We used to emit:
/** Token number,to be returned by the scanner. */
static final int NUM = 258;
/** Token number,to be returned by the scanner. */
static final int NEG = 259;
with no space after the comma. Fix that.
* data/skeletons/bison.m4 (b4_token_format): Quote where appropriate.
Unfortunately in the Java skeleton the user cannot override the way
locations are displayed, and locations don't know the structure of the
positions. So they cannot implement the tricks used in the C/C++
skeletons to display "1.1" instead of "1.1-1.2".
* tests/local.at (Java): Add support for column tracking in the
locations, as we did in examples/java/calc.
* tests/calc.at: Use AT_CALC_YYLEX.
Soon calculator tests for Java will move from java.at to calc.at.
Which implies improving the Java testing infrastructure in
local.at (for instance really tracking columns in positions, not just
token number). Detach java.at from local.at.
* tests/java.at (AT_JAVA_POSITION_DEFINE_OLD): New.
Use it.
* examples/java/calc/Calc.y: The StreamTokenizer cannot "peek" for the
next character, it reads it, and keeps it for the next call. So the
current location is one passed the end of the current token. To avoid
this, keep the previous position, and use it to end the current token.
* examples/java/calc/Calc.test: Adjust.
The C, C++ and D skeletons used to show the stack right after popping
the stack during the reduction. Now that the stack is printed after
reaching a new state, that has become useless:
Entering state 1
Stack now 0 1
Reducing stack by rule 5 (line 83):
$1 = token "number" (1)
-> $$ = nterm exp (1)
Stack now 0
Entering state 8
Stack now 0 8
Remove the "Stack now 0" line.
* data/skeletons/lalr1.cc, data/skeletons/lalr1.d,
* data/skeletons/lalr1.java, data/skeletons/yacc.c:
Here.
Currently, if we have long rules and series of shift, we stack states
without showing stack. Let's be more incremental, and do how the Java
skeleton does.
* data/skeletons/lalr1.cc, data/skeletons/lalr1.d,
* data/skeletons/yacc.c:
Here.
Adjust test cases.
* tests/torture.at (AT_DATA_STACK_TORTURE): Disable stack traces: this
test produces a very large stack, and showing the stack each time we
shift a token goes quadatric.
The Java skeleton displays
Reading a token:
Next token is token "number" (1)
while the other display
Reading a token: Next token is token "number" (1)
When generating logs in the scanner, the first part is separated from
the second, and the end of the scanner logs have the second part
pasted in. So let's propagate the Java way, but with the colon.
* data/skeletons/glr.c, data/skeletons/lalr1.cc, data/skeletons/lalr1.d,
* data/skeletons/lalr1.java, data/skeletons/yacc.c: Do it.
Adjust test cases and doc.
* tests/local.at (AT_LANG_MATCH): New.
(AT_YYERROR_DECLARE(java), AT_YYERROR_DECLARE_EXTERN(java)): New.
* tests/calc.at: The grammar file for Java is quite different for the
others, and continuing to assemble it from pieces makes the grammar
file hard to understand. Let's also dispatch on the language to
assemble it, and isolate Java from the others.
Most of this comes from java.at.
This example, so far, was tracking the current token number, not the
current column number. This is not nice for an example...
* examples/java/Calc.y (PositionReader): New.
Use it.
* examples/java/Calc.test: Check the output.
AFAICT, autoboxing/unboxing was added in Java 5 (September 30, 2004).
I think we can afford to use it. It should help us merge some Java
tests with the main ones.
However, beware that != does not unbox: it compares the object
addresses.
* examples/java/Calc.y, tests/java.at: Simplify.
* examples/java/Calc.test, tests/java.at: Improve tests.
When building the test cases, emitting code in the epilogue is very
constraining. Let's make it simpler thanks to %code epilogue.
However, I don't want to document this: it is bad style to use it (we
should avoid having too many ways to write the same thing,
TI!MTOWTDI), just put your code in the true epilogue section.
