortunately, when
there are unused symbols, the symbols are renumbered _after_ the
numbers were used in the rule actions. As a result, the evaluation of
the skeleton failed because it used non existing symbol numbers.
Which is the happy scenario: we could use numbers of other existing
symbols...
Reported by Balázs Scheidler.
http://lists.gnu.org/archive/html/bug-bison/2019-01/msg00044.html
Translating the rule actions after the symbol renumbering moves too
many parts in bison. Relying on the symbol identifiers is more
troublesome than it might first seem: some don't have an
identifier (tokens with only a literal string), some might have a
complex one (tokens with a literal string with characters special for
M4). Well, these are tokens, but nterms also have issues: "dummy"
nterms (for midrule actions) are named $@32 etc. which is risky for
M4.
Instead, let's simply give M4 the mapping between the old numbers and
the new ones. To avoid confusion between old and new numbers, always
emit pre-renumbering numbers as "orig NUM".
* data/README: Give details about "orig NUM".
* data/skeletons/bison.m4 (__b4_symbol, _b4_symbol): Resolve the
"orig NUM".
* src/output.c (prepare_symbol_definitions): Pass nterm_map to m4.
* src/reduce.h, src/reduce.c (nterm_map): Extract it from
nonterminals_reduce, to make it public.
(reduce_free): Free it.
* src/scan-code.l (handle_action_dollar): When referring to a nterm,
use "orig NUM".
* tests/reduce.at (Useless Parts): New, based Balázs Scheidler's
report.
Reported by Derek Clegg
http://lists.gnu.org/archive/html/bug-bison/2019-01/msg00021.html
aux/parser-internal.h:429:12: error: 'syntax_error' has no out-of-line virtual
method definitions; its vtable will be emitted in every translation unit
[-Werror,-Wweak-vtables]
struct syntax_error : std::runtime_error
To avoid this warning, we need syntax_error to have a virtual function
defined in a compilation unit. Let it be the destructor. To comply
with C++98, this dtor should be 'throw()'. Merely making YY_NOEXCEPT
be 'throw()' in C++98 triggers
errors (http://lists.gnu.org/archive/html/bug-bison/2019-01/msg00022.html),
so let's introduce YY_NOTHROW and flag only ~syntax_error with it.
Also, since we now have an explicit dtor, we need to provide an copy
ctor.
* configure.ac (warn_cxx): Add -Wweak-vtables.
* data/skeletons/c++.m4 (YY_NOTHROW): New.
(syntax_error): Declare the dtor, and define the copy ctor.
* data/skeletons/glr.cc, data/skeletons/lalr1.cc (~syntax_error):
Define.