* data/skeletons/glr.c, data/skeletons/lalr1.d, data/skeletons/lalr1.java,
* data/skeletons/yacc.c: Implement support for %code epilogue.
Remove useless comments.
* tests/calc.at, tests/java.at: Simplify.
Modeled after what was done in yacc.c, yet somewhat different since
yyGLRStack play the role of the full yyparse_context_t. It will
probably be defined as a alias, to make interfaces compatible.
LAC is not supported here. And is somewhat tricky.
* data/skeletons/glr.c (yyexpected_tokens, yysyntax_error_arguments): New.
* data/skeletons/glr.c: Move the handling of error messages after the
definition of types. Especially after yyGLRStack, which we will need
in forthcoming patches.
Simplify: yyGLRStack is defined at this point.
The PDF output is more consistent: some nodes were not pointed to with
their title. The HTML output becomes "see section Foo" instead of
"see Foo", but this should be addressed in the next Texinfo release.
Info output is simplified, as it uses only the node name and not its
title. But it's considered easier to read this way.
See https://lists.gnu.org/r/help-texinfo/2020-01/msg00031.html.
* doc/bison.texi: Set @xrefautomaticsectiontitle on.
Simplify all uses of ref.
Add an example to demonstrate the use of push parser. I'm pleasantly
surprised: parse.error=detailed works like a charm with push parsers.
* examples/c/local.mk, examples/c/pushcalc/Makefile
* examples/c/pushcalc/README.md, examples/c/pushcalc/calc.test,
* examples/c/pushcalc/calc.y, examples/c/pushcalc/local.mk:
New.
* examples/c/calc/calc.y: Restore to its original state, with
parse.error=detailed instead of parse.error=custom (this example
should be simple).
* examples/c/calc/calc.test: Check syntax errors.
* examples/c/lexcalc/parse.y: Add comments.
We used to display the unexpected token first:
$ bison foo.y
foo.y:1.8-13: error: syntax error, unexpected %token, expecting character literal or identifier or <tag>
1 | %token %token
| ^~~~~~
GCC uses a different format:
$ gcc-mp-9 foo.c
foo.c:1:5: error: expected identifier or '(' before ')' token
1 | int()()()
| ^
and so does Clang:
$ clang-mp-9.0 foo.c
foo.c:1:5: error: expected identifier or '('
int()()()
^
1 error generated.
They display the unexpected token last (or not at all). Also, they
don't waste width with "syntax error". Let's try that. It gives, for
the same example as above:
$ bison foo.y
foo.y:1.8-13: error: expected character literal or identifier or <tag> before %token
1 | %token %token
| ^~~~~~
* src/complain.h, src/complain.c (syntax_error): New.
* src/parse-gram.y (yyreport_syntax_error): Use it.
As a test case, support translations in Bison itself.
* src/parse-gram.y: Mark the translatable tokens.
While at it, use clearer names.
* tests/input.at: Adjust expectations.
Don't force callers of location_caret to have to deal with flags that
disable it.
* src/location.h, src/location.c (location_caret)
(location_caret_suggestion): Early return if disabled.
* src/complain.c: Simplify.
Some users would like to avoid having to "parse" the *.y file to find
the strings to translate. Let's issue the translatable tokens with N_
to allow "parsing" the generated parsers instead.
See
https://lists.gnu.org/archive/html/bison-patches/2019-01/msg00015.html
* src/output.c (prepare_symbol_names): Issue symbol_names with N_()
markup.
In addition to
%token NUM "number"
accept
%token NUM _("number")
in which case the token will be translated in error messages.
Do not use _() in the output if there are no translatable tokens.
* src/symtab.h, src/symtab.c (symbol): Add a 'translatable' member.
* src/parse-gram.y (TSTRING): New token.
(string_as_id.opt): Replace with...
(alias): this.
Use it.
* src/scan-gram.l (SC_ESCAPED_TSTRING): New start conditions, to match
TSTRINGs.
* src/output.c (prepare_symbols): Define b4_translatable if there are
translatable strings.
* data/skeletons/glr.c, data/skeletons/lalr1.cc,
* data/skeletons/yacc.c (yytnamerr): Receive b4_translatable, and use it.
* src/output.c (escape_trigraphs, xescape_trigraphs): New.
(prepare_symbol_names): Use it.
* tests/regression.at: Check the handling of trigraphs with
parse.error = detailed.
"detailed" error messages are almost like "verbose", except that we
don't double escape them, they don't get inner quotes, we don't use
yytnamerr, and we hide the table.
"custom" is exposed with the "detailed" tokens, not the "verbose"
ones: they are not double-quoted.
Because there's a risk that some people use yytname even without
"verbose", let's keep yytname (instead of yys_name) in "simple"
parse.error.
* src/output.c (prepare_symbol_names): Be ready to output symbol names
unquoted.
(prepare_symbol_names): Output both the old tname table, and the new
symbol_names one.
* data/skeletons/bison.m4: Accept 'detailed'.
* data/skeletons/yacc.c: When parse.error is 'detailed', don't emit
yytname and yytnamerr, just yysymbol_name with the table inside.
* tests/calc.at: Adjust.
Only parse.error verbose and simple will get the original yytname: the
other options will rely on a different table. So let's move on top of
the yysymbol_name function.
* data/skeletons/c.m4 (yy_symbol_print): Use yysymbol_name.
* data/skeletons/glr.c (yytokenName): Rename as...
(yysymbol_name): this.
The change of naming scheme is unfortunate, but it's definitely glr.c
which is "wrong".
Currently yy_symbol_print is defined before yytokenName, although it
should use it instead of read yytname directly. Move blocks around to
avoid this.
* data/skeletons/glr.c (yy_symbol_print): Move its definition after
that of yytokenName.
See https://github.com/nemequ/icc-travis/issues/15.
Thanks to Jeff Hammond and Evan Nemerson for their help.
* configure.ac (warn_common): Disable dubious warnings.
* .travis.yml: Use ICC again.
Currently we get warnings with GCC 4.8 when running the
maintainer-check-g++ tests:
143. skeletons.at:85: testing Installed skeleton file names ...
../../tests/skeletons.at:120: COLUMNS=1000; export COLUMNS; bison --color=no -fno-caret --skeleton=yacc.c -o input-cmd-line.c input-cmd-line.y
../../tests/skeletons.at:121: $CC $CFLAGS $CPPFLAGS $LDFLAGS -o input-cmd-line input-cmd-line.c $LIBS
stderr:
input-cmd-line.c: In function 'int yysyntax_error(long int*, char**, const yyparse_context_t*)':
input-cmd-line.c:977:52: error: conversion to 'int' from 'long int' may alter its value [-Werror=conversion]
YYSIZEOF (yyarg) / YYSIZEOF (*yyarg));
^
cc1plus: all warnings being treated as errors
stdout:
../../tests/skeletons.at:121: exit code was 1, expected 0
and
429. calc.at:823: testing Calculator parse.error=custom %locations api.prefix={calc} ...
../../tests/calc.at:823: COLUMNS=1000; export COLUMNS; bison --color=no -fno-caret -Wno-deprecated -o calc.c calc.y
../../tests/calc.at:823: $CC $CFLAGS $CPPFLAGS $LDFLAGS -o calc calc.c $LIBS
stderr:
calc.y: In function 'int yyreport_syntax_error(const yyparse_context_t*)':
calc.y:157:58: error: conversion to 'int' from 'long unsigned int' may alter its value [-Werror=conversion]
int n = yysyntax_error_arguments (ctx, arg, sizeof arg / sizeof *arg);
^
cc1plus: all warnings being treated as errors
stdout:
../../tests/calc.at:823: exit code was 1, expected 0
We could use a cast to avoid the warning, but it becomes too
cluttered. We can also use YYPTRDIFF_T, but that forces the user to
use YYPTRDIFF_T too, although this is an array of tokens, which is
limited by YYNTOKENS, an int. So let's completely avoid this warning.
* data/skeletons/yacc.c, tests/local.at (yyreport_syntax_error): Avoid
relying on sizeof to compute the array capacity.
Enhance the calculator tests: show that passing arguments to yyerror
works.
* tests/calc.at: Add a new parse-param, nerrs, which counts the number
of syntax errors in a run.
* tests/local.at: Adjust to handle the new 'nerrs' argument, when
present.
The custom error reporting function show sees the user's additional
arguments. Let's experiment with passing them as arguments to
yyreport_syntax_error, but maybe storing them in the context would be
a bettter alternative.
* data/skeletons/yacc.c (yyreport_syntax_error): Handle the
parse-params.
* tests/calc.at, tests/local.at: Adjust.
Currently the parse-params are tested in calc.at by checking that the
global variable and the parse-params have the same value. But it does
not check that value, that could remain being 0 just as well.
* tests/calc.at: Don't define the params when they are not used.
Check the final value of result and count.
Also, do count the number of line of logs.
Provide users with a means to query for the currently allowed tokens.
Could be used for autocompletion for instance.
* data/skeletons/yacc.c (yyexpected_tokens): New, extracted from
yysyntax_error_arguments.
* examples/c/calc/calc.y (PRINT_EXPECTED_TOKENS): New.
Use it.
We use a different format to check parse.error custom. Compute the
"verbose" one from it instead of forcing the test author to provide
the various formats of expected error messages.
* tests/calc.at (_AT_CHECK_CALC_ERROR): Handle this transformation
when needed.
Simplify callers.
* tests/local.at (AT_ERROR_CUSTOM_IF, AT_ERROR_VERBOSE_IF)
(AT_ERROR_SIMPLE_IF): New.
(AT_YYERROR_DEFINE(c)): Generate yyreport_syntax_error.
* tests/calc.at (_AT_CHECK_CALC_ERROR): Accept custom error messages
as additional test case.
Use it.
Add a new test case for %define parse.error custom.
When parse.error is custom, let users define a yyreport_syntax_error
function, and use it.
* data/skeletons/bison.m4 (b4_error_verbose_if): Accept 'custom'.
* data/skeletons/yacc.c: Implement it.
* examples/c/calc/calc.y: Experiment with it.
That allows users to cover more cases, such as easily filtering some
arguments they don't want to expose. But they now have to call
yysymbol_name explicitly.
* data/skeletons/yacc.c (yysyntax_error_arguments, yysyntax_error):
Deal with symbol numbers instead of symbol names.
Isolate a function that returns the list of expected and unexpected
tokens. It will be exposed to users willing to customize their error
messages.
* data/skeletons/yacc.c (yyparse_context_t): New.
(yyerror_message_arguments): New, extracted from yysyntax_error.
When not empty, AT_PARSE_PARAMS was guaranteed to end with a comma.
Remove the trailing comma, so that we can use AT_PARSE_PARAMS at the
end of the arguments, not only at the beginning.
* tests/local.at: here.
Unfortunately, m4_append relies on the macro not being defined whereas
we would have preferred it to check for emptiness. So use
m4_define/m4_undefine instead of m4_pushdef/m4_popdef.
Let's have C be the reference, and match it elsewhere. Maybe C is too
verbose and some adjustments are needed, but then that would be done
in another batch of patches.
* data/skeletons/lalr1.cc: Print the stack once we popped after
YYERROR, and before emptying the stack at the end of parsing.
Currently the C and C++ parse traces differ in the order in which the
stack is displayed: bottom up in C, top down in C++. Let's stick to
the C order.
* data/skeletons/stack.hh (stack::iterator, stack::const_iterator)
(begin, end): Be forward, not backward.
Provide the users with a public API to get the name of the tokens. A
thin wrapper around yytname.
* data/skeletons/yacc.c (yysymbol_name): New.
Use it.
I was hoping it would help us catch warnings when char is
unsigned (see 78bb152a63), but it does
not seem to help. It's a pity that the compiler is the same all over
the place, I would have preferred testing others.
* .travis.yml: here.
Akim Demaille